Estimating the Effects of Global Patent Protection in...

38
Estimating the Effects of Global Patent Protection in Pharmaceuticals: A Case Study of Quinolones in India By SHUBHAM CHAUDHURI,PINELOPI K. GOLDBERG, AND PANLE JIA* Under the Agreement on Trade-Related Intellectual Property Rights, the World Trade Organization members are required to enforce product patents for pharma- ceuticals. In this paper we empirically investigate the welfare effects of this requirement on developing countries using data for the fluoroquinolones subseg- ment of the systemic anti-bacterials segment of the Indian pharmaceuticals market. Our results suggest that concerns about the potential adverse welfare effects of TRIPS may have some basis. We estimate that the withdrawal of all domestic products in this subsegment is associated with substantial welfare losses to the Indian economy, even in the presence of price regulation. The overwhelming portion of this welfare loss derives from the loss of consumer welfare. (JEL F13, L65, O34) Under the Agreement on Trade-Related Intel- lectual Property Rights (TRIPS)—finalized during the Uruguay round of multilateral trade negotia- tions in 1995—nations must, as a condition of membership in the World Trade Organization (WTO), recognize and enforce product patents in all fields of technology, including pharmaceuti- cals. At the time the TRIPS agreement went into effect, many low- and middle-income countries made an exception for pharmaceuticals, even if they recognized product patents in other areas, because low-cost access to life-saving drugs and essential medicines was deemed to be an overrid- ing public policy priority. To meet their obliga- tions under TRIPS, however, these countries had to introduce or amend their patent legislation to include pharmaceutical product patents, with the transition and least-developed economies having until 2005 to do so. The negotiations leading up to TRIPS, and in particular the provisions relating to pharmaceu- ticals, were highly contentious. Though more than ten years have passed since TRIPS was finalized, there continues to be considerable controversy and debate regarding its merits. The main point of contention is the claim made by governments of many poor developing econo- mies that unqualified patent protection for phar- maceuticals will result in substantially higher prices for medicines, with adverse conse- quences for the health and well-being of their citizens. Countering this claim, research-based global pharmaceutical companies, which have potentially lost billions of dollars because of patent infringement by third world firms that have reverse-engineered their products, argue that the introduction of product patents is un- likely to raise prices significantly because most patented products have many therapeutic sub- stitutes. Moreover, they claim that the absence of patent protection has served as a disincentive to engage in research on diseases that dispro- portionately afflict the world’s poor, implying that patent protection for pharmaceuticals will actually benefit less-developed economies by stimulating innovation and transfer of technology. Given the scope of TRIPS and the intensity of the accompanying debate, it is remarkable how * Chaudhuri: World Bank, 1818 H Street, NW, Wash- ington, DC 20433 (e-mail: [email protected]); Goldberg: Department of Economics, Yale University, Box 208264, New Haven, CT 06520, National Bureau of Eco- nomic Research, and Bureau for Research in Economic Analysis and Development (e-mail: penny.goldberg@ yale.edu); Jia: Department of Economics, Massachusetts Institute of Technology, 50 Memorial Drive, Cambridge, MA 02142 (e-mail: [email protected]). We would like to thank Isher Ahluwalia, former director, Indian Council for Re- search on International Economic Relations, for providing us with access to the primary data used in this study. We also thank Abhijit Banerjee, Steve Berry, Anne Case, Angus Deaton, Phil Haile, Elhanan Helpman, Michael Keane, Jenny Lanjouw, Dick Nelson, Ariel Pakes, Chris Paxson, Fiona Scott-Morton, T. N. Srinivasan, a coeditor, and the referees for many detailed and constructive comments on an earlier draft, and participants at several seminars in the United States and Europe for helpful feedback. 1477

Transcript of Estimating the Effects of Global Patent Protection in...

Page 1: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

Estimating the Effects of Global Patent Protection inPharmaceuticals A Case Study of Quinolones in India

By SHUBHAM CHAUDHURI PINELOPI K GOLDBERG AND PANLE JIA

Under the Agreement on Trade-Related Intellectual Property Rights the WorldTrade Organization members are required to enforce product patents for pharma-ceuticals In this paper we empirically investigate the welfare effects of thisrequirement on developing countries using data for the fluoroquinolones subseg-ment of the systemic anti-bacterials segment of the Indian pharmaceuticals marketOur results suggest that concerns about the potential adverse welfare effects ofTRIPS may have some basis We estimate that the withdrawal of all domesticproducts in this subsegment is associated with substantial welfare losses to theIndian economy even in the presence of price regulation The overwhelming portionof this welfare loss derives from the loss of consumer welfare (JEL F13 L65 O34)

Under the Agreement on Trade-Related Intel-lectual Property Rights (TRIPS)mdashfinalized duringthe Uruguay round of multilateral trade negotia-tions in 1995mdashnations must as a condition ofmembership in the World Trade Organization(WTO) recognize and enforce product patents inall fields of technology including pharmaceuti-cals At the time the TRIPS agreement went intoeffect many low- and middle-income countriesmade an exception for pharmaceuticals even ifthey recognized product patents in other areasbecause low-cost access to life-saving drugs andessential medicines was deemed to be an overrid-ing public policy priority To meet their obliga-tions under TRIPS however these countries had

to introduce or amend their patent legislation toinclude pharmaceutical product patents with thetransition and least-developed economies havinguntil 2005 to do so

The negotiations leading up to TRIPS and inparticular the provisions relating to pharmaceu-ticals were highly contentious Though morethan ten years have passed since TRIPS wasfinalized there continues to be considerablecontroversy and debate regarding its merits Themain point of contention is the claim made bygovernments of many poor developing econo-mies that unqualified patent protection for phar-maceuticals will result in substantially higherprices for medicines with adverse conse-quences for the health and well-being of theircitizens Countering this claim research-basedglobal pharmaceutical companies which havepotentially lost billions of dollars because ofpatent infringement by third world firms thathave reverse-engineered their products arguethat the introduction of product patents is un-likely to raise prices significantly because mostpatented products have many therapeutic sub-stitutes Moreover they claim that the absenceof patent protection has served as a disincentiveto engage in research on diseases that dispro-portionately afflict the worldrsquos poor implyingthat patent protection for pharmaceuticals willactually benefit less-developed economies bystimulating innovation and transfer of technology

Given the scope of TRIPS and the intensity ofthe accompanying debate it is remarkable how

Chaudhuri World Bank 1818 H Street NW Wash-ington DC 20433 (e-mail schaudhuriworldbankorg)Goldberg Department of Economics Yale University Box208264 New Haven CT 06520 National Bureau of Eco-nomic Research and Bureau for Research in EconomicAnalysis and Development (e-mail pennygoldbergyaleedu) Jia Department of Economics MassachusettsInstitute of Technology 50 Memorial Drive CambridgeMA 02142 (e-mail pjiamitedu) We would like to thankIsher Ahluwalia former director Indian Council for Re-search on International Economic Relations for providingus with access to the primary data used in this study Wealso thank Abhijit Banerjee Steve Berry Anne Case AngusDeaton Phil Haile Elhanan Helpman Michael KeaneJenny Lanjouw Dick Nelson Ariel Pakes Chris PaxsonFiona Scott-Morton T N Srinivasan a coeditor and thereferees for many detailed and constructive comments on anearlier draft and participants at several seminars in theUnited States and Europe for helpful feedback

1477

sparse is the evidence on which these divergentclaims are based1 Apart from the findings of asmall number of studies we refer to in moredetail below little is known about the extent towhich pharmaceutical prices in less-developedeconomies might increase with the introductionof product patents and the magnitude of theassociated welfare losses2 Past empirical stud-ies on the impact of patents on prices and inno-vative activity in various sectors includingpharmaceuticals have focused almost exclu-sively on developed economies Aside from thefact that none of these studies estimates welfareeffects the conclusions from these studies arenot directly pertinent to the TRIPS debate be-cause the structure of demand for pharmaceuti-cals in less-developed economies differs fromthat in developed economies in several criticalrespects3

Any assessment of the potential price andwelfare effects of TRIPS needs therefore to bebased on a better empirically grounded under-standing of the characteristics of demand andthe structure of markets for pharmaceuticals inpoor developing economies To what extent areconsumers willing to trade off lower prices forolder possibly less effective therapies Howdoes this vary across different therapeutic seg-ments Are consumers willing to pay a pre-mium for the pedigree and brand reputation ofproducts marketed by subsidiaries of foreignmultinationals How competitive are pharma-ceutical markets The welfare of consumers

depends on the pricing strategies and decisionsof pharmaceutical firms But these in turn de-rive from the firmsrsquo assessment of the structureof market demand If consumers are unwillingto pay substantially more for newer patenteddrugs for which there exist older possiblyslightly less effective generic substitutes theability of patent-holders to charge a premiumwill be limited As mentioned above there havebeen a few studies that carefully consideredthese issues and used explicit models of con-sumer and firm behavior to simulate the welfarelosses implied by patent protection4 Their find-ings are ultimately limited however by the factthat the simulations that are used to evaluate thepotential impact of patents are in each instancebased on assumptions about demand character-istics and market structure rather than on actualestimates of the relevant parameters

This paper takes a first step toward filling thisgap We provide the first rigorously derivedestimates of the possible impact of pharmaceu-tical product patents on prices and welfare in adeveloping economy Using detailed product-level data on monthly pharmaceutical pricesand sales over a two-year period from January1999 to December 2000 we estimate key priceand expenditure elasticities and supply-side pa-rameters for the fluoroquinolone (quinolonehenceforth5) segment of systemic antibacterials(ie antibiotics) in the Indian pharmaceuticalsmarket We chose this segment both because itcontains several products that were still underpatent in the United States during our sampleperiod and because antibiotics are importantfrom a public health policy point of view (com-pared to for example Prozac Viagra or otherlifestyle drugs that were also under patent pro-tection in the United States during this period)We then use our estimates to carry out counter-factual simulations of what prices profits (ofboth domestic firms and subsidiaries of foreignmultinationals) and consumer welfare wouldhave been had the quinolone molecules westudy been under patent in India as they were in

1 There is a sizeable theoretical literature on the welfareimpact of patent protection that generally finds that theeffects of patents in a multicountry setting are substantiallymore complicated than their respective effects in a singleclosed economy where both innovating firms and innova-tion beneficiaries (ie consumers) are of the same nation-ality (see William Nordhaus 1969 Judith C Chen andGene M Grossman 1990 Ishac Diwan and Dani Rodrick1991 Alan V Deardorff 1992 Elhanan Helpman 1993and Grossman and Edwin L-C Lai 2004) Empirical workin this area is however still in its infancy

2 Even less is known about the other central questionsrelevant to the TRIPS debate namely the extent to whichpharmaceutical research and product development prioritiesare likely to shift as a result of TRIPS and how large thewelfare benefits of any therapeutically innovative drugs thatresult from this shift are likely to be The only paper that hascarefully addressed such questions is Jean O Lanjouw andIain M Cockburn (2001)

3 For a representative example of estimation of pharma-ceuticals demand in developed countries see Sara F Ellisonet al (1997)

4 See for instance Pablo M Challu (1991) Carsten Fink(2000) Keith E Maskus and Denise E Konan (1994) JulioJ Nogues (1993) Arvind Subramanian (1995) and Ja-yashree Wattal (2000)

5 Technically the term ldquofluoroquinolonesrdquo refers to thelatest generation of quinolones Older quinolones however(eg nalidixic acid) have market shares close to zero

1478 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

the United States at the time The presence ofmany therapeutic substitutes within the antibi-otics segment make this product category idealfor investigating the claim that the presence ofclose substitutes will prevent drug prices fromrising once patent protection is enforced Ofcourse to the extent that our estimates refer toantibiotics they are not directly applicable toother pharmaceutical product categories thatmay have different demand structures For ex-ample a finding of large substitution effectstoward nonpatented products would not neces-sarily apply to a market segment with only fewor possibly no therapeutic substitutes Still afinding of limited substitution toward otherdrugs and associated large price increaseswould suggest that the effects of patent enforce-ment in other pharmaceutical segments withfewer therapeutic substitutes might be evenlarger

India provides a natural setting for our anal-ysis for a number of reasons It is a leadingexample of a low-income country that did notrecognize pharmaceutical product patents at thetime the TRIPS agreement went into effect Infact during the Uruguay round of negotiationsIndia led the opposition to the TRIPS articlesmandating pharmaceutical product patents Interms of the structure of demand India is aprototypical example of a low-income countrywith a large number of poor households whichbecause health insurance coverage is nonexist-ent have to meet all medical expenses out ofpocket Moreover the disease profile of theIndian population mirrors that of many otherlow-income countries and is considerably dif-ferent from that of most developed economiesLastly the domestic Indian pharmaceutical in-dustry which as of 2002 was the largest pro-ducer of generic drugs in the world in terms ofvolume is typical of that of many middle-in-come countries with large numbers of small andmedium-sized firms with significant imitativecapabilities producing and marketing drugs do-mestically that are under patent elsewhere

During the period covered by our data sev-eral molecules in the quinolone family were stillunder patent in the United States but productscontaining these molecules were being pro-duced and distributed in India by both a numberof domestic firms and a number of local subsid-iaries of foreign multinationals We aggregatethese products into a number of mutually exclu-

sive product groups where within each productgroup all products contain the same quinolonemolecule (eg ciprofloxacin or norfloxacin)and are produced by firms with the same do-mestic or foreign status We then estimate atwo-level demand system employing the Al-most Ideal Demand System (AIDS) specifica-tion of Angus S Deaton and John Muellbauer(1980b) in both levels The higher level corre-sponds to the allocation of expenditures tovarious subsegments within the systemic anti-bacterials segment of the market At the lowerlevel we estimate the parameters relevant forthe allocation of expenditures within the quin-olone subsegment to the various productgroups within this subsegment (eg foreignciprofloxacin domestic ciprofloxacin or do-mestic norfloxacin)

With these estimates in hand we turn to thecounterfactuals The basic counterfactual sce-narios we consider all involve the withdrawal ofone or more of the domestic quinolone productgroups from the market The idea here is thathad US patents for say ciprofloxacin beenrecognized in India all domestic products con-taining ciprofloxacin would not be present in themarket That would leave only the foreign cip-rofloxacin product group in the market Usingour estimates of the own cross-price and ex-penditure elasticities of the various productgroups as well as estimates for the upper andlower bounds of marginal costs of productionwe are able to simulate the prices and marketshares that would obtain under each of the sce-narios Moreover using the expenditure func-tion associated with the higher-level AIDSspecification we are able to calculate the wel-fare lossmdashmeasured in terms of the compensat-ing variation ie the additional expenditurethat the representative Indian consumer wouldneed to incur to maintain her utility level inthe face of the domestic product withdraw-al(s) and the accompanying price and marketshare changesmdashunder each of the counterfac-tual scenarios

Apart from the fact that our counterfactualsimulations are based on estimated rather thanassumed parameter values this paper buildsupon the earlier studies in two substantive and(it turns out) empirically important ways

First by accommodating the possibility thatconsumers may differentiate between domesticand foreign products even when these products

1479VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

contain the same patentable molecule we allowfor an additional channel through which theintroduction of product patents and the conse-quent withdrawal of domestic products may ad-versely affect consumers that is through theloss of product variety In contrast previousstudies on developing countries assume thatconsumers are indifferent between foreign anddomestic products that contain the same mole-cule What this implies is that any adverse wel-fare effects are realized only through increasedprices The difference is most evident if weconsider a scenario under which domestic prod-ucts are forced to withdraw from the marketbecause of the introduction of product patentsbut strict price regulations maintain prices atprepatent levels In our approach consumerswould still experience a welfare loss whereas inthe framework adopted in earlier studies such ascenario would entail no loss of welfare

Empirically the component of the consumerwelfare loss attributable to the reduction of va-riety from the withdrawal of domestic productsturns out to be significant We interpret thiscomponent as capturing primarily an ldquoease ofaccessrdquo effect due to differences in the market-ing and distribution networks domestic prod-ucts are more readily available to Indianconsumers than products produced by foreignsubsidiaries From a policy perspective thissuggests a possible role for compulsory licens-ing in addition to or in lieu of price regulationsince the latter by itself will not alleviate thewelfare loss due to loss of variety Alterna-tively one could argue thatmdashto the extent thatthe loss we attribute to the reduction of productvariety is due to the fact that the current productportfolios and distribution networks of foreignproducers are limitedmdashit is purely a transitionalphenomenon and should thus not be included inthe welfare calculations This is a controversialpoint we discuss in detail in the results sectionIf foreign firms respond to patent enforcementby investing in distribution networks or by us-ing licensing agreements with domestic firms tomake their products more readily available toIndian consumers the ldquoease of accessrdquo effectwould indeed diminish in importance in thelonger run though of course it could be signif-icant in the first years after patent enforcementWhether these investments will materializehowever is open to debate If TRIPS is accom-panied by price regulation in order to limit price

increases in poor developing countries the in-centives of multinationals to invest in marketingand distribution in these countries may dimin-ish At any rate to take into account the possi-bility that the welfare loss due to the reductionof variety is a temporary phenomenon we alsopresent a more conservative welfare loss esti-mate by subtracting the ldquoproduct varietyrdquo com-ponent from our total loss estimate This givesus a lower bound estimate that is due to priceincreases alone Though it is only about a thirdof our upper bound estimate this lower boundestimate is still very large in absolute termsrepresenting 24 percent of antibiotic sales in2000

A second and perhaps even more importantmethodological difference between this paperand earlier studies is that we allow for andflexibly estimate a range of cross-product-group and cross-molecule substitution effectsIn contrast cross-price effects are ignored inearlier studies To see why cross-price effectsare likely to alter estimated welfare effects sig-nificantly in this context imagine a scenariowhere the introduction of patents leads to mo-nopoly pricing in the market for a particularpatentable molecule If the markets for potentialsubstitutes are imperfectly competitive then theincrease in price in the original patentable mar-ket will lead to corresponding upward priceadjustments in the related markets as producersof substitute products reoptimize in the face ofthe increased demand for their products Themagnitude of any upward adjustments will nat-urally vary with the degree of competition inrelated markets and with the strength of thecross-price effects But as long as the cross-price effects are positive and related marketsare not perfectly competitive the loss of con-sumer surplus because of monopoly pricing inone market will be multiplied through the rippleeffects of upward price adjustments in relatedmarkets

If this were just a theoretical possibility itwould not be of much interest These multipliereffects however turn out to be substantial inour counterfactual scenarios Most strikinglythe estimated loss of consumer welfare from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenar-

1480 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

ios in each of which only one of the domesticproduct groups is withdrawn What this veryclearly indicates is that past studies that haveestimated the aggregate effects of patent protec-tion by adding up the losses estimated sepa-rately in each of a number of patentablemarkets may have substantially underestimatedthe magnitude of the consumer welfare lossesfrom the introduction of pharmaceutical productpatents

In absolute terms we estimate that in theabsence of any price regulation the prices offoreign patented products would rise between100 percent and 400 percent In the more real-istic case of some form of price regulation thatwould keep drug prices fixed at their pre-TRIPSlevel the total annual welfare losses to theIndian economy from the withdrawal of all fourdomestic product groups in the quinolone sub-segment would be on the order of Rs 1370billion or about 50 percent of the sales of theentire systemic antibacterials segment in 2000At the then-prevailing exchange rate this trans-lates into a figure of US$305 million Of thisamount foregone profits of domestic producersconstitute roughly Rs 23 billion or US$50 mil-lion The overwhelming portion of the totalwelfare loss therefore derives from the loss ofconsumer welfare

The welfare loss we estimate represents onlythe static costs of patent enforcement arisingfrom pricing distortions and reduction in prod-uct variety Our approach does not address thepotential dynamic benefits of innovations thatmay result from international property rightsprotection Nevertheless we believe that esti-mating these static costs is important wheneverthere is a radical change in policymdashwhichTRIPS represents for a good part of the devel-oping world Even if there is the potential forlong-term benefits knowledge of the short-runcosts is important for designing an appropriatepolicy response that will potentially mitigate theadverse short-run impact Having said that itis worth noting that according to our esti-mates the total profit gains of patent enforce-ment to foreign producers in the absence ofany price regulation would be only aboutUS$53 million per year With price regulationthat would keep the prices of drugs suppliedby multinational subsidiaries at their pre-TRIPS level the profit gains become onlyUS$196 million per year

The remainder of this paper is organized asfollows In the next section we lay out theessential features of the Indian pharmaceuticalsmarket provide more detail about the segmentsthat we focus on in the empirical analysis andbriefly describe the primary data we use Sec-tion II describes the analytic framework and theeconometric strategy we use to estimate therelevant parameters and construct the counter-factual scenarios We discuss our results in Sec-tion III Section IV concludes

I The Setting and the Data

Between April 1972 when the Indian PatentsAct (1970) became effective and March 2005when Indiarsquos parliament passed the 3rd Amend-ment of the Patents Act India did not recognizeproduct patents for pharmaceuticals The IndianPatents Act which replaced the inherited Brit-ish colonial law regarding intellectual propertyrights specifically excluded pharmaceuticalproduct patents and admitted process patentsonly for a period of seven years In contrast thelatest amendment recognizes patents on endproducts that under the new regime will remainin force for 20 years6

The two stated objectives of the 1970 actwere the development of an indigenous pharma-ceuticals industry and the provision of low-costaccess to medicines for Indian consumers Con-sistent with these objectives and with the broaderleftward tilt in policy a number of other measureswere introducedmdashdrug price controls restrictionson capacity expansion limits on multinational eq-uity sharesmdashthat in the years since have on theone hand kept pharmaceutical prices low and onthe other encouraged the development of the In-dian pharmaceutical industry Many of these reg-ulations and restrictions have been lifted or easedsince the mid-1980s with marked acceleration inthe pace of liberalization during the 1990s

Over the last 20 years the Indian pharmaceu-tical industry has grown rapidly to the pointwhere it is now the worldrsquos largest producer offormulations in terms of volume and one of the

6 Indian companies that are now producing drugs forwhich patent applications were submitted between the sign-ing of the TRIPS agreement in 1995 and January 1 2005will be allowed to continue producing if they pay a royaltyto the patent holder

1481VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

worldrsquos largest producers of bulk drugs7 Thestructure of the industry has also evolved In1970 the industry was dominated by multina-tional subsidiaries by 2001 Indian-owned firmswere not just the leading players in the industrymany had also become major exporters

The data we use in this paper are from theretail pharmaceutical audits of ORG-MARGIndiarsquos premier market research and consultingfirm The audit provides detailed product-levelinformationmdashestimates of monthly retail salesin each of the four geographic zones of Indiaprice dosage form launch date brand namechemical name therapeutic categorizationmdashon

all pharmaceutical products sold in India byabout 300 of the largest firms representingroughly 90 percent of domestic retail sales ofpharmaceuticals The coverage of the audit isextensive reaching a representative panel ofthousands of retail chemists in over 350 citiesand towns The data collected which providethe only real source of disaggregate informationon the Indian pharmaceutical market are usedby both the government of India in formulatingpricing policy and other decisions and the In-dian pharmaceutical industry in determiningpricing and marketing strategies We have in-formation at a monthly frequency for the periodof January 1999 to December 2000 Tables 1 to3 provide a set of descriptive statistics that areessential for understanding the focus of ouranalysis and interpreting our results

As noted earlier the characteristics of de-mand for pharmaceuticals in India are likely todiffer considerably from those in developedeconomies With a share of 23 percent the

7 Bulk drugs are the therapeutically relevant active phar-maceutical ingredients that are combined with a variety ofinactive ingredients to make the formulations ultimatelyconsumed by patients Firms in the pharmaceutical sectorcan be one of three types bulk drugs producers pureformulators or integrated firms which produce both bulkdrugs and market formulations

TABLE 1mdashTHE QUINOLONES SUBSEGMENT

Molecule

Shares ()of quinolones sales Sales (Rs mill) 2000

Domesticfirms

Foreignsubsidiaries

Domesticfirms

Foreignsubsidiaries

Ciprofloxacin 530 27 3030 156Norfloxacin 112 01 640 3Ofloxacin 116 31 665 177Sparfloxacin 108 01 620 4Lomefloxacin 15 86 Pefloxacin 13 01 72 5Levofloxacin 00 0 Nalidixic acid 13 73

TABLE 2mdashBASIC INFORMATION ABOUT THE FOUR QUINOLONES MOLECULES 2000

Ciprofloxacin Norfloxacin Ofloxacin Sparfloxacin

US or European patent holder Bayer Merck Ortho-McNeil Rhone-PoulencYear of US patent expiry 2003 1998 2003 2010Year of US-FDA approval 1987 1986 1990 1996Year first introduced in India 1989 1988 1990 1996No of domestic Indian firms 75 40 17 25No of foreign subsidiaries 8 2 2 1Sales weighted average per-unit API

price of products produced byDomestic Indian firms 1124 905 9008 7884Foreign subsidiaries 1035 528 10847

API Active pharmaceutical ingredient

1482 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

antiinfectives segment ranks second in Indiawhereas in the world market it is fifth and has ashare of only 90 percent Hence antiinfectivesare important in India not only from a health

and public policy point of view but also as asource of firm revenue

With this in mind we focus on one partic-ular subsegment of antiinfectives namely the

TABLE 3mdashSUMMARY STATISTICS FOR THE QUINOLONES SUBSEGMENT 1999ndash2000

North East West South

Annual quinolones expenditure per household (Rs) 3125 1975 2764 2359(366) (367) (407) (286)

Annual antibiotics expenditure per household (Rs) 11988 8424 11052 9624(1224) (1224) (960) (996)

No of SKUsForeign ciprofloxacin 1238 1129 1308 1246

(150) (190) (102) (106)Foreign norfloxacin 183 171 200 158

(070) (075) (088) (083)Foreign ofloxacin 304 296 296 300

(086) (086) (091) (088)Domestic ciprofloxacin 10621 9763 10342 10550

(599) (434) (722) (451)Domestic norfloxacin 3896 3496 3617 3942

(271) (268) (251) (379)Domestic ofloxacin 1846 1600 1725 1725

(680) (634) (586) (635)Domestic sparfloxacin 2983 2829 3121 2929

(557) (638) (688) (657)Price per-unit API (Rs)

Foreign ciprofloxacin 958 1090 1085 1007(128) (066) (071) (058)

Foreign norfloxacin 563 509 605 435(077) (133) (139) (147)

Foreign ofloxacin 10946 10943 10886 10612(620) (664) (700) (1140)

Domestic ciprofloxacin 1143 1067 1131 1152(016) (015) (017) (013)

Domestic norfloxacin 951 907 888 873(024) (035) (037) (020)

Domestic ofloxacin 9163 8964 8565 9341(1615) (1565) (1422) (1407)

Domestic sparfloxacin 7972 7849 7688 8028(976) (1014) (1185) (1037)

Annual sales (Rs mill)Foreign ciprofloxacin 4179 2431 4520 2947

(1534) (816) (1273) (648)Foreign norfloxacin 128 100 058 073

(101) (082) (044) (057)Foreign ofloxacin 5446 3184 3522 3111

(1399) (933) (906) (703)Domestic ciprofloxacin 96229 58591 67874 70381

(10626) (13026) (12226) (8740)Domestic norfloxacin 22255 11971 14918 15829

(3884) (1945) (2691) (1626)Domestic ofloxacin 12502 9621 14936 11205

(4434) (3011) (5282) (4259)Domestic sparfloxacin 15617 12175 16130 9811

(3141) (2576) (4674) (3420)

Note Standard deviations in parentheses API Active pharmaceutical ingredient

1483VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

quinolone subsegment Quinolones fall into thesystemic antibiotics and antibacterials segmentof the Indian pharmaceuticals market whichgenerates over three-quarters of the revenues inthe antiinfectives segment8 The systemic anti-bacterials segment includes all of the originalmiracle drugs that first sparked the growth ofthe global research-based pharmaceutical indus-try in the postndashWorld War II period as well aslater generations of molecules that have beenintroduced in the last four decades

Among systemic antibacterials quinolonesare the latest generation molecules available inIndia We focus our analysis on quinolones forseveral reasons First quinolones are the drug ofchoice for a large number of bacterial infec-tions some of which are also treated by alter-native drugs (see Table A1 in the Appendixwhich outlines the spectrum of activity for eachmolecule family within the antibacterials seg-ment) Hence if there were one product groupfor which we would expect to have many sub-stitutes readily available this would be quino-lones Second with a share of 20 percent in thesales of systemic antibacterials quinolones rep-resent one of the largest subsegments within thistherapeutic category Finally several moleculeswithin the quinolone subsegment were still un-der patent in the United States at the time of ourinvestigation Table 2 details the basic informa-tion about the four quinolone molecules that arethe focus of our analysis The first row showsthe year of US patent expiry this ranges from1998 for norfloxacin to 2010 for sparfloxacinQuinolones include in principle four moremolecules that are listed at the bottom of Ta-ble 1 the market shares of these molecules arenegligible however so that we exclude thesemolecules from our analysis

Table 2 reveals several other interesting factsabout competition in the quinolone market inIndia First note the large number of firmsoperating in this subsegment The large numberof domestic firms is perhaps not that surprisinggiven that pharmaceutical product patents werenot recognized in India9 What is more surpris-ing is the number of foreign firms selling pat-ented products (eg ciprofloxacin) the fact that

multiple foreign firms sell a patented productindicates that such firms often ldquoinfringerdquo10

patent laws in India while complying with themin developed world countries The last two rowsof Table 2 further indicate that domestic prod-ucts often sell at a premium With the exceptionof ofloxacin the average prices of products of-fered by Indian firms are higher than the pricesof products offered by foreign subsidiaries Thispreliminary evidence suggests that Indian con-sumers do not place a premium on the brandname and reputation of big multinational phar-maceutical concerns Moreover the higherprice of domestic products does not seem toprevent domestic companies from capturing alarge market share This is most evident in thecase of ciprofloxacin where domestic firmshave with 53 percent the largest share in thetotal sales of quinolonesmdashthis despite the factthat the average price of these products is 10percent higher than the price of foreign productscontaining the same molecule

Table 3 provides additional summary statis-tics for our data broken down by region Thefirst two rows of the table report the averageannual household expenditure on quinolonesand antibiotics respectively Note that in bothcases the average expenditure is higher in thenorth and west these regions include states withhigher per capita incomes and tend to be moreindustrialized and urbanized than those in theeast and south Pharmaceutical products areavailable in multiple presentations that is com-binations of dosage forms (eg capsule tabletsyrup) strength (eg 100 milligrams 500 mil-ligrams) and packet sizes (eg 50 capsule bot-tle 100 tablet bottle) The various presentationsin which a product is available are often referredto as stock-keeping units or SKUs11 The num-ber of SKUs for each product group withinquinolones is reported at the top of Table 3 Aswith the more aggregate numbers on firms and

8 In addition to anti-bacterials this segment also containsantivirals

9 Accordingly the common distinction between ldquobrandedrdquoand ldquogenericrdquo products is irrelevant here

10 We emphasize here that the word ldquoinfringerdquo belongsin quotes because patent laws do not currently exist inIndia infringement in the legal sense is not possible It ishowever striking that the same firms that accuse Indianproducers of ldquopiracyrdquo sell in India products that are patentedin the United States and for which the patent is held by adifferent multinational corporation

11 For instance a 100 capsule bottle of 100 milligramcapsules of a particular branded drug and a 50 capsulebottle of 100 milligram capsules of the same branded drugwould be identified as two separate SKUs

1484 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

products reported in Table 2 the differencebetween domestic and foreign products isstriking The number of SKUs offered byIndian firms is consistently larger than thenumber offered by subsidiaries of foreignmultinationals The number of SKUs variesslightly across regions but more importantlyit varies across time as some SKUs disap-pear while new ones get introduced duringour sample period

Many pharmaceutical products in India aresubject to price controls12 While the specificsof the price regulation are too complex for anyeconomic model to capture adequately themain concern for the empirical analysis is thatprice controls may lead to a lack of price vari-ation over time so that the demand functioncannot be identified Prices at the most disag-gregate SKU level are relatively stable overtime there are variations due to occasionalchanges in the estimated cost (due for exampleto changes in exchange rates that affect the costof imported materials or bulk drugs) but suchvariations tend to be infrequent and small inmagnitude The degree of time variation ishowever substantially larger once one aggre-gates to the product level This variability stemsnot only from the fact that the SKUs over whichwe aggregate may experience changes in theirrespective prices at different points in time butalso from the fact that the range of SKUs of-fered in the market does not remain constantover time The entry and exit of presentationswithin the same product group that have differ-

ent prices effectively affects the price that con-sumers face for this drug in each period

The middle portion of Table 3 reports themean price and standard deviation for eachproduct group by region Prices vary by regionthough there is no clear pattern emerging fromthe table with regard to the cross-regional vari-ation (in the sense of some regions being sys-tematically more expensive than others) Toexamine what portion of the total price variationis due to time versus regional variation weconducted an analysis of variance of prices thatwe report in Table 4 The table is based onseparate regressions for each product group(pooling data across groups with big differencesin their average prices is not particularly infor-mative as most of the price variation is ac-counted for by product group dummies) Thelast two columns of the table show the fractionof price variation accounted for by region andtime dummies respectively As evident fromthe table a significant fraction of the total vari-ance in prices can be attributed to time varia-tion In the demand estimation we include a fullset of product groupndashspecific regional dummiesso that the price parameters are identified en-tirely based on this time variation within eachproduct group The time variation of productgroup prices is driven primarily by composi-tional changes within each group the revenueshares of the individual SKUs that compriseeach product group change over time (see therelated discussion in Section IIC) while in eachperiod there is entry and exit of SKUs into thesample To check whether this pattern reflectsgenuine entry and exit as opposed to samplingvariation we examined the revenue shares ofthe SKUs that leave the sample relative to theones that remain during the entire period Theresults are reported in Table 5 While the SKUs

12 The details of the procedures for price fixation canbe found in the official government Web site httpwwwnppaindianicinindex1html under the link ldquoDrug Price Con-trol Order 1995rdquo A new pharmaceutical policy was introducedin 2002 but our data were collected before that year

TABLE 4mdashANALYSIS OF PRODUCT PRICE VARIANCE

Product groupPartial SS

timePartial SS

region Total SSPercentage

explained by timePercentage

explained by region

Foreign ciprofloxacin 0296 0306 1047 283 292Foreign norfloxacin 1036 1002 4179 248 240Foreign ofloxacin 0429 0021 0571 752 36Domestic ciprofloxacin 0005 0088 0104 50 841Domestic norfloxacin 0059 0098 0198 296 494Domestic ofloxacin 2858 0104 3056 935 34Domestic sparfloxacin 1754 0032 1860 943 17

1485VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

that exit tend to be smaller (their average shareis 1 percent as opposed to 34 percent for thoseSKUs that are present during the entire sampleperiod) the shares of the two groups do notseem orders of magnitude apart13 In additionour data cover only the 300 largest firms sellingin the Indian market so that firms with verysmall shares are not included in our sample

II The Analytic Framework and EstimationApproach

Patent enforcement in the Indian pharmaceu-tical market will have the effect of eliminatingdomestic products whose active pharmaceuticalingredients are protected by (foreign) patentsThus assessing the effects of patent enforce-ment is tantamount to assessing the effects ofwithdrawing domestic products from the mar-ket This task is the converse of evaluating newproduct introduction accordingly the concep-tual framework we use to address the questionsof interest is similar to the one developed in theliterature for the valuation of new goods14

We start by estimating demand for quino-lones Given that the market is characterized byimperfect competition the counterfactual anal-ysis requires that we also model the supply sideas removal of one product will affect the pricesof other products especially those that are closecompetitors The existence of price regulationin the Indian pharmaceutical market imposespotential constraints on firmsrsquo maximizationproblem Given these constraints and the com-

plexity of the price regulation process the typ-ical approach of deriving estimates of actualmarginal costs and markups by exploiting thefirst-order conditions of profit maximizing firmsdoes not seem particularly promising Insteadwe use our demand estimates to place upper andlower bounds on marginal costs and markups

With demand elasticities and upper and lowerbounds for marginal costs in place we thenconduct counterfactual simulations We con-sider several alternative scenarios dependingon the number of domestic products affected bypatent enforcement For each scenario we com-pute the counterfactual prices and use them toassess the effects of domestic product with-drawal on consumer welfare (as measured bythe compensating variation) firm profits andsocial welfare As with the valuation of newproducts the big conceptual problem facing thispart of the analysis is that we need to extrapo-late from the region of the data to the point atwhich demand for the products that exit themarket becomes zero This conceptual issue ispresent in any attempt to evaluate a major pol-icy change for which no historical precedentexists like the enforcement of patent laws inIndia One advantage of the present study is thatwe have a limited set of price data for Pakistana country with similar demographics to Indiabut with a market structure that resembles theone that would emerge in India under patentenforcement (monopoly of multinational sub-sidiaries) By comparing the prices of productsoffered by multinationals in Pakistan to thosewe compute in our counterfactual simulationsfor the products that would be offered by mul-tinationals in India if patent laws were enforcedwe can get a sense of how plausible our coun-terfactual estimates are

A Demand

The demand modelling is based on the multi-stage budgeting approach Our primary motiva-tion for adopting this approach is a practical oneIn the multistage budgeting approach the depen-dent variable is defined as a revenue share whichis appealing given that the products we include inthe analysis contain different molecules (ie ac-tive pharmaceutical ingredients or APIs) Eventhough we do have data on the quantity of therelevant API (eg 100 milligrams of ciprofloxa-cin) contained in each product converting the

13 While we cannot completely rule out the possibilitythat some of the exit is due to sampling variation note thatthe latter should be reflected in low precision of the demandparameter estimates The demand parameters however areprecisely estimated

14 See Manuel Trajtenberg (1989) Jerry A Hausman(1994) and Timothy F Bresnahan (1997) for representativeexamples and a discussion of the relevant issues

TABLE 5mdashREVENUE SHARES OF THE EXITING SKUS

Revenue share ofexiting SKUs

Revenue shares ofall SKUs

Full sample 10 34Northern region 08 33Eastern region 12 36Western region 10 34Southern region 12 33

1486 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

revenue shares to physical shares is extremelydifficult if not infeasible in the case of antibioticsBecause such drugs are ldquosystemicrdquo by nature theyare used to treat a large number of infections andthe dosage of each drug depends on the particularinfection it is supposed to address (for example thedosage will differ depending on whether the anti-biotic is used to combat an ear infection or tuber-culosis) This particular feature of antibioticscomplicates the conversion of revenue to physicalshares

The basic idea of the multistage budgetingapproach is to use the therapeutic classificationof a productmdashie the therapeutic segment andsubsegment the product belongs tomdashto orga-nize all products in the systemic antibacterialssegment into a hierarchical taxonomy consist-ing of two levels At the higher level are thevarious subsegments of systemic antibacterialsThe first stage of budgeting corresponds to theallocation of expenditures across the subseg-ments in this upper level of the taxonomy

In the second stage of the budgeting processcorresponding to the lower level of the taxon-omy a flexible functional form is adopted tomodel how the expenditures allocated to eachsubsegment are distributed across the productswithin that subsegment In particular to modeldemand at the second stage we employ theAIDS specification proposed by Deaton andMuellbauer (1980b)15

While the two-stage demand estimation ap-proach offers functional form flexibility its ap-plication to the Indian systemic antibioticsmarket poses a few problems The first is thatdue to entry and exit many SKUs and evenproducts in our sample are not present in everyperiod AIDS does not have a good way ofdealing with a varying number of products as itwas developed with broad commodity catego-ries in mind which are consumed by all con-sumers every period To solve this problem weaggregate within each subsegment (eg quino-lones) SKUs into product groups where withineach product group all SKUs contain the samemolecule and are produced by firms with thesame domesticforeign status Specifically let aSKU k be indexed by its molecule (or API) M its

domesticforeign status DF indicating whether itis produced by a domestic (Indian) or a subsidiaryof a foreign (multinational) firm a particular pre-sentation s and the particular firm f that producesit We aggregate SKUs over presentations andfirms to obtain a newly defined product group iwhich is indexed only by molecule M and domes-ticforeign status DF and has revenue Ri yenfsRk with i (M DF) k (M DF f s) and pricepi yenfs kpk where k denotes the conditional(on M and DF) revenue share of this particularproduct ie

(1) k Rk

Ri

In most cases the resulting product groups arebroad enough to be present every period16

The usual concern with this aggregation pro-cedure is that it may lead us to overstate firmsrsquomarket power as we ignore competition amongfirms with the same domesticforeign statusproducing the same molecule In the presentapplication however this concern is unlikely tobe of great importance as the effect of patentenforcement is to wipe out all domestic compe-tition at once while granting foreign firms mo-nopoly power hence competition among firmsfor patented molecules becomes irrelevant Theaggregation according to the domesticforeignstatus (within a particular molecule) thus corre-sponds to the scope of our analysis and theparticular questions of interest

The second problem is that for our approachto be useful in welfare analysis the allocation oftotal expenditures to group expenditures at thehigher stage has to be modelled in a way con-sistent with utility maximization In general thesolution of this allocation problem requiresknowledge of all individual product prices

15 Representative applications of the multistage budget-ing approach include Sara F Ellison et al (1997) Hausman(1994) and Hausman and Gregory K Leonard (2002)

16 We are missing only four observations (ie monthregion combinations) all for the drug group of ForeignNorfloxacin August 1999 in the south May 2000 in thewest October 2000 in the south and November 2000 in theeast In these cases we set the revenue shares of ForeignNorfloxacin equal to zero In general with 01 percent ofquinolone sales (see Table 1 row 2) Foreign Norfloxacinhas a very small share of the market This probably explainswhy the results pertaining to this drug group are unreliableit is the only drug group for which we do not obtain asignificant price elasticity of demand while its cross-priceelasticities with other foreign drug groups often have thewrong sign

1487VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

From an empirical point of view this is notparticularly useful as it eliminates all compu-tational advantages of the two-stage approachTo address this problem we adopt an approxi-mate solution to model the higher-level expen-diture allocation along the lines suggested byDeaton and Muellbauer (1980b pp 131ndash32)This gives rise to a two-level AIDS specification

Consider the lower-level estimation firstwhich refers to the allocation of a particularsubsegmentrsquos expenditure to the product groupswithin the subsegment In our application therelevant subsegment is quinolones which weindex with Q Let the product groups within thissubsegment be indexed by i 1 N pi be theprice of product group i (where as noted abovei refers to a particular molecule and domesticforeign status combination) and XQ be the totalexpenditure on the quinolone segment The rev-enue share of each product group is given by

(2) i i j

ijln pj ilnXQ

PQ

where i the revenue share of product group iis defined as

(3) i pi qi

yenj

pj qj

xi

XQ with i j Q

XQ is the overall expenditure on the quinolonesubsegment and PQ is a price index given by

(4) ln PQ ap 0 i

iln pi

1

2 i

j

ijln piln pj

With a limited number of product groups and asufficiently large number of time-series observa-tions the flexibility implied by the AIDS modeldoes not impose too many demands on the data Inthe present application however where the num-ber of observations is limited the AIDS model isnot estimable in this general form To reduce thenumber of parameters that need to be estimatedwe impose two sets of restrictions

The first set of restrictions is implied by thetheory of utility maximization Specificallythese restrictions are

Adding-up yenk k 1 yenk k 0 yenk kj 0 j

Homogeneity yenk jk 0 j Symmetry ij 1frasl2 [ij ji] ji This last

restriction by itself reduces the number of parameters to [N(N 1)]2

The second set of restrictions we impose aimsat further reducing the number of parameters tobe estimated by exploiting our knowledge of thisparticular market Specifically for each productgroup i we allow one ij parameter for all productgroups j that have different molecules from prod-uct group i and are produced by foreign firms andone ij for all product groups j with differentmolecules produced by domestic firms We donrsquotimpose any restrictions on the ij parameter whenproduct group j has the same molecule as productgroup i (By construction product groups i and jcontain products produced by firms with differentdomesticforeign status)

To better illustrate the nature of the restric-tions we impose on the patterns of substitutionacross products some additional notation isneeded Let d(i j) be an indicator of the degreeof similarity (or difference) between productgroup i and product group j along the dimen-sions we are able to observe (molecule M anddomesticforeign status DF) For any two prod-uct groups i and j d(i j) can take on one of thefollowing three values17

(5)

di j (1 0) if Mi Mj DFi DFj

(0 1) if Mi Mj DFi DFj

(0 0) if Mi Mj DFi DFj

Let

(6) Diab j di j a b

17 The sequence (1 1) is not possible for two differentproducts in this case the parameter corresponds to theproductrsquos own-price effect that is ii

1488 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The equation at the lower level becomes

(7) i i iiln pi i10ln pj j Di10

j Di

01

i01ln pj

j Di

00

i00ln pj ilnXQ

PQ

Note that

The parameter ii captures a product grouprsquosown-price effect (there will be as many iiparameters as the number of product groups)

The parameter i10 captures the cross-priceeffect of the product group containing prod-ucts with the same molecule but produced byfirms of different nationality

The parameter i01 captures the cross-priceeffects of product groups containing productswith different molecules but produced byfirms with the same nationality

The parameter i00 captures the cross-priceeffects of product groups containing productswith different molecules produced by firms ofdifferent nationality

Before we take the demand equation to thedata we make two modifications The first oneis to let the product-specific effects i vary byregion r The resulting product-specific regionaleffects ir have two interpretations first theycontrol for the ldquoqualityrdquo of each drug withquality differences being allowed to vary acrossregions second they proxy for demographicsand other demand shifters which vary by re-gion and may affect the demand of each prod-uct group differently18 Note that by includingproduct-specific regional effects in the demandspecification we estimate the price parametersbased on the within-product-group variation ofprices in each region In an earlier version of thepaper we also estimated the demand systemwithout regional dummies and obtained similarresults

The second modification that allows us to gofrom a deterministic to a stochastic specificationof the demand equation is to include an additiveerror term in (7) The latter takes into accountthe fact that (7) is not expected to fit the dataexactly The error term irt accounts for mea-surement error (due to the fact that the productgroup prices pjrt we employ in the estimationare not exact price indices but approximationsthereof) and (potentially region-specific) de-mand shocks that may affect the demand for aproduct in a particular period Examples of suchshocks include an advertising campaign for aparticular product which temporarily increasesthe demand for this product or the outbreak ofa (potentially region-specific) epidemic whichcalls for the use of a particular drug We discussthe interpretation and properties of this errorterm in more detail in the next section

The final form of the equation we estimate atthe lower level becomes (with subscript t de-noting month and subscript r denoting region)

(8) irt i ir iiln pirt i10ln pjrt j Di10

j Di

01

i01ln pjrt

j Di

00

i00ln pjrt

ilnXQrt

PQrt irt

The analysis so far has conditioned on the ex-penditure allocated to the quinolone subseg-ment XQ The upper level of the estimationconsiders the problem of allocating total expen-diture across the different systemic antibioticssubsegments one of which is quinolones Theupper-level demand function is given by

(9) G G H

GHln PH GlnX

P

where all variables denoted by capital letters aredefined as before but now refer to subsegments(G H ) rather than product groups within asubsegment and the total expenditures on sys-temic antibiotics X are deflated by the Stone priceindex log P yenH Hlog PH When estimating thesystem above we impose all the restrictions

18 Given the short time span of our sample typicaldemand shifters such as age distribution income distribu-tion and education hardly change over our sample periodSuch shifters are therefore absorbed by the region-specificproduct fixed effects ir

1489VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

implied by utility maximization as we do withthe estimation of the lower-level AIDS We donot however impose any additional restrictionson the substitution patterns at this stage so thatthe cross-price effects across segments remainrelatively unconstrained

Estimation of the higher-level AIDS allowsus to obtain the unconditional own- and cross-price elasticities that are used in the formulationof the supply problem and welfare analysisThese are given by the formula

(10) ij ijXQXQ

ln qi

ln XQ

ln XQ

ln PQ

ln PQ

ln pj

with the conditional cross price elasticitiesgiven by

ijXQXQ ij ij jlnXQ

PQ

i

As in the lower stage we include subsegment-specific regional dummies in the specificationof the upper-stage demand system so that thefinal form of the estimating equation at theupper stage becomes

(11) Grt G Gr H

GHln PHrt

GlnXrt

Prt Grt

In sum the demand system we take to thedata is represented by equations (8) and (11)and the associated parameter restrictions im-plied by economic theory

B Modelling the Supply Side of the Market

Counterfactual simulations concerning theeffects of domestic product withdrawal requireknowledge of the marginal costs of pharmaceu-tical firms operating in the Indian market Theseare unobservable The usual approach in theNew Empirical Industrial Organization litera-ture has been to exploit the firm equilibriumconditions to infer marginal cost For exampleit is often assumed that the marginal cost ci isconstant and that the industry is an oligopoly

engaging in Bertrand competition with differen-tiated products Assuming that firms myopicallymaximize profits each period one can then de-rive the firmsrsquo first-order conditions that corre-spond to the assumptions above about costsmarket structure and firm behavior

We deviate from this procedure as the presenceof price regulation renders the assumption of un-constrained period-by-period maximization un-tenable Ideally one would like to explicitlyincorporate the price and other administrative con-trols into the firmrsquos optimization problem andderive the first-order conditions under the assump-tion of constrained maximization The complexityof the price regulation however makes this ap-proach infeasible Therefore we adopt an alterna-tive approach that does not rely on modelling theprice-setting process but is instead based on de-riving upper and lower bounds for marginal costsand markups

In particular an upper bound for marginalcosts and a lower bound for the markups (zero)can be derived under the assumption of perfectcompetition While this assumption is clearlyunrealistic in the pharmaceuticals market it isuseful in providing an upper bound for costs ci

Uwhich will be given by

(12) ciU pi

On the other hand a lower bound for marginalcosts ci

L (and upper bound for markups) can bederived by assuming that there is perfect collu-sion within each product group i (where i refershere to a moleculedomestic-foreign combina-tion)19 and ignoring price controls so thatprices are determined by the first-order condi-tion of the jointly profit-maximizing firmswithin product group i Solving this first-ordercondition for the (lower bound of) marginal costthen gives

(13) ciL pi 1

1

iipipj

19 The reason that this assumption will lead to an under-statement of marginal costs (and hence overstatement ofmarket power) is that it assumes away competition amongfirms within each drug group

1490 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

where ii(pipj) the own-price elasticity of de-

mand for product i will depend on the productrsquosown-price pi and all other productsrsquo prices pjwith i j

Once we have obtained the demand elastic-ities through estimation of the demand sys-tem we can calculate the upper and lowerbounds for marginal costs and correspondingmarkups according to (12) or (13) These willthen be employed in the counterfactual sim-ulations While our counterfactuals use boththe lower and upper bounds for costs most ofour discussion will be based on using thelower bounds for costs since this is the moreinteresting case in the policy analysis it givesus the largest possible profits for pharmaceuticalfirms and hence corresponds to the worst pos-sible scenario facing Indian firms and the bestpossible scenario facing multinationals underTRIPS

C Identification Assumptions and EstimationApproach

The discussion of the demand system has sofar abstracted from the issue of price endogene-ity The usual premise in the industrial organi-zation literature is that correlation of prices withthe error term in the demand equation arises byvirtue of the first-order conditions of profit-maximizing firms20 As we discuss in this sec-tion this source of simultaneity bias is unlikelyto be of major concern in the present contextbecause of the existence of price regulation Ourprimary concern is instead with the simultane-ity bias that is implied by the particular wayprices are constructed

To understand the sources of the simultaneitybias note first that the price of product group piwe employ in the demand estimation should bethought of as a proxy for an exact price indexthat we do not observe As such pi contains bydefinition measurement error and it will becorrelated with the error term of the demandequation Specifically the demand equation weare interested in estimating can be written insimplified form (suppressing the subscripts rand t and ignoring parameter restrictions forconvenience) as

i i j

ijln pjT ilnXQ

PQ i

where pjT denotes the true (exact) price index for

product group i and i captures unobservedvariables that may affect demand in a particularperiod (eg an advertising campaign or an ep-idemic) The price index pj

T depends on theSKU prices pk Note that in the presence ofSKU entry and exit into the sample the priceindex pj

T will vary over time even if the pricesof the individual SKUs remain stable as theproducts (SKUs) that comprise each productgroup change each period

Ideally we would like to compute the exactprice index pj

T for each group and employ theseindices in the demand estimation Under theassumption of predetermined SKU prices eachperiod the exact price indices would be uncor-related with the error term and the demandparameters could be estimated by simple OLSUnfortunately this strategy is not feasible in thecurrent context

To derive an exact price index for each prod-uct group it is necessary to model consumerrsquoschoice among SKUs conditional on the choiceof the group To this end it is necessary toexplicitly introduce a third stage in the demandspecification and estimate the parameters asso-ciated with the choice at that stage Howeverany specification based on three-stage budget-ing (that would require directly specifying ademand function associated with the SKUchoice at the third stage and imposing additiveseparability at the higher stages) would be in-feasible to estimate for three main reasonsFirst there is a large number of SKUs withinsome groups so that the demand parameters can-not be estimated even if one imposes parameterrestrictions second the SKU prices exhibit littlevariation over time lastly the choice sets at thislowest stage vary over time rendering estimationof the price parameters associated with SKUs thatare not available in all periods infeasible Alterna-tively one could abandon the multistage budget-ing altogether and adopt a discrete choiceapproach which is inherently better suited to deal-ing with varying choice sets and limited pricevariation A discrete choice approach would beparticularly appealing for modeling the SKUchoice conditional on product group choice giventhat within each group all SKUs contain the same

20 See Steven Berry et al (1995) or Aviv Nevo (2001)for a related discussion

1491VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

molecule and are hence comparable As notedearlier however a discrete choice approach at thehigher stages of the demand estimation requiresconverting revenue to physical shares which ischallenging in the case of systemic drugs such asantibiotics for which the appropriate dosage is notwell defined Finally a combination of a multi-stage budgeting approach for the higher stages anda discrete choice approach for the lowest stagewould be feasible to estimate but inconsistentwith a model of utility maximization For thesereasons we have no alternative but to adopt anapproximation for measuring the price index foreach product group

Let us denote this approximation by pjA and let

13j denote the proportional ldquoapproximationrdquo errorassociated with measuring the true price index forgroup j so that ln pj

A ln pjT ln 13j The

estimating demand equation can then be written as

i i j

ijln pjA ilnXQ

PQ

j

ijln 13j i

or

i i j

ijln pjA ilnXQ

PQ i

where the new error term i yenj ijln 13j iis comprised of two components the first onereflects ldquomeasurementrdquo error due to the fact thatthe product group prices we employ are approx-imations and not exact price indices and thesecond one captures conventional demandshocks The demand equation as written abovecorresponds to the equation we take to the data

The particular proxy we use for the exact priceindex for each product group j is the revenue-share-weighted average of the prices of multipleSKUs available within this group that is

pjA

k j

k pk

The notation above illustrates the two source ofprice variation variation in the SKU prices pkand variation in the weights k

We are concerned with potential correlation ofthe so-constructed product group prices with bothcomponents of the product group demand error

term The correlation between pjA and the first

component yenj ijln 13j is independent of theparticular way we construct pj

A and inherent in thefact that we measure the exact price index witherror The potential correlation between pj

A and thesecond component of the product group demanderror the shock i arises however because of thespecific way we construct the proxy pj

A and inparticular because of the presence of the reve-nue share weights k in pj

A since the SKUrevenue share weights k will generally dependon the product group expenditure they arelikely to be correlated with the product groupdemand shock i

To address the simultaneity bias we use in-strumental variables To this end we need vari-ables that are correlated with the proxies pj

A butuncorrelated with the error term i We use thenumber of SKUs within each product group j asan instrument as it does justice to the idea thatvariation in the product group price index stemsin part from variation in the set of SKUs that areavailable each period It is clearly correlatedwith the average group prices pj

A and plausiblyuncorrelated with the demand shock i Theassumption that the number of SKUs is uncor-related with the conventional demand error willhowever be violated if the introduction of anew SKU changes the perceived quality of adrug21 or if it is accompanied by promotionalactivities which will be reflected in i Our hopeis that this is not too often the case Assumingthat the SKU number for each group j is alsouncorrelated with the weighted sum of the mea-surement errors of all product groups yenj ijln 13j itcan be used as an instrument for the product groupprice pj

A22 If one accepts the premise that prices atthe SKU level are exogenous then SKU pricescan also be used as instruments The final list ofinstruments we use hence includes the number ofSKUs in each group the prices of the five largestSKUs for each group and all other exogenousvariables in the demand estimation such as re-gionalproduct dummy interactions and the upper-

21 Note that we already proxy for the quality of each prod-uct group through the product fixed effects i In the actualestimation we actually let these fixed effects be region-specific(ir) so that we allow quality to vary across regions

22 If the measurement error increases in the number ofincluded SKUs this assumption will be violated Unfortu-nately we have no way of checking the validity of thisassumption

1492 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

level total expenditure on antibiotics Regressionsof group prices on the instruments above yieldhigh R-squares with most regressors highly sig-nificant indicating that our instruments are highlycorrelated with prices23

Our sample includes four molecules cipro-floxacin norfloxacin ofloxacin and spar-floxacin Except for sparfloxacin all othermolecules are produced by both foreign anddomestic firms24 So we have seven products(domestic ciprofloxacin foreign ciprofloxa-cin etc) with 96 observations (two years ofmonthly data four geographical regions foreach period) for each product The parametersin the lower-level AIDS demand system asdefined in equation (8) are the product fixedeffects i the productndashregional dummy inter-actions ir the own revenue-share price elas-ticities ii the cross revenue-share priceelasticities i10 i01 i00 and the revenue-share expenditure elasticities i In estimatingthe parameters we first regress prices on allinstrumental variables and then plug the pre-dicted values for prices in the constrainedleast-square regression (For a detailed expla-nation of the constraints see the previoussection)

Given that we impose many cross-equationconstraints and employ instrumental variablesin the estimation it is difficult to derive stan-dard errors for the parameter estimates ana-lytically Our error term interpretation in (8)implies that the error terms for each productgroup are likely to be correlated across re-gions for example if a national advertisingcampaign increases the demand for a rela-tively expensive presentation in one regionsimultaneously increasing the aggregate de-mand and price index for the correspondingproduct group in that region it is likely that thesame effect will be observed in other regions Inprinciple we could exploit the cross-regionalcorrelation in the error terms to estimate thedemand system using seemingly unrelated re-

gression (SUR) This however would requireus to derive the variance-covariance matrix ofthe parameters analytically which as notedabove is cumbersome in the current contextInstead we use the bootstrap method choosingto remain agnostic about the structure of thevariance-covariance matrix The potential dis-advantage of this approach over SUR is a loss inefficiency but as will become apparent in theresults section the model parameters are fairlyprecisely estimated To maintain the marketstructure we randomly sample the periods (withreplacement) and use the same periods for allproducts Regarding the optimal number ofbootstrap repetitions ideally one would followthe three-step method proposed by Donald W KAndrews et al (2000) Empirical evidence sug-gests however that one rarely needs more than200 replications to estimate the standard er-rors25 To be safe we generate 300 bootstrapsamples (with replacement) based on the origi-nal data and estimate the standard errors usingthe standard errors of the bootstrap sampleestimates

The estimation of the top-level AIDS systemis similar The constraints imposed on this top-level demand system are adding-up homogene-ity and symmetry as defined on page 12Again bootstrapping is used to obtain the stan-dard errors of the parameter estimates

D The Counterfactual Scenarios

In assessing the effects of patent enforce-ment we start by focusing on the most extremecase in which compulsory licensing is not anoption and foreign firms are not subject to pricecontrols We use the results from the analysis ofthis case as a benchmark In reality the outcomeof the WTO negotiations is more likely to in-volve some constraints on the monopoly powerof foreign firms selling patented products indeveloping countries such as price caps orcompulsory licensing Our framework can eas-ily accommodate these cases as will becomeapparent in the next subsection

We now focus on the effects of potentialpatent enforcement in the quinolone segmentWe consider several scenarios which vary inthe number and size of domestic products that

23 Specifically the R-squares from the first-stage regres-sions of product group prices on instruments range from057 to 095 except for domestic ciprofloxacin for whichthe R-square is 017

24 Sparfloxacin is actually offered by one foreign subsid-iary in India (see Table 2) However its revenue share isminiscule We therefore treat Sparfloxacin as being pro-duced by domestic firms only 25 See Bradley Efron and Robert J Tibshirani (1993)

1493VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

will be removed from the market In particularwe consider the following five scenarios26

Withdrawal of one large domestic productgroup only domestic ciprofloxacin

Withdrawal of one relatively small domesticproduct group only domestic ofloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand norfloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand sparfloxacin

Withdrawal of all four domestic quinoloneproduct groups

As the list above suggests we proceed fromanalyzing the effects of single-product with-drawal to the analysis of eliminating the en-tire domestic segment This approach wasmotivated by early empirical results that in-dicated that the existence and extent of com-petition from domestic firms has a significantbearing on the predicted effects of patent en-forcement that is our predictions regardingprices and welfare vary substantially depend-ing on how many domestic products are af-fected by patent enforcement and on whethersome domestic competition will remain presentafter TRIPS In addition the size of the affecteddomestic groups is relevant for the welfare pre-dictions accordingly we examine both scenar-ios in which a large domestic product groupsuch as domestic ciprofloxacin is eliminatedfrom the market and scenarios in which the

eliminated domestic group is relatively small(eg domestic ofloxacin) In all scenarios wemaintain the assumption that the set of prod-ucts offered by the remaining firms in themarket does not change in response to patentenforcement

E Computation of Virtual Prices and NewEquilibrium Prices

The first step in the counterfactual analysisis to derive the new equilibrium prices underpatent enforcement27 In this context thereare two sets of prices that are relevant Thefirst set consists of the virtual prices of those(domestic) products that will not be availableonce TRIPS is put in effect To calculate thesevirtual prices we set the expenditures of therelevant products equal to zero The secondset of prices consists of the prices of thoseproducts that remain in the market In deriv-ing these prices we start by assuming profitmaximization without any form of price reg-ulation the firms remaining in the marketreoptimize in response to the policy changeand set new prices taking the prices of allother firms as given28 Of course at the equi-librium all prices change in response to thefact that some domestic products are nolonger present The new equilibrium pricesfor products that remain in the market are thuscomputed by utilizing the first-order condi-tions of profit-maximizing firms into whichthe virtual prices of the eliminated productsare substituted Hence to compute the newequilibrium prices we solve an equation sys-tem of the following form

26 As Table 2 indicates most of the patents for the drugsin the quinolone segment have expired by now so that noneof the scenarios described in this section is going to mate-rialize in practice The reason we consider these alternativescenarios in our counterfactual simulations is to get a senseof how the presence and extent of domestic competitionaffects welfare calculations Such calculations may becomerelevant in the future in other therapeutic classes in whichpatents have not expired yet and where domestic firmscompete with foreign patent holders While the particularexit scenarios will depend on the exact year of patent expiryof drugs in each therapeutic class and the specific way inwhich patents will be enforced (eg whether certain do-mestic drugs will be grandfathered in) the general conclu-sion that emerges from our calculations is that the consumerwelfare loss is substantially smaller in cases where somedomestic competition remains present in the market Hencethe particular way in which patent enforcement is imple-mented is essential for assessing its welfare impact

27 Of course until product patents are in fact introducedthese prices will not be observable Note also that we areassuming here that the range of available products will notchange with the introduction of patents

28 As mentioned above this first set of calculations ab-stracts from the existence of remaining price controls orother government regulations that would impose constraintson the firmsrsquo profit maximization problem Accordingly theresulting numbers should be interpreted as a benchmark ofwhat would happen if markets were completely unregulatedTo examine what the welfare effects of TRIPS would be inthe more realistic scenario of price regulation we subse-quently consider a scenario in which regulation keeps theprices of the products provided by patent-holders at theirpre-TRIPS level For a more detailed discussion see theresults section

1494 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

For products i that are withdrawn from themarket

(14) 0 i ir iiln pirV i10ln pjr j Di

10

j Di

01

i01ln pjr

j Di

00

i00ln pjr

ilnXQr

PQr

For products k that remain in the market

(15) pkr ckr 1 1

kk pkrpjr pirV 1

In the equations above pirV denotes the virtual

prices of the products that are removed fromthe market while pjr denotes the updatedprices of all other products Note that whensolving for the virtual prices we account forthe fact that both the price index for quino-lones PQr and the expenditure allocated tothis subsegment XQr need to be updated toreflect the fact that as a result of the pricechanges there may be substitution away fromthis subsegment To obtain the new quinoloneexpenditure XQr and the new price index PQrwe use the estimates and formulas for thehigher-level AIDS system In equation (15)ck refers to the marginal cost for product k thatwe have obtained from the previous estimationstage As mentioned in the previous subsectionwe conducted the simulations using both theupper and lower bounds for marginal cost ck

u

and ckL respectively The term kk(pkr

pjrpirV) re-

fers to the unconditional own-price elasticity forproduct k which is a function of the eliminatedproductsrsquo virtual prices and the remaining prod-uctsrsquo new equilibrium prices We conduct thecounterfactual simulations and welfare analysisat the regional level because the presence ofregion-specific product effects in the demandestimation implies region-specific demand elas-ticities and hence region-specific marginalcosts and markups The presentation and discus-sion of our results focus on national averages ofthe relevant variables that we construct by com-

puting weighted averages across regions usingthe population of each region as a weight

F Welfare Assessment

The simulation of the new equilibrium underpatent protection can provide important insightsinto how consumers and firms will respond tothe removal of domestic products in the market(for example toward which products consumerswill substitute or which prices will increase themost) To get a more precise idea of how peo-plersquos well-being will ultimately be affected byTRIPS we compute as a last step in our anal-ysis the welfare effects of the policy changeSocial welfare is defined as consumer welfareand the sum of domestic firm profits Thechange in domestic profits can easily be calcu-lated by comparing the domestic firm (variable)profits at the pre-TRIPS prices to the profitsthese firms will realize at the new simulatedprices Although foreign firm profits do notcount in domestic welfare calculations we alsocompute the effects of patent enforcement onforeign firm profits to get an idea of how largethe expected benefits of TRIPS are for thesefirms This provides in some sense an indirectway of assessing whether the claims that patentenforcement in countries like India will lead tomore research on developing-country-specificdiseases (such as malaria) have any validity iffor example we find that the effect of patentenforcement on the foreign firm profits realizedin India is small in magnitude it is unlikely thatforeign firms will engage in more developing-country-specific research in response to TRIPSIt is important to note that in all these calcula-tions we work with the lower-bound estimatesfor marginal costs since these give rise to thehighest possible markups Hence our estimateof profit loss for domestic producers most likelyoverstates this loss

On the consumer side we measure changes inconsumer welfare by the compensating variation(CV) defined as the additional expenditure thatconsumers need in order to achieve the sameutility level as before patent enforcement at thenew prices Specifically let P0 denote the pricevector before patent enforcement P the simulatedprice vector post-TRIPS (that we obtained usingthe methods described in the previous subsection)u0 the utility attained by consumers before TRIPS

1495VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

and E(u P) the higher-level expenditure functionThen the compensating variation is given by

(16) CV Eu0 P Eu0 P0

Note that the CV as computed in (16) representsthe combination of three effects

The pure product-variety effect that is theeffect that arises because one or more prod-ucts are not available to consumers anymoreholding the prices of all other remainingproducts and the total expenditure on thequinolone subsegment XQ constant

The expenditure-switching effect that is theeffect arising from substitution away fromquinolones and toward other subsegments ofthe antibiotics market again holding the pricesof all other remaining products constant

The reduced-competition effect that is theeffect that arises because the firms remainingin the market adjust (increase) their prices inresponse to the removal of domestic products

From both an analytical and a policy point ofview it is desirable to assess how large each ofthese effects is Accordingly we decompose thetotal effect on consumer welfare (the CV as givenby equation (16)) using the following procedure

To get the pure product variety effect wecompute virtual prices for the products that areremoved from the market holding the quino-lone expenditure XQ and the prices of all otherproducts fixed Let us call the resulting pricevector P1 Then the pure product variety effectis represented by E(u0 P1) E(u0 P0)

To compute the expenditure-switching effectwe compute another set of virtual prices againholding the prices of all remaining products fixedbut letting quinolone expenditure adjust in re-sponse to the new price index for the quinolonesegment (Given that the prices of the remainingproducts remain fixed the change in the priceindex arises only because of the removal of one ormore domestic products) Note that this scenariomost closely resembles the case in which patentlaws are enforced but strict price regulation keepsthe prices of the products offered in the market attheir pre-TRIP level Let us label the so-computedprice vector P2 The expenditure-switching effectis then E(u0 P2) E(u0 P1)

Finally the reduced competition effect aris-ing from higher prices for the remaining prod-

ucts is computed as the residual change in thecompensating variation once the product-varietyand expenditure-switching effects have been ac-counted for that is E(u0 P) E(u0 P2)where the price vector P is computed accordingto the formulas (14) and (15) to reflect theadjustment of prices to the new regime

To compute the standard errors associatedwith the counterfactual simulations (that is thestandard errors for the counterfactual prices andwelfare estimates) we again use bootstrappingWe first bootstrap the original sample (sales andprices of all the drugs) 300 times Next we esti-mate the AIDS model for each of these samplesand compute the counterfactual equilibrium pricesand welfare losses corresponding to each policyscenario for each of the 300 simulations In thefinal step we compute standard errors29

One limitation of our framework is that itdoes not allow for a heterogeneous response ofconsumers to the policy change Accordinglyour framework is not suited to addressing thequestion of how different groups in the popula-tion will be affected by patent law enforce-ment30 Along the same lines our frameworkdoes not accommodate the possibility of pricediscrimination which might lead to differentresults regarding welfare losses and profit gainsrelative to uniform pricing31 Still we believethat the results of the counterfactual simulationscan provide important insights into the likelyaggregate response to patent enforcement andthe factors that drive this response

III Results

A The Structure of Demand

Tables A2 and A3 in the Appendix display theresults from estimation of the lower- and upper-level AIDS systems respectively For ease of in-terpretation rather than discussing the coefficientestimates we focus our discussion on the implied

29 In cases some of the simulations do not converge wecompute the standard errors using only those simulationsthat converged

30 Addressing this question would require at a mini-mum micro data on consumer purchases as in Goldberg(1995) To our knowledge such data do not exist for theIndian pharmaceutical market

31 See Ernst R Berndt (1994) for an extensive discussionof uniform pricing

1496 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

unconditional price and expenditure elasticities re-ported in Table 6 Given that the region-specificproduct group effects ir reported in the last fourcolumns of Tables A2 and A3 imply region-spe-cific demand elasticities we report separate elas-ticities for each region in Tables 6A and 6B Asevident from the comparison of these tables theelasticities are very similar across regions so thatwe can focus the remaining discussion on theelasticities of one region only the northern regiondisplayed in Table 6A

The diagonal terms of Table 6A report theown-price elasticities which are in all but onecase negative and highly significant The oneexception is the foreign norfloxacin productgroupmdashwhose share of quinolone sales is 007percentmdashfor which we estimate a negative butinsignificant own-price elasticity For the remain-ing product groups demand appears to be highlyelastic with the estimated elasticities lower than2 in four out of the six cases The magnitude ofthe own-price elasticities matches the features ofthe Indian pharmaceutical market mentioned ear-lier which would suggest that Indian consumersare likely to be quite price-sensitive32 The elas-ticities appear especially large if one takes intoaccount that they refer to product groups (such asdomestic ciprofloxacin) and not individual drugsoffered by particular firms Their relative magni-tudes are also intuitive the drug with the largestmarket share and a relatively high price (domesticciprofloxacin) appears to be one of the least elas-tic In contrast foreign ciprofloxacin is highlyprice elastic this is plausible as ciprofloxacindrugs offered by subsidiaries of multinationalsface the stiffest competition in this market seg-ment from approximately 75 Indian firms offeringthe same molecule

The estimated expenditure elasticities appear inthe last column These are all positive indicatingthat the demand for all product groups is normalThe remaining cells display the estimated cross-price elasticities As one might perhaps expect forproducts within a therapeutic subsegment theseare mostly positive Out of a total of 49 priceelasticities we estimate there are six that do notconform to expectations these are the cross-price elasticities between different foreign prod-

uct groups which are estimated to be negativeand significant33 Fortunately these elasticitieshave negligible impact on the welfare analysisgiven that our counterfactuals focus on theeffect of withdrawing one or more domesticproducts from the market the most relevantelasticities are the ones that capture the responseof various product group shares to a change in theprice of one or more domestic groups these elas-ticities are the cross-price elasticities between var-ious domestic groups and the ones betweendomestic and foreign groups which are plausibleand precisely estimated34

Regarding these elasticities a striking aspectof our estimates is how large positive andsignificant the cross-price elasticities betweendifferent domestic product groups aremdashin factfor norfloxacin and ofloxacin we estimate thatdomestic product groups containing differentmolecules are closer substitutes for one anotherthan product groups that contain the same mol-ecule but are produced by foreign firms Incontrast for ciprofloxacin (the molecule withthe largest revenue share) we estimate a largepositive cross-price elasticity between the do-mestic and foreign versions

The fact that domestic products appear to beclose substitutes for other domestic productsthat contain different molecules truly representsan ldquoempiricalrdquo finding in the sense that we do

32 In developed economies elasticities of this magnitude havetypically been found only for generic drugs (and even then onlyrarely) or among consumers who lack health insurance

33 While these elasticities are clearly counterintuitive theyare not inconsistent with the underlying demand system whichimposes no restrictions on the sign of the cross-price elastici-ties We do not have a good explanation of why these elastic-ities are estimated to be negative A potential explanation isthat the shares of the foreign products are very small giventhis we observe very few consumers switching from one (verysmall) foreign group to another (very small) foreign groupwhen the price of the first foreign group goes up and hence theinference is not very reliable in this case

34 The cross-price elasticities between foreign drug groupscontaining different molecules will also have an effect on thewelfare estimates given that the withdrawal of domestic prod-ucts will generally lead to changes in the prices of all foreignproducts and market shares will be reallocated from someforeign products to other foreign products based on the newprices However this reallocation from ldquoforeign to foreignrdquo istruly second order in our case compared to the reallocationfrom ldquodomestic to domesticrdquo and ldquodomestic to foreignrdquo prod-ucts In scenarios in which we consider patent enforcementaccompanied by strict price regulation the cross-price elastic-ities between foreign product groups are in fact completelyirrelevant as the prices of foreign products are not allowed toincrease in these cases Still our welfare loss estimates remainsubstantial drivenmdashas beforemdashby the loss of product variety

1497VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

not impose it through any of our assumptionsregarding the demand function The questionthat naturally arises then is what might explainthis finding While we cannot formally addressthis question anecdotal accounts in various in-dustry studies suggest that the explanation maylie in the differences between domestic andforeign firms in the structure and coverage ofretail distribution networks

Distribution networks for pharmaceuticals inIndia are typically organized in a hierarchicalfashion Pharmaceutical companies deal mainlywith carrying and forwarding (CampF) agents inmany instances regionally based who each sup-ply a network of stockists (wholesalers) Thesestockists in turn deal with the retail pharma-cists through whom retail sales ultimately oc-cur35 The market share enjoyed by a particularpharmaceutical product therefore depends inpart on the number of retail pharmacists who

stock the product And it is here that thereappears to be a distinction between domesticfirms and multinational subsidiaries In particu-lar the retail reach of domestic firms as agroup tends to be much more comprehensivethan that of multinational subsidiaries (IndianCredit Rating Agency (ICRA) 1999)36

There appear to be two reasons for this Thefirst is that many of the larger Indian firmsbecause they have a much larger portfolio ofproducts over which to spread the associatedfixed costs typically have more extensive net-works of medical representatives The second issimply that there are many more domestic firms(and products) on the market At the retail levelthis would imply that local pharmacists mightbe more likely to stock domestic products con-taining two different molecules say ciprofloxa-cin and norfloxacin than they would domesticand foreign versions of the same molecule Tothe extent that patients (or their doctors) arewilling to substitute across molecules in order tosave on transport or search costs (eg going toanother pharmacy to check whether a particular

35 There are estimated to be some 300000 retail pharma-cists in India On average stockists deal with about 75 retailers(ICRA 1999) There are naturally variations in this structureand a host of specific exclusive dealing and other arrangementsexists in practice Pharmaceutical firms also maintain networksof medical representatives whose main function is to marketthe companyrsquos products to doctors who do the actual prescrib-ing of drugs In some instances firms do sell directly to thedoctors who then become the ldquoretailerrdquo as far as patients areconcerned but these are relatively rare

36 These differences were also highlighted in conversa-tions one of the authors had with CEOs and managingdirectors of several pharmaceutical firms as part of a sep-arate study

TABLE 6AmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN THE NORTHERN REGION

Product group

Elasticity with respect to

Prices of foreign productgroups Prices of domestic product groups Overall

quinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 557 013dagger 015 401 011dagger 011dagger 016 137(179) (007) (007) (184) (006) (006) (006) (029)

Foreign norfloxacin 427dagger 045 427dagger 350dagger 602 451 465 220(242) (112) (242) (210) (623) (184) (183) (105)

Foreign ofloxacin 011 010dagger 138 009 009dagger 023 011 116(005) (005) (031) (027) (005) (028) (004) (017)

Domestic ciprofloxacin 018 001 001 168 008 008 010 117(008) (000) (001) (023) (002) (002) (002) (003)

Domestic norfloxacin 004 003 004 058 223 042 040 073(001) (003) (001) (017) (011) (004) (003) (009)

Domestic ofloxacin 005 005 011 077 074 342 074 089(002) (002) (013) (028) (008) (025) (008) (021)

Domestic sparfloxacin 007 004 007 115 063 063 288 028(002) (001) (002) (015) (006) (006) (017) (012)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent significance level and dagger (dagger) denotes significance at the 10-percent level

1498 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE 6BmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN OTHER REGIONS

Demand patterns in the eastern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 594 014dagger 016 431 013 011 017 140(195) (008) (008) (201) (006) (007) (007) (031)

Foreign norfloxacin 329dagger 058 329dagger 264dagger 460 346 359 192(180) (083) (179) (153) (451) (138) (139) (082)

Foreign ofloxacin 012 010dagger 140 010 010 024 013 117(006) (006) (034) (031) (005) (030) (005) (019)

Domestic ciprofloxacin 018 001 001 172 009 008 011 117(008) (000) (002) (027) (002) (003) (002) (003)

Domestic norfloxacin 004 004 004 068 242 049 046 070(002) (004) (002) (020) (013) (005) (004) (010)

Domestic ofloxacin 004dagger 004 009 061 060 295 060 092(002) (001) (010) (028) (007) (022) (007) (017)

Domestic sparfloxacin 005 003 005 092 048 051 251 043(002) (001) (001) (017) (005) (006) (016) (010)

Demand patterns in the western region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 427 009dagger 011 286 008 006 011 126(129) (005) (005) (133) (004) (005) (004) (021)

Foreign norfloxacin 714dagger 008 710dagger 594dagger 995 742 770 299dagger

(369) (183) (368) (328) (1015) (297) (296) (178)Foreign ofloxacin 013 012dagger 145 009 012 026 013 119

(007) (006) (037) (029) (005) (033) (005) (021)Domestic ciprofloxacin 019 001 000 174 009 008 011 118

(008) (000) (001) (024) (002) (003) (002) (004)Domestic norfloxacin 004 004 004 067 243 050 047 069

(002) (004) (001) (018) (014) (005) (004) (010)Domestic ofloxacin 003 003 007 048 048 257 048 093

(002) (001) (008) (022) (005) (015) (004) (014)Domestic sparfloxacin 006 003 005 083 046 049 241 046

(002) (001) (001) (014) (004) (005) (012) (009)

Demand patterns in the southern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 560 013dagger 015 402 011 010 017 137(182) (007) (007) (187) (006) (006) (007) (029)

Foreign norfloxacin 533dagger 031 531dagger 431dagger 748 559 584 249dagger

(301) (134) (300) (245) (734) (219) (219) (128)Foreign ofloxacin 014 012dagger 148 011 012 029 016 120

(007) (007) (039) (033) (006) (035) (006) (022)Domestic ciprofloxacin 017 001 000 169 008 007 011 116

(008) (000) (001) (025) (002) (003) (002) (003)Domestic norfloxacin 004 003 003 059 225 043 040 073

(001) (003) (001) (019) (012) (004) (004) (009)Domestic ofloxacin 004dagger 004 009 062 060 296 060 091

(002) (001) (010) (026) (006) (019) (007) (017)Domestic sparfloxacin 008 005 008 138 073 075 323 016

(002) (001) (002) (018) (007) (008) (021) (014)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at 5-percent level anddagger (dagger) denotes significance at the 10-percent level

1499VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

foreign product is in stock) in aggregate datawe would expect to find precisely the substitu-tion patterns that we report in Table 6

Whether the particular explanation we provideabove is the correct one the high degree of sub-stitutability between domestic product groupsturns out to have important implications for thewelfare calculations We discuss these in moredetail below when we present the results of thecounterfactual welfare analysis Another elasticitywith important implications for the counterfactu-als is the price elasticity for the quinolone subseg-ment as a whole which indicates how likelyconsumers are to switch to other antibioticsgroups when faced with a price increase for quin-olones This elasticity is computed on the basis ofthe results in Table A3 and it is at 111 (stan-dard error 024) this is large in magnitudebutmdashas expectedmdashsmaller in absolute value thanthe own-price elasticities of the product groupswithin the quinolone subsegment

The results in Tables 6A and 6B are based onour preferred specification discussed in SectionII In Tables A4 to A6 in the Appendix weexperiment with some alternative specificationsTables A4(a)ndashA4(c) correspond to a specificationthat includes in addition to product-group-specificregional fixed effects product-group-specific (andfor the upper level antibiotics-segment-specific)seasonal effects We distinguish among three sea-sonsmdashthe summer monsoon and wintermdashandreport the unconditional demand elasticities for

the northern region for each of these seasons Asevident from the tables in the Appendix our elas-ticity estimates are robust to the inclusion of sea-sonal effects The demand elasticities in Table A5are based on estimation of the demand system byOLS Compared to the elasticities obtained by IVthe OLS elasticities are smaller in absolute valueimplying that welfare calculations based on theOLS estimates would produce larger welfare lossestimates Nevertheless some of the patterns re-garding the cross-price elasticities discussed ear-lier are also evident in the OLS results inparticular the cross-price elasticities between dif-ferent domestic product groups are all positivelarge and significant and in most instances largerthan the cross-price elasticities between drugs thatcontain the same molecule but are produced byfirms of different domesticforeign status Theclose substitutability of domestic products indi-cated by both the OLS and IV estimates seems tobe one of the most robust findings of the paper

B Cost and Markup Estimates

Table 7 displays the marginal costs markupsand profits implied by the price elasticity esti-mates of Tables 6A and 6B for each of the sevenproduct groups Given that our regional effectsimply different price elasticities for each regionour marginal cost and markup estimates alsodiffer by region Given however that based on

TABLE 7mdashUPPER AND LOWER BOUNDS FOR MARGINAL COST MARKUP AND ANNUAL PROFIT BY PRODUCT GROUPS WITHIN

THE QUINOLONE SUBSEGMENT

Product groupLower boundfor MC (Rs)

Upper boundfor markup

Upper boundfor profit(Rs mill)

Upper boundfor MC (Rs)

Lower boundfor markup

Lower boundfor profit

(Rs)

Foreign ciprofloxacin 83 19 269 103 0 00(123) (012) (1655)

Foreign norfloxacin NA NA NA 53 0 00Foreign ofloxacin 323 70 1061 1085 0 00

(2316) (021) (3185)Domestic ciprofloxacin 47 59 17019 112 0 00

(114) (010) (29858)Domestic norfloxacin 52 43 2807 90 0 00

(020) (002) (1532)Domestic ofloxacin 587 34 1612 901 0 00

(218) (002) (1280)Domestic sparfloxacin 495 37 1985 788 0 00

(157) (002) (1100)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level Estimated lower bound forforeign norfloxacinrsquos marginal cost is negative since the estimated price elasticity is less than one in absolute value

1500 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Tables 6A and 6B the price elasticities do notseem to differ substantially across regions wereport for ease of exposition only the nationalaverages for marginal costs and markups inTable 7

Table 7 has two parts In the left part (firstthree columns) we report the lower bound formarginal cost and the corresponding upperbound for markup and upper bound for totalannual profit for each product group Thesenumbers are based on the price elasticities weobtained from estimating the two-level AIDSdemand system Since we do not have a reliableestimate for the price elasticity of foreign nor-floxacin (the point estimate is negative but lessthan one in absolute value and insignificant)we cannot compute the lower bound for mar-ginal cost in this case It is important to note thatthese estimates do not reflect either the actualmarginal cost or the actual markup for thesedrugs both because the existence of price reg-ulation implies that the unconstrained first-orderconditions are unlikely to hold each period andbecause our aggregation across firms of thesame domesticforeign status supplying thesame molecule makes the interpretation of theseestimates problematic In particular the fact thatwe ignore competition among firms within eachproduct group implies that our estimates willtend to overstate market power These numberswill prove useful however in the counterfac-tual simulations as they can give us a sense ofhow large the maximum profit gains for multi-national firms and the maximum profit lossesfor domestic firms are likely to be under patentenforcement The right part of the table (lastthree columns) reports the upper bound for mar-ginal cost which we obtain by simply taking themarginal cost to be equal to the observed priceWe do not report standard errors in this casesince the numbers are based on actual data Themarkups corresponding to these marginal costupper bounds are of course zero We conductthe counterfactual simulations using both thelower and upper bounds for marginal costs

The (upper bounds for) markups on the leftside of the table are generally plausible Thedomestic ciprofloxacin product group that dom-inates the quinolone subsegment and for whichwe documented high prices a high marketshare and a relatively low elasticity of demandenjoys one of the highest markup upper bounds(60 percent) and accounts for nearly 70 percent

of all profits derived within the subsegmentForeign ciprofloxacin on the other hand whichfaces the stiffest competition from domesticfirms and for which we estimated a highly elas-tic demand has the lowest markup upper bound(19 percent)

C Counterfactual Estimates of the Impact onPrices and Welfare

With estimates of the key demand and costparameters in hand we turn to the counterfac-tuals We consider the five separate scenarioslisted in the previous section All of the scenar-ios involve the withdrawal of one or more of thedomestic product groups from the market Ta-ble 8 displays our estimates of the consumerwelfare losses that result under the differentscenarios The losses are expressed in billion ofrupees per year All numbers presented in Ta-ble 8 and subsequent tables are based on usingthe lower bounds for marginal cost and upperbounds for markup in the simulations As dis-cussed earlier these numbers are the more in-teresting to work with since they give us anupper bound for the changes in the profits ofdomestic and foreign firms that would resultfrom patent enforcement In the Appendix wealso present results based on using the upperbounds for marginal costs in the simulations inwhich case the pre-TRIPS profits of domesticand foreign firms are zero In all cases marginalcosts are assumed to be constant in outputWhile naturally the profit implications differdepending on whether one uses the upper orlower bounds for marginal costs (firm profits arezero if one assumes the upper bound of mar-ginal cost in which case price equals marginalcost) the estimated consumer welfare losses aresimilar in the two cases We discuss these re-sults in more detail at the end of this section

The first column presents our estimates of theconsumer welfare losses attributable to the pureloss of product variety effect where we fix theprices of all remaining products as well as theoverall expenditure on quinolones while with-drawing one or more of the domestic productgroups Note that in our initial specification ofthe demand system had we not allowed for thepossibility that consumers might differentiatebetween domestic and foreign products evenwhen they contain the same molecule this

1501VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

particular component of the loss of consumerwelfare would not have arisen

The estimates reported in the second columnincorporate the expenditure-switching effect ontop of the loss of product variety Here basedupon the price elasticity estimates from thehigher-level AIDS system we adjust (down-ward) the expenditures allocated to the quino-lone subsegment as the composite price ofquinolones effectively increases as a consequenceof the higher virtual prices of the domestic productgroups that are withdrawn from the market Be-cause the estimates in this column are generatedassuming that the prices of the products that re-main in the market are not adjusted upward theyprovide a sense of what consumer losses would beif the introduction of product patents were coupledwith strict price regulation aimed at maintainingprices at prepatent levels Alternatively they canbe thought of as the relevant welfare numbers ifintense competition among firms within the re-maining product groups kept the prices of theproducts that were still offered in the market closeto the firmsrsquo marginal costs

The last column of Table 8 displays the esti-mated consumer welfare losses when bothcross-segment expenditure-switching and with-in-segment upward price adjustments are takeninto account

If we compare the results across the firstsecond and third columns all the counterfac-tual scenarios produce qualitatively similar pat-terns patterns that are consistent with what we

would expect Starting from the initial loss ofwelfare attributable to the loss of product vari-ety the option of switching expenditures outof the quinolone subsegment to other subseg-ments mitigates some of the initial welfareloss But if we then incorporate the upwardprice adjustments that result in response to thereduced competition the welfare losses aremagnified

Of particular interest from a policy perspec-tive are the relative magnitudes of these threeeffects which are similar under all the counter-factual scenarios though absolute levels varyconsiderably First despite the fact that the de-mand for quinolones is quite sensitive to thecomposite price of quinolonesmdashthe upper levelprice elasticity is 111mdashthe cross-subsegmentexpenditure-switching effects are in all thecases small (in absolute value terms) relative tothe other two effects For instance under thescenario where all the domestic quinolone prod-uct groups are withdrawn from the market theoverall consumer welfare loss of Rs 1781 bil-lion per year can be decomposed into an initialloss of Rs 1176 billion (66 percent) attribut-able to the loss of product variety a slightreduction in this initial loss of Rs 041 billion(2 percent) from Rs 1176 billion to Rs1135 billion because of expenditure switchingand a subsequent additional loss of Rs 646billion (36 percent) from Rs 1135 billion toRs 1781 billion because of the reduced com-petition and consequent price increases

TABLE 8mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure loss ofvariety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 732(087) (089) (146)

Only ofloxacin 008 008 023(008) (008) (010)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1253(177) (180) (415)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1058(142) (145) (331)

All four domestic quinolones products 1176dagger 1135dagger 1781(643) (634) (1270)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent significance level and dagger (dagger)denotes significance at the 10-percent level

1502 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 2: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

sparse is the evidence on which these divergentclaims are based1 Apart from the findings of asmall number of studies we refer to in moredetail below little is known about the extent towhich pharmaceutical prices in less-developedeconomies might increase with the introductionof product patents and the magnitude of theassociated welfare losses2 Past empirical stud-ies on the impact of patents on prices and inno-vative activity in various sectors includingpharmaceuticals have focused almost exclu-sively on developed economies Aside from thefact that none of these studies estimates welfareeffects the conclusions from these studies arenot directly pertinent to the TRIPS debate be-cause the structure of demand for pharmaceuti-cals in less-developed economies differs fromthat in developed economies in several criticalrespects3

Any assessment of the potential price andwelfare effects of TRIPS needs therefore to bebased on a better empirically grounded under-standing of the characteristics of demand andthe structure of markets for pharmaceuticals inpoor developing economies To what extent areconsumers willing to trade off lower prices forolder possibly less effective therapies Howdoes this vary across different therapeutic seg-ments Are consumers willing to pay a pre-mium for the pedigree and brand reputation ofproducts marketed by subsidiaries of foreignmultinationals How competitive are pharma-ceutical markets The welfare of consumers

depends on the pricing strategies and decisionsof pharmaceutical firms But these in turn de-rive from the firmsrsquo assessment of the structureof market demand If consumers are unwillingto pay substantially more for newer patenteddrugs for which there exist older possiblyslightly less effective generic substitutes theability of patent-holders to charge a premiumwill be limited As mentioned above there havebeen a few studies that carefully consideredthese issues and used explicit models of con-sumer and firm behavior to simulate the welfarelosses implied by patent protection4 Their find-ings are ultimately limited however by the factthat the simulations that are used to evaluate thepotential impact of patents are in each instancebased on assumptions about demand character-istics and market structure rather than on actualestimates of the relevant parameters

This paper takes a first step toward filling thisgap We provide the first rigorously derivedestimates of the possible impact of pharmaceu-tical product patents on prices and welfare in adeveloping economy Using detailed product-level data on monthly pharmaceutical pricesand sales over a two-year period from January1999 to December 2000 we estimate key priceand expenditure elasticities and supply-side pa-rameters for the fluoroquinolone (quinolonehenceforth5) segment of systemic antibacterials(ie antibiotics) in the Indian pharmaceuticalsmarket We chose this segment both because itcontains several products that were still underpatent in the United States during our sampleperiod and because antibiotics are importantfrom a public health policy point of view (com-pared to for example Prozac Viagra or otherlifestyle drugs that were also under patent pro-tection in the United States during this period)We then use our estimates to carry out counter-factual simulations of what prices profits (ofboth domestic firms and subsidiaries of foreignmultinationals) and consumer welfare wouldhave been had the quinolone molecules westudy been under patent in India as they were in

1 There is a sizeable theoretical literature on the welfareimpact of patent protection that generally finds that theeffects of patents in a multicountry setting are substantiallymore complicated than their respective effects in a singleclosed economy where both innovating firms and innova-tion beneficiaries (ie consumers) are of the same nation-ality (see William Nordhaus 1969 Judith C Chen andGene M Grossman 1990 Ishac Diwan and Dani Rodrick1991 Alan V Deardorff 1992 Elhanan Helpman 1993and Grossman and Edwin L-C Lai 2004) Empirical workin this area is however still in its infancy

2 Even less is known about the other central questionsrelevant to the TRIPS debate namely the extent to whichpharmaceutical research and product development prioritiesare likely to shift as a result of TRIPS and how large thewelfare benefits of any therapeutically innovative drugs thatresult from this shift are likely to be The only paper that hascarefully addressed such questions is Jean O Lanjouw andIain M Cockburn (2001)

3 For a representative example of estimation of pharma-ceuticals demand in developed countries see Sara F Ellisonet al (1997)

4 See for instance Pablo M Challu (1991) Carsten Fink(2000) Keith E Maskus and Denise E Konan (1994) JulioJ Nogues (1993) Arvind Subramanian (1995) and Ja-yashree Wattal (2000)

5 Technically the term ldquofluoroquinolonesrdquo refers to thelatest generation of quinolones Older quinolones however(eg nalidixic acid) have market shares close to zero

1478 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

the United States at the time The presence ofmany therapeutic substitutes within the antibi-otics segment make this product category idealfor investigating the claim that the presence ofclose substitutes will prevent drug prices fromrising once patent protection is enforced Ofcourse to the extent that our estimates refer toantibiotics they are not directly applicable toother pharmaceutical product categories thatmay have different demand structures For ex-ample a finding of large substitution effectstoward nonpatented products would not neces-sarily apply to a market segment with only fewor possibly no therapeutic substitutes Still afinding of limited substitution toward otherdrugs and associated large price increaseswould suggest that the effects of patent enforce-ment in other pharmaceutical segments withfewer therapeutic substitutes might be evenlarger

India provides a natural setting for our anal-ysis for a number of reasons It is a leadingexample of a low-income country that did notrecognize pharmaceutical product patents at thetime the TRIPS agreement went into effect Infact during the Uruguay round of negotiationsIndia led the opposition to the TRIPS articlesmandating pharmaceutical product patents Interms of the structure of demand India is aprototypical example of a low-income countrywith a large number of poor households whichbecause health insurance coverage is nonexist-ent have to meet all medical expenses out ofpocket Moreover the disease profile of theIndian population mirrors that of many otherlow-income countries and is considerably dif-ferent from that of most developed economiesLastly the domestic Indian pharmaceutical in-dustry which as of 2002 was the largest pro-ducer of generic drugs in the world in terms ofvolume is typical of that of many middle-in-come countries with large numbers of small andmedium-sized firms with significant imitativecapabilities producing and marketing drugs do-mestically that are under patent elsewhere

During the period covered by our data sev-eral molecules in the quinolone family were stillunder patent in the United States but productscontaining these molecules were being pro-duced and distributed in India by both a numberof domestic firms and a number of local subsid-iaries of foreign multinationals We aggregatethese products into a number of mutually exclu-

sive product groups where within each productgroup all products contain the same quinolonemolecule (eg ciprofloxacin or norfloxacin)and are produced by firms with the same do-mestic or foreign status We then estimate atwo-level demand system employing the Al-most Ideal Demand System (AIDS) specifica-tion of Angus S Deaton and John Muellbauer(1980b) in both levels The higher level corre-sponds to the allocation of expenditures tovarious subsegments within the systemic anti-bacterials segment of the market At the lowerlevel we estimate the parameters relevant forthe allocation of expenditures within the quin-olone subsegment to the various productgroups within this subsegment (eg foreignciprofloxacin domestic ciprofloxacin or do-mestic norfloxacin)

With these estimates in hand we turn to thecounterfactuals The basic counterfactual sce-narios we consider all involve the withdrawal ofone or more of the domestic quinolone productgroups from the market The idea here is thathad US patents for say ciprofloxacin beenrecognized in India all domestic products con-taining ciprofloxacin would not be present in themarket That would leave only the foreign cip-rofloxacin product group in the market Usingour estimates of the own cross-price and ex-penditure elasticities of the various productgroups as well as estimates for the upper andlower bounds of marginal costs of productionwe are able to simulate the prices and marketshares that would obtain under each of the sce-narios Moreover using the expenditure func-tion associated with the higher-level AIDSspecification we are able to calculate the wel-fare lossmdashmeasured in terms of the compensat-ing variation ie the additional expenditurethat the representative Indian consumer wouldneed to incur to maintain her utility level inthe face of the domestic product withdraw-al(s) and the accompanying price and marketshare changesmdashunder each of the counterfac-tual scenarios

Apart from the fact that our counterfactualsimulations are based on estimated rather thanassumed parameter values this paper buildsupon the earlier studies in two substantive and(it turns out) empirically important ways

First by accommodating the possibility thatconsumers may differentiate between domesticand foreign products even when these products

1479VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

contain the same patentable molecule we allowfor an additional channel through which theintroduction of product patents and the conse-quent withdrawal of domestic products may ad-versely affect consumers that is through theloss of product variety In contrast previousstudies on developing countries assume thatconsumers are indifferent between foreign anddomestic products that contain the same mole-cule What this implies is that any adverse wel-fare effects are realized only through increasedprices The difference is most evident if weconsider a scenario under which domestic prod-ucts are forced to withdraw from the marketbecause of the introduction of product patentsbut strict price regulations maintain prices atprepatent levels In our approach consumerswould still experience a welfare loss whereas inthe framework adopted in earlier studies such ascenario would entail no loss of welfare

Empirically the component of the consumerwelfare loss attributable to the reduction of va-riety from the withdrawal of domestic productsturns out to be significant We interpret thiscomponent as capturing primarily an ldquoease ofaccessrdquo effect due to differences in the market-ing and distribution networks domestic prod-ucts are more readily available to Indianconsumers than products produced by foreignsubsidiaries From a policy perspective thissuggests a possible role for compulsory licens-ing in addition to or in lieu of price regulationsince the latter by itself will not alleviate thewelfare loss due to loss of variety Alterna-tively one could argue thatmdashto the extent thatthe loss we attribute to the reduction of productvariety is due to the fact that the current productportfolios and distribution networks of foreignproducers are limitedmdashit is purely a transitionalphenomenon and should thus not be included inthe welfare calculations This is a controversialpoint we discuss in detail in the results sectionIf foreign firms respond to patent enforcementby investing in distribution networks or by us-ing licensing agreements with domestic firms tomake their products more readily available toIndian consumers the ldquoease of accessrdquo effectwould indeed diminish in importance in thelonger run though of course it could be signif-icant in the first years after patent enforcementWhether these investments will materializehowever is open to debate If TRIPS is accom-panied by price regulation in order to limit price

increases in poor developing countries the in-centives of multinationals to invest in marketingand distribution in these countries may dimin-ish At any rate to take into account the possi-bility that the welfare loss due to the reductionof variety is a temporary phenomenon we alsopresent a more conservative welfare loss esti-mate by subtracting the ldquoproduct varietyrdquo com-ponent from our total loss estimate This givesus a lower bound estimate that is due to priceincreases alone Though it is only about a thirdof our upper bound estimate this lower boundestimate is still very large in absolute termsrepresenting 24 percent of antibiotic sales in2000

A second and perhaps even more importantmethodological difference between this paperand earlier studies is that we allow for andflexibly estimate a range of cross-product-group and cross-molecule substitution effectsIn contrast cross-price effects are ignored inearlier studies To see why cross-price effectsare likely to alter estimated welfare effects sig-nificantly in this context imagine a scenariowhere the introduction of patents leads to mo-nopoly pricing in the market for a particularpatentable molecule If the markets for potentialsubstitutes are imperfectly competitive then theincrease in price in the original patentable mar-ket will lead to corresponding upward priceadjustments in the related markets as producersof substitute products reoptimize in the face ofthe increased demand for their products Themagnitude of any upward adjustments will nat-urally vary with the degree of competition inrelated markets and with the strength of thecross-price effects But as long as the cross-price effects are positive and related marketsare not perfectly competitive the loss of con-sumer surplus because of monopoly pricing inone market will be multiplied through the rippleeffects of upward price adjustments in relatedmarkets

If this were just a theoretical possibility itwould not be of much interest These multipliereffects however turn out to be substantial inour counterfactual scenarios Most strikinglythe estimated loss of consumer welfare from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenar-

1480 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

ios in each of which only one of the domesticproduct groups is withdrawn What this veryclearly indicates is that past studies that haveestimated the aggregate effects of patent protec-tion by adding up the losses estimated sepa-rately in each of a number of patentablemarkets may have substantially underestimatedthe magnitude of the consumer welfare lossesfrom the introduction of pharmaceutical productpatents

In absolute terms we estimate that in theabsence of any price regulation the prices offoreign patented products would rise between100 percent and 400 percent In the more real-istic case of some form of price regulation thatwould keep drug prices fixed at their pre-TRIPSlevel the total annual welfare losses to theIndian economy from the withdrawal of all fourdomestic product groups in the quinolone sub-segment would be on the order of Rs 1370billion or about 50 percent of the sales of theentire systemic antibacterials segment in 2000At the then-prevailing exchange rate this trans-lates into a figure of US$305 million Of thisamount foregone profits of domestic producersconstitute roughly Rs 23 billion or US$50 mil-lion The overwhelming portion of the totalwelfare loss therefore derives from the loss ofconsumer welfare

The welfare loss we estimate represents onlythe static costs of patent enforcement arisingfrom pricing distortions and reduction in prod-uct variety Our approach does not address thepotential dynamic benefits of innovations thatmay result from international property rightsprotection Nevertheless we believe that esti-mating these static costs is important wheneverthere is a radical change in policymdashwhichTRIPS represents for a good part of the devel-oping world Even if there is the potential forlong-term benefits knowledge of the short-runcosts is important for designing an appropriatepolicy response that will potentially mitigate theadverse short-run impact Having said that itis worth noting that according to our esti-mates the total profit gains of patent enforce-ment to foreign producers in the absence ofany price regulation would be only aboutUS$53 million per year With price regulationthat would keep the prices of drugs suppliedby multinational subsidiaries at their pre-TRIPS level the profit gains become onlyUS$196 million per year

The remainder of this paper is organized asfollows In the next section we lay out theessential features of the Indian pharmaceuticalsmarket provide more detail about the segmentsthat we focus on in the empirical analysis andbriefly describe the primary data we use Sec-tion II describes the analytic framework and theeconometric strategy we use to estimate therelevant parameters and construct the counter-factual scenarios We discuss our results in Sec-tion III Section IV concludes

I The Setting and the Data

Between April 1972 when the Indian PatentsAct (1970) became effective and March 2005when Indiarsquos parliament passed the 3rd Amend-ment of the Patents Act India did not recognizeproduct patents for pharmaceuticals The IndianPatents Act which replaced the inherited Brit-ish colonial law regarding intellectual propertyrights specifically excluded pharmaceuticalproduct patents and admitted process patentsonly for a period of seven years In contrast thelatest amendment recognizes patents on endproducts that under the new regime will remainin force for 20 years6

The two stated objectives of the 1970 actwere the development of an indigenous pharma-ceuticals industry and the provision of low-costaccess to medicines for Indian consumers Con-sistent with these objectives and with the broaderleftward tilt in policy a number of other measureswere introducedmdashdrug price controls restrictionson capacity expansion limits on multinational eq-uity sharesmdashthat in the years since have on theone hand kept pharmaceutical prices low and onthe other encouraged the development of the In-dian pharmaceutical industry Many of these reg-ulations and restrictions have been lifted or easedsince the mid-1980s with marked acceleration inthe pace of liberalization during the 1990s

Over the last 20 years the Indian pharmaceu-tical industry has grown rapidly to the pointwhere it is now the worldrsquos largest producer offormulations in terms of volume and one of the

6 Indian companies that are now producing drugs forwhich patent applications were submitted between the sign-ing of the TRIPS agreement in 1995 and January 1 2005will be allowed to continue producing if they pay a royaltyto the patent holder

1481VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

worldrsquos largest producers of bulk drugs7 Thestructure of the industry has also evolved In1970 the industry was dominated by multina-tional subsidiaries by 2001 Indian-owned firmswere not just the leading players in the industrymany had also become major exporters

The data we use in this paper are from theretail pharmaceutical audits of ORG-MARGIndiarsquos premier market research and consultingfirm The audit provides detailed product-levelinformationmdashestimates of monthly retail salesin each of the four geographic zones of Indiaprice dosage form launch date brand namechemical name therapeutic categorizationmdashon

all pharmaceutical products sold in India byabout 300 of the largest firms representingroughly 90 percent of domestic retail sales ofpharmaceuticals The coverage of the audit isextensive reaching a representative panel ofthousands of retail chemists in over 350 citiesand towns The data collected which providethe only real source of disaggregate informationon the Indian pharmaceutical market are usedby both the government of India in formulatingpricing policy and other decisions and the In-dian pharmaceutical industry in determiningpricing and marketing strategies We have in-formation at a monthly frequency for the periodof January 1999 to December 2000 Tables 1 to3 provide a set of descriptive statistics that areessential for understanding the focus of ouranalysis and interpreting our results

As noted earlier the characteristics of de-mand for pharmaceuticals in India are likely todiffer considerably from those in developedeconomies With a share of 23 percent the

7 Bulk drugs are the therapeutically relevant active phar-maceutical ingredients that are combined with a variety ofinactive ingredients to make the formulations ultimatelyconsumed by patients Firms in the pharmaceutical sectorcan be one of three types bulk drugs producers pureformulators or integrated firms which produce both bulkdrugs and market formulations

TABLE 1mdashTHE QUINOLONES SUBSEGMENT

Molecule

Shares ()of quinolones sales Sales (Rs mill) 2000

Domesticfirms

Foreignsubsidiaries

Domesticfirms

Foreignsubsidiaries

Ciprofloxacin 530 27 3030 156Norfloxacin 112 01 640 3Ofloxacin 116 31 665 177Sparfloxacin 108 01 620 4Lomefloxacin 15 86 Pefloxacin 13 01 72 5Levofloxacin 00 0 Nalidixic acid 13 73

TABLE 2mdashBASIC INFORMATION ABOUT THE FOUR QUINOLONES MOLECULES 2000

Ciprofloxacin Norfloxacin Ofloxacin Sparfloxacin

US or European patent holder Bayer Merck Ortho-McNeil Rhone-PoulencYear of US patent expiry 2003 1998 2003 2010Year of US-FDA approval 1987 1986 1990 1996Year first introduced in India 1989 1988 1990 1996No of domestic Indian firms 75 40 17 25No of foreign subsidiaries 8 2 2 1Sales weighted average per-unit API

price of products produced byDomestic Indian firms 1124 905 9008 7884Foreign subsidiaries 1035 528 10847

API Active pharmaceutical ingredient

1482 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

antiinfectives segment ranks second in Indiawhereas in the world market it is fifth and has ashare of only 90 percent Hence antiinfectivesare important in India not only from a health

and public policy point of view but also as asource of firm revenue

With this in mind we focus on one partic-ular subsegment of antiinfectives namely the

TABLE 3mdashSUMMARY STATISTICS FOR THE QUINOLONES SUBSEGMENT 1999ndash2000

North East West South

Annual quinolones expenditure per household (Rs) 3125 1975 2764 2359(366) (367) (407) (286)

Annual antibiotics expenditure per household (Rs) 11988 8424 11052 9624(1224) (1224) (960) (996)

No of SKUsForeign ciprofloxacin 1238 1129 1308 1246

(150) (190) (102) (106)Foreign norfloxacin 183 171 200 158

(070) (075) (088) (083)Foreign ofloxacin 304 296 296 300

(086) (086) (091) (088)Domestic ciprofloxacin 10621 9763 10342 10550

(599) (434) (722) (451)Domestic norfloxacin 3896 3496 3617 3942

(271) (268) (251) (379)Domestic ofloxacin 1846 1600 1725 1725

(680) (634) (586) (635)Domestic sparfloxacin 2983 2829 3121 2929

(557) (638) (688) (657)Price per-unit API (Rs)

Foreign ciprofloxacin 958 1090 1085 1007(128) (066) (071) (058)

Foreign norfloxacin 563 509 605 435(077) (133) (139) (147)

Foreign ofloxacin 10946 10943 10886 10612(620) (664) (700) (1140)

Domestic ciprofloxacin 1143 1067 1131 1152(016) (015) (017) (013)

Domestic norfloxacin 951 907 888 873(024) (035) (037) (020)

Domestic ofloxacin 9163 8964 8565 9341(1615) (1565) (1422) (1407)

Domestic sparfloxacin 7972 7849 7688 8028(976) (1014) (1185) (1037)

Annual sales (Rs mill)Foreign ciprofloxacin 4179 2431 4520 2947

(1534) (816) (1273) (648)Foreign norfloxacin 128 100 058 073

(101) (082) (044) (057)Foreign ofloxacin 5446 3184 3522 3111

(1399) (933) (906) (703)Domestic ciprofloxacin 96229 58591 67874 70381

(10626) (13026) (12226) (8740)Domestic norfloxacin 22255 11971 14918 15829

(3884) (1945) (2691) (1626)Domestic ofloxacin 12502 9621 14936 11205

(4434) (3011) (5282) (4259)Domestic sparfloxacin 15617 12175 16130 9811

(3141) (2576) (4674) (3420)

Note Standard deviations in parentheses API Active pharmaceutical ingredient

1483VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

quinolone subsegment Quinolones fall into thesystemic antibiotics and antibacterials segmentof the Indian pharmaceuticals market whichgenerates over three-quarters of the revenues inthe antiinfectives segment8 The systemic anti-bacterials segment includes all of the originalmiracle drugs that first sparked the growth ofthe global research-based pharmaceutical indus-try in the postndashWorld War II period as well aslater generations of molecules that have beenintroduced in the last four decades

Among systemic antibacterials quinolonesare the latest generation molecules available inIndia We focus our analysis on quinolones forseveral reasons First quinolones are the drug ofchoice for a large number of bacterial infec-tions some of which are also treated by alter-native drugs (see Table A1 in the Appendixwhich outlines the spectrum of activity for eachmolecule family within the antibacterials seg-ment) Hence if there were one product groupfor which we would expect to have many sub-stitutes readily available this would be quino-lones Second with a share of 20 percent in thesales of systemic antibacterials quinolones rep-resent one of the largest subsegments within thistherapeutic category Finally several moleculeswithin the quinolone subsegment were still un-der patent in the United States at the time of ourinvestigation Table 2 details the basic informa-tion about the four quinolone molecules that arethe focus of our analysis The first row showsthe year of US patent expiry this ranges from1998 for norfloxacin to 2010 for sparfloxacinQuinolones include in principle four moremolecules that are listed at the bottom of Ta-ble 1 the market shares of these molecules arenegligible however so that we exclude thesemolecules from our analysis

Table 2 reveals several other interesting factsabout competition in the quinolone market inIndia First note the large number of firmsoperating in this subsegment The large numberof domestic firms is perhaps not that surprisinggiven that pharmaceutical product patents werenot recognized in India9 What is more surpris-ing is the number of foreign firms selling pat-ented products (eg ciprofloxacin) the fact that

multiple foreign firms sell a patented productindicates that such firms often ldquoinfringerdquo10

patent laws in India while complying with themin developed world countries The last two rowsof Table 2 further indicate that domestic prod-ucts often sell at a premium With the exceptionof ofloxacin the average prices of products of-fered by Indian firms are higher than the pricesof products offered by foreign subsidiaries Thispreliminary evidence suggests that Indian con-sumers do not place a premium on the brandname and reputation of big multinational phar-maceutical concerns Moreover the higherprice of domestic products does not seem toprevent domestic companies from capturing alarge market share This is most evident in thecase of ciprofloxacin where domestic firmshave with 53 percent the largest share in thetotal sales of quinolonesmdashthis despite the factthat the average price of these products is 10percent higher than the price of foreign productscontaining the same molecule

Table 3 provides additional summary statis-tics for our data broken down by region Thefirst two rows of the table report the averageannual household expenditure on quinolonesand antibiotics respectively Note that in bothcases the average expenditure is higher in thenorth and west these regions include states withhigher per capita incomes and tend to be moreindustrialized and urbanized than those in theeast and south Pharmaceutical products areavailable in multiple presentations that is com-binations of dosage forms (eg capsule tabletsyrup) strength (eg 100 milligrams 500 mil-ligrams) and packet sizes (eg 50 capsule bot-tle 100 tablet bottle) The various presentationsin which a product is available are often referredto as stock-keeping units or SKUs11 The num-ber of SKUs for each product group withinquinolones is reported at the top of Table 3 Aswith the more aggregate numbers on firms and

8 In addition to anti-bacterials this segment also containsantivirals

9 Accordingly the common distinction between ldquobrandedrdquoand ldquogenericrdquo products is irrelevant here

10 We emphasize here that the word ldquoinfringerdquo belongsin quotes because patent laws do not currently exist inIndia infringement in the legal sense is not possible It ishowever striking that the same firms that accuse Indianproducers of ldquopiracyrdquo sell in India products that are patentedin the United States and for which the patent is held by adifferent multinational corporation

11 For instance a 100 capsule bottle of 100 milligramcapsules of a particular branded drug and a 50 capsulebottle of 100 milligram capsules of the same branded drugwould be identified as two separate SKUs

1484 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

products reported in Table 2 the differencebetween domestic and foreign products isstriking The number of SKUs offered byIndian firms is consistently larger than thenumber offered by subsidiaries of foreignmultinationals The number of SKUs variesslightly across regions but more importantlyit varies across time as some SKUs disap-pear while new ones get introduced duringour sample period

Many pharmaceutical products in India aresubject to price controls12 While the specificsof the price regulation are too complex for anyeconomic model to capture adequately themain concern for the empirical analysis is thatprice controls may lead to a lack of price vari-ation over time so that the demand functioncannot be identified Prices at the most disag-gregate SKU level are relatively stable overtime there are variations due to occasionalchanges in the estimated cost (due for exampleto changes in exchange rates that affect the costof imported materials or bulk drugs) but suchvariations tend to be infrequent and small inmagnitude The degree of time variation ishowever substantially larger once one aggre-gates to the product level This variability stemsnot only from the fact that the SKUs over whichwe aggregate may experience changes in theirrespective prices at different points in time butalso from the fact that the range of SKUs of-fered in the market does not remain constantover time The entry and exit of presentationswithin the same product group that have differ-

ent prices effectively affects the price that con-sumers face for this drug in each period

The middle portion of Table 3 reports themean price and standard deviation for eachproduct group by region Prices vary by regionthough there is no clear pattern emerging fromthe table with regard to the cross-regional vari-ation (in the sense of some regions being sys-tematically more expensive than others) Toexamine what portion of the total price variationis due to time versus regional variation weconducted an analysis of variance of prices thatwe report in Table 4 The table is based onseparate regressions for each product group(pooling data across groups with big differencesin their average prices is not particularly infor-mative as most of the price variation is ac-counted for by product group dummies) Thelast two columns of the table show the fractionof price variation accounted for by region andtime dummies respectively As evident fromthe table a significant fraction of the total vari-ance in prices can be attributed to time varia-tion In the demand estimation we include a fullset of product groupndashspecific regional dummiesso that the price parameters are identified en-tirely based on this time variation within eachproduct group The time variation of productgroup prices is driven primarily by composi-tional changes within each group the revenueshares of the individual SKUs that compriseeach product group change over time (see therelated discussion in Section IIC) while in eachperiod there is entry and exit of SKUs into thesample To check whether this pattern reflectsgenuine entry and exit as opposed to samplingvariation we examined the revenue shares ofthe SKUs that leave the sample relative to theones that remain during the entire period Theresults are reported in Table 5 While the SKUs

12 The details of the procedures for price fixation canbe found in the official government Web site httpwwwnppaindianicinindex1html under the link ldquoDrug Price Con-trol Order 1995rdquo A new pharmaceutical policy was introducedin 2002 but our data were collected before that year

TABLE 4mdashANALYSIS OF PRODUCT PRICE VARIANCE

Product groupPartial SS

timePartial SS

region Total SSPercentage

explained by timePercentage

explained by region

Foreign ciprofloxacin 0296 0306 1047 283 292Foreign norfloxacin 1036 1002 4179 248 240Foreign ofloxacin 0429 0021 0571 752 36Domestic ciprofloxacin 0005 0088 0104 50 841Domestic norfloxacin 0059 0098 0198 296 494Domestic ofloxacin 2858 0104 3056 935 34Domestic sparfloxacin 1754 0032 1860 943 17

1485VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

that exit tend to be smaller (their average shareis 1 percent as opposed to 34 percent for thoseSKUs that are present during the entire sampleperiod) the shares of the two groups do notseem orders of magnitude apart13 In additionour data cover only the 300 largest firms sellingin the Indian market so that firms with verysmall shares are not included in our sample

II The Analytic Framework and EstimationApproach

Patent enforcement in the Indian pharmaceu-tical market will have the effect of eliminatingdomestic products whose active pharmaceuticalingredients are protected by (foreign) patentsThus assessing the effects of patent enforce-ment is tantamount to assessing the effects ofwithdrawing domestic products from the mar-ket This task is the converse of evaluating newproduct introduction accordingly the concep-tual framework we use to address the questionsof interest is similar to the one developed in theliterature for the valuation of new goods14

We start by estimating demand for quino-lones Given that the market is characterized byimperfect competition the counterfactual anal-ysis requires that we also model the supply sideas removal of one product will affect the pricesof other products especially those that are closecompetitors The existence of price regulationin the Indian pharmaceutical market imposespotential constraints on firmsrsquo maximizationproblem Given these constraints and the com-

plexity of the price regulation process the typ-ical approach of deriving estimates of actualmarginal costs and markups by exploiting thefirst-order conditions of profit maximizing firmsdoes not seem particularly promising Insteadwe use our demand estimates to place upper andlower bounds on marginal costs and markups

With demand elasticities and upper and lowerbounds for marginal costs in place we thenconduct counterfactual simulations We con-sider several alternative scenarios dependingon the number of domestic products affected bypatent enforcement For each scenario we com-pute the counterfactual prices and use them toassess the effects of domestic product with-drawal on consumer welfare (as measured bythe compensating variation) firm profits andsocial welfare As with the valuation of newproducts the big conceptual problem facing thispart of the analysis is that we need to extrapo-late from the region of the data to the point atwhich demand for the products that exit themarket becomes zero This conceptual issue ispresent in any attempt to evaluate a major pol-icy change for which no historical precedentexists like the enforcement of patent laws inIndia One advantage of the present study is thatwe have a limited set of price data for Pakistana country with similar demographics to Indiabut with a market structure that resembles theone that would emerge in India under patentenforcement (monopoly of multinational sub-sidiaries) By comparing the prices of productsoffered by multinationals in Pakistan to thosewe compute in our counterfactual simulationsfor the products that would be offered by mul-tinationals in India if patent laws were enforcedwe can get a sense of how plausible our coun-terfactual estimates are

A Demand

The demand modelling is based on the multi-stage budgeting approach Our primary motiva-tion for adopting this approach is a practical oneIn the multistage budgeting approach the depen-dent variable is defined as a revenue share whichis appealing given that the products we include inthe analysis contain different molecules (ie ac-tive pharmaceutical ingredients or APIs) Eventhough we do have data on the quantity of therelevant API (eg 100 milligrams of ciprofloxa-cin) contained in each product converting the

13 While we cannot completely rule out the possibilitythat some of the exit is due to sampling variation note thatthe latter should be reflected in low precision of the demandparameter estimates The demand parameters however areprecisely estimated

14 See Manuel Trajtenberg (1989) Jerry A Hausman(1994) and Timothy F Bresnahan (1997) for representativeexamples and a discussion of the relevant issues

TABLE 5mdashREVENUE SHARES OF THE EXITING SKUS

Revenue share ofexiting SKUs

Revenue shares ofall SKUs

Full sample 10 34Northern region 08 33Eastern region 12 36Western region 10 34Southern region 12 33

1486 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

revenue shares to physical shares is extremelydifficult if not infeasible in the case of antibioticsBecause such drugs are ldquosystemicrdquo by nature theyare used to treat a large number of infections andthe dosage of each drug depends on the particularinfection it is supposed to address (for example thedosage will differ depending on whether the anti-biotic is used to combat an ear infection or tuber-culosis) This particular feature of antibioticscomplicates the conversion of revenue to physicalshares

The basic idea of the multistage budgetingapproach is to use the therapeutic classificationof a productmdashie the therapeutic segment andsubsegment the product belongs tomdashto orga-nize all products in the systemic antibacterialssegment into a hierarchical taxonomy consist-ing of two levels At the higher level are thevarious subsegments of systemic antibacterialsThe first stage of budgeting corresponds to theallocation of expenditures across the subseg-ments in this upper level of the taxonomy

In the second stage of the budgeting processcorresponding to the lower level of the taxon-omy a flexible functional form is adopted tomodel how the expenditures allocated to eachsubsegment are distributed across the productswithin that subsegment In particular to modeldemand at the second stage we employ theAIDS specification proposed by Deaton andMuellbauer (1980b)15

While the two-stage demand estimation ap-proach offers functional form flexibility its ap-plication to the Indian systemic antibioticsmarket poses a few problems The first is thatdue to entry and exit many SKUs and evenproducts in our sample are not present in everyperiod AIDS does not have a good way ofdealing with a varying number of products as itwas developed with broad commodity catego-ries in mind which are consumed by all con-sumers every period To solve this problem weaggregate within each subsegment (eg quino-lones) SKUs into product groups where withineach product group all SKUs contain the samemolecule and are produced by firms with thesame domesticforeign status Specifically let aSKU k be indexed by its molecule (or API) M its

domesticforeign status DF indicating whether itis produced by a domestic (Indian) or a subsidiaryof a foreign (multinational) firm a particular pre-sentation s and the particular firm f that producesit We aggregate SKUs over presentations andfirms to obtain a newly defined product group iwhich is indexed only by molecule M and domes-ticforeign status DF and has revenue Ri yenfsRk with i (M DF) k (M DF f s) and pricepi yenfs kpk where k denotes the conditional(on M and DF) revenue share of this particularproduct ie

(1) k Rk

Ri

In most cases the resulting product groups arebroad enough to be present every period16

The usual concern with this aggregation pro-cedure is that it may lead us to overstate firmsrsquomarket power as we ignore competition amongfirms with the same domesticforeign statusproducing the same molecule In the presentapplication however this concern is unlikely tobe of great importance as the effect of patentenforcement is to wipe out all domestic compe-tition at once while granting foreign firms mo-nopoly power hence competition among firmsfor patented molecules becomes irrelevant Theaggregation according to the domesticforeignstatus (within a particular molecule) thus corre-sponds to the scope of our analysis and theparticular questions of interest

The second problem is that for our approachto be useful in welfare analysis the allocation oftotal expenditures to group expenditures at thehigher stage has to be modelled in a way con-sistent with utility maximization In general thesolution of this allocation problem requiresknowledge of all individual product prices

15 Representative applications of the multistage budget-ing approach include Sara F Ellison et al (1997) Hausman(1994) and Hausman and Gregory K Leonard (2002)

16 We are missing only four observations (ie monthregion combinations) all for the drug group of ForeignNorfloxacin August 1999 in the south May 2000 in thewest October 2000 in the south and November 2000 in theeast In these cases we set the revenue shares of ForeignNorfloxacin equal to zero In general with 01 percent ofquinolone sales (see Table 1 row 2) Foreign Norfloxacinhas a very small share of the market This probably explainswhy the results pertaining to this drug group are unreliableit is the only drug group for which we do not obtain asignificant price elasticity of demand while its cross-priceelasticities with other foreign drug groups often have thewrong sign

1487VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

From an empirical point of view this is notparticularly useful as it eliminates all compu-tational advantages of the two-stage approachTo address this problem we adopt an approxi-mate solution to model the higher-level expen-diture allocation along the lines suggested byDeaton and Muellbauer (1980b pp 131ndash32)This gives rise to a two-level AIDS specification

Consider the lower-level estimation firstwhich refers to the allocation of a particularsubsegmentrsquos expenditure to the product groupswithin the subsegment In our application therelevant subsegment is quinolones which weindex with Q Let the product groups within thissubsegment be indexed by i 1 N pi be theprice of product group i (where as noted abovei refers to a particular molecule and domesticforeign status combination) and XQ be the totalexpenditure on the quinolone segment The rev-enue share of each product group is given by

(2) i i j

ijln pj ilnXQ

PQ

where i the revenue share of product group iis defined as

(3) i pi qi

yenj

pj qj

xi

XQ with i j Q

XQ is the overall expenditure on the quinolonesubsegment and PQ is a price index given by

(4) ln PQ ap 0 i

iln pi

1

2 i

j

ijln piln pj

With a limited number of product groups and asufficiently large number of time-series observa-tions the flexibility implied by the AIDS modeldoes not impose too many demands on the data Inthe present application however where the num-ber of observations is limited the AIDS model isnot estimable in this general form To reduce thenumber of parameters that need to be estimatedwe impose two sets of restrictions

The first set of restrictions is implied by thetheory of utility maximization Specificallythese restrictions are

Adding-up yenk k 1 yenk k 0 yenk kj 0 j

Homogeneity yenk jk 0 j Symmetry ij 1frasl2 [ij ji] ji This last

restriction by itself reduces the number of parameters to [N(N 1)]2

The second set of restrictions we impose aimsat further reducing the number of parameters tobe estimated by exploiting our knowledge of thisparticular market Specifically for each productgroup i we allow one ij parameter for all productgroups j that have different molecules from prod-uct group i and are produced by foreign firms andone ij for all product groups j with differentmolecules produced by domestic firms We donrsquotimpose any restrictions on the ij parameter whenproduct group j has the same molecule as productgroup i (By construction product groups i and jcontain products produced by firms with differentdomesticforeign status)

To better illustrate the nature of the restric-tions we impose on the patterns of substitutionacross products some additional notation isneeded Let d(i j) be an indicator of the degreeof similarity (or difference) between productgroup i and product group j along the dimen-sions we are able to observe (molecule M anddomesticforeign status DF) For any two prod-uct groups i and j d(i j) can take on one of thefollowing three values17

(5)

di j (1 0) if Mi Mj DFi DFj

(0 1) if Mi Mj DFi DFj

(0 0) if Mi Mj DFi DFj

Let

(6) Diab j di j a b

17 The sequence (1 1) is not possible for two differentproducts in this case the parameter corresponds to theproductrsquos own-price effect that is ii

1488 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The equation at the lower level becomes

(7) i i iiln pi i10ln pj j Di10

j Di

01

i01ln pj

j Di

00

i00ln pj ilnXQ

PQ

Note that

The parameter ii captures a product grouprsquosown-price effect (there will be as many iiparameters as the number of product groups)

The parameter i10 captures the cross-priceeffect of the product group containing prod-ucts with the same molecule but produced byfirms of different nationality

The parameter i01 captures the cross-priceeffects of product groups containing productswith different molecules but produced byfirms with the same nationality

The parameter i00 captures the cross-priceeffects of product groups containing productswith different molecules produced by firms ofdifferent nationality

Before we take the demand equation to thedata we make two modifications The first oneis to let the product-specific effects i vary byregion r The resulting product-specific regionaleffects ir have two interpretations first theycontrol for the ldquoqualityrdquo of each drug withquality differences being allowed to vary acrossregions second they proxy for demographicsand other demand shifters which vary by re-gion and may affect the demand of each prod-uct group differently18 Note that by includingproduct-specific regional effects in the demandspecification we estimate the price parametersbased on the within-product-group variation ofprices in each region In an earlier version of thepaper we also estimated the demand systemwithout regional dummies and obtained similarresults

The second modification that allows us to gofrom a deterministic to a stochastic specificationof the demand equation is to include an additiveerror term in (7) The latter takes into accountthe fact that (7) is not expected to fit the dataexactly The error term irt accounts for mea-surement error (due to the fact that the productgroup prices pjrt we employ in the estimationare not exact price indices but approximationsthereof) and (potentially region-specific) de-mand shocks that may affect the demand for aproduct in a particular period Examples of suchshocks include an advertising campaign for aparticular product which temporarily increasesthe demand for this product or the outbreak ofa (potentially region-specific) epidemic whichcalls for the use of a particular drug We discussthe interpretation and properties of this errorterm in more detail in the next section

The final form of the equation we estimate atthe lower level becomes (with subscript t de-noting month and subscript r denoting region)

(8) irt i ir iiln pirt i10ln pjrt j Di10

j Di

01

i01ln pjrt

j Di

00

i00ln pjrt

ilnXQrt

PQrt irt

The analysis so far has conditioned on the ex-penditure allocated to the quinolone subseg-ment XQ The upper level of the estimationconsiders the problem of allocating total expen-diture across the different systemic antibioticssubsegments one of which is quinolones Theupper-level demand function is given by

(9) G G H

GHln PH GlnX

P

where all variables denoted by capital letters aredefined as before but now refer to subsegments(G H ) rather than product groups within asubsegment and the total expenditures on sys-temic antibiotics X are deflated by the Stone priceindex log P yenH Hlog PH When estimating thesystem above we impose all the restrictions

18 Given the short time span of our sample typicaldemand shifters such as age distribution income distribu-tion and education hardly change over our sample periodSuch shifters are therefore absorbed by the region-specificproduct fixed effects ir

1489VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

implied by utility maximization as we do withthe estimation of the lower-level AIDS We donot however impose any additional restrictionson the substitution patterns at this stage so thatthe cross-price effects across segments remainrelatively unconstrained

Estimation of the higher-level AIDS allowsus to obtain the unconditional own- and cross-price elasticities that are used in the formulationof the supply problem and welfare analysisThese are given by the formula

(10) ij ijXQXQ

ln qi

ln XQ

ln XQ

ln PQ

ln PQ

ln pj

with the conditional cross price elasticitiesgiven by

ijXQXQ ij ij jlnXQ

PQ

i

As in the lower stage we include subsegment-specific regional dummies in the specificationof the upper-stage demand system so that thefinal form of the estimating equation at theupper stage becomes

(11) Grt G Gr H

GHln PHrt

GlnXrt

Prt Grt

In sum the demand system we take to thedata is represented by equations (8) and (11)and the associated parameter restrictions im-plied by economic theory

B Modelling the Supply Side of the Market

Counterfactual simulations concerning theeffects of domestic product withdrawal requireknowledge of the marginal costs of pharmaceu-tical firms operating in the Indian market Theseare unobservable The usual approach in theNew Empirical Industrial Organization litera-ture has been to exploit the firm equilibriumconditions to infer marginal cost For exampleit is often assumed that the marginal cost ci isconstant and that the industry is an oligopoly

engaging in Bertrand competition with differen-tiated products Assuming that firms myopicallymaximize profits each period one can then de-rive the firmsrsquo first-order conditions that corre-spond to the assumptions above about costsmarket structure and firm behavior

We deviate from this procedure as the presenceof price regulation renders the assumption of un-constrained period-by-period maximization un-tenable Ideally one would like to explicitlyincorporate the price and other administrative con-trols into the firmrsquos optimization problem andderive the first-order conditions under the assump-tion of constrained maximization The complexityof the price regulation however makes this ap-proach infeasible Therefore we adopt an alterna-tive approach that does not rely on modelling theprice-setting process but is instead based on de-riving upper and lower bounds for marginal costsand markups

In particular an upper bound for marginalcosts and a lower bound for the markups (zero)can be derived under the assumption of perfectcompetition While this assumption is clearlyunrealistic in the pharmaceuticals market it isuseful in providing an upper bound for costs ci

Uwhich will be given by

(12) ciU pi

On the other hand a lower bound for marginalcosts ci

L (and upper bound for markups) can bederived by assuming that there is perfect collu-sion within each product group i (where i refershere to a moleculedomestic-foreign combina-tion)19 and ignoring price controls so thatprices are determined by the first-order condi-tion of the jointly profit-maximizing firmswithin product group i Solving this first-ordercondition for the (lower bound of) marginal costthen gives

(13) ciL pi 1

1

iipipj

19 The reason that this assumption will lead to an under-statement of marginal costs (and hence overstatement ofmarket power) is that it assumes away competition amongfirms within each drug group

1490 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

where ii(pipj) the own-price elasticity of de-

mand for product i will depend on the productrsquosown-price pi and all other productsrsquo prices pjwith i j

Once we have obtained the demand elastic-ities through estimation of the demand sys-tem we can calculate the upper and lowerbounds for marginal costs and correspondingmarkups according to (12) or (13) These willthen be employed in the counterfactual sim-ulations While our counterfactuals use boththe lower and upper bounds for costs most ofour discussion will be based on using thelower bounds for costs since this is the moreinteresting case in the policy analysis it givesus the largest possible profits for pharmaceuticalfirms and hence corresponds to the worst pos-sible scenario facing Indian firms and the bestpossible scenario facing multinationals underTRIPS

C Identification Assumptions and EstimationApproach

The discussion of the demand system has sofar abstracted from the issue of price endogene-ity The usual premise in the industrial organi-zation literature is that correlation of prices withthe error term in the demand equation arises byvirtue of the first-order conditions of profit-maximizing firms20 As we discuss in this sec-tion this source of simultaneity bias is unlikelyto be of major concern in the present contextbecause of the existence of price regulation Ourprimary concern is instead with the simultane-ity bias that is implied by the particular wayprices are constructed

To understand the sources of the simultaneitybias note first that the price of product group piwe employ in the demand estimation should bethought of as a proxy for an exact price indexthat we do not observe As such pi contains bydefinition measurement error and it will becorrelated with the error term of the demandequation Specifically the demand equation weare interested in estimating can be written insimplified form (suppressing the subscripts rand t and ignoring parameter restrictions forconvenience) as

i i j

ijln pjT ilnXQ

PQ i

where pjT denotes the true (exact) price index for

product group i and i captures unobservedvariables that may affect demand in a particularperiod (eg an advertising campaign or an ep-idemic) The price index pj

T depends on theSKU prices pk Note that in the presence ofSKU entry and exit into the sample the priceindex pj

T will vary over time even if the pricesof the individual SKUs remain stable as theproducts (SKUs) that comprise each productgroup change each period

Ideally we would like to compute the exactprice index pj

T for each group and employ theseindices in the demand estimation Under theassumption of predetermined SKU prices eachperiod the exact price indices would be uncor-related with the error term and the demandparameters could be estimated by simple OLSUnfortunately this strategy is not feasible in thecurrent context

To derive an exact price index for each prod-uct group it is necessary to model consumerrsquoschoice among SKUs conditional on the choiceof the group To this end it is necessary toexplicitly introduce a third stage in the demandspecification and estimate the parameters asso-ciated with the choice at that stage Howeverany specification based on three-stage budget-ing (that would require directly specifying ademand function associated with the SKUchoice at the third stage and imposing additiveseparability at the higher stages) would be in-feasible to estimate for three main reasonsFirst there is a large number of SKUs withinsome groups so that the demand parameters can-not be estimated even if one imposes parameterrestrictions second the SKU prices exhibit littlevariation over time lastly the choice sets at thislowest stage vary over time rendering estimationof the price parameters associated with SKUs thatare not available in all periods infeasible Alterna-tively one could abandon the multistage budget-ing altogether and adopt a discrete choiceapproach which is inherently better suited to deal-ing with varying choice sets and limited pricevariation A discrete choice approach would beparticularly appealing for modeling the SKUchoice conditional on product group choice giventhat within each group all SKUs contain the same

20 See Steven Berry et al (1995) or Aviv Nevo (2001)for a related discussion

1491VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

molecule and are hence comparable As notedearlier however a discrete choice approach at thehigher stages of the demand estimation requiresconverting revenue to physical shares which ischallenging in the case of systemic drugs such asantibiotics for which the appropriate dosage is notwell defined Finally a combination of a multi-stage budgeting approach for the higher stages anda discrete choice approach for the lowest stagewould be feasible to estimate but inconsistentwith a model of utility maximization For thesereasons we have no alternative but to adopt anapproximation for measuring the price index foreach product group

Let us denote this approximation by pjA and let

13j denote the proportional ldquoapproximationrdquo errorassociated with measuring the true price index forgroup j so that ln pj

A ln pjT ln 13j The

estimating demand equation can then be written as

i i j

ijln pjA ilnXQ

PQ

j

ijln 13j i

or

i i j

ijln pjA ilnXQ

PQ i

where the new error term i yenj ijln 13j iis comprised of two components the first onereflects ldquomeasurementrdquo error due to the fact thatthe product group prices we employ are approx-imations and not exact price indices and thesecond one captures conventional demandshocks The demand equation as written abovecorresponds to the equation we take to the data

The particular proxy we use for the exact priceindex for each product group j is the revenue-share-weighted average of the prices of multipleSKUs available within this group that is

pjA

k j

k pk

The notation above illustrates the two source ofprice variation variation in the SKU prices pkand variation in the weights k

We are concerned with potential correlation ofthe so-constructed product group prices with bothcomponents of the product group demand error

term The correlation between pjA and the first

component yenj ijln 13j is independent of theparticular way we construct pj

A and inherent in thefact that we measure the exact price index witherror The potential correlation between pj

A and thesecond component of the product group demanderror the shock i arises however because of thespecific way we construct the proxy pj

A and inparticular because of the presence of the reve-nue share weights k in pj

A since the SKUrevenue share weights k will generally dependon the product group expenditure they arelikely to be correlated with the product groupdemand shock i

To address the simultaneity bias we use in-strumental variables To this end we need vari-ables that are correlated with the proxies pj

A butuncorrelated with the error term i We use thenumber of SKUs within each product group j asan instrument as it does justice to the idea thatvariation in the product group price index stemsin part from variation in the set of SKUs that areavailable each period It is clearly correlatedwith the average group prices pj

A and plausiblyuncorrelated with the demand shock i Theassumption that the number of SKUs is uncor-related with the conventional demand error willhowever be violated if the introduction of anew SKU changes the perceived quality of adrug21 or if it is accompanied by promotionalactivities which will be reflected in i Our hopeis that this is not too often the case Assumingthat the SKU number for each group j is alsouncorrelated with the weighted sum of the mea-surement errors of all product groups yenj ijln 13j itcan be used as an instrument for the product groupprice pj

A22 If one accepts the premise that prices atthe SKU level are exogenous then SKU pricescan also be used as instruments The final list ofinstruments we use hence includes the number ofSKUs in each group the prices of the five largestSKUs for each group and all other exogenousvariables in the demand estimation such as re-gionalproduct dummy interactions and the upper-

21 Note that we already proxy for the quality of each prod-uct group through the product fixed effects i In the actualestimation we actually let these fixed effects be region-specific(ir) so that we allow quality to vary across regions

22 If the measurement error increases in the number ofincluded SKUs this assumption will be violated Unfortu-nately we have no way of checking the validity of thisassumption

1492 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

level total expenditure on antibiotics Regressionsof group prices on the instruments above yieldhigh R-squares with most regressors highly sig-nificant indicating that our instruments are highlycorrelated with prices23

Our sample includes four molecules cipro-floxacin norfloxacin ofloxacin and spar-floxacin Except for sparfloxacin all othermolecules are produced by both foreign anddomestic firms24 So we have seven products(domestic ciprofloxacin foreign ciprofloxa-cin etc) with 96 observations (two years ofmonthly data four geographical regions foreach period) for each product The parametersin the lower-level AIDS demand system asdefined in equation (8) are the product fixedeffects i the productndashregional dummy inter-actions ir the own revenue-share price elas-ticities ii the cross revenue-share priceelasticities i10 i01 i00 and the revenue-share expenditure elasticities i In estimatingthe parameters we first regress prices on allinstrumental variables and then plug the pre-dicted values for prices in the constrainedleast-square regression (For a detailed expla-nation of the constraints see the previoussection)

Given that we impose many cross-equationconstraints and employ instrumental variablesin the estimation it is difficult to derive stan-dard errors for the parameter estimates ana-lytically Our error term interpretation in (8)implies that the error terms for each productgroup are likely to be correlated across re-gions for example if a national advertisingcampaign increases the demand for a rela-tively expensive presentation in one regionsimultaneously increasing the aggregate de-mand and price index for the correspondingproduct group in that region it is likely that thesame effect will be observed in other regions Inprinciple we could exploit the cross-regionalcorrelation in the error terms to estimate thedemand system using seemingly unrelated re-

gression (SUR) This however would requireus to derive the variance-covariance matrix ofthe parameters analytically which as notedabove is cumbersome in the current contextInstead we use the bootstrap method choosingto remain agnostic about the structure of thevariance-covariance matrix The potential dis-advantage of this approach over SUR is a loss inefficiency but as will become apparent in theresults section the model parameters are fairlyprecisely estimated To maintain the marketstructure we randomly sample the periods (withreplacement) and use the same periods for allproducts Regarding the optimal number ofbootstrap repetitions ideally one would followthe three-step method proposed by Donald W KAndrews et al (2000) Empirical evidence sug-gests however that one rarely needs more than200 replications to estimate the standard er-rors25 To be safe we generate 300 bootstrapsamples (with replacement) based on the origi-nal data and estimate the standard errors usingthe standard errors of the bootstrap sampleestimates

The estimation of the top-level AIDS systemis similar The constraints imposed on this top-level demand system are adding-up homogene-ity and symmetry as defined on page 12Again bootstrapping is used to obtain the stan-dard errors of the parameter estimates

D The Counterfactual Scenarios

In assessing the effects of patent enforce-ment we start by focusing on the most extremecase in which compulsory licensing is not anoption and foreign firms are not subject to pricecontrols We use the results from the analysis ofthis case as a benchmark In reality the outcomeof the WTO negotiations is more likely to in-volve some constraints on the monopoly powerof foreign firms selling patented products indeveloping countries such as price caps orcompulsory licensing Our framework can eas-ily accommodate these cases as will becomeapparent in the next subsection

We now focus on the effects of potentialpatent enforcement in the quinolone segmentWe consider several scenarios which vary inthe number and size of domestic products that

23 Specifically the R-squares from the first-stage regres-sions of product group prices on instruments range from057 to 095 except for domestic ciprofloxacin for whichthe R-square is 017

24 Sparfloxacin is actually offered by one foreign subsid-iary in India (see Table 2) However its revenue share isminiscule We therefore treat Sparfloxacin as being pro-duced by domestic firms only 25 See Bradley Efron and Robert J Tibshirani (1993)

1493VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

will be removed from the market In particularwe consider the following five scenarios26

Withdrawal of one large domestic productgroup only domestic ciprofloxacin

Withdrawal of one relatively small domesticproduct group only domestic ofloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand norfloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand sparfloxacin

Withdrawal of all four domestic quinoloneproduct groups

As the list above suggests we proceed fromanalyzing the effects of single-product with-drawal to the analysis of eliminating the en-tire domestic segment This approach wasmotivated by early empirical results that in-dicated that the existence and extent of com-petition from domestic firms has a significantbearing on the predicted effects of patent en-forcement that is our predictions regardingprices and welfare vary substantially depend-ing on how many domestic products are af-fected by patent enforcement and on whethersome domestic competition will remain presentafter TRIPS In addition the size of the affecteddomestic groups is relevant for the welfare pre-dictions accordingly we examine both scenar-ios in which a large domestic product groupsuch as domestic ciprofloxacin is eliminatedfrom the market and scenarios in which the

eliminated domestic group is relatively small(eg domestic ofloxacin) In all scenarios wemaintain the assumption that the set of prod-ucts offered by the remaining firms in themarket does not change in response to patentenforcement

E Computation of Virtual Prices and NewEquilibrium Prices

The first step in the counterfactual analysisis to derive the new equilibrium prices underpatent enforcement27 In this context thereare two sets of prices that are relevant Thefirst set consists of the virtual prices of those(domestic) products that will not be availableonce TRIPS is put in effect To calculate thesevirtual prices we set the expenditures of therelevant products equal to zero The secondset of prices consists of the prices of thoseproducts that remain in the market In deriv-ing these prices we start by assuming profitmaximization without any form of price reg-ulation the firms remaining in the marketreoptimize in response to the policy changeand set new prices taking the prices of allother firms as given28 Of course at the equi-librium all prices change in response to thefact that some domestic products are nolonger present The new equilibrium pricesfor products that remain in the market are thuscomputed by utilizing the first-order condi-tions of profit-maximizing firms into whichthe virtual prices of the eliminated productsare substituted Hence to compute the newequilibrium prices we solve an equation sys-tem of the following form

26 As Table 2 indicates most of the patents for the drugsin the quinolone segment have expired by now so that noneof the scenarios described in this section is going to mate-rialize in practice The reason we consider these alternativescenarios in our counterfactual simulations is to get a senseof how the presence and extent of domestic competitionaffects welfare calculations Such calculations may becomerelevant in the future in other therapeutic classes in whichpatents have not expired yet and where domestic firmscompete with foreign patent holders While the particularexit scenarios will depend on the exact year of patent expiryof drugs in each therapeutic class and the specific way inwhich patents will be enforced (eg whether certain do-mestic drugs will be grandfathered in) the general conclu-sion that emerges from our calculations is that the consumerwelfare loss is substantially smaller in cases where somedomestic competition remains present in the market Hencethe particular way in which patent enforcement is imple-mented is essential for assessing its welfare impact

27 Of course until product patents are in fact introducedthese prices will not be observable Note also that we areassuming here that the range of available products will notchange with the introduction of patents

28 As mentioned above this first set of calculations ab-stracts from the existence of remaining price controls orother government regulations that would impose constraintson the firmsrsquo profit maximization problem Accordingly theresulting numbers should be interpreted as a benchmark ofwhat would happen if markets were completely unregulatedTo examine what the welfare effects of TRIPS would be inthe more realistic scenario of price regulation we subse-quently consider a scenario in which regulation keeps theprices of the products provided by patent-holders at theirpre-TRIPS level For a more detailed discussion see theresults section

1494 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

For products i that are withdrawn from themarket

(14) 0 i ir iiln pirV i10ln pjr j Di

10

j Di

01

i01ln pjr

j Di

00

i00ln pjr

ilnXQr

PQr

For products k that remain in the market

(15) pkr ckr 1 1

kk pkrpjr pirV 1

In the equations above pirV denotes the virtual

prices of the products that are removed fromthe market while pjr denotes the updatedprices of all other products Note that whensolving for the virtual prices we account forthe fact that both the price index for quino-lones PQr and the expenditure allocated tothis subsegment XQr need to be updated toreflect the fact that as a result of the pricechanges there may be substitution away fromthis subsegment To obtain the new quinoloneexpenditure XQr and the new price index PQrwe use the estimates and formulas for thehigher-level AIDS system In equation (15)ck refers to the marginal cost for product k thatwe have obtained from the previous estimationstage As mentioned in the previous subsectionwe conducted the simulations using both theupper and lower bounds for marginal cost ck

u

and ckL respectively The term kk(pkr

pjrpirV) re-

fers to the unconditional own-price elasticity forproduct k which is a function of the eliminatedproductsrsquo virtual prices and the remaining prod-uctsrsquo new equilibrium prices We conduct thecounterfactual simulations and welfare analysisat the regional level because the presence ofregion-specific product effects in the demandestimation implies region-specific demand elas-ticities and hence region-specific marginalcosts and markups The presentation and discus-sion of our results focus on national averages ofthe relevant variables that we construct by com-

puting weighted averages across regions usingthe population of each region as a weight

F Welfare Assessment

The simulation of the new equilibrium underpatent protection can provide important insightsinto how consumers and firms will respond tothe removal of domestic products in the market(for example toward which products consumerswill substitute or which prices will increase themost) To get a more precise idea of how peo-plersquos well-being will ultimately be affected byTRIPS we compute as a last step in our anal-ysis the welfare effects of the policy changeSocial welfare is defined as consumer welfareand the sum of domestic firm profits Thechange in domestic profits can easily be calcu-lated by comparing the domestic firm (variable)profits at the pre-TRIPS prices to the profitsthese firms will realize at the new simulatedprices Although foreign firm profits do notcount in domestic welfare calculations we alsocompute the effects of patent enforcement onforeign firm profits to get an idea of how largethe expected benefits of TRIPS are for thesefirms This provides in some sense an indirectway of assessing whether the claims that patentenforcement in countries like India will lead tomore research on developing-country-specificdiseases (such as malaria) have any validity iffor example we find that the effect of patentenforcement on the foreign firm profits realizedin India is small in magnitude it is unlikely thatforeign firms will engage in more developing-country-specific research in response to TRIPSIt is important to note that in all these calcula-tions we work with the lower-bound estimatesfor marginal costs since these give rise to thehighest possible markups Hence our estimateof profit loss for domestic producers most likelyoverstates this loss

On the consumer side we measure changes inconsumer welfare by the compensating variation(CV) defined as the additional expenditure thatconsumers need in order to achieve the sameutility level as before patent enforcement at thenew prices Specifically let P0 denote the pricevector before patent enforcement P the simulatedprice vector post-TRIPS (that we obtained usingthe methods described in the previous subsection)u0 the utility attained by consumers before TRIPS

1495VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

and E(u P) the higher-level expenditure functionThen the compensating variation is given by

(16) CV Eu0 P Eu0 P0

Note that the CV as computed in (16) representsthe combination of three effects

The pure product-variety effect that is theeffect that arises because one or more prod-ucts are not available to consumers anymoreholding the prices of all other remainingproducts and the total expenditure on thequinolone subsegment XQ constant

The expenditure-switching effect that is theeffect arising from substitution away fromquinolones and toward other subsegments ofthe antibiotics market again holding the pricesof all other remaining products constant

The reduced-competition effect that is theeffect that arises because the firms remainingin the market adjust (increase) their prices inresponse to the removal of domestic products

From both an analytical and a policy point ofview it is desirable to assess how large each ofthese effects is Accordingly we decompose thetotal effect on consumer welfare (the CV as givenby equation (16)) using the following procedure

To get the pure product variety effect wecompute virtual prices for the products that areremoved from the market holding the quino-lone expenditure XQ and the prices of all otherproducts fixed Let us call the resulting pricevector P1 Then the pure product variety effectis represented by E(u0 P1) E(u0 P0)

To compute the expenditure-switching effectwe compute another set of virtual prices againholding the prices of all remaining products fixedbut letting quinolone expenditure adjust in re-sponse to the new price index for the quinolonesegment (Given that the prices of the remainingproducts remain fixed the change in the priceindex arises only because of the removal of one ormore domestic products) Note that this scenariomost closely resembles the case in which patentlaws are enforced but strict price regulation keepsthe prices of the products offered in the market attheir pre-TRIP level Let us label the so-computedprice vector P2 The expenditure-switching effectis then E(u0 P2) E(u0 P1)

Finally the reduced competition effect aris-ing from higher prices for the remaining prod-

ucts is computed as the residual change in thecompensating variation once the product-varietyand expenditure-switching effects have been ac-counted for that is E(u0 P) E(u0 P2)where the price vector P is computed accordingto the formulas (14) and (15) to reflect theadjustment of prices to the new regime

To compute the standard errors associatedwith the counterfactual simulations (that is thestandard errors for the counterfactual prices andwelfare estimates) we again use bootstrappingWe first bootstrap the original sample (sales andprices of all the drugs) 300 times Next we esti-mate the AIDS model for each of these samplesand compute the counterfactual equilibrium pricesand welfare losses corresponding to each policyscenario for each of the 300 simulations In thefinal step we compute standard errors29

One limitation of our framework is that itdoes not allow for a heterogeneous response ofconsumers to the policy change Accordinglyour framework is not suited to addressing thequestion of how different groups in the popula-tion will be affected by patent law enforce-ment30 Along the same lines our frameworkdoes not accommodate the possibility of pricediscrimination which might lead to differentresults regarding welfare losses and profit gainsrelative to uniform pricing31 Still we believethat the results of the counterfactual simulationscan provide important insights into the likelyaggregate response to patent enforcement andthe factors that drive this response

III Results

A The Structure of Demand

Tables A2 and A3 in the Appendix display theresults from estimation of the lower- and upper-level AIDS systems respectively For ease of in-terpretation rather than discussing the coefficientestimates we focus our discussion on the implied

29 In cases some of the simulations do not converge wecompute the standard errors using only those simulationsthat converged

30 Addressing this question would require at a mini-mum micro data on consumer purchases as in Goldberg(1995) To our knowledge such data do not exist for theIndian pharmaceutical market

31 See Ernst R Berndt (1994) for an extensive discussionof uniform pricing

1496 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

unconditional price and expenditure elasticities re-ported in Table 6 Given that the region-specificproduct group effects ir reported in the last fourcolumns of Tables A2 and A3 imply region-spe-cific demand elasticities we report separate elas-ticities for each region in Tables 6A and 6B Asevident from the comparison of these tables theelasticities are very similar across regions so thatwe can focus the remaining discussion on theelasticities of one region only the northern regiondisplayed in Table 6A

The diagonal terms of Table 6A report theown-price elasticities which are in all but onecase negative and highly significant The oneexception is the foreign norfloxacin productgroupmdashwhose share of quinolone sales is 007percentmdashfor which we estimate a negative butinsignificant own-price elasticity For the remain-ing product groups demand appears to be highlyelastic with the estimated elasticities lower than2 in four out of the six cases The magnitude ofthe own-price elasticities matches the features ofthe Indian pharmaceutical market mentioned ear-lier which would suggest that Indian consumersare likely to be quite price-sensitive32 The elas-ticities appear especially large if one takes intoaccount that they refer to product groups (such asdomestic ciprofloxacin) and not individual drugsoffered by particular firms Their relative magni-tudes are also intuitive the drug with the largestmarket share and a relatively high price (domesticciprofloxacin) appears to be one of the least elas-tic In contrast foreign ciprofloxacin is highlyprice elastic this is plausible as ciprofloxacindrugs offered by subsidiaries of multinationalsface the stiffest competition in this market seg-ment from approximately 75 Indian firms offeringthe same molecule

The estimated expenditure elasticities appear inthe last column These are all positive indicatingthat the demand for all product groups is normalThe remaining cells display the estimated cross-price elasticities As one might perhaps expect forproducts within a therapeutic subsegment theseare mostly positive Out of a total of 49 priceelasticities we estimate there are six that do notconform to expectations these are the cross-price elasticities between different foreign prod-

uct groups which are estimated to be negativeand significant33 Fortunately these elasticitieshave negligible impact on the welfare analysisgiven that our counterfactuals focus on theeffect of withdrawing one or more domesticproducts from the market the most relevantelasticities are the ones that capture the responseof various product group shares to a change in theprice of one or more domestic groups these elas-ticities are the cross-price elasticities between var-ious domestic groups and the ones betweendomestic and foreign groups which are plausibleand precisely estimated34

Regarding these elasticities a striking aspectof our estimates is how large positive andsignificant the cross-price elasticities betweendifferent domestic product groups aremdashin factfor norfloxacin and ofloxacin we estimate thatdomestic product groups containing differentmolecules are closer substitutes for one anotherthan product groups that contain the same mol-ecule but are produced by foreign firms Incontrast for ciprofloxacin (the molecule withthe largest revenue share) we estimate a largepositive cross-price elasticity between the do-mestic and foreign versions

The fact that domestic products appear to beclose substitutes for other domestic productsthat contain different molecules truly representsan ldquoempiricalrdquo finding in the sense that we do

32 In developed economies elasticities of this magnitude havetypically been found only for generic drugs (and even then onlyrarely) or among consumers who lack health insurance

33 While these elasticities are clearly counterintuitive theyare not inconsistent with the underlying demand system whichimposes no restrictions on the sign of the cross-price elastici-ties We do not have a good explanation of why these elastic-ities are estimated to be negative A potential explanation isthat the shares of the foreign products are very small giventhis we observe very few consumers switching from one (verysmall) foreign group to another (very small) foreign groupwhen the price of the first foreign group goes up and hence theinference is not very reliable in this case

34 The cross-price elasticities between foreign drug groupscontaining different molecules will also have an effect on thewelfare estimates given that the withdrawal of domestic prod-ucts will generally lead to changes in the prices of all foreignproducts and market shares will be reallocated from someforeign products to other foreign products based on the newprices However this reallocation from ldquoforeign to foreignrdquo istruly second order in our case compared to the reallocationfrom ldquodomestic to domesticrdquo and ldquodomestic to foreignrdquo prod-ucts In scenarios in which we consider patent enforcementaccompanied by strict price regulation the cross-price elastic-ities between foreign product groups are in fact completelyirrelevant as the prices of foreign products are not allowed toincrease in these cases Still our welfare loss estimates remainsubstantial drivenmdashas beforemdashby the loss of product variety

1497VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

not impose it through any of our assumptionsregarding the demand function The questionthat naturally arises then is what might explainthis finding While we cannot formally addressthis question anecdotal accounts in various in-dustry studies suggest that the explanation maylie in the differences between domestic andforeign firms in the structure and coverage ofretail distribution networks

Distribution networks for pharmaceuticals inIndia are typically organized in a hierarchicalfashion Pharmaceutical companies deal mainlywith carrying and forwarding (CampF) agents inmany instances regionally based who each sup-ply a network of stockists (wholesalers) Thesestockists in turn deal with the retail pharma-cists through whom retail sales ultimately oc-cur35 The market share enjoyed by a particularpharmaceutical product therefore depends inpart on the number of retail pharmacists who

stock the product And it is here that thereappears to be a distinction between domesticfirms and multinational subsidiaries In particu-lar the retail reach of domestic firms as agroup tends to be much more comprehensivethan that of multinational subsidiaries (IndianCredit Rating Agency (ICRA) 1999)36

There appear to be two reasons for this Thefirst is that many of the larger Indian firmsbecause they have a much larger portfolio ofproducts over which to spread the associatedfixed costs typically have more extensive net-works of medical representatives The second issimply that there are many more domestic firms(and products) on the market At the retail levelthis would imply that local pharmacists mightbe more likely to stock domestic products con-taining two different molecules say ciprofloxa-cin and norfloxacin than they would domesticand foreign versions of the same molecule Tothe extent that patients (or their doctors) arewilling to substitute across molecules in order tosave on transport or search costs (eg going toanother pharmacy to check whether a particular

35 There are estimated to be some 300000 retail pharma-cists in India On average stockists deal with about 75 retailers(ICRA 1999) There are naturally variations in this structureand a host of specific exclusive dealing and other arrangementsexists in practice Pharmaceutical firms also maintain networksof medical representatives whose main function is to marketthe companyrsquos products to doctors who do the actual prescrib-ing of drugs In some instances firms do sell directly to thedoctors who then become the ldquoretailerrdquo as far as patients areconcerned but these are relatively rare

36 These differences were also highlighted in conversa-tions one of the authors had with CEOs and managingdirectors of several pharmaceutical firms as part of a sep-arate study

TABLE 6AmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN THE NORTHERN REGION

Product group

Elasticity with respect to

Prices of foreign productgroups Prices of domestic product groups Overall

quinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 557 013dagger 015 401 011dagger 011dagger 016 137(179) (007) (007) (184) (006) (006) (006) (029)

Foreign norfloxacin 427dagger 045 427dagger 350dagger 602 451 465 220(242) (112) (242) (210) (623) (184) (183) (105)

Foreign ofloxacin 011 010dagger 138 009 009dagger 023 011 116(005) (005) (031) (027) (005) (028) (004) (017)

Domestic ciprofloxacin 018 001 001 168 008 008 010 117(008) (000) (001) (023) (002) (002) (002) (003)

Domestic norfloxacin 004 003 004 058 223 042 040 073(001) (003) (001) (017) (011) (004) (003) (009)

Domestic ofloxacin 005 005 011 077 074 342 074 089(002) (002) (013) (028) (008) (025) (008) (021)

Domestic sparfloxacin 007 004 007 115 063 063 288 028(002) (001) (002) (015) (006) (006) (017) (012)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent significance level and dagger (dagger) denotes significance at the 10-percent level

1498 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE 6BmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN OTHER REGIONS

Demand patterns in the eastern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 594 014dagger 016 431 013 011 017 140(195) (008) (008) (201) (006) (007) (007) (031)

Foreign norfloxacin 329dagger 058 329dagger 264dagger 460 346 359 192(180) (083) (179) (153) (451) (138) (139) (082)

Foreign ofloxacin 012 010dagger 140 010 010 024 013 117(006) (006) (034) (031) (005) (030) (005) (019)

Domestic ciprofloxacin 018 001 001 172 009 008 011 117(008) (000) (002) (027) (002) (003) (002) (003)

Domestic norfloxacin 004 004 004 068 242 049 046 070(002) (004) (002) (020) (013) (005) (004) (010)

Domestic ofloxacin 004dagger 004 009 061 060 295 060 092(002) (001) (010) (028) (007) (022) (007) (017)

Domestic sparfloxacin 005 003 005 092 048 051 251 043(002) (001) (001) (017) (005) (006) (016) (010)

Demand patterns in the western region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 427 009dagger 011 286 008 006 011 126(129) (005) (005) (133) (004) (005) (004) (021)

Foreign norfloxacin 714dagger 008 710dagger 594dagger 995 742 770 299dagger

(369) (183) (368) (328) (1015) (297) (296) (178)Foreign ofloxacin 013 012dagger 145 009 012 026 013 119

(007) (006) (037) (029) (005) (033) (005) (021)Domestic ciprofloxacin 019 001 000 174 009 008 011 118

(008) (000) (001) (024) (002) (003) (002) (004)Domestic norfloxacin 004 004 004 067 243 050 047 069

(002) (004) (001) (018) (014) (005) (004) (010)Domestic ofloxacin 003 003 007 048 048 257 048 093

(002) (001) (008) (022) (005) (015) (004) (014)Domestic sparfloxacin 006 003 005 083 046 049 241 046

(002) (001) (001) (014) (004) (005) (012) (009)

Demand patterns in the southern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 560 013dagger 015 402 011 010 017 137(182) (007) (007) (187) (006) (006) (007) (029)

Foreign norfloxacin 533dagger 031 531dagger 431dagger 748 559 584 249dagger

(301) (134) (300) (245) (734) (219) (219) (128)Foreign ofloxacin 014 012dagger 148 011 012 029 016 120

(007) (007) (039) (033) (006) (035) (006) (022)Domestic ciprofloxacin 017 001 000 169 008 007 011 116

(008) (000) (001) (025) (002) (003) (002) (003)Domestic norfloxacin 004 003 003 059 225 043 040 073

(001) (003) (001) (019) (012) (004) (004) (009)Domestic ofloxacin 004dagger 004 009 062 060 296 060 091

(002) (001) (010) (026) (006) (019) (007) (017)Domestic sparfloxacin 008 005 008 138 073 075 323 016

(002) (001) (002) (018) (007) (008) (021) (014)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at 5-percent level anddagger (dagger) denotes significance at the 10-percent level

1499VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

foreign product is in stock) in aggregate datawe would expect to find precisely the substitu-tion patterns that we report in Table 6

Whether the particular explanation we provideabove is the correct one the high degree of sub-stitutability between domestic product groupsturns out to have important implications for thewelfare calculations We discuss these in moredetail below when we present the results of thecounterfactual welfare analysis Another elasticitywith important implications for the counterfactu-als is the price elasticity for the quinolone subseg-ment as a whole which indicates how likelyconsumers are to switch to other antibioticsgroups when faced with a price increase for quin-olones This elasticity is computed on the basis ofthe results in Table A3 and it is at 111 (stan-dard error 024) this is large in magnitudebutmdashas expectedmdashsmaller in absolute value thanthe own-price elasticities of the product groupswithin the quinolone subsegment

The results in Tables 6A and 6B are based onour preferred specification discussed in SectionII In Tables A4 to A6 in the Appendix weexperiment with some alternative specificationsTables A4(a)ndashA4(c) correspond to a specificationthat includes in addition to product-group-specificregional fixed effects product-group-specific (andfor the upper level antibiotics-segment-specific)seasonal effects We distinguish among three sea-sonsmdashthe summer monsoon and wintermdashandreport the unconditional demand elasticities for

the northern region for each of these seasons Asevident from the tables in the Appendix our elas-ticity estimates are robust to the inclusion of sea-sonal effects The demand elasticities in Table A5are based on estimation of the demand system byOLS Compared to the elasticities obtained by IVthe OLS elasticities are smaller in absolute valueimplying that welfare calculations based on theOLS estimates would produce larger welfare lossestimates Nevertheless some of the patterns re-garding the cross-price elasticities discussed ear-lier are also evident in the OLS results inparticular the cross-price elasticities between dif-ferent domestic product groups are all positivelarge and significant and in most instances largerthan the cross-price elasticities between drugs thatcontain the same molecule but are produced byfirms of different domesticforeign status Theclose substitutability of domestic products indi-cated by both the OLS and IV estimates seems tobe one of the most robust findings of the paper

B Cost and Markup Estimates

Table 7 displays the marginal costs markupsand profits implied by the price elasticity esti-mates of Tables 6A and 6B for each of the sevenproduct groups Given that our regional effectsimply different price elasticities for each regionour marginal cost and markup estimates alsodiffer by region Given however that based on

TABLE 7mdashUPPER AND LOWER BOUNDS FOR MARGINAL COST MARKUP AND ANNUAL PROFIT BY PRODUCT GROUPS WITHIN

THE QUINOLONE SUBSEGMENT

Product groupLower boundfor MC (Rs)

Upper boundfor markup

Upper boundfor profit(Rs mill)

Upper boundfor MC (Rs)

Lower boundfor markup

Lower boundfor profit

(Rs)

Foreign ciprofloxacin 83 19 269 103 0 00(123) (012) (1655)

Foreign norfloxacin NA NA NA 53 0 00Foreign ofloxacin 323 70 1061 1085 0 00

(2316) (021) (3185)Domestic ciprofloxacin 47 59 17019 112 0 00

(114) (010) (29858)Domestic norfloxacin 52 43 2807 90 0 00

(020) (002) (1532)Domestic ofloxacin 587 34 1612 901 0 00

(218) (002) (1280)Domestic sparfloxacin 495 37 1985 788 0 00

(157) (002) (1100)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level Estimated lower bound forforeign norfloxacinrsquos marginal cost is negative since the estimated price elasticity is less than one in absolute value

1500 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Tables 6A and 6B the price elasticities do notseem to differ substantially across regions wereport for ease of exposition only the nationalaverages for marginal costs and markups inTable 7

Table 7 has two parts In the left part (firstthree columns) we report the lower bound formarginal cost and the corresponding upperbound for markup and upper bound for totalannual profit for each product group Thesenumbers are based on the price elasticities weobtained from estimating the two-level AIDSdemand system Since we do not have a reliableestimate for the price elasticity of foreign nor-floxacin (the point estimate is negative but lessthan one in absolute value and insignificant)we cannot compute the lower bound for mar-ginal cost in this case It is important to note thatthese estimates do not reflect either the actualmarginal cost or the actual markup for thesedrugs both because the existence of price reg-ulation implies that the unconstrained first-orderconditions are unlikely to hold each period andbecause our aggregation across firms of thesame domesticforeign status supplying thesame molecule makes the interpretation of theseestimates problematic In particular the fact thatwe ignore competition among firms within eachproduct group implies that our estimates willtend to overstate market power These numberswill prove useful however in the counterfac-tual simulations as they can give us a sense ofhow large the maximum profit gains for multi-national firms and the maximum profit lossesfor domestic firms are likely to be under patentenforcement The right part of the table (lastthree columns) reports the upper bound for mar-ginal cost which we obtain by simply taking themarginal cost to be equal to the observed priceWe do not report standard errors in this casesince the numbers are based on actual data Themarkups corresponding to these marginal costupper bounds are of course zero We conductthe counterfactual simulations using both thelower and upper bounds for marginal costs

The (upper bounds for) markups on the leftside of the table are generally plausible Thedomestic ciprofloxacin product group that dom-inates the quinolone subsegment and for whichwe documented high prices a high marketshare and a relatively low elasticity of demandenjoys one of the highest markup upper bounds(60 percent) and accounts for nearly 70 percent

of all profits derived within the subsegmentForeign ciprofloxacin on the other hand whichfaces the stiffest competition from domesticfirms and for which we estimated a highly elas-tic demand has the lowest markup upper bound(19 percent)

C Counterfactual Estimates of the Impact onPrices and Welfare

With estimates of the key demand and costparameters in hand we turn to the counterfac-tuals We consider the five separate scenarioslisted in the previous section All of the scenar-ios involve the withdrawal of one or more of thedomestic product groups from the market Ta-ble 8 displays our estimates of the consumerwelfare losses that result under the differentscenarios The losses are expressed in billion ofrupees per year All numbers presented in Ta-ble 8 and subsequent tables are based on usingthe lower bounds for marginal cost and upperbounds for markup in the simulations As dis-cussed earlier these numbers are the more in-teresting to work with since they give us anupper bound for the changes in the profits ofdomestic and foreign firms that would resultfrom patent enforcement In the Appendix wealso present results based on using the upperbounds for marginal costs in the simulations inwhich case the pre-TRIPS profits of domesticand foreign firms are zero In all cases marginalcosts are assumed to be constant in outputWhile naturally the profit implications differdepending on whether one uses the upper orlower bounds for marginal costs (firm profits arezero if one assumes the upper bound of mar-ginal cost in which case price equals marginalcost) the estimated consumer welfare losses aresimilar in the two cases We discuss these re-sults in more detail at the end of this section

The first column presents our estimates of theconsumer welfare losses attributable to the pureloss of product variety effect where we fix theprices of all remaining products as well as theoverall expenditure on quinolones while with-drawing one or more of the domestic productgroups Note that in our initial specification ofthe demand system had we not allowed for thepossibility that consumers might differentiatebetween domestic and foreign products evenwhen they contain the same molecule this

1501VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

particular component of the loss of consumerwelfare would not have arisen

The estimates reported in the second columnincorporate the expenditure-switching effect ontop of the loss of product variety Here basedupon the price elasticity estimates from thehigher-level AIDS system we adjust (down-ward) the expenditures allocated to the quino-lone subsegment as the composite price ofquinolones effectively increases as a consequenceof the higher virtual prices of the domestic productgroups that are withdrawn from the market Be-cause the estimates in this column are generatedassuming that the prices of the products that re-main in the market are not adjusted upward theyprovide a sense of what consumer losses would beif the introduction of product patents were coupledwith strict price regulation aimed at maintainingprices at prepatent levels Alternatively they canbe thought of as the relevant welfare numbers ifintense competition among firms within the re-maining product groups kept the prices of theproducts that were still offered in the market closeto the firmsrsquo marginal costs

The last column of Table 8 displays the esti-mated consumer welfare losses when bothcross-segment expenditure-switching and with-in-segment upward price adjustments are takeninto account

If we compare the results across the firstsecond and third columns all the counterfac-tual scenarios produce qualitatively similar pat-terns patterns that are consistent with what we

would expect Starting from the initial loss ofwelfare attributable to the loss of product vari-ety the option of switching expenditures outof the quinolone subsegment to other subseg-ments mitigates some of the initial welfareloss But if we then incorporate the upwardprice adjustments that result in response to thereduced competition the welfare losses aremagnified

Of particular interest from a policy perspec-tive are the relative magnitudes of these threeeffects which are similar under all the counter-factual scenarios though absolute levels varyconsiderably First despite the fact that the de-mand for quinolones is quite sensitive to thecomposite price of quinolonesmdashthe upper levelprice elasticity is 111mdashthe cross-subsegmentexpenditure-switching effects are in all thecases small (in absolute value terms) relative tothe other two effects For instance under thescenario where all the domestic quinolone prod-uct groups are withdrawn from the market theoverall consumer welfare loss of Rs 1781 bil-lion per year can be decomposed into an initialloss of Rs 1176 billion (66 percent) attribut-able to the loss of product variety a slightreduction in this initial loss of Rs 041 billion(2 percent) from Rs 1176 billion to Rs1135 billion because of expenditure switchingand a subsequent additional loss of Rs 646billion (36 percent) from Rs 1135 billion toRs 1781 billion because of the reduced com-petition and consequent price increases

TABLE 8mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure loss ofvariety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 732(087) (089) (146)

Only ofloxacin 008 008 023(008) (008) (010)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1253(177) (180) (415)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1058(142) (145) (331)

All four domestic quinolones products 1176dagger 1135dagger 1781(643) (634) (1270)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent significance level and dagger (dagger)denotes significance at the 10-percent level

1502 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 3: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

the United States at the time The presence ofmany therapeutic substitutes within the antibi-otics segment make this product category idealfor investigating the claim that the presence ofclose substitutes will prevent drug prices fromrising once patent protection is enforced Ofcourse to the extent that our estimates refer toantibiotics they are not directly applicable toother pharmaceutical product categories thatmay have different demand structures For ex-ample a finding of large substitution effectstoward nonpatented products would not neces-sarily apply to a market segment with only fewor possibly no therapeutic substitutes Still afinding of limited substitution toward otherdrugs and associated large price increaseswould suggest that the effects of patent enforce-ment in other pharmaceutical segments withfewer therapeutic substitutes might be evenlarger

India provides a natural setting for our anal-ysis for a number of reasons It is a leadingexample of a low-income country that did notrecognize pharmaceutical product patents at thetime the TRIPS agreement went into effect Infact during the Uruguay round of negotiationsIndia led the opposition to the TRIPS articlesmandating pharmaceutical product patents Interms of the structure of demand India is aprototypical example of a low-income countrywith a large number of poor households whichbecause health insurance coverage is nonexist-ent have to meet all medical expenses out ofpocket Moreover the disease profile of theIndian population mirrors that of many otherlow-income countries and is considerably dif-ferent from that of most developed economiesLastly the domestic Indian pharmaceutical in-dustry which as of 2002 was the largest pro-ducer of generic drugs in the world in terms ofvolume is typical of that of many middle-in-come countries with large numbers of small andmedium-sized firms with significant imitativecapabilities producing and marketing drugs do-mestically that are under patent elsewhere

During the period covered by our data sev-eral molecules in the quinolone family were stillunder patent in the United States but productscontaining these molecules were being pro-duced and distributed in India by both a numberof domestic firms and a number of local subsid-iaries of foreign multinationals We aggregatethese products into a number of mutually exclu-

sive product groups where within each productgroup all products contain the same quinolonemolecule (eg ciprofloxacin or norfloxacin)and are produced by firms with the same do-mestic or foreign status We then estimate atwo-level demand system employing the Al-most Ideal Demand System (AIDS) specifica-tion of Angus S Deaton and John Muellbauer(1980b) in both levels The higher level corre-sponds to the allocation of expenditures tovarious subsegments within the systemic anti-bacterials segment of the market At the lowerlevel we estimate the parameters relevant forthe allocation of expenditures within the quin-olone subsegment to the various productgroups within this subsegment (eg foreignciprofloxacin domestic ciprofloxacin or do-mestic norfloxacin)

With these estimates in hand we turn to thecounterfactuals The basic counterfactual sce-narios we consider all involve the withdrawal ofone or more of the domestic quinolone productgroups from the market The idea here is thathad US patents for say ciprofloxacin beenrecognized in India all domestic products con-taining ciprofloxacin would not be present in themarket That would leave only the foreign cip-rofloxacin product group in the market Usingour estimates of the own cross-price and ex-penditure elasticities of the various productgroups as well as estimates for the upper andlower bounds of marginal costs of productionwe are able to simulate the prices and marketshares that would obtain under each of the sce-narios Moreover using the expenditure func-tion associated with the higher-level AIDSspecification we are able to calculate the wel-fare lossmdashmeasured in terms of the compensat-ing variation ie the additional expenditurethat the representative Indian consumer wouldneed to incur to maintain her utility level inthe face of the domestic product withdraw-al(s) and the accompanying price and marketshare changesmdashunder each of the counterfac-tual scenarios

Apart from the fact that our counterfactualsimulations are based on estimated rather thanassumed parameter values this paper buildsupon the earlier studies in two substantive and(it turns out) empirically important ways

First by accommodating the possibility thatconsumers may differentiate between domesticand foreign products even when these products

1479VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

contain the same patentable molecule we allowfor an additional channel through which theintroduction of product patents and the conse-quent withdrawal of domestic products may ad-versely affect consumers that is through theloss of product variety In contrast previousstudies on developing countries assume thatconsumers are indifferent between foreign anddomestic products that contain the same mole-cule What this implies is that any adverse wel-fare effects are realized only through increasedprices The difference is most evident if weconsider a scenario under which domestic prod-ucts are forced to withdraw from the marketbecause of the introduction of product patentsbut strict price regulations maintain prices atprepatent levels In our approach consumerswould still experience a welfare loss whereas inthe framework adopted in earlier studies such ascenario would entail no loss of welfare

Empirically the component of the consumerwelfare loss attributable to the reduction of va-riety from the withdrawal of domestic productsturns out to be significant We interpret thiscomponent as capturing primarily an ldquoease ofaccessrdquo effect due to differences in the market-ing and distribution networks domestic prod-ucts are more readily available to Indianconsumers than products produced by foreignsubsidiaries From a policy perspective thissuggests a possible role for compulsory licens-ing in addition to or in lieu of price regulationsince the latter by itself will not alleviate thewelfare loss due to loss of variety Alterna-tively one could argue thatmdashto the extent thatthe loss we attribute to the reduction of productvariety is due to the fact that the current productportfolios and distribution networks of foreignproducers are limitedmdashit is purely a transitionalphenomenon and should thus not be included inthe welfare calculations This is a controversialpoint we discuss in detail in the results sectionIf foreign firms respond to patent enforcementby investing in distribution networks or by us-ing licensing agreements with domestic firms tomake their products more readily available toIndian consumers the ldquoease of accessrdquo effectwould indeed diminish in importance in thelonger run though of course it could be signif-icant in the first years after patent enforcementWhether these investments will materializehowever is open to debate If TRIPS is accom-panied by price regulation in order to limit price

increases in poor developing countries the in-centives of multinationals to invest in marketingand distribution in these countries may dimin-ish At any rate to take into account the possi-bility that the welfare loss due to the reductionof variety is a temporary phenomenon we alsopresent a more conservative welfare loss esti-mate by subtracting the ldquoproduct varietyrdquo com-ponent from our total loss estimate This givesus a lower bound estimate that is due to priceincreases alone Though it is only about a thirdof our upper bound estimate this lower boundestimate is still very large in absolute termsrepresenting 24 percent of antibiotic sales in2000

A second and perhaps even more importantmethodological difference between this paperand earlier studies is that we allow for andflexibly estimate a range of cross-product-group and cross-molecule substitution effectsIn contrast cross-price effects are ignored inearlier studies To see why cross-price effectsare likely to alter estimated welfare effects sig-nificantly in this context imagine a scenariowhere the introduction of patents leads to mo-nopoly pricing in the market for a particularpatentable molecule If the markets for potentialsubstitutes are imperfectly competitive then theincrease in price in the original patentable mar-ket will lead to corresponding upward priceadjustments in the related markets as producersof substitute products reoptimize in the face ofthe increased demand for their products Themagnitude of any upward adjustments will nat-urally vary with the degree of competition inrelated markets and with the strength of thecross-price effects But as long as the cross-price effects are positive and related marketsare not perfectly competitive the loss of con-sumer surplus because of monopoly pricing inone market will be multiplied through the rippleeffects of upward price adjustments in relatedmarkets

If this were just a theoretical possibility itwould not be of much interest These multipliereffects however turn out to be substantial inour counterfactual scenarios Most strikinglythe estimated loss of consumer welfare from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenar-

1480 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

ios in each of which only one of the domesticproduct groups is withdrawn What this veryclearly indicates is that past studies that haveestimated the aggregate effects of patent protec-tion by adding up the losses estimated sepa-rately in each of a number of patentablemarkets may have substantially underestimatedthe magnitude of the consumer welfare lossesfrom the introduction of pharmaceutical productpatents

In absolute terms we estimate that in theabsence of any price regulation the prices offoreign patented products would rise between100 percent and 400 percent In the more real-istic case of some form of price regulation thatwould keep drug prices fixed at their pre-TRIPSlevel the total annual welfare losses to theIndian economy from the withdrawal of all fourdomestic product groups in the quinolone sub-segment would be on the order of Rs 1370billion or about 50 percent of the sales of theentire systemic antibacterials segment in 2000At the then-prevailing exchange rate this trans-lates into a figure of US$305 million Of thisamount foregone profits of domestic producersconstitute roughly Rs 23 billion or US$50 mil-lion The overwhelming portion of the totalwelfare loss therefore derives from the loss ofconsumer welfare

The welfare loss we estimate represents onlythe static costs of patent enforcement arisingfrom pricing distortions and reduction in prod-uct variety Our approach does not address thepotential dynamic benefits of innovations thatmay result from international property rightsprotection Nevertheless we believe that esti-mating these static costs is important wheneverthere is a radical change in policymdashwhichTRIPS represents for a good part of the devel-oping world Even if there is the potential forlong-term benefits knowledge of the short-runcosts is important for designing an appropriatepolicy response that will potentially mitigate theadverse short-run impact Having said that itis worth noting that according to our esti-mates the total profit gains of patent enforce-ment to foreign producers in the absence ofany price regulation would be only aboutUS$53 million per year With price regulationthat would keep the prices of drugs suppliedby multinational subsidiaries at their pre-TRIPS level the profit gains become onlyUS$196 million per year

The remainder of this paper is organized asfollows In the next section we lay out theessential features of the Indian pharmaceuticalsmarket provide more detail about the segmentsthat we focus on in the empirical analysis andbriefly describe the primary data we use Sec-tion II describes the analytic framework and theeconometric strategy we use to estimate therelevant parameters and construct the counter-factual scenarios We discuss our results in Sec-tion III Section IV concludes

I The Setting and the Data

Between April 1972 when the Indian PatentsAct (1970) became effective and March 2005when Indiarsquos parliament passed the 3rd Amend-ment of the Patents Act India did not recognizeproduct patents for pharmaceuticals The IndianPatents Act which replaced the inherited Brit-ish colonial law regarding intellectual propertyrights specifically excluded pharmaceuticalproduct patents and admitted process patentsonly for a period of seven years In contrast thelatest amendment recognizes patents on endproducts that under the new regime will remainin force for 20 years6

The two stated objectives of the 1970 actwere the development of an indigenous pharma-ceuticals industry and the provision of low-costaccess to medicines for Indian consumers Con-sistent with these objectives and with the broaderleftward tilt in policy a number of other measureswere introducedmdashdrug price controls restrictionson capacity expansion limits on multinational eq-uity sharesmdashthat in the years since have on theone hand kept pharmaceutical prices low and onthe other encouraged the development of the In-dian pharmaceutical industry Many of these reg-ulations and restrictions have been lifted or easedsince the mid-1980s with marked acceleration inthe pace of liberalization during the 1990s

Over the last 20 years the Indian pharmaceu-tical industry has grown rapidly to the pointwhere it is now the worldrsquos largest producer offormulations in terms of volume and one of the

6 Indian companies that are now producing drugs forwhich patent applications were submitted between the sign-ing of the TRIPS agreement in 1995 and January 1 2005will be allowed to continue producing if they pay a royaltyto the patent holder

1481VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

worldrsquos largest producers of bulk drugs7 Thestructure of the industry has also evolved In1970 the industry was dominated by multina-tional subsidiaries by 2001 Indian-owned firmswere not just the leading players in the industrymany had also become major exporters

The data we use in this paper are from theretail pharmaceutical audits of ORG-MARGIndiarsquos premier market research and consultingfirm The audit provides detailed product-levelinformationmdashestimates of monthly retail salesin each of the four geographic zones of Indiaprice dosage form launch date brand namechemical name therapeutic categorizationmdashon

all pharmaceutical products sold in India byabout 300 of the largest firms representingroughly 90 percent of domestic retail sales ofpharmaceuticals The coverage of the audit isextensive reaching a representative panel ofthousands of retail chemists in over 350 citiesand towns The data collected which providethe only real source of disaggregate informationon the Indian pharmaceutical market are usedby both the government of India in formulatingpricing policy and other decisions and the In-dian pharmaceutical industry in determiningpricing and marketing strategies We have in-formation at a monthly frequency for the periodof January 1999 to December 2000 Tables 1 to3 provide a set of descriptive statistics that areessential for understanding the focus of ouranalysis and interpreting our results

As noted earlier the characteristics of de-mand for pharmaceuticals in India are likely todiffer considerably from those in developedeconomies With a share of 23 percent the

7 Bulk drugs are the therapeutically relevant active phar-maceutical ingredients that are combined with a variety ofinactive ingredients to make the formulations ultimatelyconsumed by patients Firms in the pharmaceutical sectorcan be one of three types bulk drugs producers pureformulators or integrated firms which produce both bulkdrugs and market formulations

TABLE 1mdashTHE QUINOLONES SUBSEGMENT

Molecule

Shares ()of quinolones sales Sales (Rs mill) 2000

Domesticfirms

Foreignsubsidiaries

Domesticfirms

Foreignsubsidiaries

Ciprofloxacin 530 27 3030 156Norfloxacin 112 01 640 3Ofloxacin 116 31 665 177Sparfloxacin 108 01 620 4Lomefloxacin 15 86 Pefloxacin 13 01 72 5Levofloxacin 00 0 Nalidixic acid 13 73

TABLE 2mdashBASIC INFORMATION ABOUT THE FOUR QUINOLONES MOLECULES 2000

Ciprofloxacin Norfloxacin Ofloxacin Sparfloxacin

US or European patent holder Bayer Merck Ortho-McNeil Rhone-PoulencYear of US patent expiry 2003 1998 2003 2010Year of US-FDA approval 1987 1986 1990 1996Year first introduced in India 1989 1988 1990 1996No of domestic Indian firms 75 40 17 25No of foreign subsidiaries 8 2 2 1Sales weighted average per-unit API

price of products produced byDomestic Indian firms 1124 905 9008 7884Foreign subsidiaries 1035 528 10847

API Active pharmaceutical ingredient

1482 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

antiinfectives segment ranks second in Indiawhereas in the world market it is fifth and has ashare of only 90 percent Hence antiinfectivesare important in India not only from a health

and public policy point of view but also as asource of firm revenue

With this in mind we focus on one partic-ular subsegment of antiinfectives namely the

TABLE 3mdashSUMMARY STATISTICS FOR THE QUINOLONES SUBSEGMENT 1999ndash2000

North East West South

Annual quinolones expenditure per household (Rs) 3125 1975 2764 2359(366) (367) (407) (286)

Annual antibiotics expenditure per household (Rs) 11988 8424 11052 9624(1224) (1224) (960) (996)

No of SKUsForeign ciprofloxacin 1238 1129 1308 1246

(150) (190) (102) (106)Foreign norfloxacin 183 171 200 158

(070) (075) (088) (083)Foreign ofloxacin 304 296 296 300

(086) (086) (091) (088)Domestic ciprofloxacin 10621 9763 10342 10550

(599) (434) (722) (451)Domestic norfloxacin 3896 3496 3617 3942

(271) (268) (251) (379)Domestic ofloxacin 1846 1600 1725 1725

(680) (634) (586) (635)Domestic sparfloxacin 2983 2829 3121 2929

(557) (638) (688) (657)Price per-unit API (Rs)

Foreign ciprofloxacin 958 1090 1085 1007(128) (066) (071) (058)

Foreign norfloxacin 563 509 605 435(077) (133) (139) (147)

Foreign ofloxacin 10946 10943 10886 10612(620) (664) (700) (1140)

Domestic ciprofloxacin 1143 1067 1131 1152(016) (015) (017) (013)

Domestic norfloxacin 951 907 888 873(024) (035) (037) (020)

Domestic ofloxacin 9163 8964 8565 9341(1615) (1565) (1422) (1407)

Domestic sparfloxacin 7972 7849 7688 8028(976) (1014) (1185) (1037)

Annual sales (Rs mill)Foreign ciprofloxacin 4179 2431 4520 2947

(1534) (816) (1273) (648)Foreign norfloxacin 128 100 058 073

(101) (082) (044) (057)Foreign ofloxacin 5446 3184 3522 3111

(1399) (933) (906) (703)Domestic ciprofloxacin 96229 58591 67874 70381

(10626) (13026) (12226) (8740)Domestic norfloxacin 22255 11971 14918 15829

(3884) (1945) (2691) (1626)Domestic ofloxacin 12502 9621 14936 11205

(4434) (3011) (5282) (4259)Domestic sparfloxacin 15617 12175 16130 9811

(3141) (2576) (4674) (3420)

Note Standard deviations in parentheses API Active pharmaceutical ingredient

1483VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

quinolone subsegment Quinolones fall into thesystemic antibiotics and antibacterials segmentof the Indian pharmaceuticals market whichgenerates over three-quarters of the revenues inthe antiinfectives segment8 The systemic anti-bacterials segment includes all of the originalmiracle drugs that first sparked the growth ofthe global research-based pharmaceutical indus-try in the postndashWorld War II period as well aslater generations of molecules that have beenintroduced in the last four decades

Among systemic antibacterials quinolonesare the latest generation molecules available inIndia We focus our analysis on quinolones forseveral reasons First quinolones are the drug ofchoice for a large number of bacterial infec-tions some of which are also treated by alter-native drugs (see Table A1 in the Appendixwhich outlines the spectrum of activity for eachmolecule family within the antibacterials seg-ment) Hence if there were one product groupfor which we would expect to have many sub-stitutes readily available this would be quino-lones Second with a share of 20 percent in thesales of systemic antibacterials quinolones rep-resent one of the largest subsegments within thistherapeutic category Finally several moleculeswithin the quinolone subsegment were still un-der patent in the United States at the time of ourinvestigation Table 2 details the basic informa-tion about the four quinolone molecules that arethe focus of our analysis The first row showsthe year of US patent expiry this ranges from1998 for norfloxacin to 2010 for sparfloxacinQuinolones include in principle four moremolecules that are listed at the bottom of Ta-ble 1 the market shares of these molecules arenegligible however so that we exclude thesemolecules from our analysis

Table 2 reveals several other interesting factsabout competition in the quinolone market inIndia First note the large number of firmsoperating in this subsegment The large numberof domestic firms is perhaps not that surprisinggiven that pharmaceutical product patents werenot recognized in India9 What is more surpris-ing is the number of foreign firms selling pat-ented products (eg ciprofloxacin) the fact that

multiple foreign firms sell a patented productindicates that such firms often ldquoinfringerdquo10

patent laws in India while complying with themin developed world countries The last two rowsof Table 2 further indicate that domestic prod-ucts often sell at a premium With the exceptionof ofloxacin the average prices of products of-fered by Indian firms are higher than the pricesof products offered by foreign subsidiaries Thispreliminary evidence suggests that Indian con-sumers do not place a premium on the brandname and reputation of big multinational phar-maceutical concerns Moreover the higherprice of domestic products does not seem toprevent domestic companies from capturing alarge market share This is most evident in thecase of ciprofloxacin where domestic firmshave with 53 percent the largest share in thetotal sales of quinolonesmdashthis despite the factthat the average price of these products is 10percent higher than the price of foreign productscontaining the same molecule

Table 3 provides additional summary statis-tics for our data broken down by region Thefirst two rows of the table report the averageannual household expenditure on quinolonesand antibiotics respectively Note that in bothcases the average expenditure is higher in thenorth and west these regions include states withhigher per capita incomes and tend to be moreindustrialized and urbanized than those in theeast and south Pharmaceutical products areavailable in multiple presentations that is com-binations of dosage forms (eg capsule tabletsyrup) strength (eg 100 milligrams 500 mil-ligrams) and packet sizes (eg 50 capsule bot-tle 100 tablet bottle) The various presentationsin which a product is available are often referredto as stock-keeping units or SKUs11 The num-ber of SKUs for each product group withinquinolones is reported at the top of Table 3 Aswith the more aggregate numbers on firms and

8 In addition to anti-bacterials this segment also containsantivirals

9 Accordingly the common distinction between ldquobrandedrdquoand ldquogenericrdquo products is irrelevant here

10 We emphasize here that the word ldquoinfringerdquo belongsin quotes because patent laws do not currently exist inIndia infringement in the legal sense is not possible It ishowever striking that the same firms that accuse Indianproducers of ldquopiracyrdquo sell in India products that are patentedin the United States and for which the patent is held by adifferent multinational corporation

11 For instance a 100 capsule bottle of 100 milligramcapsules of a particular branded drug and a 50 capsulebottle of 100 milligram capsules of the same branded drugwould be identified as two separate SKUs

1484 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

products reported in Table 2 the differencebetween domestic and foreign products isstriking The number of SKUs offered byIndian firms is consistently larger than thenumber offered by subsidiaries of foreignmultinationals The number of SKUs variesslightly across regions but more importantlyit varies across time as some SKUs disap-pear while new ones get introduced duringour sample period

Many pharmaceutical products in India aresubject to price controls12 While the specificsof the price regulation are too complex for anyeconomic model to capture adequately themain concern for the empirical analysis is thatprice controls may lead to a lack of price vari-ation over time so that the demand functioncannot be identified Prices at the most disag-gregate SKU level are relatively stable overtime there are variations due to occasionalchanges in the estimated cost (due for exampleto changes in exchange rates that affect the costof imported materials or bulk drugs) but suchvariations tend to be infrequent and small inmagnitude The degree of time variation ishowever substantially larger once one aggre-gates to the product level This variability stemsnot only from the fact that the SKUs over whichwe aggregate may experience changes in theirrespective prices at different points in time butalso from the fact that the range of SKUs of-fered in the market does not remain constantover time The entry and exit of presentationswithin the same product group that have differ-

ent prices effectively affects the price that con-sumers face for this drug in each period

The middle portion of Table 3 reports themean price and standard deviation for eachproduct group by region Prices vary by regionthough there is no clear pattern emerging fromthe table with regard to the cross-regional vari-ation (in the sense of some regions being sys-tematically more expensive than others) Toexamine what portion of the total price variationis due to time versus regional variation weconducted an analysis of variance of prices thatwe report in Table 4 The table is based onseparate regressions for each product group(pooling data across groups with big differencesin their average prices is not particularly infor-mative as most of the price variation is ac-counted for by product group dummies) Thelast two columns of the table show the fractionof price variation accounted for by region andtime dummies respectively As evident fromthe table a significant fraction of the total vari-ance in prices can be attributed to time varia-tion In the demand estimation we include a fullset of product groupndashspecific regional dummiesso that the price parameters are identified en-tirely based on this time variation within eachproduct group The time variation of productgroup prices is driven primarily by composi-tional changes within each group the revenueshares of the individual SKUs that compriseeach product group change over time (see therelated discussion in Section IIC) while in eachperiod there is entry and exit of SKUs into thesample To check whether this pattern reflectsgenuine entry and exit as opposed to samplingvariation we examined the revenue shares ofthe SKUs that leave the sample relative to theones that remain during the entire period Theresults are reported in Table 5 While the SKUs

12 The details of the procedures for price fixation canbe found in the official government Web site httpwwwnppaindianicinindex1html under the link ldquoDrug Price Con-trol Order 1995rdquo A new pharmaceutical policy was introducedin 2002 but our data were collected before that year

TABLE 4mdashANALYSIS OF PRODUCT PRICE VARIANCE

Product groupPartial SS

timePartial SS

region Total SSPercentage

explained by timePercentage

explained by region

Foreign ciprofloxacin 0296 0306 1047 283 292Foreign norfloxacin 1036 1002 4179 248 240Foreign ofloxacin 0429 0021 0571 752 36Domestic ciprofloxacin 0005 0088 0104 50 841Domestic norfloxacin 0059 0098 0198 296 494Domestic ofloxacin 2858 0104 3056 935 34Domestic sparfloxacin 1754 0032 1860 943 17

1485VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

that exit tend to be smaller (their average shareis 1 percent as opposed to 34 percent for thoseSKUs that are present during the entire sampleperiod) the shares of the two groups do notseem orders of magnitude apart13 In additionour data cover only the 300 largest firms sellingin the Indian market so that firms with verysmall shares are not included in our sample

II The Analytic Framework and EstimationApproach

Patent enforcement in the Indian pharmaceu-tical market will have the effect of eliminatingdomestic products whose active pharmaceuticalingredients are protected by (foreign) patentsThus assessing the effects of patent enforce-ment is tantamount to assessing the effects ofwithdrawing domestic products from the mar-ket This task is the converse of evaluating newproduct introduction accordingly the concep-tual framework we use to address the questionsof interest is similar to the one developed in theliterature for the valuation of new goods14

We start by estimating demand for quino-lones Given that the market is characterized byimperfect competition the counterfactual anal-ysis requires that we also model the supply sideas removal of one product will affect the pricesof other products especially those that are closecompetitors The existence of price regulationin the Indian pharmaceutical market imposespotential constraints on firmsrsquo maximizationproblem Given these constraints and the com-

plexity of the price regulation process the typ-ical approach of deriving estimates of actualmarginal costs and markups by exploiting thefirst-order conditions of profit maximizing firmsdoes not seem particularly promising Insteadwe use our demand estimates to place upper andlower bounds on marginal costs and markups

With demand elasticities and upper and lowerbounds for marginal costs in place we thenconduct counterfactual simulations We con-sider several alternative scenarios dependingon the number of domestic products affected bypatent enforcement For each scenario we com-pute the counterfactual prices and use them toassess the effects of domestic product with-drawal on consumer welfare (as measured bythe compensating variation) firm profits andsocial welfare As with the valuation of newproducts the big conceptual problem facing thispart of the analysis is that we need to extrapo-late from the region of the data to the point atwhich demand for the products that exit themarket becomes zero This conceptual issue ispresent in any attempt to evaluate a major pol-icy change for which no historical precedentexists like the enforcement of patent laws inIndia One advantage of the present study is thatwe have a limited set of price data for Pakistana country with similar demographics to Indiabut with a market structure that resembles theone that would emerge in India under patentenforcement (monopoly of multinational sub-sidiaries) By comparing the prices of productsoffered by multinationals in Pakistan to thosewe compute in our counterfactual simulationsfor the products that would be offered by mul-tinationals in India if patent laws were enforcedwe can get a sense of how plausible our coun-terfactual estimates are

A Demand

The demand modelling is based on the multi-stage budgeting approach Our primary motiva-tion for adopting this approach is a practical oneIn the multistage budgeting approach the depen-dent variable is defined as a revenue share whichis appealing given that the products we include inthe analysis contain different molecules (ie ac-tive pharmaceutical ingredients or APIs) Eventhough we do have data on the quantity of therelevant API (eg 100 milligrams of ciprofloxa-cin) contained in each product converting the

13 While we cannot completely rule out the possibilitythat some of the exit is due to sampling variation note thatthe latter should be reflected in low precision of the demandparameter estimates The demand parameters however areprecisely estimated

14 See Manuel Trajtenberg (1989) Jerry A Hausman(1994) and Timothy F Bresnahan (1997) for representativeexamples and a discussion of the relevant issues

TABLE 5mdashREVENUE SHARES OF THE EXITING SKUS

Revenue share ofexiting SKUs

Revenue shares ofall SKUs

Full sample 10 34Northern region 08 33Eastern region 12 36Western region 10 34Southern region 12 33

1486 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

revenue shares to physical shares is extremelydifficult if not infeasible in the case of antibioticsBecause such drugs are ldquosystemicrdquo by nature theyare used to treat a large number of infections andthe dosage of each drug depends on the particularinfection it is supposed to address (for example thedosage will differ depending on whether the anti-biotic is used to combat an ear infection or tuber-culosis) This particular feature of antibioticscomplicates the conversion of revenue to physicalshares

The basic idea of the multistage budgetingapproach is to use the therapeutic classificationof a productmdashie the therapeutic segment andsubsegment the product belongs tomdashto orga-nize all products in the systemic antibacterialssegment into a hierarchical taxonomy consist-ing of two levels At the higher level are thevarious subsegments of systemic antibacterialsThe first stage of budgeting corresponds to theallocation of expenditures across the subseg-ments in this upper level of the taxonomy

In the second stage of the budgeting processcorresponding to the lower level of the taxon-omy a flexible functional form is adopted tomodel how the expenditures allocated to eachsubsegment are distributed across the productswithin that subsegment In particular to modeldemand at the second stage we employ theAIDS specification proposed by Deaton andMuellbauer (1980b)15

While the two-stage demand estimation ap-proach offers functional form flexibility its ap-plication to the Indian systemic antibioticsmarket poses a few problems The first is thatdue to entry and exit many SKUs and evenproducts in our sample are not present in everyperiod AIDS does not have a good way ofdealing with a varying number of products as itwas developed with broad commodity catego-ries in mind which are consumed by all con-sumers every period To solve this problem weaggregate within each subsegment (eg quino-lones) SKUs into product groups where withineach product group all SKUs contain the samemolecule and are produced by firms with thesame domesticforeign status Specifically let aSKU k be indexed by its molecule (or API) M its

domesticforeign status DF indicating whether itis produced by a domestic (Indian) or a subsidiaryof a foreign (multinational) firm a particular pre-sentation s and the particular firm f that producesit We aggregate SKUs over presentations andfirms to obtain a newly defined product group iwhich is indexed only by molecule M and domes-ticforeign status DF and has revenue Ri yenfsRk with i (M DF) k (M DF f s) and pricepi yenfs kpk where k denotes the conditional(on M and DF) revenue share of this particularproduct ie

(1) k Rk

Ri

In most cases the resulting product groups arebroad enough to be present every period16

The usual concern with this aggregation pro-cedure is that it may lead us to overstate firmsrsquomarket power as we ignore competition amongfirms with the same domesticforeign statusproducing the same molecule In the presentapplication however this concern is unlikely tobe of great importance as the effect of patentenforcement is to wipe out all domestic compe-tition at once while granting foreign firms mo-nopoly power hence competition among firmsfor patented molecules becomes irrelevant Theaggregation according to the domesticforeignstatus (within a particular molecule) thus corre-sponds to the scope of our analysis and theparticular questions of interest

The second problem is that for our approachto be useful in welfare analysis the allocation oftotal expenditures to group expenditures at thehigher stage has to be modelled in a way con-sistent with utility maximization In general thesolution of this allocation problem requiresknowledge of all individual product prices

15 Representative applications of the multistage budget-ing approach include Sara F Ellison et al (1997) Hausman(1994) and Hausman and Gregory K Leonard (2002)

16 We are missing only four observations (ie monthregion combinations) all for the drug group of ForeignNorfloxacin August 1999 in the south May 2000 in thewest October 2000 in the south and November 2000 in theeast In these cases we set the revenue shares of ForeignNorfloxacin equal to zero In general with 01 percent ofquinolone sales (see Table 1 row 2) Foreign Norfloxacinhas a very small share of the market This probably explainswhy the results pertaining to this drug group are unreliableit is the only drug group for which we do not obtain asignificant price elasticity of demand while its cross-priceelasticities with other foreign drug groups often have thewrong sign

1487VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

From an empirical point of view this is notparticularly useful as it eliminates all compu-tational advantages of the two-stage approachTo address this problem we adopt an approxi-mate solution to model the higher-level expen-diture allocation along the lines suggested byDeaton and Muellbauer (1980b pp 131ndash32)This gives rise to a two-level AIDS specification

Consider the lower-level estimation firstwhich refers to the allocation of a particularsubsegmentrsquos expenditure to the product groupswithin the subsegment In our application therelevant subsegment is quinolones which weindex with Q Let the product groups within thissubsegment be indexed by i 1 N pi be theprice of product group i (where as noted abovei refers to a particular molecule and domesticforeign status combination) and XQ be the totalexpenditure on the quinolone segment The rev-enue share of each product group is given by

(2) i i j

ijln pj ilnXQ

PQ

where i the revenue share of product group iis defined as

(3) i pi qi

yenj

pj qj

xi

XQ with i j Q

XQ is the overall expenditure on the quinolonesubsegment and PQ is a price index given by

(4) ln PQ ap 0 i

iln pi

1

2 i

j

ijln piln pj

With a limited number of product groups and asufficiently large number of time-series observa-tions the flexibility implied by the AIDS modeldoes not impose too many demands on the data Inthe present application however where the num-ber of observations is limited the AIDS model isnot estimable in this general form To reduce thenumber of parameters that need to be estimatedwe impose two sets of restrictions

The first set of restrictions is implied by thetheory of utility maximization Specificallythese restrictions are

Adding-up yenk k 1 yenk k 0 yenk kj 0 j

Homogeneity yenk jk 0 j Symmetry ij 1frasl2 [ij ji] ji This last

restriction by itself reduces the number of parameters to [N(N 1)]2

The second set of restrictions we impose aimsat further reducing the number of parameters tobe estimated by exploiting our knowledge of thisparticular market Specifically for each productgroup i we allow one ij parameter for all productgroups j that have different molecules from prod-uct group i and are produced by foreign firms andone ij for all product groups j with differentmolecules produced by domestic firms We donrsquotimpose any restrictions on the ij parameter whenproduct group j has the same molecule as productgroup i (By construction product groups i and jcontain products produced by firms with differentdomesticforeign status)

To better illustrate the nature of the restric-tions we impose on the patterns of substitutionacross products some additional notation isneeded Let d(i j) be an indicator of the degreeof similarity (or difference) between productgroup i and product group j along the dimen-sions we are able to observe (molecule M anddomesticforeign status DF) For any two prod-uct groups i and j d(i j) can take on one of thefollowing three values17

(5)

di j (1 0) if Mi Mj DFi DFj

(0 1) if Mi Mj DFi DFj

(0 0) if Mi Mj DFi DFj

Let

(6) Diab j di j a b

17 The sequence (1 1) is not possible for two differentproducts in this case the parameter corresponds to theproductrsquos own-price effect that is ii

1488 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The equation at the lower level becomes

(7) i i iiln pi i10ln pj j Di10

j Di

01

i01ln pj

j Di

00

i00ln pj ilnXQ

PQ

Note that

The parameter ii captures a product grouprsquosown-price effect (there will be as many iiparameters as the number of product groups)

The parameter i10 captures the cross-priceeffect of the product group containing prod-ucts with the same molecule but produced byfirms of different nationality

The parameter i01 captures the cross-priceeffects of product groups containing productswith different molecules but produced byfirms with the same nationality

The parameter i00 captures the cross-priceeffects of product groups containing productswith different molecules produced by firms ofdifferent nationality

Before we take the demand equation to thedata we make two modifications The first oneis to let the product-specific effects i vary byregion r The resulting product-specific regionaleffects ir have two interpretations first theycontrol for the ldquoqualityrdquo of each drug withquality differences being allowed to vary acrossregions second they proxy for demographicsand other demand shifters which vary by re-gion and may affect the demand of each prod-uct group differently18 Note that by includingproduct-specific regional effects in the demandspecification we estimate the price parametersbased on the within-product-group variation ofprices in each region In an earlier version of thepaper we also estimated the demand systemwithout regional dummies and obtained similarresults

The second modification that allows us to gofrom a deterministic to a stochastic specificationof the demand equation is to include an additiveerror term in (7) The latter takes into accountthe fact that (7) is not expected to fit the dataexactly The error term irt accounts for mea-surement error (due to the fact that the productgroup prices pjrt we employ in the estimationare not exact price indices but approximationsthereof) and (potentially region-specific) de-mand shocks that may affect the demand for aproduct in a particular period Examples of suchshocks include an advertising campaign for aparticular product which temporarily increasesthe demand for this product or the outbreak ofa (potentially region-specific) epidemic whichcalls for the use of a particular drug We discussthe interpretation and properties of this errorterm in more detail in the next section

The final form of the equation we estimate atthe lower level becomes (with subscript t de-noting month and subscript r denoting region)

(8) irt i ir iiln pirt i10ln pjrt j Di10

j Di

01

i01ln pjrt

j Di

00

i00ln pjrt

ilnXQrt

PQrt irt

The analysis so far has conditioned on the ex-penditure allocated to the quinolone subseg-ment XQ The upper level of the estimationconsiders the problem of allocating total expen-diture across the different systemic antibioticssubsegments one of which is quinolones Theupper-level demand function is given by

(9) G G H

GHln PH GlnX

P

where all variables denoted by capital letters aredefined as before but now refer to subsegments(G H ) rather than product groups within asubsegment and the total expenditures on sys-temic antibiotics X are deflated by the Stone priceindex log P yenH Hlog PH When estimating thesystem above we impose all the restrictions

18 Given the short time span of our sample typicaldemand shifters such as age distribution income distribu-tion and education hardly change over our sample periodSuch shifters are therefore absorbed by the region-specificproduct fixed effects ir

1489VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

implied by utility maximization as we do withthe estimation of the lower-level AIDS We donot however impose any additional restrictionson the substitution patterns at this stage so thatthe cross-price effects across segments remainrelatively unconstrained

Estimation of the higher-level AIDS allowsus to obtain the unconditional own- and cross-price elasticities that are used in the formulationof the supply problem and welfare analysisThese are given by the formula

(10) ij ijXQXQ

ln qi

ln XQ

ln XQ

ln PQ

ln PQ

ln pj

with the conditional cross price elasticitiesgiven by

ijXQXQ ij ij jlnXQ

PQ

i

As in the lower stage we include subsegment-specific regional dummies in the specificationof the upper-stage demand system so that thefinal form of the estimating equation at theupper stage becomes

(11) Grt G Gr H

GHln PHrt

GlnXrt

Prt Grt

In sum the demand system we take to thedata is represented by equations (8) and (11)and the associated parameter restrictions im-plied by economic theory

B Modelling the Supply Side of the Market

Counterfactual simulations concerning theeffects of domestic product withdrawal requireknowledge of the marginal costs of pharmaceu-tical firms operating in the Indian market Theseare unobservable The usual approach in theNew Empirical Industrial Organization litera-ture has been to exploit the firm equilibriumconditions to infer marginal cost For exampleit is often assumed that the marginal cost ci isconstant and that the industry is an oligopoly

engaging in Bertrand competition with differen-tiated products Assuming that firms myopicallymaximize profits each period one can then de-rive the firmsrsquo first-order conditions that corre-spond to the assumptions above about costsmarket structure and firm behavior

We deviate from this procedure as the presenceof price regulation renders the assumption of un-constrained period-by-period maximization un-tenable Ideally one would like to explicitlyincorporate the price and other administrative con-trols into the firmrsquos optimization problem andderive the first-order conditions under the assump-tion of constrained maximization The complexityof the price regulation however makes this ap-proach infeasible Therefore we adopt an alterna-tive approach that does not rely on modelling theprice-setting process but is instead based on de-riving upper and lower bounds for marginal costsand markups

In particular an upper bound for marginalcosts and a lower bound for the markups (zero)can be derived under the assumption of perfectcompetition While this assumption is clearlyunrealistic in the pharmaceuticals market it isuseful in providing an upper bound for costs ci

Uwhich will be given by

(12) ciU pi

On the other hand a lower bound for marginalcosts ci

L (and upper bound for markups) can bederived by assuming that there is perfect collu-sion within each product group i (where i refershere to a moleculedomestic-foreign combina-tion)19 and ignoring price controls so thatprices are determined by the first-order condi-tion of the jointly profit-maximizing firmswithin product group i Solving this first-ordercondition for the (lower bound of) marginal costthen gives

(13) ciL pi 1

1

iipipj

19 The reason that this assumption will lead to an under-statement of marginal costs (and hence overstatement ofmarket power) is that it assumes away competition amongfirms within each drug group

1490 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

where ii(pipj) the own-price elasticity of de-

mand for product i will depend on the productrsquosown-price pi and all other productsrsquo prices pjwith i j

Once we have obtained the demand elastic-ities through estimation of the demand sys-tem we can calculate the upper and lowerbounds for marginal costs and correspondingmarkups according to (12) or (13) These willthen be employed in the counterfactual sim-ulations While our counterfactuals use boththe lower and upper bounds for costs most ofour discussion will be based on using thelower bounds for costs since this is the moreinteresting case in the policy analysis it givesus the largest possible profits for pharmaceuticalfirms and hence corresponds to the worst pos-sible scenario facing Indian firms and the bestpossible scenario facing multinationals underTRIPS

C Identification Assumptions and EstimationApproach

The discussion of the demand system has sofar abstracted from the issue of price endogene-ity The usual premise in the industrial organi-zation literature is that correlation of prices withthe error term in the demand equation arises byvirtue of the first-order conditions of profit-maximizing firms20 As we discuss in this sec-tion this source of simultaneity bias is unlikelyto be of major concern in the present contextbecause of the existence of price regulation Ourprimary concern is instead with the simultane-ity bias that is implied by the particular wayprices are constructed

To understand the sources of the simultaneitybias note first that the price of product group piwe employ in the demand estimation should bethought of as a proxy for an exact price indexthat we do not observe As such pi contains bydefinition measurement error and it will becorrelated with the error term of the demandequation Specifically the demand equation weare interested in estimating can be written insimplified form (suppressing the subscripts rand t and ignoring parameter restrictions forconvenience) as

i i j

ijln pjT ilnXQ

PQ i

where pjT denotes the true (exact) price index for

product group i and i captures unobservedvariables that may affect demand in a particularperiod (eg an advertising campaign or an ep-idemic) The price index pj

T depends on theSKU prices pk Note that in the presence ofSKU entry and exit into the sample the priceindex pj

T will vary over time even if the pricesof the individual SKUs remain stable as theproducts (SKUs) that comprise each productgroup change each period

Ideally we would like to compute the exactprice index pj

T for each group and employ theseindices in the demand estimation Under theassumption of predetermined SKU prices eachperiod the exact price indices would be uncor-related with the error term and the demandparameters could be estimated by simple OLSUnfortunately this strategy is not feasible in thecurrent context

To derive an exact price index for each prod-uct group it is necessary to model consumerrsquoschoice among SKUs conditional on the choiceof the group To this end it is necessary toexplicitly introduce a third stage in the demandspecification and estimate the parameters asso-ciated with the choice at that stage Howeverany specification based on three-stage budget-ing (that would require directly specifying ademand function associated with the SKUchoice at the third stage and imposing additiveseparability at the higher stages) would be in-feasible to estimate for three main reasonsFirst there is a large number of SKUs withinsome groups so that the demand parameters can-not be estimated even if one imposes parameterrestrictions second the SKU prices exhibit littlevariation over time lastly the choice sets at thislowest stage vary over time rendering estimationof the price parameters associated with SKUs thatare not available in all periods infeasible Alterna-tively one could abandon the multistage budget-ing altogether and adopt a discrete choiceapproach which is inherently better suited to deal-ing with varying choice sets and limited pricevariation A discrete choice approach would beparticularly appealing for modeling the SKUchoice conditional on product group choice giventhat within each group all SKUs contain the same

20 See Steven Berry et al (1995) or Aviv Nevo (2001)for a related discussion

1491VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

molecule and are hence comparable As notedearlier however a discrete choice approach at thehigher stages of the demand estimation requiresconverting revenue to physical shares which ischallenging in the case of systemic drugs such asantibiotics for which the appropriate dosage is notwell defined Finally a combination of a multi-stage budgeting approach for the higher stages anda discrete choice approach for the lowest stagewould be feasible to estimate but inconsistentwith a model of utility maximization For thesereasons we have no alternative but to adopt anapproximation for measuring the price index foreach product group

Let us denote this approximation by pjA and let

13j denote the proportional ldquoapproximationrdquo errorassociated with measuring the true price index forgroup j so that ln pj

A ln pjT ln 13j The

estimating demand equation can then be written as

i i j

ijln pjA ilnXQ

PQ

j

ijln 13j i

or

i i j

ijln pjA ilnXQ

PQ i

where the new error term i yenj ijln 13j iis comprised of two components the first onereflects ldquomeasurementrdquo error due to the fact thatthe product group prices we employ are approx-imations and not exact price indices and thesecond one captures conventional demandshocks The demand equation as written abovecorresponds to the equation we take to the data

The particular proxy we use for the exact priceindex for each product group j is the revenue-share-weighted average of the prices of multipleSKUs available within this group that is

pjA

k j

k pk

The notation above illustrates the two source ofprice variation variation in the SKU prices pkand variation in the weights k

We are concerned with potential correlation ofthe so-constructed product group prices with bothcomponents of the product group demand error

term The correlation between pjA and the first

component yenj ijln 13j is independent of theparticular way we construct pj

A and inherent in thefact that we measure the exact price index witherror The potential correlation between pj

A and thesecond component of the product group demanderror the shock i arises however because of thespecific way we construct the proxy pj

A and inparticular because of the presence of the reve-nue share weights k in pj

A since the SKUrevenue share weights k will generally dependon the product group expenditure they arelikely to be correlated with the product groupdemand shock i

To address the simultaneity bias we use in-strumental variables To this end we need vari-ables that are correlated with the proxies pj

A butuncorrelated with the error term i We use thenumber of SKUs within each product group j asan instrument as it does justice to the idea thatvariation in the product group price index stemsin part from variation in the set of SKUs that areavailable each period It is clearly correlatedwith the average group prices pj

A and plausiblyuncorrelated with the demand shock i Theassumption that the number of SKUs is uncor-related with the conventional demand error willhowever be violated if the introduction of anew SKU changes the perceived quality of adrug21 or if it is accompanied by promotionalactivities which will be reflected in i Our hopeis that this is not too often the case Assumingthat the SKU number for each group j is alsouncorrelated with the weighted sum of the mea-surement errors of all product groups yenj ijln 13j itcan be used as an instrument for the product groupprice pj

A22 If one accepts the premise that prices atthe SKU level are exogenous then SKU pricescan also be used as instruments The final list ofinstruments we use hence includes the number ofSKUs in each group the prices of the five largestSKUs for each group and all other exogenousvariables in the demand estimation such as re-gionalproduct dummy interactions and the upper-

21 Note that we already proxy for the quality of each prod-uct group through the product fixed effects i In the actualestimation we actually let these fixed effects be region-specific(ir) so that we allow quality to vary across regions

22 If the measurement error increases in the number ofincluded SKUs this assumption will be violated Unfortu-nately we have no way of checking the validity of thisassumption

1492 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

level total expenditure on antibiotics Regressionsof group prices on the instruments above yieldhigh R-squares with most regressors highly sig-nificant indicating that our instruments are highlycorrelated with prices23

Our sample includes four molecules cipro-floxacin norfloxacin ofloxacin and spar-floxacin Except for sparfloxacin all othermolecules are produced by both foreign anddomestic firms24 So we have seven products(domestic ciprofloxacin foreign ciprofloxa-cin etc) with 96 observations (two years ofmonthly data four geographical regions foreach period) for each product The parametersin the lower-level AIDS demand system asdefined in equation (8) are the product fixedeffects i the productndashregional dummy inter-actions ir the own revenue-share price elas-ticities ii the cross revenue-share priceelasticities i10 i01 i00 and the revenue-share expenditure elasticities i In estimatingthe parameters we first regress prices on allinstrumental variables and then plug the pre-dicted values for prices in the constrainedleast-square regression (For a detailed expla-nation of the constraints see the previoussection)

Given that we impose many cross-equationconstraints and employ instrumental variablesin the estimation it is difficult to derive stan-dard errors for the parameter estimates ana-lytically Our error term interpretation in (8)implies that the error terms for each productgroup are likely to be correlated across re-gions for example if a national advertisingcampaign increases the demand for a rela-tively expensive presentation in one regionsimultaneously increasing the aggregate de-mand and price index for the correspondingproduct group in that region it is likely that thesame effect will be observed in other regions Inprinciple we could exploit the cross-regionalcorrelation in the error terms to estimate thedemand system using seemingly unrelated re-

gression (SUR) This however would requireus to derive the variance-covariance matrix ofthe parameters analytically which as notedabove is cumbersome in the current contextInstead we use the bootstrap method choosingto remain agnostic about the structure of thevariance-covariance matrix The potential dis-advantage of this approach over SUR is a loss inefficiency but as will become apparent in theresults section the model parameters are fairlyprecisely estimated To maintain the marketstructure we randomly sample the periods (withreplacement) and use the same periods for allproducts Regarding the optimal number ofbootstrap repetitions ideally one would followthe three-step method proposed by Donald W KAndrews et al (2000) Empirical evidence sug-gests however that one rarely needs more than200 replications to estimate the standard er-rors25 To be safe we generate 300 bootstrapsamples (with replacement) based on the origi-nal data and estimate the standard errors usingthe standard errors of the bootstrap sampleestimates

The estimation of the top-level AIDS systemis similar The constraints imposed on this top-level demand system are adding-up homogene-ity and symmetry as defined on page 12Again bootstrapping is used to obtain the stan-dard errors of the parameter estimates

D The Counterfactual Scenarios

In assessing the effects of patent enforce-ment we start by focusing on the most extremecase in which compulsory licensing is not anoption and foreign firms are not subject to pricecontrols We use the results from the analysis ofthis case as a benchmark In reality the outcomeof the WTO negotiations is more likely to in-volve some constraints on the monopoly powerof foreign firms selling patented products indeveloping countries such as price caps orcompulsory licensing Our framework can eas-ily accommodate these cases as will becomeapparent in the next subsection

We now focus on the effects of potentialpatent enforcement in the quinolone segmentWe consider several scenarios which vary inthe number and size of domestic products that

23 Specifically the R-squares from the first-stage regres-sions of product group prices on instruments range from057 to 095 except for domestic ciprofloxacin for whichthe R-square is 017

24 Sparfloxacin is actually offered by one foreign subsid-iary in India (see Table 2) However its revenue share isminiscule We therefore treat Sparfloxacin as being pro-duced by domestic firms only 25 See Bradley Efron and Robert J Tibshirani (1993)

1493VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

will be removed from the market In particularwe consider the following five scenarios26

Withdrawal of one large domestic productgroup only domestic ciprofloxacin

Withdrawal of one relatively small domesticproduct group only domestic ofloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand norfloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand sparfloxacin

Withdrawal of all four domestic quinoloneproduct groups

As the list above suggests we proceed fromanalyzing the effects of single-product with-drawal to the analysis of eliminating the en-tire domestic segment This approach wasmotivated by early empirical results that in-dicated that the existence and extent of com-petition from domestic firms has a significantbearing on the predicted effects of patent en-forcement that is our predictions regardingprices and welfare vary substantially depend-ing on how many domestic products are af-fected by patent enforcement and on whethersome domestic competition will remain presentafter TRIPS In addition the size of the affecteddomestic groups is relevant for the welfare pre-dictions accordingly we examine both scenar-ios in which a large domestic product groupsuch as domestic ciprofloxacin is eliminatedfrom the market and scenarios in which the

eliminated domestic group is relatively small(eg domestic ofloxacin) In all scenarios wemaintain the assumption that the set of prod-ucts offered by the remaining firms in themarket does not change in response to patentenforcement

E Computation of Virtual Prices and NewEquilibrium Prices

The first step in the counterfactual analysisis to derive the new equilibrium prices underpatent enforcement27 In this context thereare two sets of prices that are relevant Thefirst set consists of the virtual prices of those(domestic) products that will not be availableonce TRIPS is put in effect To calculate thesevirtual prices we set the expenditures of therelevant products equal to zero The secondset of prices consists of the prices of thoseproducts that remain in the market In deriv-ing these prices we start by assuming profitmaximization without any form of price reg-ulation the firms remaining in the marketreoptimize in response to the policy changeand set new prices taking the prices of allother firms as given28 Of course at the equi-librium all prices change in response to thefact that some domestic products are nolonger present The new equilibrium pricesfor products that remain in the market are thuscomputed by utilizing the first-order condi-tions of profit-maximizing firms into whichthe virtual prices of the eliminated productsare substituted Hence to compute the newequilibrium prices we solve an equation sys-tem of the following form

26 As Table 2 indicates most of the patents for the drugsin the quinolone segment have expired by now so that noneof the scenarios described in this section is going to mate-rialize in practice The reason we consider these alternativescenarios in our counterfactual simulations is to get a senseof how the presence and extent of domestic competitionaffects welfare calculations Such calculations may becomerelevant in the future in other therapeutic classes in whichpatents have not expired yet and where domestic firmscompete with foreign patent holders While the particularexit scenarios will depend on the exact year of patent expiryof drugs in each therapeutic class and the specific way inwhich patents will be enforced (eg whether certain do-mestic drugs will be grandfathered in) the general conclu-sion that emerges from our calculations is that the consumerwelfare loss is substantially smaller in cases where somedomestic competition remains present in the market Hencethe particular way in which patent enforcement is imple-mented is essential for assessing its welfare impact

27 Of course until product patents are in fact introducedthese prices will not be observable Note also that we areassuming here that the range of available products will notchange with the introduction of patents

28 As mentioned above this first set of calculations ab-stracts from the existence of remaining price controls orother government regulations that would impose constraintson the firmsrsquo profit maximization problem Accordingly theresulting numbers should be interpreted as a benchmark ofwhat would happen if markets were completely unregulatedTo examine what the welfare effects of TRIPS would be inthe more realistic scenario of price regulation we subse-quently consider a scenario in which regulation keeps theprices of the products provided by patent-holders at theirpre-TRIPS level For a more detailed discussion see theresults section

1494 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

For products i that are withdrawn from themarket

(14) 0 i ir iiln pirV i10ln pjr j Di

10

j Di

01

i01ln pjr

j Di

00

i00ln pjr

ilnXQr

PQr

For products k that remain in the market

(15) pkr ckr 1 1

kk pkrpjr pirV 1

In the equations above pirV denotes the virtual

prices of the products that are removed fromthe market while pjr denotes the updatedprices of all other products Note that whensolving for the virtual prices we account forthe fact that both the price index for quino-lones PQr and the expenditure allocated tothis subsegment XQr need to be updated toreflect the fact that as a result of the pricechanges there may be substitution away fromthis subsegment To obtain the new quinoloneexpenditure XQr and the new price index PQrwe use the estimates and formulas for thehigher-level AIDS system In equation (15)ck refers to the marginal cost for product k thatwe have obtained from the previous estimationstage As mentioned in the previous subsectionwe conducted the simulations using both theupper and lower bounds for marginal cost ck

u

and ckL respectively The term kk(pkr

pjrpirV) re-

fers to the unconditional own-price elasticity forproduct k which is a function of the eliminatedproductsrsquo virtual prices and the remaining prod-uctsrsquo new equilibrium prices We conduct thecounterfactual simulations and welfare analysisat the regional level because the presence ofregion-specific product effects in the demandestimation implies region-specific demand elas-ticities and hence region-specific marginalcosts and markups The presentation and discus-sion of our results focus on national averages ofthe relevant variables that we construct by com-

puting weighted averages across regions usingthe population of each region as a weight

F Welfare Assessment

The simulation of the new equilibrium underpatent protection can provide important insightsinto how consumers and firms will respond tothe removal of domestic products in the market(for example toward which products consumerswill substitute or which prices will increase themost) To get a more precise idea of how peo-plersquos well-being will ultimately be affected byTRIPS we compute as a last step in our anal-ysis the welfare effects of the policy changeSocial welfare is defined as consumer welfareand the sum of domestic firm profits Thechange in domestic profits can easily be calcu-lated by comparing the domestic firm (variable)profits at the pre-TRIPS prices to the profitsthese firms will realize at the new simulatedprices Although foreign firm profits do notcount in domestic welfare calculations we alsocompute the effects of patent enforcement onforeign firm profits to get an idea of how largethe expected benefits of TRIPS are for thesefirms This provides in some sense an indirectway of assessing whether the claims that patentenforcement in countries like India will lead tomore research on developing-country-specificdiseases (such as malaria) have any validity iffor example we find that the effect of patentenforcement on the foreign firm profits realizedin India is small in magnitude it is unlikely thatforeign firms will engage in more developing-country-specific research in response to TRIPSIt is important to note that in all these calcula-tions we work with the lower-bound estimatesfor marginal costs since these give rise to thehighest possible markups Hence our estimateof profit loss for domestic producers most likelyoverstates this loss

On the consumer side we measure changes inconsumer welfare by the compensating variation(CV) defined as the additional expenditure thatconsumers need in order to achieve the sameutility level as before patent enforcement at thenew prices Specifically let P0 denote the pricevector before patent enforcement P the simulatedprice vector post-TRIPS (that we obtained usingthe methods described in the previous subsection)u0 the utility attained by consumers before TRIPS

1495VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

and E(u P) the higher-level expenditure functionThen the compensating variation is given by

(16) CV Eu0 P Eu0 P0

Note that the CV as computed in (16) representsthe combination of three effects

The pure product-variety effect that is theeffect that arises because one or more prod-ucts are not available to consumers anymoreholding the prices of all other remainingproducts and the total expenditure on thequinolone subsegment XQ constant

The expenditure-switching effect that is theeffect arising from substitution away fromquinolones and toward other subsegments ofthe antibiotics market again holding the pricesof all other remaining products constant

The reduced-competition effect that is theeffect that arises because the firms remainingin the market adjust (increase) their prices inresponse to the removal of domestic products

From both an analytical and a policy point ofview it is desirable to assess how large each ofthese effects is Accordingly we decompose thetotal effect on consumer welfare (the CV as givenby equation (16)) using the following procedure

To get the pure product variety effect wecompute virtual prices for the products that areremoved from the market holding the quino-lone expenditure XQ and the prices of all otherproducts fixed Let us call the resulting pricevector P1 Then the pure product variety effectis represented by E(u0 P1) E(u0 P0)

To compute the expenditure-switching effectwe compute another set of virtual prices againholding the prices of all remaining products fixedbut letting quinolone expenditure adjust in re-sponse to the new price index for the quinolonesegment (Given that the prices of the remainingproducts remain fixed the change in the priceindex arises only because of the removal of one ormore domestic products) Note that this scenariomost closely resembles the case in which patentlaws are enforced but strict price regulation keepsthe prices of the products offered in the market attheir pre-TRIP level Let us label the so-computedprice vector P2 The expenditure-switching effectis then E(u0 P2) E(u0 P1)

Finally the reduced competition effect aris-ing from higher prices for the remaining prod-

ucts is computed as the residual change in thecompensating variation once the product-varietyand expenditure-switching effects have been ac-counted for that is E(u0 P) E(u0 P2)where the price vector P is computed accordingto the formulas (14) and (15) to reflect theadjustment of prices to the new regime

To compute the standard errors associatedwith the counterfactual simulations (that is thestandard errors for the counterfactual prices andwelfare estimates) we again use bootstrappingWe first bootstrap the original sample (sales andprices of all the drugs) 300 times Next we esti-mate the AIDS model for each of these samplesand compute the counterfactual equilibrium pricesand welfare losses corresponding to each policyscenario for each of the 300 simulations In thefinal step we compute standard errors29

One limitation of our framework is that itdoes not allow for a heterogeneous response ofconsumers to the policy change Accordinglyour framework is not suited to addressing thequestion of how different groups in the popula-tion will be affected by patent law enforce-ment30 Along the same lines our frameworkdoes not accommodate the possibility of pricediscrimination which might lead to differentresults regarding welfare losses and profit gainsrelative to uniform pricing31 Still we believethat the results of the counterfactual simulationscan provide important insights into the likelyaggregate response to patent enforcement andthe factors that drive this response

III Results

A The Structure of Demand

Tables A2 and A3 in the Appendix display theresults from estimation of the lower- and upper-level AIDS systems respectively For ease of in-terpretation rather than discussing the coefficientestimates we focus our discussion on the implied

29 In cases some of the simulations do not converge wecompute the standard errors using only those simulationsthat converged

30 Addressing this question would require at a mini-mum micro data on consumer purchases as in Goldberg(1995) To our knowledge such data do not exist for theIndian pharmaceutical market

31 See Ernst R Berndt (1994) for an extensive discussionof uniform pricing

1496 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

unconditional price and expenditure elasticities re-ported in Table 6 Given that the region-specificproduct group effects ir reported in the last fourcolumns of Tables A2 and A3 imply region-spe-cific demand elasticities we report separate elas-ticities for each region in Tables 6A and 6B Asevident from the comparison of these tables theelasticities are very similar across regions so thatwe can focus the remaining discussion on theelasticities of one region only the northern regiondisplayed in Table 6A

The diagonal terms of Table 6A report theown-price elasticities which are in all but onecase negative and highly significant The oneexception is the foreign norfloxacin productgroupmdashwhose share of quinolone sales is 007percentmdashfor which we estimate a negative butinsignificant own-price elasticity For the remain-ing product groups demand appears to be highlyelastic with the estimated elasticities lower than2 in four out of the six cases The magnitude ofthe own-price elasticities matches the features ofthe Indian pharmaceutical market mentioned ear-lier which would suggest that Indian consumersare likely to be quite price-sensitive32 The elas-ticities appear especially large if one takes intoaccount that they refer to product groups (such asdomestic ciprofloxacin) and not individual drugsoffered by particular firms Their relative magni-tudes are also intuitive the drug with the largestmarket share and a relatively high price (domesticciprofloxacin) appears to be one of the least elas-tic In contrast foreign ciprofloxacin is highlyprice elastic this is plausible as ciprofloxacindrugs offered by subsidiaries of multinationalsface the stiffest competition in this market seg-ment from approximately 75 Indian firms offeringthe same molecule

The estimated expenditure elasticities appear inthe last column These are all positive indicatingthat the demand for all product groups is normalThe remaining cells display the estimated cross-price elasticities As one might perhaps expect forproducts within a therapeutic subsegment theseare mostly positive Out of a total of 49 priceelasticities we estimate there are six that do notconform to expectations these are the cross-price elasticities between different foreign prod-

uct groups which are estimated to be negativeand significant33 Fortunately these elasticitieshave negligible impact on the welfare analysisgiven that our counterfactuals focus on theeffect of withdrawing one or more domesticproducts from the market the most relevantelasticities are the ones that capture the responseof various product group shares to a change in theprice of one or more domestic groups these elas-ticities are the cross-price elasticities between var-ious domestic groups and the ones betweendomestic and foreign groups which are plausibleand precisely estimated34

Regarding these elasticities a striking aspectof our estimates is how large positive andsignificant the cross-price elasticities betweendifferent domestic product groups aremdashin factfor norfloxacin and ofloxacin we estimate thatdomestic product groups containing differentmolecules are closer substitutes for one anotherthan product groups that contain the same mol-ecule but are produced by foreign firms Incontrast for ciprofloxacin (the molecule withthe largest revenue share) we estimate a largepositive cross-price elasticity between the do-mestic and foreign versions

The fact that domestic products appear to beclose substitutes for other domestic productsthat contain different molecules truly representsan ldquoempiricalrdquo finding in the sense that we do

32 In developed economies elasticities of this magnitude havetypically been found only for generic drugs (and even then onlyrarely) or among consumers who lack health insurance

33 While these elasticities are clearly counterintuitive theyare not inconsistent with the underlying demand system whichimposes no restrictions on the sign of the cross-price elastici-ties We do not have a good explanation of why these elastic-ities are estimated to be negative A potential explanation isthat the shares of the foreign products are very small giventhis we observe very few consumers switching from one (verysmall) foreign group to another (very small) foreign groupwhen the price of the first foreign group goes up and hence theinference is not very reliable in this case

34 The cross-price elasticities between foreign drug groupscontaining different molecules will also have an effect on thewelfare estimates given that the withdrawal of domestic prod-ucts will generally lead to changes in the prices of all foreignproducts and market shares will be reallocated from someforeign products to other foreign products based on the newprices However this reallocation from ldquoforeign to foreignrdquo istruly second order in our case compared to the reallocationfrom ldquodomestic to domesticrdquo and ldquodomestic to foreignrdquo prod-ucts In scenarios in which we consider patent enforcementaccompanied by strict price regulation the cross-price elastic-ities between foreign product groups are in fact completelyirrelevant as the prices of foreign products are not allowed toincrease in these cases Still our welfare loss estimates remainsubstantial drivenmdashas beforemdashby the loss of product variety

1497VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

not impose it through any of our assumptionsregarding the demand function The questionthat naturally arises then is what might explainthis finding While we cannot formally addressthis question anecdotal accounts in various in-dustry studies suggest that the explanation maylie in the differences between domestic andforeign firms in the structure and coverage ofretail distribution networks

Distribution networks for pharmaceuticals inIndia are typically organized in a hierarchicalfashion Pharmaceutical companies deal mainlywith carrying and forwarding (CampF) agents inmany instances regionally based who each sup-ply a network of stockists (wholesalers) Thesestockists in turn deal with the retail pharma-cists through whom retail sales ultimately oc-cur35 The market share enjoyed by a particularpharmaceutical product therefore depends inpart on the number of retail pharmacists who

stock the product And it is here that thereappears to be a distinction between domesticfirms and multinational subsidiaries In particu-lar the retail reach of domestic firms as agroup tends to be much more comprehensivethan that of multinational subsidiaries (IndianCredit Rating Agency (ICRA) 1999)36

There appear to be two reasons for this Thefirst is that many of the larger Indian firmsbecause they have a much larger portfolio ofproducts over which to spread the associatedfixed costs typically have more extensive net-works of medical representatives The second issimply that there are many more domestic firms(and products) on the market At the retail levelthis would imply that local pharmacists mightbe more likely to stock domestic products con-taining two different molecules say ciprofloxa-cin and norfloxacin than they would domesticand foreign versions of the same molecule Tothe extent that patients (or their doctors) arewilling to substitute across molecules in order tosave on transport or search costs (eg going toanother pharmacy to check whether a particular

35 There are estimated to be some 300000 retail pharma-cists in India On average stockists deal with about 75 retailers(ICRA 1999) There are naturally variations in this structureand a host of specific exclusive dealing and other arrangementsexists in practice Pharmaceutical firms also maintain networksof medical representatives whose main function is to marketthe companyrsquos products to doctors who do the actual prescrib-ing of drugs In some instances firms do sell directly to thedoctors who then become the ldquoretailerrdquo as far as patients areconcerned but these are relatively rare

36 These differences were also highlighted in conversa-tions one of the authors had with CEOs and managingdirectors of several pharmaceutical firms as part of a sep-arate study

TABLE 6AmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN THE NORTHERN REGION

Product group

Elasticity with respect to

Prices of foreign productgroups Prices of domestic product groups Overall

quinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 557 013dagger 015 401 011dagger 011dagger 016 137(179) (007) (007) (184) (006) (006) (006) (029)

Foreign norfloxacin 427dagger 045 427dagger 350dagger 602 451 465 220(242) (112) (242) (210) (623) (184) (183) (105)

Foreign ofloxacin 011 010dagger 138 009 009dagger 023 011 116(005) (005) (031) (027) (005) (028) (004) (017)

Domestic ciprofloxacin 018 001 001 168 008 008 010 117(008) (000) (001) (023) (002) (002) (002) (003)

Domestic norfloxacin 004 003 004 058 223 042 040 073(001) (003) (001) (017) (011) (004) (003) (009)

Domestic ofloxacin 005 005 011 077 074 342 074 089(002) (002) (013) (028) (008) (025) (008) (021)

Domestic sparfloxacin 007 004 007 115 063 063 288 028(002) (001) (002) (015) (006) (006) (017) (012)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent significance level and dagger (dagger) denotes significance at the 10-percent level

1498 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE 6BmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN OTHER REGIONS

Demand patterns in the eastern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 594 014dagger 016 431 013 011 017 140(195) (008) (008) (201) (006) (007) (007) (031)

Foreign norfloxacin 329dagger 058 329dagger 264dagger 460 346 359 192(180) (083) (179) (153) (451) (138) (139) (082)

Foreign ofloxacin 012 010dagger 140 010 010 024 013 117(006) (006) (034) (031) (005) (030) (005) (019)

Domestic ciprofloxacin 018 001 001 172 009 008 011 117(008) (000) (002) (027) (002) (003) (002) (003)

Domestic norfloxacin 004 004 004 068 242 049 046 070(002) (004) (002) (020) (013) (005) (004) (010)

Domestic ofloxacin 004dagger 004 009 061 060 295 060 092(002) (001) (010) (028) (007) (022) (007) (017)

Domestic sparfloxacin 005 003 005 092 048 051 251 043(002) (001) (001) (017) (005) (006) (016) (010)

Demand patterns in the western region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 427 009dagger 011 286 008 006 011 126(129) (005) (005) (133) (004) (005) (004) (021)

Foreign norfloxacin 714dagger 008 710dagger 594dagger 995 742 770 299dagger

(369) (183) (368) (328) (1015) (297) (296) (178)Foreign ofloxacin 013 012dagger 145 009 012 026 013 119

(007) (006) (037) (029) (005) (033) (005) (021)Domestic ciprofloxacin 019 001 000 174 009 008 011 118

(008) (000) (001) (024) (002) (003) (002) (004)Domestic norfloxacin 004 004 004 067 243 050 047 069

(002) (004) (001) (018) (014) (005) (004) (010)Domestic ofloxacin 003 003 007 048 048 257 048 093

(002) (001) (008) (022) (005) (015) (004) (014)Domestic sparfloxacin 006 003 005 083 046 049 241 046

(002) (001) (001) (014) (004) (005) (012) (009)

Demand patterns in the southern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 560 013dagger 015 402 011 010 017 137(182) (007) (007) (187) (006) (006) (007) (029)

Foreign norfloxacin 533dagger 031 531dagger 431dagger 748 559 584 249dagger

(301) (134) (300) (245) (734) (219) (219) (128)Foreign ofloxacin 014 012dagger 148 011 012 029 016 120

(007) (007) (039) (033) (006) (035) (006) (022)Domestic ciprofloxacin 017 001 000 169 008 007 011 116

(008) (000) (001) (025) (002) (003) (002) (003)Domestic norfloxacin 004 003 003 059 225 043 040 073

(001) (003) (001) (019) (012) (004) (004) (009)Domestic ofloxacin 004dagger 004 009 062 060 296 060 091

(002) (001) (010) (026) (006) (019) (007) (017)Domestic sparfloxacin 008 005 008 138 073 075 323 016

(002) (001) (002) (018) (007) (008) (021) (014)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at 5-percent level anddagger (dagger) denotes significance at the 10-percent level

1499VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

foreign product is in stock) in aggregate datawe would expect to find precisely the substitu-tion patterns that we report in Table 6

Whether the particular explanation we provideabove is the correct one the high degree of sub-stitutability between domestic product groupsturns out to have important implications for thewelfare calculations We discuss these in moredetail below when we present the results of thecounterfactual welfare analysis Another elasticitywith important implications for the counterfactu-als is the price elasticity for the quinolone subseg-ment as a whole which indicates how likelyconsumers are to switch to other antibioticsgroups when faced with a price increase for quin-olones This elasticity is computed on the basis ofthe results in Table A3 and it is at 111 (stan-dard error 024) this is large in magnitudebutmdashas expectedmdashsmaller in absolute value thanthe own-price elasticities of the product groupswithin the quinolone subsegment

The results in Tables 6A and 6B are based onour preferred specification discussed in SectionII In Tables A4 to A6 in the Appendix weexperiment with some alternative specificationsTables A4(a)ndashA4(c) correspond to a specificationthat includes in addition to product-group-specificregional fixed effects product-group-specific (andfor the upper level antibiotics-segment-specific)seasonal effects We distinguish among three sea-sonsmdashthe summer monsoon and wintermdashandreport the unconditional demand elasticities for

the northern region for each of these seasons Asevident from the tables in the Appendix our elas-ticity estimates are robust to the inclusion of sea-sonal effects The demand elasticities in Table A5are based on estimation of the demand system byOLS Compared to the elasticities obtained by IVthe OLS elasticities are smaller in absolute valueimplying that welfare calculations based on theOLS estimates would produce larger welfare lossestimates Nevertheless some of the patterns re-garding the cross-price elasticities discussed ear-lier are also evident in the OLS results inparticular the cross-price elasticities between dif-ferent domestic product groups are all positivelarge and significant and in most instances largerthan the cross-price elasticities between drugs thatcontain the same molecule but are produced byfirms of different domesticforeign status Theclose substitutability of domestic products indi-cated by both the OLS and IV estimates seems tobe one of the most robust findings of the paper

B Cost and Markup Estimates

Table 7 displays the marginal costs markupsand profits implied by the price elasticity esti-mates of Tables 6A and 6B for each of the sevenproduct groups Given that our regional effectsimply different price elasticities for each regionour marginal cost and markup estimates alsodiffer by region Given however that based on

TABLE 7mdashUPPER AND LOWER BOUNDS FOR MARGINAL COST MARKUP AND ANNUAL PROFIT BY PRODUCT GROUPS WITHIN

THE QUINOLONE SUBSEGMENT

Product groupLower boundfor MC (Rs)

Upper boundfor markup

Upper boundfor profit(Rs mill)

Upper boundfor MC (Rs)

Lower boundfor markup

Lower boundfor profit

(Rs)

Foreign ciprofloxacin 83 19 269 103 0 00(123) (012) (1655)

Foreign norfloxacin NA NA NA 53 0 00Foreign ofloxacin 323 70 1061 1085 0 00

(2316) (021) (3185)Domestic ciprofloxacin 47 59 17019 112 0 00

(114) (010) (29858)Domestic norfloxacin 52 43 2807 90 0 00

(020) (002) (1532)Domestic ofloxacin 587 34 1612 901 0 00

(218) (002) (1280)Domestic sparfloxacin 495 37 1985 788 0 00

(157) (002) (1100)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level Estimated lower bound forforeign norfloxacinrsquos marginal cost is negative since the estimated price elasticity is less than one in absolute value

1500 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Tables 6A and 6B the price elasticities do notseem to differ substantially across regions wereport for ease of exposition only the nationalaverages for marginal costs and markups inTable 7

Table 7 has two parts In the left part (firstthree columns) we report the lower bound formarginal cost and the corresponding upperbound for markup and upper bound for totalannual profit for each product group Thesenumbers are based on the price elasticities weobtained from estimating the two-level AIDSdemand system Since we do not have a reliableestimate for the price elasticity of foreign nor-floxacin (the point estimate is negative but lessthan one in absolute value and insignificant)we cannot compute the lower bound for mar-ginal cost in this case It is important to note thatthese estimates do not reflect either the actualmarginal cost or the actual markup for thesedrugs both because the existence of price reg-ulation implies that the unconstrained first-orderconditions are unlikely to hold each period andbecause our aggregation across firms of thesame domesticforeign status supplying thesame molecule makes the interpretation of theseestimates problematic In particular the fact thatwe ignore competition among firms within eachproduct group implies that our estimates willtend to overstate market power These numberswill prove useful however in the counterfac-tual simulations as they can give us a sense ofhow large the maximum profit gains for multi-national firms and the maximum profit lossesfor domestic firms are likely to be under patentenforcement The right part of the table (lastthree columns) reports the upper bound for mar-ginal cost which we obtain by simply taking themarginal cost to be equal to the observed priceWe do not report standard errors in this casesince the numbers are based on actual data Themarkups corresponding to these marginal costupper bounds are of course zero We conductthe counterfactual simulations using both thelower and upper bounds for marginal costs

The (upper bounds for) markups on the leftside of the table are generally plausible Thedomestic ciprofloxacin product group that dom-inates the quinolone subsegment and for whichwe documented high prices a high marketshare and a relatively low elasticity of demandenjoys one of the highest markup upper bounds(60 percent) and accounts for nearly 70 percent

of all profits derived within the subsegmentForeign ciprofloxacin on the other hand whichfaces the stiffest competition from domesticfirms and for which we estimated a highly elas-tic demand has the lowest markup upper bound(19 percent)

C Counterfactual Estimates of the Impact onPrices and Welfare

With estimates of the key demand and costparameters in hand we turn to the counterfac-tuals We consider the five separate scenarioslisted in the previous section All of the scenar-ios involve the withdrawal of one or more of thedomestic product groups from the market Ta-ble 8 displays our estimates of the consumerwelfare losses that result under the differentscenarios The losses are expressed in billion ofrupees per year All numbers presented in Ta-ble 8 and subsequent tables are based on usingthe lower bounds for marginal cost and upperbounds for markup in the simulations As dis-cussed earlier these numbers are the more in-teresting to work with since they give us anupper bound for the changes in the profits ofdomestic and foreign firms that would resultfrom patent enforcement In the Appendix wealso present results based on using the upperbounds for marginal costs in the simulations inwhich case the pre-TRIPS profits of domesticand foreign firms are zero In all cases marginalcosts are assumed to be constant in outputWhile naturally the profit implications differdepending on whether one uses the upper orlower bounds for marginal costs (firm profits arezero if one assumes the upper bound of mar-ginal cost in which case price equals marginalcost) the estimated consumer welfare losses aresimilar in the two cases We discuss these re-sults in more detail at the end of this section

The first column presents our estimates of theconsumer welfare losses attributable to the pureloss of product variety effect where we fix theprices of all remaining products as well as theoverall expenditure on quinolones while with-drawing one or more of the domestic productgroups Note that in our initial specification ofthe demand system had we not allowed for thepossibility that consumers might differentiatebetween domestic and foreign products evenwhen they contain the same molecule this

1501VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

particular component of the loss of consumerwelfare would not have arisen

The estimates reported in the second columnincorporate the expenditure-switching effect ontop of the loss of product variety Here basedupon the price elasticity estimates from thehigher-level AIDS system we adjust (down-ward) the expenditures allocated to the quino-lone subsegment as the composite price ofquinolones effectively increases as a consequenceof the higher virtual prices of the domestic productgroups that are withdrawn from the market Be-cause the estimates in this column are generatedassuming that the prices of the products that re-main in the market are not adjusted upward theyprovide a sense of what consumer losses would beif the introduction of product patents were coupledwith strict price regulation aimed at maintainingprices at prepatent levels Alternatively they canbe thought of as the relevant welfare numbers ifintense competition among firms within the re-maining product groups kept the prices of theproducts that were still offered in the market closeto the firmsrsquo marginal costs

The last column of Table 8 displays the esti-mated consumer welfare losses when bothcross-segment expenditure-switching and with-in-segment upward price adjustments are takeninto account

If we compare the results across the firstsecond and third columns all the counterfac-tual scenarios produce qualitatively similar pat-terns patterns that are consistent with what we

would expect Starting from the initial loss ofwelfare attributable to the loss of product vari-ety the option of switching expenditures outof the quinolone subsegment to other subseg-ments mitigates some of the initial welfareloss But if we then incorporate the upwardprice adjustments that result in response to thereduced competition the welfare losses aremagnified

Of particular interest from a policy perspec-tive are the relative magnitudes of these threeeffects which are similar under all the counter-factual scenarios though absolute levels varyconsiderably First despite the fact that the de-mand for quinolones is quite sensitive to thecomposite price of quinolonesmdashthe upper levelprice elasticity is 111mdashthe cross-subsegmentexpenditure-switching effects are in all thecases small (in absolute value terms) relative tothe other two effects For instance under thescenario where all the domestic quinolone prod-uct groups are withdrawn from the market theoverall consumer welfare loss of Rs 1781 bil-lion per year can be decomposed into an initialloss of Rs 1176 billion (66 percent) attribut-able to the loss of product variety a slightreduction in this initial loss of Rs 041 billion(2 percent) from Rs 1176 billion to Rs1135 billion because of expenditure switchingand a subsequent additional loss of Rs 646billion (36 percent) from Rs 1135 billion toRs 1781 billion because of the reduced com-petition and consequent price increases

TABLE 8mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure loss ofvariety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 732(087) (089) (146)

Only ofloxacin 008 008 023(008) (008) (010)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1253(177) (180) (415)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1058(142) (145) (331)

All four domestic quinolones products 1176dagger 1135dagger 1781(643) (634) (1270)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent significance level and dagger (dagger)denotes significance at the 10-percent level

1502 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 4: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

contain the same patentable molecule we allowfor an additional channel through which theintroduction of product patents and the conse-quent withdrawal of domestic products may ad-versely affect consumers that is through theloss of product variety In contrast previousstudies on developing countries assume thatconsumers are indifferent between foreign anddomestic products that contain the same mole-cule What this implies is that any adverse wel-fare effects are realized only through increasedprices The difference is most evident if weconsider a scenario under which domestic prod-ucts are forced to withdraw from the marketbecause of the introduction of product patentsbut strict price regulations maintain prices atprepatent levels In our approach consumerswould still experience a welfare loss whereas inthe framework adopted in earlier studies such ascenario would entail no loss of welfare

Empirically the component of the consumerwelfare loss attributable to the reduction of va-riety from the withdrawal of domestic productsturns out to be significant We interpret thiscomponent as capturing primarily an ldquoease ofaccessrdquo effect due to differences in the market-ing and distribution networks domestic prod-ucts are more readily available to Indianconsumers than products produced by foreignsubsidiaries From a policy perspective thissuggests a possible role for compulsory licens-ing in addition to or in lieu of price regulationsince the latter by itself will not alleviate thewelfare loss due to loss of variety Alterna-tively one could argue thatmdashto the extent thatthe loss we attribute to the reduction of productvariety is due to the fact that the current productportfolios and distribution networks of foreignproducers are limitedmdashit is purely a transitionalphenomenon and should thus not be included inthe welfare calculations This is a controversialpoint we discuss in detail in the results sectionIf foreign firms respond to patent enforcementby investing in distribution networks or by us-ing licensing agreements with domestic firms tomake their products more readily available toIndian consumers the ldquoease of accessrdquo effectwould indeed diminish in importance in thelonger run though of course it could be signif-icant in the first years after patent enforcementWhether these investments will materializehowever is open to debate If TRIPS is accom-panied by price regulation in order to limit price

increases in poor developing countries the in-centives of multinationals to invest in marketingand distribution in these countries may dimin-ish At any rate to take into account the possi-bility that the welfare loss due to the reductionof variety is a temporary phenomenon we alsopresent a more conservative welfare loss esti-mate by subtracting the ldquoproduct varietyrdquo com-ponent from our total loss estimate This givesus a lower bound estimate that is due to priceincreases alone Though it is only about a thirdof our upper bound estimate this lower boundestimate is still very large in absolute termsrepresenting 24 percent of antibiotic sales in2000

A second and perhaps even more importantmethodological difference between this paperand earlier studies is that we allow for andflexibly estimate a range of cross-product-group and cross-molecule substitution effectsIn contrast cross-price effects are ignored inearlier studies To see why cross-price effectsare likely to alter estimated welfare effects sig-nificantly in this context imagine a scenariowhere the introduction of patents leads to mo-nopoly pricing in the market for a particularpatentable molecule If the markets for potentialsubstitutes are imperfectly competitive then theincrease in price in the original patentable mar-ket will lead to corresponding upward priceadjustments in the related markets as producersof substitute products reoptimize in the face ofthe increased demand for their products Themagnitude of any upward adjustments will nat-urally vary with the degree of competition inrelated markets and with the strength of thecross-price effects But as long as the cross-price effects are positive and related marketsare not perfectly competitive the loss of con-sumer surplus because of monopoly pricing inone market will be multiplied through the rippleeffects of upward price adjustments in relatedmarkets

If this were just a theoretical possibility itwould not be of much interest These multipliereffects however turn out to be substantial inour counterfactual scenarios Most strikinglythe estimated loss of consumer welfare from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenar-

1480 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

ios in each of which only one of the domesticproduct groups is withdrawn What this veryclearly indicates is that past studies that haveestimated the aggregate effects of patent protec-tion by adding up the losses estimated sepa-rately in each of a number of patentablemarkets may have substantially underestimatedthe magnitude of the consumer welfare lossesfrom the introduction of pharmaceutical productpatents

In absolute terms we estimate that in theabsence of any price regulation the prices offoreign patented products would rise between100 percent and 400 percent In the more real-istic case of some form of price regulation thatwould keep drug prices fixed at their pre-TRIPSlevel the total annual welfare losses to theIndian economy from the withdrawal of all fourdomestic product groups in the quinolone sub-segment would be on the order of Rs 1370billion or about 50 percent of the sales of theentire systemic antibacterials segment in 2000At the then-prevailing exchange rate this trans-lates into a figure of US$305 million Of thisamount foregone profits of domestic producersconstitute roughly Rs 23 billion or US$50 mil-lion The overwhelming portion of the totalwelfare loss therefore derives from the loss ofconsumer welfare

The welfare loss we estimate represents onlythe static costs of patent enforcement arisingfrom pricing distortions and reduction in prod-uct variety Our approach does not address thepotential dynamic benefits of innovations thatmay result from international property rightsprotection Nevertheless we believe that esti-mating these static costs is important wheneverthere is a radical change in policymdashwhichTRIPS represents for a good part of the devel-oping world Even if there is the potential forlong-term benefits knowledge of the short-runcosts is important for designing an appropriatepolicy response that will potentially mitigate theadverse short-run impact Having said that itis worth noting that according to our esti-mates the total profit gains of patent enforce-ment to foreign producers in the absence ofany price regulation would be only aboutUS$53 million per year With price regulationthat would keep the prices of drugs suppliedby multinational subsidiaries at their pre-TRIPS level the profit gains become onlyUS$196 million per year

The remainder of this paper is organized asfollows In the next section we lay out theessential features of the Indian pharmaceuticalsmarket provide more detail about the segmentsthat we focus on in the empirical analysis andbriefly describe the primary data we use Sec-tion II describes the analytic framework and theeconometric strategy we use to estimate therelevant parameters and construct the counter-factual scenarios We discuss our results in Sec-tion III Section IV concludes

I The Setting and the Data

Between April 1972 when the Indian PatentsAct (1970) became effective and March 2005when Indiarsquos parliament passed the 3rd Amend-ment of the Patents Act India did not recognizeproduct patents for pharmaceuticals The IndianPatents Act which replaced the inherited Brit-ish colonial law regarding intellectual propertyrights specifically excluded pharmaceuticalproduct patents and admitted process patentsonly for a period of seven years In contrast thelatest amendment recognizes patents on endproducts that under the new regime will remainin force for 20 years6

The two stated objectives of the 1970 actwere the development of an indigenous pharma-ceuticals industry and the provision of low-costaccess to medicines for Indian consumers Con-sistent with these objectives and with the broaderleftward tilt in policy a number of other measureswere introducedmdashdrug price controls restrictionson capacity expansion limits on multinational eq-uity sharesmdashthat in the years since have on theone hand kept pharmaceutical prices low and onthe other encouraged the development of the In-dian pharmaceutical industry Many of these reg-ulations and restrictions have been lifted or easedsince the mid-1980s with marked acceleration inthe pace of liberalization during the 1990s

Over the last 20 years the Indian pharmaceu-tical industry has grown rapidly to the pointwhere it is now the worldrsquos largest producer offormulations in terms of volume and one of the

6 Indian companies that are now producing drugs forwhich patent applications were submitted between the sign-ing of the TRIPS agreement in 1995 and January 1 2005will be allowed to continue producing if they pay a royaltyto the patent holder

1481VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

worldrsquos largest producers of bulk drugs7 Thestructure of the industry has also evolved In1970 the industry was dominated by multina-tional subsidiaries by 2001 Indian-owned firmswere not just the leading players in the industrymany had also become major exporters

The data we use in this paper are from theretail pharmaceutical audits of ORG-MARGIndiarsquos premier market research and consultingfirm The audit provides detailed product-levelinformationmdashestimates of monthly retail salesin each of the four geographic zones of Indiaprice dosage form launch date brand namechemical name therapeutic categorizationmdashon

all pharmaceutical products sold in India byabout 300 of the largest firms representingroughly 90 percent of domestic retail sales ofpharmaceuticals The coverage of the audit isextensive reaching a representative panel ofthousands of retail chemists in over 350 citiesand towns The data collected which providethe only real source of disaggregate informationon the Indian pharmaceutical market are usedby both the government of India in formulatingpricing policy and other decisions and the In-dian pharmaceutical industry in determiningpricing and marketing strategies We have in-formation at a monthly frequency for the periodof January 1999 to December 2000 Tables 1 to3 provide a set of descriptive statistics that areessential for understanding the focus of ouranalysis and interpreting our results

As noted earlier the characteristics of de-mand for pharmaceuticals in India are likely todiffer considerably from those in developedeconomies With a share of 23 percent the

7 Bulk drugs are the therapeutically relevant active phar-maceutical ingredients that are combined with a variety ofinactive ingredients to make the formulations ultimatelyconsumed by patients Firms in the pharmaceutical sectorcan be one of three types bulk drugs producers pureformulators or integrated firms which produce both bulkdrugs and market formulations

TABLE 1mdashTHE QUINOLONES SUBSEGMENT

Molecule

Shares ()of quinolones sales Sales (Rs mill) 2000

Domesticfirms

Foreignsubsidiaries

Domesticfirms

Foreignsubsidiaries

Ciprofloxacin 530 27 3030 156Norfloxacin 112 01 640 3Ofloxacin 116 31 665 177Sparfloxacin 108 01 620 4Lomefloxacin 15 86 Pefloxacin 13 01 72 5Levofloxacin 00 0 Nalidixic acid 13 73

TABLE 2mdashBASIC INFORMATION ABOUT THE FOUR QUINOLONES MOLECULES 2000

Ciprofloxacin Norfloxacin Ofloxacin Sparfloxacin

US or European patent holder Bayer Merck Ortho-McNeil Rhone-PoulencYear of US patent expiry 2003 1998 2003 2010Year of US-FDA approval 1987 1986 1990 1996Year first introduced in India 1989 1988 1990 1996No of domestic Indian firms 75 40 17 25No of foreign subsidiaries 8 2 2 1Sales weighted average per-unit API

price of products produced byDomestic Indian firms 1124 905 9008 7884Foreign subsidiaries 1035 528 10847

API Active pharmaceutical ingredient

1482 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

antiinfectives segment ranks second in Indiawhereas in the world market it is fifth and has ashare of only 90 percent Hence antiinfectivesare important in India not only from a health

and public policy point of view but also as asource of firm revenue

With this in mind we focus on one partic-ular subsegment of antiinfectives namely the

TABLE 3mdashSUMMARY STATISTICS FOR THE QUINOLONES SUBSEGMENT 1999ndash2000

North East West South

Annual quinolones expenditure per household (Rs) 3125 1975 2764 2359(366) (367) (407) (286)

Annual antibiotics expenditure per household (Rs) 11988 8424 11052 9624(1224) (1224) (960) (996)

No of SKUsForeign ciprofloxacin 1238 1129 1308 1246

(150) (190) (102) (106)Foreign norfloxacin 183 171 200 158

(070) (075) (088) (083)Foreign ofloxacin 304 296 296 300

(086) (086) (091) (088)Domestic ciprofloxacin 10621 9763 10342 10550

(599) (434) (722) (451)Domestic norfloxacin 3896 3496 3617 3942

(271) (268) (251) (379)Domestic ofloxacin 1846 1600 1725 1725

(680) (634) (586) (635)Domestic sparfloxacin 2983 2829 3121 2929

(557) (638) (688) (657)Price per-unit API (Rs)

Foreign ciprofloxacin 958 1090 1085 1007(128) (066) (071) (058)

Foreign norfloxacin 563 509 605 435(077) (133) (139) (147)

Foreign ofloxacin 10946 10943 10886 10612(620) (664) (700) (1140)

Domestic ciprofloxacin 1143 1067 1131 1152(016) (015) (017) (013)

Domestic norfloxacin 951 907 888 873(024) (035) (037) (020)

Domestic ofloxacin 9163 8964 8565 9341(1615) (1565) (1422) (1407)

Domestic sparfloxacin 7972 7849 7688 8028(976) (1014) (1185) (1037)

Annual sales (Rs mill)Foreign ciprofloxacin 4179 2431 4520 2947

(1534) (816) (1273) (648)Foreign norfloxacin 128 100 058 073

(101) (082) (044) (057)Foreign ofloxacin 5446 3184 3522 3111

(1399) (933) (906) (703)Domestic ciprofloxacin 96229 58591 67874 70381

(10626) (13026) (12226) (8740)Domestic norfloxacin 22255 11971 14918 15829

(3884) (1945) (2691) (1626)Domestic ofloxacin 12502 9621 14936 11205

(4434) (3011) (5282) (4259)Domestic sparfloxacin 15617 12175 16130 9811

(3141) (2576) (4674) (3420)

Note Standard deviations in parentheses API Active pharmaceutical ingredient

1483VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

quinolone subsegment Quinolones fall into thesystemic antibiotics and antibacterials segmentof the Indian pharmaceuticals market whichgenerates over three-quarters of the revenues inthe antiinfectives segment8 The systemic anti-bacterials segment includes all of the originalmiracle drugs that first sparked the growth ofthe global research-based pharmaceutical indus-try in the postndashWorld War II period as well aslater generations of molecules that have beenintroduced in the last four decades

Among systemic antibacterials quinolonesare the latest generation molecules available inIndia We focus our analysis on quinolones forseveral reasons First quinolones are the drug ofchoice for a large number of bacterial infec-tions some of which are also treated by alter-native drugs (see Table A1 in the Appendixwhich outlines the spectrum of activity for eachmolecule family within the antibacterials seg-ment) Hence if there were one product groupfor which we would expect to have many sub-stitutes readily available this would be quino-lones Second with a share of 20 percent in thesales of systemic antibacterials quinolones rep-resent one of the largest subsegments within thistherapeutic category Finally several moleculeswithin the quinolone subsegment were still un-der patent in the United States at the time of ourinvestigation Table 2 details the basic informa-tion about the four quinolone molecules that arethe focus of our analysis The first row showsthe year of US patent expiry this ranges from1998 for norfloxacin to 2010 for sparfloxacinQuinolones include in principle four moremolecules that are listed at the bottom of Ta-ble 1 the market shares of these molecules arenegligible however so that we exclude thesemolecules from our analysis

Table 2 reveals several other interesting factsabout competition in the quinolone market inIndia First note the large number of firmsoperating in this subsegment The large numberof domestic firms is perhaps not that surprisinggiven that pharmaceutical product patents werenot recognized in India9 What is more surpris-ing is the number of foreign firms selling pat-ented products (eg ciprofloxacin) the fact that

multiple foreign firms sell a patented productindicates that such firms often ldquoinfringerdquo10

patent laws in India while complying with themin developed world countries The last two rowsof Table 2 further indicate that domestic prod-ucts often sell at a premium With the exceptionof ofloxacin the average prices of products of-fered by Indian firms are higher than the pricesof products offered by foreign subsidiaries Thispreliminary evidence suggests that Indian con-sumers do not place a premium on the brandname and reputation of big multinational phar-maceutical concerns Moreover the higherprice of domestic products does not seem toprevent domestic companies from capturing alarge market share This is most evident in thecase of ciprofloxacin where domestic firmshave with 53 percent the largest share in thetotal sales of quinolonesmdashthis despite the factthat the average price of these products is 10percent higher than the price of foreign productscontaining the same molecule

Table 3 provides additional summary statis-tics for our data broken down by region Thefirst two rows of the table report the averageannual household expenditure on quinolonesand antibiotics respectively Note that in bothcases the average expenditure is higher in thenorth and west these regions include states withhigher per capita incomes and tend to be moreindustrialized and urbanized than those in theeast and south Pharmaceutical products areavailable in multiple presentations that is com-binations of dosage forms (eg capsule tabletsyrup) strength (eg 100 milligrams 500 mil-ligrams) and packet sizes (eg 50 capsule bot-tle 100 tablet bottle) The various presentationsin which a product is available are often referredto as stock-keeping units or SKUs11 The num-ber of SKUs for each product group withinquinolones is reported at the top of Table 3 Aswith the more aggregate numbers on firms and

8 In addition to anti-bacterials this segment also containsantivirals

9 Accordingly the common distinction between ldquobrandedrdquoand ldquogenericrdquo products is irrelevant here

10 We emphasize here that the word ldquoinfringerdquo belongsin quotes because patent laws do not currently exist inIndia infringement in the legal sense is not possible It ishowever striking that the same firms that accuse Indianproducers of ldquopiracyrdquo sell in India products that are patentedin the United States and for which the patent is held by adifferent multinational corporation

11 For instance a 100 capsule bottle of 100 milligramcapsules of a particular branded drug and a 50 capsulebottle of 100 milligram capsules of the same branded drugwould be identified as two separate SKUs

1484 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

products reported in Table 2 the differencebetween domestic and foreign products isstriking The number of SKUs offered byIndian firms is consistently larger than thenumber offered by subsidiaries of foreignmultinationals The number of SKUs variesslightly across regions but more importantlyit varies across time as some SKUs disap-pear while new ones get introduced duringour sample period

Many pharmaceutical products in India aresubject to price controls12 While the specificsof the price regulation are too complex for anyeconomic model to capture adequately themain concern for the empirical analysis is thatprice controls may lead to a lack of price vari-ation over time so that the demand functioncannot be identified Prices at the most disag-gregate SKU level are relatively stable overtime there are variations due to occasionalchanges in the estimated cost (due for exampleto changes in exchange rates that affect the costof imported materials or bulk drugs) but suchvariations tend to be infrequent and small inmagnitude The degree of time variation ishowever substantially larger once one aggre-gates to the product level This variability stemsnot only from the fact that the SKUs over whichwe aggregate may experience changes in theirrespective prices at different points in time butalso from the fact that the range of SKUs of-fered in the market does not remain constantover time The entry and exit of presentationswithin the same product group that have differ-

ent prices effectively affects the price that con-sumers face for this drug in each period

The middle portion of Table 3 reports themean price and standard deviation for eachproduct group by region Prices vary by regionthough there is no clear pattern emerging fromthe table with regard to the cross-regional vari-ation (in the sense of some regions being sys-tematically more expensive than others) Toexamine what portion of the total price variationis due to time versus regional variation weconducted an analysis of variance of prices thatwe report in Table 4 The table is based onseparate regressions for each product group(pooling data across groups with big differencesin their average prices is not particularly infor-mative as most of the price variation is ac-counted for by product group dummies) Thelast two columns of the table show the fractionof price variation accounted for by region andtime dummies respectively As evident fromthe table a significant fraction of the total vari-ance in prices can be attributed to time varia-tion In the demand estimation we include a fullset of product groupndashspecific regional dummiesso that the price parameters are identified en-tirely based on this time variation within eachproduct group The time variation of productgroup prices is driven primarily by composi-tional changes within each group the revenueshares of the individual SKUs that compriseeach product group change over time (see therelated discussion in Section IIC) while in eachperiod there is entry and exit of SKUs into thesample To check whether this pattern reflectsgenuine entry and exit as opposed to samplingvariation we examined the revenue shares ofthe SKUs that leave the sample relative to theones that remain during the entire period Theresults are reported in Table 5 While the SKUs

12 The details of the procedures for price fixation canbe found in the official government Web site httpwwwnppaindianicinindex1html under the link ldquoDrug Price Con-trol Order 1995rdquo A new pharmaceutical policy was introducedin 2002 but our data were collected before that year

TABLE 4mdashANALYSIS OF PRODUCT PRICE VARIANCE

Product groupPartial SS

timePartial SS

region Total SSPercentage

explained by timePercentage

explained by region

Foreign ciprofloxacin 0296 0306 1047 283 292Foreign norfloxacin 1036 1002 4179 248 240Foreign ofloxacin 0429 0021 0571 752 36Domestic ciprofloxacin 0005 0088 0104 50 841Domestic norfloxacin 0059 0098 0198 296 494Domestic ofloxacin 2858 0104 3056 935 34Domestic sparfloxacin 1754 0032 1860 943 17

1485VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

that exit tend to be smaller (their average shareis 1 percent as opposed to 34 percent for thoseSKUs that are present during the entire sampleperiod) the shares of the two groups do notseem orders of magnitude apart13 In additionour data cover only the 300 largest firms sellingin the Indian market so that firms with verysmall shares are not included in our sample

II The Analytic Framework and EstimationApproach

Patent enforcement in the Indian pharmaceu-tical market will have the effect of eliminatingdomestic products whose active pharmaceuticalingredients are protected by (foreign) patentsThus assessing the effects of patent enforce-ment is tantamount to assessing the effects ofwithdrawing domestic products from the mar-ket This task is the converse of evaluating newproduct introduction accordingly the concep-tual framework we use to address the questionsof interest is similar to the one developed in theliterature for the valuation of new goods14

We start by estimating demand for quino-lones Given that the market is characterized byimperfect competition the counterfactual anal-ysis requires that we also model the supply sideas removal of one product will affect the pricesof other products especially those that are closecompetitors The existence of price regulationin the Indian pharmaceutical market imposespotential constraints on firmsrsquo maximizationproblem Given these constraints and the com-

plexity of the price regulation process the typ-ical approach of deriving estimates of actualmarginal costs and markups by exploiting thefirst-order conditions of profit maximizing firmsdoes not seem particularly promising Insteadwe use our demand estimates to place upper andlower bounds on marginal costs and markups

With demand elasticities and upper and lowerbounds for marginal costs in place we thenconduct counterfactual simulations We con-sider several alternative scenarios dependingon the number of domestic products affected bypatent enforcement For each scenario we com-pute the counterfactual prices and use them toassess the effects of domestic product with-drawal on consumer welfare (as measured bythe compensating variation) firm profits andsocial welfare As with the valuation of newproducts the big conceptual problem facing thispart of the analysis is that we need to extrapo-late from the region of the data to the point atwhich demand for the products that exit themarket becomes zero This conceptual issue ispresent in any attempt to evaluate a major pol-icy change for which no historical precedentexists like the enforcement of patent laws inIndia One advantage of the present study is thatwe have a limited set of price data for Pakistana country with similar demographics to Indiabut with a market structure that resembles theone that would emerge in India under patentenforcement (monopoly of multinational sub-sidiaries) By comparing the prices of productsoffered by multinationals in Pakistan to thosewe compute in our counterfactual simulationsfor the products that would be offered by mul-tinationals in India if patent laws were enforcedwe can get a sense of how plausible our coun-terfactual estimates are

A Demand

The demand modelling is based on the multi-stage budgeting approach Our primary motiva-tion for adopting this approach is a practical oneIn the multistage budgeting approach the depen-dent variable is defined as a revenue share whichis appealing given that the products we include inthe analysis contain different molecules (ie ac-tive pharmaceutical ingredients or APIs) Eventhough we do have data on the quantity of therelevant API (eg 100 milligrams of ciprofloxa-cin) contained in each product converting the

13 While we cannot completely rule out the possibilitythat some of the exit is due to sampling variation note thatthe latter should be reflected in low precision of the demandparameter estimates The demand parameters however areprecisely estimated

14 See Manuel Trajtenberg (1989) Jerry A Hausman(1994) and Timothy F Bresnahan (1997) for representativeexamples and a discussion of the relevant issues

TABLE 5mdashREVENUE SHARES OF THE EXITING SKUS

Revenue share ofexiting SKUs

Revenue shares ofall SKUs

Full sample 10 34Northern region 08 33Eastern region 12 36Western region 10 34Southern region 12 33

1486 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

revenue shares to physical shares is extremelydifficult if not infeasible in the case of antibioticsBecause such drugs are ldquosystemicrdquo by nature theyare used to treat a large number of infections andthe dosage of each drug depends on the particularinfection it is supposed to address (for example thedosage will differ depending on whether the anti-biotic is used to combat an ear infection or tuber-culosis) This particular feature of antibioticscomplicates the conversion of revenue to physicalshares

The basic idea of the multistage budgetingapproach is to use the therapeutic classificationof a productmdashie the therapeutic segment andsubsegment the product belongs tomdashto orga-nize all products in the systemic antibacterialssegment into a hierarchical taxonomy consist-ing of two levels At the higher level are thevarious subsegments of systemic antibacterialsThe first stage of budgeting corresponds to theallocation of expenditures across the subseg-ments in this upper level of the taxonomy

In the second stage of the budgeting processcorresponding to the lower level of the taxon-omy a flexible functional form is adopted tomodel how the expenditures allocated to eachsubsegment are distributed across the productswithin that subsegment In particular to modeldemand at the second stage we employ theAIDS specification proposed by Deaton andMuellbauer (1980b)15

While the two-stage demand estimation ap-proach offers functional form flexibility its ap-plication to the Indian systemic antibioticsmarket poses a few problems The first is thatdue to entry and exit many SKUs and evenproducts in our sample are not present in everyperiod AIDS does not have a good way ofdealing with a varying number of products as itwas developed with broad commodity catego-ries in mind which are consumed by all con-sumers every period To solve this problem weaggregate within each subsegment (eg quino-lones) SKUs into product groups where withineach product group all SKUs contain the samemolecule and are produced by firms with thesame domesticforeign status Specifically let aSKU k be indexed by its molecule (or API) M its

domesticforeign status DF indicating whether itis produced by a domestic (Indian) or a subsidiaryof a foreign (multinational) firm a particular pre-sentation s and the particular firm f that producesit We aggregate SKUs over presentations andfirms to obtain a newly defined product group iwhich is indexed only by molecule M and domes-ticforeign status DF and has revenue Ri yenfsRk with i (M DF) k (M DF f s) and pricepi yenfs kpk where k denotes the conditional(on M and DF) revenue share of this particularproduct ie

(1) k Rk

Ri

In most cases the resulting product groups arebroad enough to be present every period16

The usual concern with this aggregation pro-cedure is that it may lead us to overstate firmsrsquomarket power as we ignore competition amongfirms with the same domesticforeign statusproducing the same molecule In the presentapplication however this concern is unlikely tobe of great importance as the effect of patentenforcement is to wipe out all domestic compe-tition at once while granting foreign firms mo-nopoly power hence competition among firmsfor patented molecules becomes irrelevant Theaggregation according to the domesticforeignstatus (within a particular molecule) thus corre-sponds to the scope of our analysis and theparticular questions of interest

The second problem is that for our approachto be useful in welfare analysis the allocation oftotal expenditures to group expenditures at thehigher stage has to be modelled in a way con-sistent with utility maximization In general thesolution of this allocation problem requiresknowledge of all individual product prices

15 Representative applications of the multistage budget-ing approach include Sara F Ellison et al (1997) Hausman(1994) and Hausman and Gregory K Leonard (2002)

16 We are missing only four observations (ie monthregion combinations) all for the drug group of ForeignNorfloxacin August 1999 in the south May 2000 in thewest October 2000 in the south and November 2000 in theeast In these cases we set the revenue shares of ForeignNorfloxacin equal to zero In general with 01 percent ofquinolone sales (see Table 1 row 2) Foreign Norfloxacinhas a very small share of the market This probably explainswhy the results pertaining to this drug group are unreliableit is the only drug group for which we do not obtain asignificant price elasticity of demand while its cross-priceelasticities with other foreign drug groups often have thewrong sign

1487VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

From an empirical point of view this is notparticularly useful as it eliminates all compu-tational advantages of the two-stage approachTo address this problem we adopt an approxi-mate solution to model the higher-level expen-diture allocation along the lines suggested byDeaton and Muellbauer (1980b pp 131ndash32)This gives rise to a two-level AIDS specification

Consider the lower-level estimation firstwhich refers to the allocation of a particularsubsegmentrsquos expenditure to the product groupswithin the subsegment In our application therelevant subsegment is quinolones which weindex with Q Let the product groups within thissubsegment be indexed by i 1 N pi be theprice of product group i (where as noted abovei refers to a particular molecule and domesticforeign status combination) and XQ be the totalexpenditure on the quinolone segment The rev-enue share of each product group is given by

(2) i i j

ijln pj ilnXQ

PQ

where i the revenue share of product group iis defined as

(3) i pi qi

yenj

pj qj

xi

XQ with i j Q

XQ is the overall expenditure on the quinolonesubsegment and PQ is a price index given by

(4) ln PQ ap 0 i

iln pi

1

2 i

j

ijln piln pj

With a limited number of product groups and asufficiently large number of time-series observa-tions the flexibility implied by the AIDS modeldoes not impose too many demands on the data Inthe present application however where the num-ber of observations is limited the AIDS model isnot estimable in this general form To reduce thenumber of parameters that need to be estimatedwe impose two sets of restrictions

The first set of restrictions is implied by thetheory of utility maximization Specificallythese restrictions are

Adding-up yenk k 1 yenk k 0 yenk kj 0 j

Homogeneity yenk jk 0 j Symmetry ij 1frasl2 [ij ji] ji This last

restriction by itself reduces the number of parameters to [N(N 1)]2

The second set of restrictions we impose aimsat further reducing the number of parameters tobe estimated by exploiting our knowledge of thisparticular market Specifically for each productgroup i we allow one ij parameter for all productgroups j that have different molecules from prod-uct group i and are produced by foreign firms andone ij for all product groups j with differentmolecules produced by domestic firms We donrsquotimpose any restrictions on the ij parameter whenproduct group j has the same molecule as productgroup i (By construction product groups i and jcontain products produced by firms with differentdomesticforeign status)

To better illustrate the nature of the restric-tions we impose on the patterns of substitutionacross products some additional notation isneeded Let d(i j) be an indicator of the degreeof similarity (or difference) between productgroup i and product group j along the dimen-sions we are able to observe (molecule M anddomesticforeign status DF) For any two prod-uct groups i and j d(i j) can take on one of thefollowing three values17

(5)

di j (1 0) if Mi Mj DFi DFj

(0 1) if Mi Mj DFi DFj

(0 0) if Mi Mj DFi DFj

Let

(6) Diab j di j a b

17 The sequence (1 1) is not possible for two differentproducts in this case the parameter corresponds to theproductrsquos own-price effect that is ii

1488 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The equation at the lower level becomes

(7) i i iiln pi i10ln pj j Di10

j Di

01

i01ln pj

j Di

00

i00ln pj ilnXQ

PQ

Note that

The parameter ii captures a product grouprsquosown-price effect (there will be as many iiparameters as the number of product groups)

The parameter i10 captures the cross-priceeffect of the product group containing prod-ucts with the same molecule but produced byfirms of different nationality

The parameter i01 captures the cross-priceeffects of product groups containing productswith different molecules but produced byfirms with the same nationality

The parameter i00 captures the cross-priceeffects of product groups containing productswith different molecules produced by firms ofdifferent nationality

Before we take the demand equation to thedata we make two modifications The first oneis to let the product-specific effects i vary byregion r The resulting product-specific regionaleffects ir have two interpretations first theycontrol for the ldquoqualityrdquo of each drug withquality differences being allowed to vary acrossregions second they proxy for demographicsand other demand shifters which vary by re-gion and may affect the demand of each prod-uct group differently18 Note that by includingproduct-specific regional effects in the demandspecification we estimate the price parametersbased on the within-product-group variation ofprices in each region In an earlier version of thepaper we also estimated the demand systemwithout regional dummies and obtained similarresults

The second modification that allows us to gofrom a deterministic to a stochastic specificationof the demand equation is to include an additiveerror term in (7) The latter takes into accountthe fact that (7) is not expected to fit the dataexactly The error term irt accounts for mea-surement error (due to the fact that the productgroup prices pjrt we employ in the estimationare not exact price indices but approximationsthereof) and (potentially region-specific) de-mand shocks that may affect the demand for aproduct in a particular period Examples of suchshocks include an advertising campaign for aparticular product which temporarily increasesthe demand for this product or the outbreak ofa (potentially region-specific) epidemic whichcalls for the use of a particular drug We discussthe interpretation and properties of this errorterm in more detail in the next section

The final form of the equation we estimate atthe lower level becomes (with subscript t de-noting month and subscript r denoting region)

(8) irt i ir iiln pirt i10ln pjrt j Di10

j Di

01

i01ln pjrt

j Di

00

i00ln pjrt

ilnXQrt

PQrt irt

The analysis so far has conditioned on the ex-penditure allocated to the quinolone subseg-ment XQ The upper level of the estimationconsiders the problem of allocating total expen-diture across the different systemic antibioticssubsegments one of which is quinolones Theupper-level demand function is given by

(9) G G H

GHln PH GlnX

P

where all variables denoted by capital letters aredefined as before but now refer to subsegments(G H ) rather than product groups within asubsegment and the total expenditures on sys-temic antibiotics X are deflated by the Stone priceindex log P yenH Hlog PH When estimating thesystem above we impose all the restrictions

18 Given the short time span of our sample typicaldemand shifters such as age distribution income distribu-tion and education hardly change over our sample periodSuch shifters are therefore absorbed by the region-specificproduct fixed effects ir

1489VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

implied by utility maximization as we do withthe estimation of the lower-level AIDS We donot however impose any additional restrictionson the substitution patterns at this stage so thatthe cross-price effects across segments remainrelatively unconstrained

Estimation of the higher-level AIDS allowsus to obtain the unconditional own- and cross-price elasticities that are used in the formulationof the supply problem and welfare analysisThese are given by the formula

(10) ij ijXQXQ

ln qi

ln XQ

ln XQ

ln PQ

ln PQ

ln pj

with the conditional cross price elasticitiesgiven by

ijXQXQ ij ij jlnXQ

PQ

i

As in the lower stage we include subsegment-specific regional dummies in the specificationof the upper-stage demand system so that thefinal form of the estimating equation at theupper stage becomes

(11) Grt G Gr H

GHln PHrt

GlnXrt

Prt Grt

In sum the demand system we take to thedata is represented by equations (8) and (11)and the associated parameter restrictions im-plied by economic theory

B Modelling the Supply Side of the Market

Counterfactual simulations concerning theeffects of domestic product withdrawal requireknowledge of the marginal costs of pharmaceu-tical firms operating in the Indian market Theseare unobservable The usual approach in theNew Empirical Industrial Organization litera-ture has been to exploit the firm equilibriumconditions to infer marginal cost For exampleit is often assumed that the marginal cost ci isconstant and that the industry is an oligopoly

engaging in Bertrand competition with differen-tiated products Assuming that firms myopicallymaximize profits each period one can then de-rive the firmsrsquo first-order conditions that corre-spond to the assumptions above about costsmarket structure and firm behavior

We deviate from this procedure as the presenceof price regulation renders the assumption of un-constrained period-by-period maximization un-tenable Ideally one would like to explicitlyincorporate the price and other administrative con-trols into the firmrsquos optimization problem andderive the first-order conditions under the assump-tion of constrained maximization The complexityof the price regulation however makes this ap-proach infeasible Therefore we adopt an alterna-tive approach that does not rely on modelling theprice-setting process but is instead based on de-riving upper and lower bounds for marginal costsand markups

In particular an upper bound for marginalcosts and a lower bound for the markups (zero)can be derived under the assumption of perfectcompetition While this assumption is clearlyunrealistic in the pharmaceuticals market it isuseful in providing an upper bound for costs ci

Uwhich will be given by

(12) ciU pi

On the other hand a lower bound for marginalcosts ci

L (and upper bound for markups) can bederived by assuming that there is perfect collu-sion within each product group i (where i refershere to a moleculedomestic-foreign combina-tion)19 and ignoring price controls so thatprices are determined by the first-order condi-tion of the jointly profit-maximizing firmswithin product group i Solving this first-ordercondition for the (lower bound of) marginal costthen gives

(13) ciL pi 1

1

iipipj

19 The reason that this assumption will lead to an under-statement of marginal costs (and hence overstatement ofmarket power) is that it assumes away competition amongfirms within each drug group

1490 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

where ii(pipj) the own-price elasticity of de-

mand for product i will depend on the productrsquosown-price pi and all other productsrsquo prices pjwith i j

Once we have obtained the demand elastic-ities through estimation of the demand sys-tem we can calculate the upper and lowerbounds for marginal costs and correspondingmarkups according to (12) or (13) These willthen be employed in the counterfactual sim-ulations While our counterfactuals use boththe lower and upper bounds for costs most ofour discussion will be based on using thelower bounds for costs since this is the moreinteresting case in the policy analysis it givesus the largest possible profits for pharmaceuticalfirms and hence corresponds to the worst pos-sible scenario facing Indian firms and the bestpossible scenario facing multinationals underTRIPS

C Identification Assumptions and EstimationApproach

The discussion of the demand system has sofar abstracted from the issue of price endogene-ity The usual premise in the industrial organi-zation literature is that correlation of prices withthe error term in the demand equation arises byvirtue of the first-order conditions of profit-maximizing firms20 As we discuss in this sec-tion this source of simultaneity bias is unlikelyto be of major concern in the present contextbecause of the existence of price regulation Ourprimary concern is instead with the simultane-ity bias that is implied by the particular wayprices are constructed

To understand the sources of the simultaneitybias note first that the price of product group piwe employ in the demand estimation should bethought of as a proxy for an exact price indexthat we do not observe As such pi contains bydefinition measurement error and it will becorrelated with the error term of the demandequation Specifically the demand equation weare interested in estimating can be written insimplified form (suppressing the subscripts rand t and ignoring parameter restrictions forconvenience) as

i i j

ijln pjT ilnXQ

PQ i

where pjT denotes the true (exact) price index for

product group i and i captures unobservedvariables that may affect demand in a particularperiod (eg an advertising campaign or an ep-idemic) The price index pj

T depends on theSKU prices pk Note that in the presence ofSKU entry and exit into the sample the priceindex pj

T will vary over time even if the pricesof the individual SKUs remain stable as theproducts (SKUs) that comprise each productgroup change each period

Ideally we would like to compute the exactprice index pj

T for each group and employ theseindices in the demand estimation Under theassumption of predetermined SKU prices eachperiod the exact price indices would be uncor-related with the error term and the demandparameters could be estimated by simple OLSUnfortunately this strategy is not feasible in thecurrent context

To derive an exact price index for each prod-uct group it is necessary to model consumerrsquoschoice among SKUs conditional on the choiceof the group To this end it is necessary toexplicitly introduce a third stage in the demandspecification and estimate the parameters asso-ciated with the choice at that stage Howeverany specification based on three-stage budget-ing (that would require directly specifying ademand function associated with the SKUchoice at the third stage and imposing additiveseparability at the higher stages) would be in-feasible to estimate for three main reasonsFirst there is a large number of SKUs withinsome groups so that the demand parameters can-not be estimated even if one imposes parameterrestrictions second the SKU prices exhibit littlevariation over time lastly the choice sets at thislowest stage vary over time rendering estimationof the price parameters associated with SKUs thatare not available in all periods infeasible Alterna-tively one could abandon the multistage budget-ing altogether and adopt a discrete choiceapproach which is inherently better suited to deal-ing with varying choice sets and limited pricevariation A discrete choice approach would beparticularly appealing for modeling the SKUchoice conditional on product group choice giventhat within each group all SKUs contain the same

20 See Steven Berry et al (1995) or Aviv Nevo (2001)for a related discussion

1491VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

molecule and are hence comparable As notedearlier however a discrete choice approach at thehigher stages of the demand estimation requiresconverting revenue to physical shares which ischallenging in the case of systemic drugs such asantibiotics for which the appropriate dosage is notwell defined Finally a combination of a multi-stage budgeting approach for the higher stages anda discrete choice approach for the lowest stagewould be feasible to estimate but inconsistentwith a model of utility maximization For thesereasons we have no alternative but to adopt anapproximation for measuring the price index foreach product group

Let us denote this approximation by pjA and let

13j denote the proportional ldquoapproximationrdquo errorassociated with measuring the true price index forgroup j so that ln pj

A ln pjT ln 13j The

estimating demand equation can then be written as

i i j

ijln pjA ilnXQ

PQ

j

ijln 13j i

or

i i j

ijln pjA ilnXQ

PQ i

where the new error term i yenj ijln 13j iis comprised of two components the first onereflects ldquomeasurementrdquo error due to the fact thatthe product group prices we employ are approx-imations and not exact price indices and thesecond one captures conventional demandshocks The demand equation as written abovecorresponds to the equation we take to the data

The particular proxy we use for the exact priceindex for each product group j is the revenue-share-weighted average of the prices of multipleSKUs available within this group that is

pjA

k j

k pk

The notation above illustrates the two source ofprice variation variation in the SKU prices pkand variation in the weights k

We are concerned with potential correlation ofthe so-constructed product group prices with bothcomponents of the product group demand error

term The correlation between pjA and the first

component yenj ijln 13j is independent of theparticular way we construct pj

A and inherent in thefact that we measure the exact price index witherror The potential correlation between pj

A and thesecond component of the product group demanderror the shock i arises however because of thespecific way we construct the proxy pj

A and inparticular because of the presence of the reve-nue share weights k in pj

A since the SKUrevenue share weights k will generally dependon the product group expenditure they arelikely to be correlated with the product groupdemand shock i

To address the simultaneity bias we use in-strumental variables To this end we need vari-ables that are correlated with the proxies pj

A butuncorrelated with the error term i We use thenumber of SKUs within each product group j asan instrument as it does justice to the idea thatvariation in the product group price index stemsin part from variation in the set of SKUs that areavailable each period It is clearly correlatedwith the average group prices pj

A and plausiblyuncorrelated with the demand shock i Theassumption that the number of SKUs is uncor-related with the conventional demand error willhowever be violated if the introduction of anew SKU changes the perceived quality of adrug21 or if it is accompanied by promotionalactivities which will be reflected in i Our hopeis that this is not too often the case Assumingthat the SKU number for each group j is alsouncorrelated with the weighted sum of the mea-surement errors of all product groups yenj ijln 13j itcan be used as an instrument for the product groupprice pj

A22 If one accepts the premise that prices atthe SKU level are exogenous then SKU pricescan also be used as instruments The final list ofinstruments we use hence includes the number ofSKUs in each group the prices of the five largestSKUs for each group and all other exogenousvariables in the demand estimation such as re-gionalproduct dummy interactions and the upper-

21 Note that we already proxy for the quality of each prod-uct group through the product fixed effects i In the actualestimation we actually let these fixed effects be region-specific(ir) so that we allow quality to vary across regions

22 If the measurement error increases in the number ofincluded SKUs this assumption will be violated Unfortu-nately we have no way of checking the validity of thisassumption

1492 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

level total expenditure on antibiotics Regressionsof group prices on the instruments above yieldhigh R-squares with most regressors highly sig-nificant indicating that our instruments are highlycorrelated with prices23

Our sample includes four molecules cipro-floxacin norfloxacin ofloxacin and spar-floxacin Except for sparfloxacin all othermolecules are produced by both foreign anddomestic firms24 So we have seven products(domestic ciprofloxacin foreign ciprofloxa-cin etc) with 96 observations (two years ofmonthly data four geographical regions foreach period) for each product The parametersin the lower-level AIDS demand system asdefined in equation (8) are the product fixedeffects i the productndashregional dummy inter-actions ir the own revenue-share price elas-ticities ii the cross revenue-share priceelasticities i10 i01 i00 and the revenue-share expenditure elasticities i In estimatingthe parameters we first regress prices on allinstrumental variables and then plug the pre-dicted values for prices in the constrainedleast-square regression (For a detailed expla-nation of the constraints see the previoussection)

Given that we impose many cross-equationconstraints and employ instrumental variablesin the estimation it is difficult to derive stan-dard errors for the parameter estimates ana-lytically Our error term interpretation in (8)implies that the error terms for each productgroup are likely to be correlated across re-gions for example if a national advertisingcampaign increases the demand for a rela-tively expensive presentation in one regionsimultaneously increasing the aggregate de-mand and price index for the correspondingproduct group in that region it is likely that thesame effect will be observed in other regions Inprinciple we could exploit the cross-regionalcorrelation in the error terms to estimate thedemand system using seemingly unrelated re-

gression (SUR) This however would requireus to derive the variance-covariance matrix ofthe parameters analytically which as notedabove is cumbersome in the current contextInstead we use the bootstrap method choosingto remain agnostic about the structure of thevariance-covariance matrix The potential dis-advantage of this approach over SUR is a loss inefficiency but as will become apparent in theresults section the model parameters are fairlyprecisely estimated To maintain the marketstructure we randomly sample the periods (withreplacement) and use the same periods for allproducts Regarding the optimal number ofbootstrap repetitions ideally one would followthe three-step method proposed by Donald W KAndrews et al (2000) Empirical evidence sug-gests however that one rarely needs more than200 replications to estimate the standard er-rors25 To be safe we generate 300 bootstrapsamples (with replacement) based on the origi-nal data and estimate the standard errors usingthe standard errors of the bootstrap sampleestimates

The estimation of the top-level AIDS systemis similar The constraints imposed on this top-level demand system are adding-up homogene-ity and symmetry as defined on page 12Again bootstrapping is used to obtain the stan-dard errors of the parameter estimates

D The Counterfactual Scenarios

In assessing the effects of patent enforce-ment we start by focusing on the most extremecase in which compulsory licensing is not anoption and foreign firms are not subject to pricecontrols We use the results from the analysis ofthis case as a benchmark In reality the outcomeof the WTO negotiations is more likely to in-volve some constraints on the monopoly powerof foreign firms selling patented products indeveloping countries such as price caps orcompulsory licensing Our framework can eas-ily accommodate these cases as will becomeapparent in the next subsection

We now focus on the effects of potentialpatent enforcement in the quinolone segmentWe consider several scenarios which vary inthe number and size of domestic products that

23 Specifically the R-squares from the first-stage regres-sions of product group prices on instruments range from057 to 095 except for domestic ciprofloxacin for whichthe R-square is 017

24 Sparfloxacin is actually offered by one foreign subsid-iary in India (see Table 2) However its revenue share isminiscule We therefore treat Sparfloxacin as being pro-duced by domestic firms only 25 See Bradley Efron and Robert J Tibshirani (1993)

1493VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

will be removed from the market In particularwe consider the following five scenarios26

Withdrawal of one large domestic productgroup only domestic ciprofloxacin

Withdrawal of one relatively small domesticproduct group only domestic ofloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand norfloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand sparfloxacin

Withdrawal of all four domestic quinoloneproduct groups

As the list above suggests we proceed fromanalyzing the effects of single-product with-drawal to the analysis of eliminating the en-tire domestic segment This approach wasmotivated by early empirical results that in-dicated that the existence and extent of com-petition from domestic firms has a significantbearing on the predicted effects of patent en-forcement that is our predictions regardingprices and welfare vary substantially depend-ing on how many domestic products are af-fected by patent enforcement and on whethersome domestic competition will remain presentafter TRIPS In addition the size of the affecteddomestic groups is relevant for the welfare pre-dictions accordingly we examine both scenar-ios in which a large domestic product groupsuch as domestic ciprofloxacin is eliminatedfrom the market and scenarios in which the

eliminated domestic group is relatively small(eg domestic ofloxacin) In all scenarios wemaintain the assumption that the set of prod-ucts offered by the remaining firms in themarket does not change in response to patentenforcement

E Computation of Virtual Prices and NewEquilibrium Prices

The first step in the counterfactual analysisis to derive the new equilibrium prices underpatent enforcement27 In this context thereare two sets of prices that are relevant Thefirst set consists of the virtual prices of those(domestic) products that will not be availableonce TRIPS is put in effect To calculate thesevirtual prices we set the expenditures of therelevant products equal to zero The secondset of prices consists of the prices of thoseproducts that remain in the market In deriv-ing these prices we start by assuming profitmaximization without any form of price reg-ulation the firms remaining in the marketreoptimize in response to the policy changeand set new prices taking the prices of allother firms as given28 Of course at the equi-librium all prices change in response to thefact that some domestic products are nolonger present The new equilibrium pricesfor products that remain in the market are thuscomputed by utilizing the first-order condi-tions of profit-maximizing firms into whichthe virtual prices of the eliminated productsare substituted Hence to compute the newequilibrium prices we solve an equation sys-tem of the following form

26 As Table 2 indicates most of the patents for the drugsin the quinolone segment have expired by now so that noneof the scenarios described in this section is going to mate-rialize in practice The reason we consider these alternativescenarios in our counterfactual simulations is to get a senseof how the presence and extent of domestic competitionaffects welfare calculations Such calculations may becomerelevant in the future in other therapeutic classes in whichpatents have not expired yet and where domestic firmscompete with foreign patent holders While the particularexit scenarios will depend on the exact year of patent expiryof drugs in each therapeutic class and the specific way inwhich patents will be enforced (eg whether certain do-mestic drugs will be grandfathered in) the general conclu-sion that emerges from our calculations is that the consumerwelfare loss is substantially smaller in cases where somedomestic competition remains present in the market Hencethe particular way in which patent enforcement is imple-mented is essential for assessing its welfare impact

27 Of course until product patents are in fact introducedthese prices will not be observable Note also that we areassuming here that the range of available products will notchange with the introduction of patents

28 As mentioned above this first set of calculations ab-stracts from the existence of remaining price controls orother government regulations that would impose constraintson the firmsrsquo profit maximization problem Accordingly theresulting numbers should be interpreted as a benchmark ofwhat would happen if markets were completely unregulatedTo examine what the welfare effects of TRIPS would be inthe more realistic scenario of price regulation we subse-quently consider a scenario in which regulation keeps theprices of the products provided by patent-holders at theirpre-TRIPS level For a more detailed discussion see theresults section

1494 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

For products i that are withdrawn from themarket

(14) 0 i ir iiln pirV i10ln pjr j Di

10

j Di

01

i01ln pjr

j Di

00

i00ln pjr

ilnXQr

PQr

For products k that remain in the market

(15) pkr ckr 1 1

kk pkrpjr pirV 1

In the equations above pirV denotes the virtual

prices of the products that are removed fromthe market while pjr denotes the updatedprices of all other products Note that whensolving for the virtual prices we account forthe fact that both the price index for quino-lones PQr and the expenditure allocated tothis subsegment XQr need to be updated toreflect the fact that as a result of the pricechanges there may be substitution away fromthis subsegment To obtain the new quinoloneexpenditure XQr and the new price index PQrwe use the estimates and formulas for thehigher-level AIDS system In equation (15)ck refers to the marginal cost for product k thatwe have obtained from the previous estimationstage As mentioned in the previous subsectionwe conducted the simulations using both theupper and lower bounds for marginal cost ck

u

and ckL respectively The term kk(pkr

pjrpirV) re-

fers to the unconditional own-price elasticity forproduct k which is a function of the eliminatedproductsrsquo virtual prices and the remaining prod-uctsrsquo new equilibrium prices We conduct thecounterfactual simulations and welfare analysisat the regional level because the presence ofregion-specific product effects in the demandestimation implies region-specific demand elas-ticities and hence region-specific marginalcosts and markups The presentation and discus-sion of our results focus on national averages ofthe relevant variables that we construct by com-

puting weighted averages across regions usingthe population of each region as a weight

F Welfare Assessment

The simulation of the new equilibrium underpatent protection can provide important insightsinto how consumers and firms will respond tothe removal of domestic products in the market(for example toward which products consumerswill substitute or which prices will increase themost) To get a more precise idea of how peo-plersquos well-being will ultimately be affected byTRIPS we compute as a last step in our anal-ysis the welfare effects of the policy changeSocial welfare is defined as consumer welfareand the sum of domestic firm profits Thechange in domestic profits can easily be calcu-lated by comparing the domestic firm (variable)profits at the pre-TRIPS prices to the profitsthese firms will realize at the new simulatedprices Although foreign firm profits do notcount in domestic welfare calculations we alsocompute the effects of patent enforcement onforeign firm profits to get an idea of how largethe expected benefits of TRIPS are for thesefirms This provides in some sense an indirectway of assessing whether the claims that patentenforcement in countries like India will lead tomore research on developing-country-specificdiseases (such as malaria) have any validity iffor example we find that the effect of patentenforcement on the foreign firm profits realizedin India is small in magnitude it is unlikely thatforeign firms will engage in more developing-country-specific research in response to TRIPSIt is important to note that in all these calcula-tions we work with the lower-bound estimatesfor marginal costs since these give rise to thehighest possible markups Hence our estimateof profit loss for domestic producers most likelyoverstates this loss

On the consumer side we measure changes inconsumer welfare by the compensating variation(CV) defined as the additional expenditure thatconsumers need in order to achieve the sameutility level as before patent enforcement at thenew prices Specifically let P0 denote the pricevector before patent enforcement P the simulatedprice vector post-TRIPS (that we obtained usingthe methods described in the previous subsection)u0 the utility attained by consumers before TRIPS

1495VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

and E(u P) the higher-level expenditure functionThen the compensating variation is given by

(16) CV Eu0 P Eu0 P0

Note that the CV as computed in (16) representsthe combination of three effects

The pure product-variety effect that is theeffect that arises because one or more prod-ucts are not available to consumers anymoreholding the prices of all other remainingproducts and the total expenditure on thequinolone subsegment XQ constant

The expenditure-switching effect that is theeffect arising from substitution away fromquinolones and toward other subsegments ofthe antibiotics market again holding the pricesof all other remaining products constant

The reduced-competition effect that is theeffect that arises because the firms remainingin the market adjust (increase) their prices inresponse to the removal of domestic products

From both an analytical and a policy point ofview it is desirable to assess how large each ofthese effects is Accordingly we decompose thetotal effect on consumer welfare (the CV as givenby equation (16)) using the following procedure

To get the pure product variety effect wecompute virtual prices for the products that areremoved from the market holding the quino-lone expenditure XQ and the prices of all otherproducts fixed Let us call the resulting pricevector P1 Then the pure product variety effectis represented by E(u0 P1) E(u0 P0)

To compute the expenditure-switching effectwe compute another set of virtual prices againholding the prices of all remaining products fixedbut letting quinolone expenditure adjust in re-sponse to the new price index for the quinolonesegment (Given that the prices of the remainingproducts remain fixed the change in the priceindex arises only because of the removal of one ormore domestic products) Note that this scenariomost closely resembles the case in which patentlaws are enforced but strict price regulation keepsthe prices of the products offered in the market attheir pre-TRIP level Let us label the so-computedprice vector P2 The expenditure-switching effectis then E(u0 P2) E(u0 P1)

Finally the reduced competition effect aris-ing from higher prices for the remaining prod-

ucts is computed as the residual change in thecompensating variation once the product-varietyand expenditure-switching effects have been ac-counted for that is E(u0 P) E(u0 P2)where the price vector P is computed accordingto the formulas (14) and (15) to reflect theadjustment of prices to the new regime

To compute the standard errors associatedwith the counterfactual simulations (that is thestandard errors for the counterfactual prices andwelfare estimates) we again use bootstrappingWe first bootstrap the original sample (sales andprices of all the drugs) 300 times Next we esti-mate the AIDS model for each of these samplesand compute the counterfactual equilibrium pricesand welfare losses corresponding to each policyscenario for each of the 300 simulations In thefinal step we compute standard errors29

One limitation of our framework is that itdoes not allow for a heterogeneous response ofconsumers to the policy change Accordinglyour framework is not suited to addressing thequestion of how different groups in the popula-tion will be affected by patent law enforce-ment30 Along the same lines our frameworkdoes not accommodate the possibility of pricediscrimination which might lead to differentresults regarding welfare losses and profit gainsrelative to uniform pricing31 Still we believethat the results of the counterfactual simulationscan provide important insights into the likelyaggregate response to patent enforcement andthe factors that drive this response

III Results

A The Structure of Demand

Tables A2 and A3 in the Appendix display theresults from estimation of the lower- and upper-level AIDS systems respectively For ease of in-terpretation rather than discussing the coefficientestimates we focus our discussion on the implied

29 In cases some of the simulations do not converge wecompute the standard errors using only those simulationsthat converged

30 Addressing this question would require at a mini-mum micro data on consumer purchases as in Goldberg(1995) To our knowledge such data do not exist for theIndian pharmaceutical market

31 See Ernst R Berndt (1994) for an extensive discussionof uniform pricing

1496 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

unconditional price and expenditure elasticities re-ported in Table 6 Given that the region-specificproduct group effects ir reported in the last fourcolumns of Tables A2 and A3 imply region-spe-cific demand elasticities we report separate elas-ticities for each region in Tables 6A and 6B Asevident from the comparison of these tables theelasticities are very similar across regions so thatwe can focus the remaining discussion on theelasticities of one region only the northern regiondisplayed in Table 6A

The diagonal terms of Table 6A report theown-price elasticities which are in all but onecase negative and highly significant The oneexception is the foreign norfloxacin productgroupmdashwhose share of quinolone sales is 007percentmdashfor which we estimate a negative butinsignificant own-price elasticity For the remain-ing product groups demand appears to be highlyelastic with the estimated elasticities lower than2 in four out of the six cases The magnitude ofthe own-price elasticities matches the features ofthe Indian pharmaceutical market mentioned ear-lier which would suggest that Indian consumersare likely to be quite price-sensitive32 The elas-ticities appear especially large if one takes intoaccount that they refer to product groups (such asdomestic ciprofloxacin) and not individual drugsoffered by particular firms Their relative magni-tudes are also intuitive the drug with the largestmarket share and a relatively high price (domesticciprofloxacin) appears to be one of the least elas-tic In contrast foreign ciprofloxacin is highlyprice elastic this is plausible as ciprofloxacindrugs offered by subsidiaries of multinationalsface the stiffest competition in this market seg-ment from approximately 75 Indian firms offeringthe same molecule

The estimated expenditure elasticities appear inthe last column These are all positive indicatingthat the demand for all product groups is normalThe remaining cells display the estimated cross-price elasticities As one might perhaps expect forproducts within a therapeutic subsegment theseare mostly positive Out of a total of 49 priceelasticities we estimate there are six that do notconform to expectations these are the cross-price elasticities between different foreign prod-

uct groups which are estimated to be negativeand significant33 Fortunately these elasticitieshave negligible impact on the welfare analysisgiven that our counterfactuals focus on theeffect of withdrawing one or more domesticproducts from the market the most relevantelasticities are the ones that capture the responseof various product group shares to a change in theprice of one or more domestic groups these elas-ticities are the cross-price elasticities between var-ious domestic groups and the ones betweendomestic and foreign groups which are plausibleand precisely estimated34

Regarding these elasticities a striking aspectof our estimates is how large positive andsignificant the cross-price elasticities betweendifferent domestic product groups aremdashin factfor norfloxacin and ofloxacin we estimate thatdomestic product groups containing differentmolecules are closer substitutes for one anotherthan product groups that contain the same mol-ecule but are produced by foreign firms Incontrast for ciprofloxacin (the molecule withthe largest revenue share) we estimate a largepositive cross-price elasticity between the do-mestic and foreign versions

The fact that domestic products appear to beclose substitutes for other domestic productsthat contain different molecules truly representsan ldquoempiricalrdquo finding in the sense that we do

32 In developed economies elasticities of this magnitude havetypically been found only for generic drugs (and even then onlyrarely) or among consumers who lack health insurance

33 While these elasticities are clearly counterintuitive theyare not inconsistent with the underlying demand system whichimposes no restrictions on the sign of the cross-price elastici-ties We do not have a good explanation of why these elastic-ities are estimated to be negative A potential explanation isthat the shares of the foreign products are very small giventhis we observe very few consumers switching from one (verysmall) foreign group to another (very small) foreign groupwhen the price of the first foreign group goes up and hence theinference is not very reliable in this case

34 The cross-price elasticities between foreign drug groupscontaining different molecules will also have an effect on thewelfare estimates given that the withdrawal of domestic prod-ucts will generally lead to changes in the prices of all foreignproducts and market shares will be reallocated from someforeign products to other foreign products based on the newprices However this reallocation from ldquoforeign to foreignrdquo istruly second order in our case compared to the reallocationfrom ldquodomestic to domesticrdquo and ldquodomestic to foreignrdquo prod-ucts In scenarios in which we consider patent enforcementaccompanied by strict price regulation the cross-price elastic-ities between foreign product groups are in fact completelyirrelevant as the prices of foreign products are not allowed toincrease in these cases Still our welfare loss estimates remainsubstantial drivenmdashas beforemdashby the loss of product variety

1497VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

not impose it through any of our assumptionsregarding the demand function The questionthat naturally arises then is what might explainthis finding While we cannot formally addressthis question anecdotal accounts in various in-dustry studies suggest that the explanation maylie in the differences between domestic andforeign firms in the structure and coverage ofretail distribution networks

Distribution networks for pharmaceuticals inIndia are typically organized in a hierarchicalfashion Pharmaceutical companies deal mainlywith carrying and forwarding (CampF) agents inmany instances regionally based who each sup-ply a network of stockists (wholesalers) Thesestockists in turn deal with the retail pharma-cists through whom retail sales ultimately oc-cur35 The market share enjoyed by a particularpharmaceutical product therefore depends inpart on the number of retail pharmacists who

stock the product And it is here that thereappears to be a distinction between domesticfirms and multinational subsidiaries In particu-lar the retail reach of domestic firms as agroup tends to be much more comprehensivethan that of multinational subsidiaries (IndianCredit Rating Agency (ICRA) 1999)36

There appear to be two reasons for this Thefirst is that many of the larger Indian firmsbecause they have a much larger portfolio ofproducts over which to spread the associatedfixed costs typically have more extensive net-works of medical representatives The second issimply that there are many more domestic firms(and products) on the market At the retail levelthis would imply that local pharmacists mightbe more likely to stock domestic products con-taining two different molecules say ciprofloxa-cin and norfloxacin than they would domesticand foreign versions of the same molecule Tothe extent that patients (or their doctors) arewilling to substitute across molecules in order tosave on transport or search costs (eg going toanother pharmacy to check whether a particular

35 There are estimated to be some 300000 retail pharma-cists in India On average stockists deal with about 75 retailers(ICRA 1999) There are naturally variations in this structureand a host of specific exclusive dealing and other arrangementsexists in practice Pharmaceutical firms also maintain networksof medical representatives whose main function is to marketthe companyrsquos products to doctors who do the actual prescrib-ing of drugs In some instances firms do sell directly to thedoctors who then become the ldquoretailerrdquo as far as patients areconcerned but these are relatively rare

36 These differences were also highlighted in conversa-tions one of the authors had with CEOs and managingdirectors of several pharmaceutical firms as part of a sep-arate study

TABLE 6AmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN THE NORTHERN REGION

Product group

Elasticity with respect to

Prices of foreign productgroups Prices of domestic product groups Overall

quinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 557 013dagger 015 401 011dagger 011dagger 016 137(179) (007) (007) (184) (006) (006) (006) (029)

Foreign norfloxacin 427dagger 045 427dagger 350dagger 602 451 465 220(242) (112) (242) (210) (623) (184) (183) (105)

Foreign ofloxacin 011 010dagger 138 009 009dagger 023 011 116(005) (005) (031) (027) (005) (028) (004) (017)

Domestic ciprofloxacin 018 001 001 168 008 008 010 117(008) (000) (001) (023) (002) (002) (002) (003)

Domestic norfloxacin 004 003 004 058 223 042 040 073(001) (003) (001) (017) (011) (004) (003) (009)

Domestic ofloxacin 005 005 011 077 074 342 074 089(002) (002) (013) (028) (008) (025) (008) (021)

Domestic sparfloxacin 007 004 007 115 063 063 288 028(002) (001) (002) (015) (006) (006) (017) (012)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent significance level and dagger (dagger) denotes significance at the 10-percent level

1498 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE 6BmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN OTHER REGIONS

Demand patterns in the eastern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 594 014dagger 016 431 013 011 017 140(195) (008) (008) (201) (006) (007) (007) (031)

Foreign norfloxacin 329dagger 058 329dagger 264dagger 460 346 359 192(180) (083) (179) (153) (451) (138) (139) (082)

Foreign ofloxacin 012 010dagger 140 010 010 024 013 117(006) (006) (034) (031) (005) (030) (005) (019)

Domestic ciprofloxacin 018 001 001 172 009 008 011 117(008) (000) (002) (027) (002) (003) (002) (003)

Domestic norfloxacin 004 004 004 068 242 049 046 070(002) (004) (002) (020) (013) (005) (004) (010)

Domestic ofloxacin 004dagger 004 009 061 060 295 060 092(002) (001) (010) (028) (007) (022) (007) (017)

Domestic sparfloxacin 005 003 005 092 048 051 251 043(002) (001) (001) (017) (005) (006) (016) (010)

Demand patterns in the western region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 427 009dagger 011 286 008 006 011 126(129) (005) (005) (133) (004) (005) (004) (021)

Foreign norfloxacin 714dagger 008 710dagger 594dagger 995 742 770 299dagger

(369) (183) (368) (328) (1015) (297) (296) (178)Foreign ofloxacin 013 012dagger 145 009 012 026 013 119

(007) (006) (037) (029) (005) (033) (005) (021)Domestic ciprofloxacin 019 001 000 174 009 008 011 118

(008) (000) (001) (024) (002) (003) (002) (004)Domestic norfloxacin 004 004 004 067 243 050 047 069

(002) (004) (001) (018) (014) (005) (004) (010)Domestic ofloxacin 003 003 007 048 048 257 048 093

(002) (001) (008) (022) (005) (015) (004) (014)Domestic sparfloxacin 006 003 005 083 046 049 241 046

(002) (001) (001) (014) (004) (005) (012) (009)

Demand patterns in the southern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 560 013dagger 015 402 011 010 017 137(182) (007) (007) (187) (006) (006) (007) (029)

Foreign norfloxacin 533dagger 031 531dagger 431dagger 748 559 584 249dagger

(301) (134) (300) (245) (734) (219) (219) (128)Foreign ofloxacin 014 012dagger 148 011 012 029 016 120

(007) (007) (039) (033) (006) (035) (006) (022)Domestic ciprofloxacin 017 001 000 169 008 007 011 116

(008) (000) (001) (025) (002) (003) (002) (003)Domestic norfloxacin 004 003 003 059 225 043 040 073

(001) (003) (001) (019) (012) (004) (004) (009)Domestic ofloxacin 004dagger 004 009 062 060 296 060 091

(002) (001) (010) (026) (006) (019) (007) (017)Domestic sparfloxacin 008 005 008 138 073 075 323 016

(002) (001) (002) (018) (007) (008) (021) (014)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at 5-percent level anddagger (dagger) denotes significance at the 10-percent level

1499VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

foreign product is in stock) in aggregate datawe would expect to find precisely the substitu-tion patterns that we report in Table 6

Whether the particular explanation we provideabove is the correct one the high degree of sub-stitutability between domestic product groupsturns out to have important implications for thewelfare calculations We discuss these in moredetail below when we present the results of thecounterfactual welfare analysis Another elasticitywith important implications for the counterfactu-als is the price elasticity for the quinolone subseg-ment as a whole which indicates how likelyconsumers are to switch to other antibioticsgroups when faced with a price increase for quin-olones This elasticity is computed on the basis ofthe results in Table A3 and it is at 111 (stan-dard error 024) this is large in magnitudebutmdashas expectedmdashsmaller in absolute value thanthe own-price elasticities of the product groupswithin the quinolone subsegment

The results in Tables 6A and 6B are based onour preferred specification discussed in SectionII In Tables A4 to A6 in the Appendix weexperiment with some alternative specificationsTables A4(a)ndashA4(c) correspond to a specificationthat includes in addition to product-group-specificregional fixed effects product-group-specific (andfor the upper level antibiotics-segment-specific)seasonal effects We distinguish among three sea-sonsmdashthe summer monsoon and wintermdashandreport the unconditional demand elasticities for

the northern region for each of these seasons Asevident from the tables in the Appendix our elas-ticity estimates are robust to the inclusion of sea-sonal effects The demand elasticities in Table A5are based on estimation of the demand system byOLS Compared to the elasticities obtained by IVthe OLS elasticities are smaller in absolute valueimplying that welfare calculations based on theOLS estimates would produce larger welfare lossestimates Nevertheless some of the patterns re-garding the cross-price elasticities discussed ear-lier are also evident in the OLS results inparticular the cross-price elasticities between dif-ferent domestic product groups are all positivelarge and significant and in most instances largerthan the cross-price elasticities between drugs thatcontain the same molecule but are produced byfirms of different domesticforeign status Theclose substitutability of domestic products indi-cated by both the OLS and IV estimates seems tobe one of the most robust findings of the paper

B Cost and Markup Estimates

Table 7 displays the marginal costs markupsand profits implied by the price elasticity esti-mates of Tables 6A and 6B for each of the sevenproduct groups Given that our regional effectsimply different price elasticities for each regionour marginal cost and markup estimates alsodiffer by region Given however that based on

TABLE 7mdashUPPER AND LOWER BOUNDS FOR MARGINAL COST MARKUP AND ANNUAL PROFIT BY PRODUCT GROUPS WITHIN

THE QUINOLONE SUBSEGMENT

Product groupLower boundfor MC (Rs)

Upper boundfor markup

Upper boundfor profit(Rs mill)

Upper boundfor MC (Rs)

Lower boundfor markup

Lower boundfor profit

(Rs)

Foreign ciprofloxacin 83 19 269 103 0 00(123) (012) (1655)

Foreign norfloxacin NA NA NA 53 0 00Foreign ofloxacin 323 70 1061 1085 0 00

(2316) (021) (3185)Domestic ciprofloxacin 47 59 17019 112 0 00

(114) (010) (29858)Domestic norfloxacin 52 43 2807 90 0 00

(020) (002) (1532)Domestic ofloxacin 587 34 1612 901 0 00

(218) (002) (1280)Domestic sparfloxacin 495 37 1985 788 0 00

(157) (002) (1100)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level Estimated lower bound forforeign norfloxacinrsquos marginal cost is negative since the estimated price elasticity is less than one in absolute value

1500 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Tables 6A and 6B the price elasticities do notseem to differ substantially across regions wereport for ease of exposition only the nationalaverages for marginal costs and markups inTable 7

Table 7 has two parts In the left part (firstthree columns) we report the lower bound formarginal cost and the corresponding upperbound for markup and upper bound for totalannual profit for each product group Thesenumbers are based on the price elasticities weobtained from estimating the two-level AIDSdemand system Since we do not have a reliableestimate for the price elasticity of foreign nor-floxacin (the point estimate is negative but lessthan one in absolute value and insignificant)we cannot compute the lower bound for mar-ginal cost in this case It is important to note thatthese estimates do not reflect either the actualmarginal cost or the actual markup for thesedrugs both because the existence of price reg-ulation implies that the unconstrained first-orderconditions are unlikely to hold each period andbecause our aggregation across firms of thesame domesticforeign status supplying thesame molecule makes the interpretation of theseestimates problematic In particular the fact thatwe ignore competition among firms within eachproduct group implies that our estimates willtend to overstate market power These numberswill prove useful however in the counterfac-tual simulations as they can give us a sense ofhow large the maximum profit gains for multi-national firms and the maximum profit lossesfor domestic firms are likely to be under patentenforcement The right part of the table (lastthree columns) reports the upper bound for mar-ginal cost which we obtain by simply taking themarginal cost to be equal to the observed priceWe do not report standard errors in this casesince the numbers are based on actual data Themarkups corresponding to these marginal costupper bounds are of course zero We conductthe counterfactual simulations using both thelower and upper bounds for marginal costs

The (upper bounds for) markups on the leftside of the table are generally plausible Thedomestic ciprofloxacin product group that dom-inates the quinolone subsegment and for whichwe documented high prices a high marketshare and a relatively low elasticity of demandenjoys one of the highest markup upper bounds(60 percent) and accounts for nearly 70 percent

of all profits derived within the subsegmentForeign ciprofloxacin on the other hand whichfaces the stiffest competition from domesticfirms and for which we estimated a highly elas-tic demand has the lowest markup upper bound(19 percent)

C Counterfactual Estimates of the Impact onPrices and Welfare

With estimates of the key demand and costparameters in hand we turn to the counterfac-tuals We consider the five separate scenarioslisted in the previous section All of the scenar-ios involve the withdrawal of one or more of thedomestic product groups from the market Ta-ble 8 displays our estimates of the consumerwelfare losses that result under the differentscenarios The losses are expressed in billion ofrupees per year All numbers presented in Ta-ble 8 and subsequent tables are based on usingthe lower bounds for marginal cost and upperbounds for markup in the simulations As dis-cussed earlier these numbers are the more in-teresting to work with since they give us anupper bound for the changes in the profits ofdomestic and foreign firms that would resultfrom patent enforcement In the Appendix wealso present results based on using the upperbounds for marginal costs in the simulations inwhich case the pre-TRIPS profits of domesticand foreign firms are zero In all cases marginalcosts are assumed to be constant in outputWhile naturally the profit implications differdepending on whether one uses the upper orlower bounds for marginal costs (firm profits arezero if one assumes the upper bound of mar-ginal cost in which case price equals marginalcost) the estimated consumer welfare losses aresimilar in the two cases We discuss these re-sults in more detail at the end of this section

The first column presents our estimates of theconsumer welfare losses attributable to the pureloss of product variety effect where we fix theprices of all remaining products as well as theoverall expenditure on quinolones while with-drawing one or more of the domestic productgroups Note that in our initial specification ofthe demand system had we not allowed for thepossibility that consumers might differentiatebetween domestic and foreign products evenwhen they contain the same molecule this

1501VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

particular component of the loss of consumerwelfare would not have arisen

The estimates reported in the second columnincorporate the expenditure-switching effect ontop of the loss of product variety Here basedupon the price elasticity estimates from thehigher-level AIDS system we adjust (down-ward) the expenditures allocated to the quino-lone subsegment as the composite price ofquinolones effectively increases as a consequenceof the higher virtual prices of the domestic productgroups that are withdrawn from the market Be-cause the estimates in this column are generatedassuming that the prices of the products that re-main in the market are not adjusted upward theyprovide a sense of what consumer losses would beif the introduction of product patents were coupledwith strict price regulation aimed at maintainingprices at prepatent levels Alternatively they canbe thought of as the relevant welfare numbers ifintense competition among firms within the re-maining product groups kept the prices of theproducts that were still offered in the market closeto the firmsrsquo marginal costs

The last column of Table 8 displays the esti-mated consumer welfare losses when bothcross-segment expenditure-switching and with-in-segment upward price adjustments are takeninto account

If we compare the results across the firstsecond and third columns all the counterfac-tual scenarios produce qualitatively similar pat-terns patterns that are consistent with what we

would expect Starting from the initial loss ofwelfare attributable to the loss of product vari-ety the option of switching expenditures outof the quinolone subsegment to other subseg-ments mitigates some of the initial welfareloss But if we then incorporate the upwardprice adjustments that result in response to thereduced competition the welfare losses aremagnified

Of particular interest from a policy perspec-tive are the relative magnitudes of these threeeffects which are similar under all the counter-factual scenarios though absolute levels varyconsiderably First despite the fact that the de-mand for quinolones is quite sensitive to thecomposite price of quinolonesmdashthe upper levelprice elasticity is 111mdashthe cross-subsegmentexpenditure-switching effects are in all thecases small (in absolute value terms) relative tothe other two effects For instance under thescenario where all the domestic quinolone prod-uct groups are withdrawn from the market theoverall consumer welfare loss of Rs 1781 bil-lion per year can be decomposed into an initialloss of Rs 1176 billion (66 percent) attribut-able to the loss of product variety a slightreduction in this initial loss of Rs 041 billion(2 percent) from Rs 1176 billion to Rs1135 billion because of expenditure switchingand a subsequent additional loss of Rs 646billion (36 percent) from Rs 1135 billion toRs 1781 billion because of the reduced com-petition and consequent price increases

TABLE 8mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure loss ofvariety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 732(087) (089) (146)

Only ofloxacin 008 008 023(008) (008) (010)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1253(177) (180) (415)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1058(142) (145) (331)

All four domestic quinolones products 1176dagger 1135dagger 1781(643) (634) (1270)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent significance level and dagger (dagger)denotes significance at the 10-percent level

1502 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 5: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

ios in each of which only one of the domesticproduct groups is withdrawn What this veryclearly indicates is that past studies that haveestimated the aggregate effects of patent protec-tion by adding up the losses estimated sepa-rately in each of a number of patentablemarkets may have substantially underestimatedthe magnitude of the consumer welfare lossesfrom the introduction of pharmaceutical productpatents

In absolute terms we estimate that in theabsence of any price regulation the prices offoreign patented products would rise between100 percent and 400 percent In the more real-istic case of some form of price regulation thatwould keep drug prices fixed at their pre-TRIPSlevel the total annual welfare losses to theIndian economy from the withdrawal of all fourdomestic product groups in the quinolone sub-segment would be on the order of Rs 1370billion or about 50 percent of the sales of theentire systemic antibacterials segment in 2000At the then-prevailing exchange rate this trans-lates into a figure of US$305 million Of thisamount foregone profits of domestic producersconstitute roughly Rs 23 billion or US$50 mil-lion The overwhelming portion of the totalwelfare loss therefore derives from the loss ofconsumer welfare

The welfare loss we estimate represents onlythe static costs of patent enforcement arisingfrom pricing distortions and reduction in prod-uct variety Our approach does not address thepotential dynamic benefits of innovations thatmay result from international property rightsprotection Nevertheless we believe that esti-mating these static costs is important wheneverthere is a radical change in policymdashwhichTRIPS represents for a good part of the devel-oping world Even if there is the potential forlong-term benefits knowledge of the short-runcosts is important for designing an appropriatepolicy response that will potentially mitigate theadverse short-run impact Having said that itis worth noting that according to our esti-mates the total profit gains of patent enforce-ment to foreign producers in the absence ofany price regulation would be only aboutUS$53 million per year With price regulationthat would keep the prices of drugs suppliedby multinational subsidiaries at their pre-TRIPS level the profit gains become onlyUS$196 million per year

The remainder of this paper is organized asfollows In the next section we lay out theessential features of the Indian pharmaceuticalsmarket provide more detail about the segmentsthat we focus on in the empirical analysis andbriefly describe the primary data we use Sec-tion II describes the analytic framework and theeconometric strategy we use to estimate therelevant parameters and construct the counter-factual scenarios We discuss our results in Sec-tion III Section IV concludes

I The Setting and the Data

Between April 1972 when the Indian PatentsAct (1970) became effective and March 2005when Indiarsquos parliament passed the 3rd Amend-ment of the Patents Act India did not recognizeproduct patents for pharmaceuticals The IndianPatents Act which replaced the inherited Brit-ish colonial law regarding intellectual propertyrights specifically excluded pharmaceuticalproduct patents and admitted process patentsonly for a period of seven years In contrast thelatest amendment recognizes patents on endproducts that under the new regime will remainin force for 20 years6

The two stated objectives of the 1970 actwere the development of an indigenous pharma-ceuticals industry and the provision of low-costaccess to medicines for Indian consumers Con-sistent with these objectives and with the broaderleftward tilt in policy a number of other measureswere introducedmdashdrug price controls restrictionson capacity expansion limits on multinational eq-uity sharesmdashthat in the years since have on theone hand kept pharmaceutical prices low and onthe other encouraged the development of the In-dian pharmaceutical industry Many of these reg-ulations and restrictions have been lifted or easedsince the mid-1980s with marked acceleration inthe pace of liberalization during the 1990s

Over the last 20 years the Indian pharmaceu-tical industry has grown rapidly to the pointwhere it is now the worldrsquos largest producer offormulations in terms of volume and one of the

6 Indian companies that are now producing drugs forwhich patent applications were submitted between the sign-ing of the TRIPS agreement in 1995 and January 1 2005will be allowed to continue producing if they pay a royaltyto the patent holder

1481VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

worldrsquos largest producers of bulk drugs7 Thestructure of the industry has also evolved In1970 the industry was dominated by multina-tional subsidiaries by 2001 Indian-owned firmswere not just the leading players in the industrymany had also become major exporters

The data we use in this paper are from theretail pharmaceutical audits of ORG-MARGIndiarsquos premier market research and consultingfirm The audit provides detailed product-levelinformationmdashestimates of monthly retail salesin each of the four geographic zones of Indiaprice dosage form launch date brand namechemical name therapeutic categorizationmdashon

all pharmaceutical products sold in India byabout 300 of the largest firms representingroughly 90 percent of domestic retail sales ofpharmaceuticals The coverage of the audit isextensive reaching a representative panel ofthousands of retail chemists in over 350 citiesand towns The data collected which providethe only real source of disaggregate informationon the Indian pharmaceutical market are usedby both the government of India in formulatingpricing policy and other decisions and the In-dian pharmaceutical industry in determiningpricing and marketing strategies We have in-formation at a monthly frequency for the periodof January 1999 to December 2000 Tables 1 to3 provide a set of descriptive statistics that areessential for understanding the focus of ouranalysis and interpreting our results

As noted earlier the characteristics of de-mand for pharmaceuticals in India are likely todiffer considerably from those in developedeconomies With a share of 23 percent the

7 Bulk drugs are the therapeutically relevant active phar-maceutical ingredients that are combined with a variety ofinactive ingredients to make the formulations ultimatelyconsumed by patients Firms in the pharmaceutical sectorcan be one of three types bulk drugs producers pureformulators or integrated firms which produce both bulkdrugs and market formulations

TABLE 1mdashTHE QUINOLONES SUBSEGMENT

Molecule

Shares ()of quinolones sales Sales (Rs mill) 2000

Domesticfirms

Foreignsubsidiaries

Domesticfirms

Foreignsubsidiaries

Ciprofloxacin 530 27 3030 156Norfloxacin 112 01 640 3Ofloxacin 116 31 665 177Sparfloxacin 108 01 620 4Lomefloxacin 15 86 Pefloxacin 13 01 72 5Levofloxacin 00 0 Nalidixic acid 13 73

TABLE 2mdashBASIC INFORMATION ABOUT THE FOUR QUINOLONES MOLECULES 2000

Ciprofloxacin Norfloxacin Ofloxacin Sparfloxacin

US or European patent holder Bayer Merck Ortho-McNeil Rhone-PoulencYear of US patent expiry 2003 1998 2003 2010Year of US-FDA approval 1987 1986 1990 1996Year first introduced in India 1989 1988 1990 1996No of domestic Indian firms 75 40 17 25No of foreign subsidiaries 8 2 2 1Sales weighted average per-unit API

price of products produced byDomestic Indian firms 1124 905 9008 7884Foreign subsidiaries 1035 528 10847

API Active pharmaceutical ingredient

1482 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

antiinfectives segment ranks second in Indiawhereas in the world market it is fifth and has ashare of only 90 percent Hence antiinfectivesare important in India not only from a health

and public policy point of view but also as asource of firm revenue

With this in mind we focus on one partic-ular subsegment of antiinfectives namely the

TABLE 3mdashSUMMARY STATISTICS FOR THE QUINOLONES SUBSEGMENT 1999ndash2000

North East West South

Annual quinolones expenditure per household (Rs) 3125 1975 2764 2359(366) (367) (407) (286)

Annual antibiotics expenditure per household (Rs) 11988 8424 11052 9624(1224) (1224) (960) (996)

No of SKUsForeign ciprofloxacin 1238 1129 1308 1246

(150) (190) (102) (106)Foreign norfloxacin 183 171 200 158

(070) (075) (088) (083)Foreign ofloxacin 304 296 296 300

(086) (086) (091) (088)Domestic ciprofloxacin 10621 9763 10342 10550

(599) (434) (722) (451)Domestic norfloxacin 3896 3496 3617 3942

(271) (268) (251) (379)Domestic ofloxacin 1846 1600 1725 1725

(680) (634) (586) (635)Domestic sparfloxacin 2983 2829 3121 2929

(557) (638) (688) (657)Price per-unit API (Rs)

Foreign ciprofloxacin 958 1090 1085 1007(128) (066) (071) (058)

Foreign norfloxacin 563 509 605 435(077) (133) (139) (147)

Foreign ofloxacin 10946 10943 10886 10612(620) (664) (700) (1140)

Domestic ciprofloxacin 1143 1067 1131 1152(016) (015) (017) (013)

Domestic norfloxacin 951 907 888 873(024) (035) (037) (020)

Domestic ofloxacin 9163 8964 8565 9341(1615) (1565) (1422) (1407)

Domestic sparfloxacin 7972 7849 7688 8028(976) (1014) (1185) (1037)

Annual sales (Rs mill)Foreign ciprofloxacin 4179 2431 4520 2947

(1534) (816) (1273) (648)Foreign norfloxacin 128 100 058 073

(101) (082) (044) (057)Foreign ofloxacin 5446 3184 3522 3111

(1399) (933) (906) (703)Domestic ciprofloxacin 96229 58591 67874 70381

(10626) (13026) (12226) (8740)Domestic norfloxacin 22255 11971 14918 15829

(3884) (1945) (2691) (1626)Domestic ofloxacin 12502 9621 14936 11205

(4434) (3011) (5282) (4259)Domestic sparfloxacin 15617 12175 16130 9811

(3141) (2576) (4674) (3420)

Note Standard deviations in parentheses API Active pharmaceutical ingredient

1483VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

quinolone subsegment Quinolones fall into thesystemic antibiotics and antibacterials segmentof the Indian pharmaceuticals market whichgenerates over three-quarters of the revenues inthe antiinfectives segment8 The systemic anti-bacterials segment includes all of the originalmiracle drugs that first sparked the growth ofthe global research-based pharmaceutical indus-try in the postndashWorld War II period as well aslater generations of molecules that have beenintroduced in the last four decades

Among systemic antibacterials quinolonesare the latest generation molecules available inIndia We focus our analysis on quinolones forseveral reasons First quinolones are the drug ofchoice for a large number of bacterial infec-tions some of which are also treated by alter-native drugs (see Table A1 in the Appendixwhich outlines the spectrum of activity for eachmolecule family within the antibacterials seg-ment) Hence if there were one product groupfor which we would expect to have many sub-stitutes readily available this would be quino-lones Second with a share of 20 percent in thesales of systemic antibacterials quinolones rep-resent one of the largest subsegments within thistherapeutic category Finally several moleculeswithin the quinolone subsegment were still un-der patent in the United States at the time of ourinvestigation Table 2 details the basic informa-tion about the four quinolone molecules that arethe focus of our analysis The first row showsthe year of US patent expiry this ranges from1998 for norfloxacin to 2010 for sparfloxacinQuinolones include in principle four moremolecules that are listed at the bottom of Ta-ble 1 the market shares of these molecules arenegligible however so that we exclude thesemolecules from our analysis

Table 2 reveals several other interesting factsabout competition in the quinolone market inIndia First note the large number of firmsoperating in this subsegment The large numberof domestic firms is perhaps not that surprisinggiven that pharmaceutical product patents werenot recognized in India9 What is more surpris-ing is the number of foreign firms selling pat-ented products (eg ciprofloxacin) the fact that

multiple foreign firms sell a patented productindicates that such firms often ldquoinfringerdquo10

patent laws in India while complying with themin developed world countries The last two rowsof Table 2 further indicate that domestic prod-ucts often sell at a premium With the exceptionof ofloxacin the average prices of products of-fered by Indian firms are higher than the pricesof products offered by foreign subsidiaries Thispreliminary evidence suggests that Indian con-sumers do not place a premium on the brandname and reputation of big multinational phar-maceutical concerns Moreover the higherprice of domestic products does not seem toprevent domestic companies from capturing alarge market share This is most evident in thecase of ciprofloxacin where domestic firmshave with 53 percent the largest share in thetotal sales of quinolonesmdashthis despite the factthat the average price of these products is 10percent higher than the price of foreign productscontaining the same molecule

Table 3 provides additional summary statis-tics for our data broken down by region Thefirst two rows of the table report the averageannual household expenditure on quinolonesand antibiotics respectively Note that in bothcases the average expenditure is higher in thenorth and west these regions include states withhigher per capita incomes and tend to be moreindustrialized and urbanized than those in theeast and south Pharmaceutical products areavailable in multiple presentations that is com-binations of dosage forms (eg capsule tabletsyrup) strength (eg 100 milligrams 500 mil-ligrams) and packet sizes (eg 50 capsule bot-tle 100 tablet bottle) The various presentationsin which a product is available are often referredto as stock-keeping units or SKUs11 The num-ber of SKUs for each product group withinquinolones is reported at the top of Table 3 Aswith the more aggregate numbers on firms and

8 In addition to anti-bacterials this segment also containsantivirals

9 Accordingly the common distinction between ldquobrandedrdquoand ldquogenericrdquo products is irrelevant here

10 We emphasize here that the word ldquoinfringerdquo belongsin quotes because patent laws do not currently exist inIndia infringement in the legal sense is not possible It ishowever striking that the same firms that accuse Indianproducers of ldquopiracyrdquo sell in India products that are patentedin the United States and for which the patent is held by adifferent multinational corporation

11 For instance a 100 capsule bottle of 100 milligramcapsules of a particular branded drug and a 50 capsulebottle of 100 milligram capsules of the same branded drugwould be identified as two separate SKUs

1484 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

products reported in Table 2 the differencebetween domestic and foreign products isstriking The number of SKUs offered byIndian firms is consistently larger than thenumber offered by subsidiaries of foreignmultinationals The number of SKUs variesslightly across regions but more importantlyit varies across time as some SKUs disap-pear while new ones get introduced duringour sample period

Many pharmaceutical products in India aresubject to price controls12 While the specificsof the price regulation are too complex for anyeconomic model to capture adequately themain concern for the empirical analysis is thatprice controls may lead to a lack of price vari-ation over time so that the demand functioncannot be identified Prices at the most disag-gregate SKU level are relatively stable overtime there are variations due to occasionalchanges in the estimated cost (due for exampleto changes in exchange rates that affect the costof imported materials or bulk drugs) but suchvariations tend to be infrequent and small inmagnitude The degree of time variation ishowever substantially larger once one aggre-gates to the product level This variability stemsnot only from the fact that the SKUs over whichwe aggregate may experience changes in theirrespective prices at different points in time butalso from the fact that the range of SKUs of-fered in the market does not remain constantover time The entry and exit of presentationswithin the same product group that have differ-

ent prices effectively affects the price that con-sumers face for this drug in each period

The middle portion of Table 3 reports themean price and standard deviation for eachproduct group by region Prices vary by regionthough there is no clear pattern emerging fromthe table with regard to the cross-regional vari-ation (in the sense of some regions being sys-tematically more expensive than others) Toexamine what portion of the total price variationis due to time versus regional variation weconducted an analysis of variance of prices thatwe report in Table 4 The table is based onseparate regressions for each product group(pooling data across groups with big differencesin their average prices is not particularly infor-mative as most of the price variation is ac-counted for by product group dummies) Thelast two columns of the table show the fractionof price variation accounted for by region andtime dummies respectively As evident fromthe table a significant fraction of the total vari-ance in prices can be attributed to time varia-tion In the demand estimation we include a fullset of product groupndashspecific regional dummiesso that the price parameters are identified en-tirely based on this time variation within eachproduct group The time variation of productgroup prices is driven primarily by composi-tional changes within each group the revenueshares of the individual SKUs that compriseeach product group change over time (see therelated discussion in Section IIC) while in eachperiod there is entry and exit of SKUs into thesample To check whether this pattern reflectsgenuine entry and exit as opposed to samplingvariation we examined the revenue shares ofthe SKUs that leave the sample relative to theones that remain during the entire period Theresults are reported in Table 5 While the SKUs

12 The details of the procedures for price fixation canbe found in the official government Web site httpwwwnppaindianicinindex1html under the link ldquoDrug Price Con-trol Order 1995rdquo A new pharmaceutical policy was introducedin 2002 but our data were collected before that year

TABLE 4mdashANALYSIS OF PRODUCT PRICE VARIANCE

Product groupPartial SS

timePartial SS

region Total SSPercentage

explained by timePercentage

explained by region

Foreign ciprofloxacin 0296 0306 1047 283 292Foreign norfloxacin 1036 1002 4179 248 240Foreign ofloxacin 0429 0021 0571 752 36Domestic ciprofloxacin 0005 0088 0104 50 841Domestic norfloxacin 0059 0098 0198 296 494Domestic ofloxacin 2858 0104 3056 935 34Domestic sparfloxacin 1754 0032 1860 943 17

1485VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

that exit tend to be smaller (their average shareis 1 percent as opposed to 34 percent for thoseSKUs that are present during the entire sampleperiod) the shares of the two groups do notseem orders of magnitude apart13 In additionour data cover only the 300 largest firms sellingin the Indian market so that firms with verysmall shares are not included in our sample

II The Analytic Framework and EstimationApproach

Patent enforcement in the Indian pharmaceu-tical market will have the effect of eliminatingdomestic products whose active pharmaceuticalingredients are protected by (foreign) patentsThus assessing the effects of patent enforce-ment is tantamount to assessing the effects ofwithdrawing domestic products from the mar-ket This task is the converse of evaluating newproduct introduction accordingly the concep-tual framework we use to address the questionsof interest is similar to the one developed in theliterature for the valuation of new goods14

We start by estimating demand for quino-lones Given that the market is characterized byimperfect competition the counterfactual anal-ysis requires that we also model the supply sideas removal of one product will affect the pricesof other products especially those that are closecompetitors The existence of price regulationin the Indian pharmaceutical market imposespotential constraints on firmsrsquo maximizationproblem Given these constraints and the com-

plexity of the price regulation process the typ-ical approach of deriving estimates of actualmarginal costs and markups by exploiting thefirst-order conditions of profit maximizing firmsdoes not seem particularly promising Insteadwe use our demand estimates to place upper andlower bounds on marginal costs and markups

With demand elasticities and upper and lowerbounds for marginal costs in place we thenconduct counterfactual simulations We con-sider several alternative scenarios dependingon the number of domestic products affected bypatent enforcement For each scenario we com-pute the counterfactual prices and use them toassess the effects of domestic product with-drawal on consumer welfare (as measured bythe compensating variation) firm profits andsocial welfare As with the valuation of newproducts the big conceptual problem facing thispart of the analysis is that we need to extrapo-late from the region of the data to the point atwhich demand for the products that exit themarket becomes zero This conceptual issue ispresent in any attempt to evaluate a major pol-icy change for which no historical precedentexists like the enforcement of patent laws inIndia One advantage of the present study is thatwe have a limited set of price data for Pakistana country with similar demographics to Indiabut with a market structure that resembles theone that would emerge in India under patentenforcement (monopoly of multinational sub-sidiaries) By comparing the prices of productsoffered by multinationals in Pakistan to thosewe compute in our counterfactual simulationsfor the products that would be offered by mul-tinationals in India if patent laws were enforcedwe can get a sense of how plausible our coun-terfactual estimates are

A Demand

The demand modelling is based on the multi-stage budgeting approach Our primary motiva-tion for adopting this approach is a practical oneIn the multistage budgeting approach the depen-dent variable is defined as a revenue share whichis appealing given that the products we include inthe analysis contain different molecules (ie ac-tive pharmaceutical ingredients or APIs) Eventhough we do have data on the quantity of therelevant API (eg 100 milligrams of ciprofloxa-cin) contained in each product converting the

13 While we cannot completely rule out the possibilitythat some of the exit is due to sampling variation note thatthe latter should be reflected in low precision of the demandparameter estimates The demand parameters however areprecisely estimated

14 See Manuel Trajtenberg (1989) Jerry A Hausman(1994) and Timothy F Bresnahan (1997) for representativeexamples and a discussion of the relevant issues

TABLE 5mdashREVENUE SHARES OF THE EXITING SKUS

Revenue share ofexiting SKUs

Revenue shares ofall SKUs

Full sample 10 34Northern region 08 33Eastern region 12 36Western region 10 34Southern region 12 33

1486 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

revenue shares to physical shares is extremelydifficult if not infeasible in the case of antibioticsBecause such drugs are ldquosystemicrdquo by nature theyare used to treat a large number of infections andthe dosage of each drug depends on the particularinfection it is supposed to address (for example thedosage will differ depending on whether the anti-biotic is used to combat an ear infection or tuber-culosis) This particular feature of antibioticscomplicates the conversion of revenue to physicalshares

The basic idea of the multistage budgetingapproach is to use the therapeutic classificationof a productmdashie the therapeutic segment andsubsegment the product belongs tomdashto orga-nize all products in the systemic antibacterialssegment into a hierarchical taxonomy consist-ing of two levels At the higher level are thevarious subsegments of systemic antibacterialsThe first stage of budgeting corresponds to theallocation of expenditures across the subseg-ments in this upper level of the taxonomy

In the second stage of the budgeting processcorresponding to the lower level of the taxon-omy a flexible functional form is adopted tomodel how the expenditures allocated to eachsubsegment are distributed across the productswithin that subsegment In particular to modeldemand at the second stage we employ theAIDS specification proposed by Deaton andMuellbauer (1980b)15

While the two-stage demand estimation ap-proach offers functional form flexibility its ap-plication to the Indian systemic antibioticsmarket poses a few problems The first is thatdue to entry and exit many SKUs and evenproducts in our sample are not present in everyperiod AIDS does not have a good way ofdealing with a varying number of products as itwas developed with broad commodity catego-ries in mind which are consumed by all con-sumers every period To solve this problem weaggregate within each subsegment (eg quino-lones) SKUs into product groups where withineach product group all SKUs contain the samemolecule and are produced by firms with thesame domesticforeign status Specifically let aSKU k be indexed by its molecule (or API) M its

domesticforeign status DF indicating whether itis produced by a domestic (Indian) or a subsidiaryof a foreign (multinational) firm a particular pre-sentation s and the particular firm f that producesit We aggregate SKUs over presentations andfirms to obtain a newly defined product group iwhich is indexed only by molecule M and domes-ticforeign status DF and has revenue Ri yenfsRk with i (M DF) k (M DF f s) and pricepi yenfs kpk where k denotes the conditional(on M and DF) revenue share of this particularproduct ie

(1) k Rk

Ri

In most cases the resulting product groups arebroad enough to be present every period16

The usual concern with this aggregation pro-cedure is that it may lead us to overstate firmsrsquomarket power as we ignore competition amongfirms with the same domesticforeign statusproducing the same molecule In the presentapplication however this concern is unlikely tobe of great importance as the effect of patentenforcement is to wipe out all domestic compe-tition at once while granting foreign firms mo-nopoly power hence competition among firmsfor patented molecules becomes irrelevant Theaggregation according to the domesticforeignstatus (within a particular molecule) thus corre-sponds to the scope of our analysis and theparticular questions of interest

The second problem is that for our approachto be useful in welfare analysis the allocation oftotal expenditures to group expenditures at thehigher stage has to be modelled in a way con-sistent with utility maximization In general thesolution of this allocation problem requiresknowledge of all individual product prices

15 Representative applications of the multistage budget-ing approach include Sara F Ellison et al (1997) Hausman(1994) and Hausman and Gregory K Leonard (2002)

16 We are missing only four observations (ie monthregion combinations) all for the drug group of ForeignNorfloxacin August 1999 in the south May 2000 in thewest October 2000 in the south and November 2000 in theeast In these cases we set the revenue shares of ForeignNorfloxacin equal to zero In general with 01 percent ofquinolone sales (see Table 1 row 2) Foreign Norfloxacinhas a very small share of the market This probably explainswhy the results pertaining to this drug group are unreliableit is the only drug group for which we do not obtain asignificant price elasticity of demand while its cross-priceelasticities with other foreign drug groups often have thewrong sign

1487VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

From an empirical point of view this is notparticularly useful as it eliminates all compu-tational advantages of the two-stage approachTo address this problem we adopt an approxi-mate solution to model the higher-level expen-diture allocation along the lines suggested byDeaton and Muellbauer (1980b pp 131ndash32)This gives rise to a two-level AIDS specification

Consider the lower-level estimation firstwhich refers to the allocation of a particularsubsegmentrsquos expenditure to the product groupswithin the subsegment In our application therelevant subsegment is quinolones which weindex with Q Let the product groups within thissubsegment be indexed by i 1 N pi be theprice of product group i (where as noted abovei refers to a particular molecule and domesticforeign status combination) and XQ be the totalexpenditure on the quinolone segment The rev-enue share of each product group is given by

(2) i i j

ijln pj ilnXQ

PQ

where i the revenue share of product group iis defined as

(3) i pi qi

yenj

pj qj

xi

XQ with i j Q

XQ is the overall expenditure on the quinolonesubsegment and PQ is a price index given by

(4) ln PQ ap 0 i

iln pi

1

2 i

j

ijln piln pj

With a limited number of product groups and asufficiently large number of time-series observa-tions the flexibility implied by the AIDS modeldoes not impose too many demands on the data Inthe present application however where the num-ber of observations is limited the AIDS model isnot estimable in this general form To reduce thenumber of parameters that need to be estimatedwe impose two sets of restrictions

The first set of restrictions is implied by thetheory of utility maximization Specificallythese restrictions are

Adding-up yenk k 1 yenk k 0 yenk kj 0 j

Homogeneity yenk jk 0 j Symmetry ij 1frasl2 [ij ji] ji This last

restriction by itself reduces the number of parameters to [N(N 1)]2

The second set of restrictions we impose aimsat further reducing the number of parameters tobe estimated by exploiting our knowledge of thisparticular market Specifically for each productgroup i we allow one ij parameter for all productgroups j that have different molecules from prod-uct group i and are produced by foreign firms andone ij for all product groups j with differentmolecules produced by domestic firms We donrsquotimpose any restrictions on the ij parameter whenproduct group j has the same molecule as productgroup i (By construction product groups i and jcontain products produced by firms with differentdomesticforeign status)

To better illustrate the nature of the restric-tions we impose on the patterns of substitutionacross products some additional notation isneeded Let d(i j) be an indicator of the degreeof similarity (or difference) between productgroup i and product group j along the dimen-sions we are able to observe (molecule M anddomesticforeign status DF) For any two prod-uct groups i and j d(i j) can take on one of thefollowing three values17

(5)

di j (1 0) if Mi Mj DFi DFj

(0 1) if Mi Mj DFi DFj

(0 0) if Mi Mj DFi DFj

Let

(6) Diab j di j a b

17 The sequence (1 1) is not possible for two differentproducts in this case the parameter corresponds to theproductrsquos own-price effect that is ii

1488 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The equation at the lower level becomes

(7) i i iiln pi i10ln pj j Di10

j Di

01

i01ln pj

j Di

00

i00ln pj ilnXQ

PQ

Note that

The parameter ii captures a product grouprsquosown-price effect (there will be as many iiparameters as the number of product groups)

The parameter i10 captures the cross-priceeffect of the product group containing prod-ucts with the same molecule but produced byfirms of different nationality

The parameter i01 captures the cross-priceeffects of product groups containing productswith different molecules but produced byfirms with the same nationality

The parameter i00 captures the cross-priceeffects of product groups containing productswith different molecules produced by firms ofdifferent nationality

Before we take the demand equation to thedata we make two modifications The first oneis to let the product-specific effects i vary byregion r The resulting product-specific regionaleffects ir have two interpretations first theycontrol for the ldquoqualityrdquo of each drug withquality differences being allowed to vary acrossregions second they proxy for demographicsand other demand shifters which vary by re-gion and may affect the demand of each prod-uct group differently18 Note that by includingproduct-specific regional effects in the demandspecification we estimate the price parametersbased on the within-product-group variation ofprices in each region In an earlier version of thepaper we also estimated the demand systemwithout regional dummies and obtained similarresults

The second modification that allows us to gofrom a deterministic to a stochastic specificationof the demand equation is to include an additiveerror term in (7) The latter takes into accountthe fact that (7) is not expected to fit the dataexactly The error term irt accounts for mea-surement error (due to the fact that the productgroup prices pjrt we employ in the estimationare not exact price indices but approximationsthereof) and (potentially region-specific) de-mand shocks that may affect the demand for aproduct in a particular period Examples of suchshocks include an advertising campaign for aparticular product which temporarily increasesthe demand for this product or the outbreak ofa (potentially region-specific) epidemic whichcalls for the use of a particular drug We discussthe interpretation and properties of this errorterm in more detail in the next section

The final form of the equation we estimate atthe lower level becomes (with subscript t de-noting month and subscript r denoting region)

(8) irt i ir iiln pirt i10ln pjrt j Di10

j Di

01

i01ln pjrt

j Di

00

i00ln pjrt

ilnXQrt

PQrt irt

The analysis so far has conditioned on the ex-penditure allocated to the quinolone subseg-ment XQ The upper level of the estimationconsiders the problem of allocating total expen-diture across the different systemic antibioticssubsegments one of which is quinolones Theupper-level demand function is given by

(9) G G H

GHln PH GlnX

P

where all variables denoted by capital letters aredefined as before but now refer to subsegments(G H ) rather than product groups within asubsegment and the total expenditures on sys-temic antibiotics X are deflated by the Stone priceindex log P yenH Hlog PH When estimating thesystem above we impose all the restrictions

18 Given the short time span of our sample typicaldemand shifters such as age distribution income distribu-tion and education hardly change over our sample periodSuch shifters are therefore absorbed by the region-specificproduct fixed effects ir

1489VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

implied by utility maximization as we do withthe estimation of the lower-level AIDS We donot however impose any additional restrictionson the substitution patterns at this stage so thatthe cross-price effects across segments remainrelatively unconstrained

Estimation of the higher-level AIDS allowsus to obtain the unconditional own- and cross-price elasticities that are used in the formulationof the supply problem and welfare analysisThese are given by the formula

(10) ij ijXQXQ

ln qi

ln XQ

ln XQ

ln PQ

ln PQ

ln pj

with the conditional cross price elasticitiesgiven by

ijXQXQ ij ij jlnXQ

PQ

i

As in the lower stage we include subsegment-specific regional dummies in the specificationof the upper-stage demand system so that thefinal form of the estimating equation at theupper stage becomes

(11) Grt G Gr H

GHln PHrt

GlnXrt

Prt Grt

In sum the demand system we take to thedata is represented by equations (8) and (11)and the associated parameter restrictions im-plied by economic theory

B Modelling the Supply Side of the Market

Counterfactual simulations concerning theeffects of domestic product withdrawal requireknowledge of the marginal costs of pharmaceu-tical firms operating in the Indian market Theseare unobservable The usual approach in theNew Empirical Industrial Organization litera-ture has been to exploit the firm equilibriumconditions to infer marginal cost For exampleit is often assumed that the marginal cost ci isconstant and that the industry is an oligopoly

engaging in Bertrand competition with differen-tiated products Assuming that firms myopicallymaximize profits each period one can then de-rive the firmsrsquo first-order conditions that corre-spond to the assumptions above about costsmarket structure and firm behavior

We deviate from this procedure as the presenceof price regulation renders the assumption of un-constrained period-by-period maximization un-tenable Ideally one would like to explicitlyincorporate the price and other administrative con-trols into the firmrsquos optimization problem andderive the first-order conditions under the assump-tion of constrained maximization The complexityof the price regulation however makes this ap-proach infeasible Therefore we adopt an alterna-tive approach that does not rely on modelling theprice-setting process but is instead based on de-riving upper and lower bounds for marginal costsand markups

In particular an upper bound for marginalcosts and a lower bound for the markups (zero)can be derived under the assumption of perfectcompetition While this assumption is clearlyunrealistic in the pharmaceuticals market it isuseful in providing an upper bound for costs ci

Uwhich will be given by

(12) ciU pi

On the other hand a lower bound for marginalcosts ci

L (and upper bound for markups) can bederived by assuming that there is perfect collu-sion within each product group i (where i refershere to a moleculedomestic-foreign combina-tion)19 and ignoring price controls so thatprices are determined by the first-order condi-tion of the jointly profit-maximizing firmswithin product group i Solving this first-ordercondition for the (lower bound of) marginal costthen gives

(13) ciL pi 1

1

iipipj

19 The reason that this assumption will lead to an under-statement of marginal costs (and hence overstatement ofmarket power) is that it assumes away competition amongfirms within each drug group

1490 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

where ii(pipj) the own-price elasticity of de-

mand for product i will depend on the productrsquosown-price pi and all other productsrsquo prices pjwith i j

Once we have obtained the demand elastic-ities through estimation of the demand sys-tem we can calculate the upper and lowerbounds for marginal costs and correspondingmarkups according to (12) or (13) These willthen be employed in the counterfactual sim-ulations While our counterfactuals use boththe lower and upper bounds for costs most ofour discussion will be based on using thelower bounds for costs since this is the moreinteresting case in the policy analysis it givesus the largest possible profits for pharmaceuticalfirms and hence corresponds to the worst pos-sible scenario facing Indian firms and the bestpossible scenario facing multinationals underTRIPS

C Identification Assumptions and EstimationApproach

The discussion of the demand system has sofar abstracted from the issue of price endogene-ity The usual premise in the industrial organi-zation literature is that correlation of prices withthe error term in the demand equation arises byvirtue of the first-order conditions of profit-maximizing firms20 As we discuss in this sec-tion this source of simultaneity bias is unlikelyto be of major concern in the present contextbecause of the existence of price regulation Ourprimary concern is instead with the simultane-ity bias that is implied by the particular wayprices are constructed

To understand the sources of the simultaneitybias note first that the price of product group piwe employ in the demand estimation should bethought of as a proxy for an exact price indexthat we do not observe As such pi contains bydefinition measurement error and it will becorrelated with the error term of the demandequation Specifically the demand equation weare interested in estimating can be written insimplified form (suppressing the subscripts rand t and ignoring parameter restrictions forconvenience) as

i i j

ijln pjT ilnXQ

PQ i

where pjT denotes the true (exact) price index for

product group i and i captures unobservedvariables that may affect demand in a particularperiod (eg an advertising campaign or an ep-idemic) The price index pj

T depends on theSKU prices pk Note that in the presence ofSKU entry and exit into the sample the priceindex pj

T will vary over time even if the pricesof the individual SKUs remain stable as theproducts (SKUs) that comprise each productgroup change each period

Ideally we would like to compute the exactprice index pj

T for each group and employ theseindices in the demand estimation Under theassumption of predetermined SKU prices eachperiod the exact price indices would be uncor-related with the error term and the demandparameters could be estimated by simple OLSUnfortunately this strategy is not feasible in thecurrent context

To derive an exact price index for each prod-uct group it is necessary to model consumerrsquoschoice among SKUs conditional on the choiceof the group To this end it is necessary toexplicitly introduce a third stage in the demandspecification and estimate the parameters asso-ciated with the choice at that stage Howeverany specification based on three-stage budget-ing (that would require directly specifying ademand function associated with the SKUchoice at the third stage and imposing additiveseparability at the higher stages) would be in-feasible to estimate for three main reasonsFirst there is a large number of SKUs withinsome groups so that the demand parameters can-not be estimated even if one imposes parameterrestrictions second the SKU prices exhibit littlevariation over time lastly the choice sets at thislowest stage vary over time rendering estimationof the price parameters associated with SKUs thatare not available in all periods infeasible Alterna-tively one could abandon the multistage budget-ing altogether and adopt a discrete choiceapproach which is inherently better suited to deal-ing with varying choice sets and limited pricevariation A discrete choice approach would beparticularly appealing for modeling the SKUchoice conditional on product group choice giventhat within each group all SKUs contain the same

20 See Steven Berry et al (1995) or Aviv Nevo (2001)for a related discussion

1491VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

molecule and are hence comparable As notedearlier however a discrete choice approach at thehigher stages of the demand estimation requiresconverting revenue to physical shares which ischallenging in the case of systemic drugs such asantibiotics for which the appropriate dosage is notwell defined Finally a combination of a multi-stage budgeting approach for the higher stages anda discrete choice approach for the lowest stagewould be feasible to estimate but inconsistentwith a model of utility maximization For thesereasons we have no alternative but to adopt anapproximation for measuring the price index foreach product group

Let us denote this approximation by pjA and let

13j denote the proportional ldquoapproximationrdquo errorassociated with measuring the true price index forgroup j so that ln pj

A ln pjT ln 13j The

estimating demand equation can then be written as

i i j

ijln pjA ilnXQ

PQ

j

ijln 13j i

or

i i j

ijln pjA ilnXQ

PQ i

where the new error term i yenj ijln 13j iis comprised of two components the first onereflects ldquomeasurementrdquo error due to the fact thatthe product group prices we employ are approx-imations and not exact price indices and thesecond one captures conventional demandshocks The demand equation as written abovecorresponds to the equation we take to the data

The particular proxy we use for the exact priceindex for each product group j is the revenue-share-weighted average of the prices of multipleSKUs available within this group that is

pjA

k j

k pk

The notation above illustrates the two source ofprice variation variation in the SKU prices pkand variation in the weights k

We are concerned with potential correlation ofthe so-constructed product group prices with bothcomponents of the product group demand error

term The correlation between pjA and the first

component yenj ijln 13j is independent of theparticular way we construct pj

A and inherent in thefact that we measure the exact price index witherror The potential correlation between pj

A and thesecond component of the product group demanderror the shock i arises however because of thespecific way we construct the proxy pj

A and inparticular because of the presence of the reve-nue share weights k in pj

A since the SKUrevenue share weights k will generally dependon the product group expenditure they arelikely to be correlated with the product groupdemand shock i

To address the simultaneity bias we use in-strumental variables To this end we need vari-ables that are correlated with the proxies pj

A butuncorrelated with the error term i We use thenumber of SKUs within each product group j asan instrument as it does justice to the idea thatvariation in the product group price index stemsin part from variation in the set of SKUs that areavailable each period It is clearly correlatedwith the average group prices pj

A and plausiblyuncorrelated with the demand shock i Theassumption that the number of SKUs is uncor-related with the conventional demand error willhowever be violated if the introduction of anew SKU changes the perceived quality of adrug21 or if it is accompanied by promotionalactivities which will be reflected in i Our hopeis that this is not too often the case Assumingthat the SKU number for each group j is alsouncorrelated with the weighted sum of the mea-surement errors of all product groups yenj ijln 13j itcan be used as an instrument for the product groupprice pj

A22 If one accepts the premise that prices atthe SKU level are exogenous then SKU pricescan also be used as instruments The final list ofinstruments we use hence includes the number ofSKUs in each group the prices of the five largestSKUs for each group and all other exogenousvariables in the demand estimation such as re-gionalproduct dummy interactions and the upper-

21 Note that we already proxy for the quality of each prod-uct group through the product fixed effects i In the actualestimation we actually let these fixed effects be region-specific(ir) so that we allow quality to vary across regions

22 If the measurement error increases in the number ofincluded SKUs this assumption will be violated Unfortu-nately we have no way of checking the validity of thisassumption

1492 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

level total expenditure on antibiotics Regressionsof group prices on the instruments above yieldhigh R-squares with most regressors highly sig-nificant indicating that our instruments are highlycorrelated with prices23

Our sample includes four molecules cipro-floxacin norfloxacin ofloxacin and spar-floxacin Except for sparfloxacin all othermolecules are produced by both foreign anddomestic firms24 So we have seven products(domestic ciprofloxacin foreign ciprofloxa-cin etc) with 96 observations (two years ofmonthly data four geographical regions foreach period) for each product The parametersin the lower-level AIDS demand system asdefined in equation (8) are the product fixedeffects i the productndashregional dummy inter-actions ir the own revenue-share price elas-ticities ii the cross revenue-share priceelasticities i10 i01 i00 and the revenue-share expenditure elasticities i In estimatingthe parameters we first regress prices on allinstrumental variables and then plug the pre-dicted values for prices in the constrainedleast-square regression (For a detailed expla-nation of the constraints see the previoussection)

Given that we impose many cross-equationconstraints and employ instrumental variablesin the estimation it is difficult to derive stan-dard errors for the parameter estimates ana-lytically Our error term interpretation in (8)implies that the error terms for each productgroup are likely to be correlated across re-gions for example if a national advertisingcampaign increases the demand for a rela-tively expensive presentation in one regionsimultaneously increasing the aggregate de-mand and price index for the correspondingproduct group in that region it is likely that thesame effect will be observed in other regions Inprinciple we could exploit the cross-regionalcorrelation in the error terms to estimate thedemand system using seemingly unrelated re-

gression (SUR) This however would requireus to derive the variance-covariance matrix ofthe parameters analytically which as notedabove is cumbersome in the current contextInstead we use the bootstrap method choosingto remain agnostic about the structure of thevariance-covariance matrix The potential dis-advantage of this approach over SUR is a loss inefficiency but as will become apparent in theresults section the model parameters are fairlyprecisely estimated To maintain the marketstructure we randomly sample the periods (withreplacement) and use the same periods for allproducts Regarding the optimal number ofbootstrap repetitions ideally one would followthe three-step method proposed by Donald W KAndrews et al (2000) Empirical evidence sug-gests however that one rarely needs more than200 replications to estimate the standard er-rors25 To be safe we generate 300 bootstrapsamples (with replacement) based on the origi-nal data and estimate the standard errors usingthe standard errors of the bootstrap sampleestimates

The estimation of the top-level AIDS systemis similar The constraints imposed on this top-level demand system are adding-up homogene-ity and symmetry as defined on page 12Again bootstrapping is used to obtain the stan-dard errors of the parameter estimates

D The Counterfactual Scenarios

In assessing the effects of patent enforce-ment we start by focusing on the most extremecase in which compulsory licensing is not anoption and foreign firms are not subject to pricecontrols We use the results from the analysis ofthis case as a benchmark In reality the outcomeof the WTO negotiations is more likely to in-volve some constraints on the monopoly powerof foreign firms selling patented products indeveloping countries such as price caps orcompulsory licensing Our framework can eas-ily accommodate these cases as will becomeapparent in the next subsection

We now focus on the effects of potentialpatent enforcement in the quinolone segmentWe consider several scenarios which vary inthe number and size of domestic products that

23 Specifically the R-squares from the first-stage regres-sions of product group prices on instruments range from057 to 095 except for domestic ciprofloxacin for whichthe R-square is 017

24 Sparfloxacin is actually offered by one foreign subsid-iary in India (see Table 2) However its revenue share isminiscule We therefore treat Sparfloxacin as being pro-duced by domestic firms only 25 See Bradley Efron and Robert J Tibshirani (1993)

1493VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

will be removed from the market In particularwe consider the following five scenarios26

Withdrawal of one large domestic productgroup only domestic ciprofloxacin

Withdrawal of one relatively small domesticproduct group only domestic ofloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand norfloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand sparfloxacin

Withdrawal of all four domestic quinoloneproduct groups

As the list above suggests we proceed fromanalyzing the effects of single-product with-drawal to the analysis of eliminating the en-tire domestic segment This approach wasmotivated by early empirical results that in-dicated that the existence and extent of com-petition from domestic firms has a significantbearing on the predicted effects of patent en-forcement that is our predictions regardingprices and welfare vary substantially depend-ing on how many domestic products are af-fected by patent enforcement and on whethersome domestic competition will remain presentafter TRIPS In addition the size of the affecteddomestic groups is relevant for the welfare pre-dictions accordingly we examine both scenar-ios in which a large domestic product groupsuch as domestic ciprofloxacin is eliminatedfrom the market and scenarios in which the

eliminated domestic group is relatively small(eg domestic ofloxacin) In all scenarios wemaintain the assumption that the set of prod-ucts offered by the remaining firms in themarket does not change in response to patentenforcement

E Computation of Virtual Prices and NewEquilibrium Prices

The first step in the counterfactual analysisis to derive the new equilibrium prices underpatent enforcement27 In this context thereare two sets of prices that are relevant Thefirst set consists of the virtual prices of those(domestic) products that will not be availableonce TRIPS is put in effect To calculate thesevirtual prices we set the expenditures of therelevant products equal to zero The secondset of prices consists of the prices of thoseproducts that remain in the market In deriv-ing these prices we start by assuming profitmaximization without any form of price reg-ulation the firms remaining in the marketreoptimize in response to the policy changeand set new prices taking the prices of allother firms as given28 Of course at the equi-librium all prices change in response to thefact that some domestic products are nolonger present The new equilibrium pricesfor products that remain in the market are thuscomputed by utilizing the first-order condi-tions of profit-maximizing firms into whichthe virtual prices of the eliminated productsare substituted Hence to compute the newequilibrium prices we solve an equation sys-tem of the following form

26 As Table 2 indicates most of the patents for the drugsin the quinolone segment have expired by now so that noneof the scenarios described in this section is going to mate-rialize in practice The reason we consider these alternativescenarios in our counterfactual simulations is to get a senseof how the presence and extent of domestic competitionaffects welfare calculations Such calculations may becomerelevant in the future in other therapeutic classes in whichpatents have not expired yet and where domestic firmscompete with foreign patent holders While the particularexit scenarios will depend on the exact year of patent expiryof drugs in each therapeutic class and the specific way inwhich patents will be enforced (eg whether certain do-mestic drugs will be grandfathered in) the general conclu-sion that emerges from our calculations is that the consumerwelfare loss is substantially smaller in cases where somedomestic competition remains present in the market Hencethe particular way in which patent enforcement is imple-mented is essential for assessing its welfare impact

27 Of course until product patents are in fact introducedthese prices will not be observable Note also that we areassuming here that the range of available products will notchange with the introduction of patents

28 As mentioned above this first set of calculations ab-stracts from the existence of remaining price controls orother government regulations that would impose constraintson the firmsrsquo profit maximization problem Accordingly theresulting numbers should be interpreted as a benchmark ofwhat would happen if markets were completely unregulatedTo examine what the welfare effects of TRIPS would be inthe more realistic scenario of price regulation we subse-quently consider a scenario in which regulation keeps theprices of the products provided by patent-holders at theirpre-TRIPS level For a more detailed discussion see theresults section

1494 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

For products i that are withdrawn from themarket

(14) 0 i ir iiln pirV i10ln pjr j Di

10

j Di

01

i01ln pjr

j Di

00

i00ln pjr

ilnXQr

PQr

For products k that remain in the market

(15) pkr ckr 1 1

kk pkrpjr pirV 1

In the equations above pirV denotes the virtual

prices of the products that are removed fromthe market while pjr denotes the updatedprices of all other products Note that whensolving for the virtual prices we account forthe fact that both the price index for quino-lones PQr and the expenditure allocated tothis subsegment XQr need to be updated toreflect the fact that as a result of the pricechanges there may be substitution away fromthis subsegment To obtain the new quinoloneexpenditure XQr and the new price index PQrwe use the estimates and formulas for thehigher-level AIDS system In equation (15)ck refers to the marginal cost for product k thatwe have obtained from the previous estimationstage As mentioned in the previous subsectionwe conducted the simulations using both theupper and lower bounds for marginal cost ck

u

and ckL respectively The term kk(pkr

pjrpirV) re-

fers to the unconditional own-price elasticity forproduct k which is a function of the eliminatedproductsrsquo virtual prices and the remaining prod-uctsrsquo new equilibrium prices We conduct thecounterfactual simulations and welfare analysisat the regional level because the presence ofregion-specific product effects in the demandestimation implies region-specific demand elas-ticities and hence region-specific marginalcosts and markups The presentation and discus-sion of our results focus on national averages ofthe relevant variables that we construct by com-

puting weighted averages across regions usingthe population of each region as a weight

F Welfare Assessment

The simulation of the new equilibrium underpatent protection can provide important insightsinto how consumers and firms will respond tothe removal of domestic products in the market(for example toward which products consumerswill substitute or which prices will increase themost) To get a more precise idea of how peo-plersquos well-being will ultimately be affected byTRIPS we compute as a last step in our anal-ysis the welfare effects of the policy changeSocial welfare is defined as consumer welfareand the sum of domestic firm profits Thechange in domestic profits can easily be calcu-lated by comparing the domestic firm (variable)profits at the pre-TRIPS prices to the profitsthese firms will realize at the new simulatedprices Although foreign firm profits do notcount in domestic welfare calculations we alsocompute the effects of patent enforcement onforeign firm profits to get an idea of how largethe expected benefits of TRIPS are for thesefirms This provides in some sense an indirectway of assessing whether the claims that patentenforcement in countries like India will lead tomore research on developing-country-specificdiseases (such as malaria) have any validity iffor example we find that the effect of patentenforcement on the foreign firm profits realizedin India is small in magnitude it is unlikely thatforeign firms will engage in more developing-country-specific research in response to TRIPSIt is important to note that in all these calcula-tions we work with the lower-bound estimatesfor marginal costs since these give rise to thehighest possible markups Hence our estimateof profit loss for domestic producers most likelyoverstates this loss

On the consumer side we measure changes inconsumer welfare by the compensating variation(CV) defined as the additional expenditure thatconsumers need in order to achieve the sameutility level as before patent enforcement at thenew prices Specifically let P0 denote the pricevector before patent enforcement P the simulatedprice vector post-TRIPS (that we obtained usingthe methods described in the previous subsection)u0 the utility attained by consumers before TRIPS

1495VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

and E(u P) the higher-level expenditure functionThen the compensating variation is given by

(16) CV Eu0 P Eu0 P0

Note that the CV as computed in (16) representsthe combination of three effects

The pure product-variety effect that is theeffect that arises because one or more prod-ucts are not available to consumers anymoreholding the prices of all other remainingproducts and the total expenditure on thequinolone subsegment XQ constant

The expenditure-switching effect that is theeffect arising from substitution away fromquinolones and toward other subsegments ofthe antibiotics market again holding the pricesof all other remaining products constant

The reduced-competition effect that is theeffect that arises because the firms remainingin the market adjust (increase) their prices inresponse to the removal of domestic products

From both an analytical and a policy point ofview it is desirable to assess how large each ofthese effects is Accordingly we decompose thetotal effect on consumer welfare (the CV as givenby equation (16)) using the following procedure

To get the pure product variety effect wecompute virtual prices for the products that areremoved from the market holding the quino-lone expenditure XQ and the prices of all otherproducts fixed Let us call the resulting pricevector P1 Then the pure product variety effectis represented by E(u0 P1) E(u0 P0)

To compute the expenditure-switching effectwe compute another set of virtual prices againholding the prices of all remaining products fixedbut letting quinolone expenditure adjust in re-sponse to the new price index for the quinolonesegment (Given that the prices of the remainingproducts remain fixed the change in the priceindex arises only because of the removal of one ormore domestic products) Note that this scenariomost closely resembles the case in which patentlaws are enforced but strict price regulation keepsthe prices of the products offered in the market attheir pre-TRIP level Let us label the so-computedprice vector P2 The expenditure-switching effectis then E(u0 P2) E(u0 P1)

Finally the reduced competition effect aris-ing from higher prices for the remaining prod-

ucts is computed as the residual change in thecompensating variation once the product-varietyand expenditure-switching effects have been ac-counted for that is E(u0 P) E(u0 P2)where the price vector P is computed accordingto the formulas (14) and (15) to reflect theadjustment of prices to the new regime

To compute the standard errors associatedwith the counterfactual simulations (that is thestandard errors for the counterfactual prices andwelfare estimates) we again use bootstrappingWe first bootstrap the original sample (sales andprices of all the drugs) 300 times Next we esti-mate the AIDS model for each of these samplesand compute the counterfactual equilibrium pricesand welfare losses corresponding to each policyscenario for each of the 300 simulations In thefinal step we compute standard errors29

One limitation of our framework is that itdoes not allow for a heterogeneous response ofconsumers to the policy change Accordinglyour framework is not suited to addressing thequestion of how different groups in the popula-tion will be affected by patent law enforce-ment30 Along the same lines our frameworkdoes not accommodate the possibility of pricediscrimination which might lead to differentresults regarding welfare losses and profit gainsrelative to uniform pricing31 Still we believethat the results of the counterfactual simulationscan provide important insights into the likelyaggregate response to patent enforcement andthe factors that drive this response

III Results

A The Structure of Demand

Tables A2 and A3 in the Appendix display theresults from estimation of the lower- and upper-level AIDS systems respectively For ease of in-terpretation rather than discussing the coefficientestimates we focus our discussion on the implied

29 In cases some of the simulations do not converge wecompute the standard errors using only those simulationsthat converged

30 Addressing this question would require at a mini-mum micro data on consumer purchases as in Goldberg(1995) To our knowledge such data do not exist for theIndian pharmaceutical market

31 See Ernst R Berndt (1994) for an extensive discussionof uniform pricing

1496 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

unconditional price and expenditure elasticities re-ported in Table 6 Given that the region-specificproduct group effects ir reported in the last fourcolumns of Tables A2 and A3 imply region-spe-cific demand elasticities we report separate elas-ticities for each region in Tables 6A and 6B Asevident from the comparison of these tables theelasticities are very similar across regions so thatwe can focus the remaining discussion on theelasticities of one region only the northern regiondisplayed in Table 6A

The diagonal terms of Table 6A report theown-price elasticities which are in all but onecase negative and highly significant The oneexception is the foreign norfloxacin productgroupmdashwhose share of quinolone sales is 007percentmdashfor which we estimate a negative butinsignificant own-price elasticity For the remain-ing product groups demand appears to be highlyelastic with the estimated elasticities lower than2 in four out of the six cases The magnitude ofthe own-price elasticities matches the features ofthe Indian pharmaceutical market mentioned ear-lier which would suggest that Indian consumersare likely to be quite price-sensitive32 The elas-ticities appear especially large if one takes intoaccount that they refer to product groups (such asdomestic ciprofloxacin) and not individual drugsoffered by particular firms Their relative magni-tudes are also intuitive the drug with the largestmarket share and a relatively high price (domesticciprofloxacin) appears to be one of the least elas-tic In contrast foreign ciprofloxacin is highlyprice elastic this is plausible as ciprofloxacindrugs offered by subsidiaries of multinationalsface the stiffest competition in this market seg-ment from approximately 75 Indian firms offeringthe same molecule

The estimated expenditure elasticities appear inthe last column These are all positive indicatingthat the demand for all product groups is normalThe remaining cells display the estimated cross-price elasticities As one might perhaps expect forproducts within a therapeutic subsegment theseare mostly positive Out of a total of 49 priceelasticities we estimate there are six that do notconform to expectations these are the cross-price elasticities between different foreign prod-

uct groups which are estimated to be negativeand significant33 Fortunately these elasticitieshave negligible impact on the welfare analysisgiven that our counterfactuals focus on theeffect of withdrawing one or more domesticproducts from the market the most relevantelasticities are the ones that capture the responseof various product group shares to a change in theprice of one or more domestic groups these elas-ticities are the cross-price elasticities between var-ious domestic groups and the ones betweendomestic and foreign groups which are plausibleand precisely estimated34

Regarding these elasticities a striking aspectof our estimates is how large positive andsignificant the cross-price elasticities betweendifferent domestic product groups aremdashin factfor norfloxacin and ofloxacin we estimate thatdomestic product groups containing differentmolecules are closer substitutes for one anotherthan product groups that contain the same mol-ecule but are produced by foreign firms Incontrast for ciprofloxacin (the molecule withthe largest revenue share) we estimate a largepositive cross-price elasticity between the do-mestic and foreign versions

The fact that domestic products appear to beclose substitutes for other domestic productsthat contain different molecules truly representsan ldquoempiricalrdquo finding in the sense that we do

32 In developed economies elasticities of this magnitude havetypically been found only for generic drugs (and even then onlyrarely) or among consumers who lack health insurance

33 While these elasticities are clearly counterintuitive theyare not inconsistent with the underlying demand system whichimposes no restrictions on the sign of the cross-price elastici-ties We do not have a good explanation of why these elastic-ities are estimated to be negative A potential explanation isthat the shares of the foreign products are very small giventhis we observe very few consumers switching from one (verysmall) foreign group to another (very small) foreign groupwhen the price of the first foreign group goes up and hence theinference is not very reliable in this case

34 The cross-price elasticities between foreign drug groupscontaining different molecules will also have an effect on thewelfare estimates given that the withdrawal of domestic prod-ucts will generally lead to changes in the prices of all foreignproducts and market shares will be reallocated from someforeign products to other foreign products based on the newprices However this reallocation from ldquoforeign to foreignrdquo istruly second order in our case compared to the reallocationfrom ldquodomestic to domesticrdquo and ldquodomestic to foreignrdquo prod-ucts In scenarios in which we consider patent enforcementaccompanied by strict price regulation the cross-price elastic-ities between foreign product groups are in fact completelyirrelevant as the prices of foreign products are not allowed toincrease in these cases Still our welfare loss estimates remainsubstantial drivenmdashas beforemdashby the loss of product variety

1497VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

not impose it through any of our assumptionsregarding the demand function The questionthat naturally arises then is what might explainthis finding While we cannot formally addressthis question anecdotal accounts in various in-dustry studies suggest that the explanation maylie in the differences between domestic andforeign firms in the structure and coverage ofretail distribution networks

Distribution networks for pharmaceuticals inIndia are typically organized in a hierarchicalfashion Pharmaceutical companies deal mainlywith carrying and forwarding (CampF) agents inmany instances regionally based who each sup-ply a network of stockists (wholesalers) Thesestockists in turn deal with the retail pharma-cists through whom retail sales ultimately oc-cur35 The market share enjoyed by a particularpharmaceutical product therefore depends inpart on the number of retail pharmacists who

stock the product And it is here that thereappears to be a distinction between domesticfirms and multinational subsidiaries In particu-lar the retail reach of domestic firms as agroup tends to be much more comprehensivethan that of multinational subsidiaries (IndianCredit Rating Agency (ICRA) 1999)36

There appear to be two reasons for this Thefirst is that many of the larger Indian firmsbecause they have a much larger portfolio ofproducts over which to spread the associatedfixed costs typically have more extensive net-works of medical representatives The second issimply that there are many more domestic firms(and products) on the market At the retail levelthis would imply that local pharmacists mightbe more likely to stock domestic products con-taining two different molecules say ciprofloxa-cin and norfloxacin than they would domesticand foreign versions of the same molecule Tothe extent that patients (or their doctors) arewilling to substitute across molecules in order tosave on transport or search costs (eg going toanother pharmacy to check whether a particular

35 There are estimated to be some 300000 retail pharma-cists in India On average stockists deal with about 75 retailers(ICRA 1999) There are naturally variations in this structureand a host of specific exclusive dealing and other arrangementsexists in practice Pharmaceutical firms also maintain networksof medical representatives whose main function is to marketthe companyrsquos products to doctors who do the actual prescrib-ing of drugs In some instances firms do sell directly to thedoctors who then become the ldquoretailerrdquo as far as patients areconcerned but these are relatively rare

36 These differences were also highlighted in conversa-tions one of the authors had with CEOs and managingdirectors of several pharmaceutical firms as part of a sep-arate study

TABLE 6AmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN THE NORTHERN REGION

Product group

Elasticity with respect to

Prices of foreign productgroups Prices of domestic product groups Overall

quinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 557 013dagger 015 401 011dagger 011dagger 016 137(179) (007) (007) (184) (006) (006) (006) (029)

Foreign norfloxacin 427dagger 045 427dagger 350dagger 602 451 465 220(242) (112) (242) (210) (623) (184) (183) (105)

Foreign ofloxacin 011 010dagger 138 009 009dagger 023 011 116(005) (005) (031) (027) (005) (028) (004) (017)

Domestic ciprofloxacin 018 001 001 168 008 008 010 117(008) (000) (001) (023) (002) (002) (002) (003)

Domestic norfloxacin 004 003 004 058 223 042 040 073(001) (003) (001) (017) (011) (004) (003) (009)

Domestic ofloxacin 005 005 011 077 074 342 074 089(002) (002) (013) (028) (008) (025) (008) (021)

Domestic sparfloxacin 007 004 007 115 063 063 288 028(002) (001) (002) (015) (006) (006) (017) (012)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent significance level and dagger (dagger) denotes significance at the 10-percent level

1498 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE 6BmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN OTHER REGIONS

Demand patterns in the eastern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 594 014dagger 016 431 013 011 017 140(195) (008) (008) (201) (006) (007) (007) (031)

Foreign norfloxacin 329dagger 058 329dagger 264dagger 460 346 359 192(180) (083) (179) (153) (451) (138) (139) (082)

Foreign ofloxacin 012 010dagger 140 010 010 024 013 117(006) (006) (034) (031) (005) (030) (005) (019)

Domestic ciprofloxacin 018 001 001 172 009 008 011 117(008) (000) (002) (027) (002) (003) (002) (003)

Domestic norfloxacin 004 004 004 068 242 049 046 070(002) (004) (002) (020) (013) (005) (004) (010)

Domestic ofloxacin 004dagger 004 009 061 060 295 060 092(002) (001) (010) (028) (007) (022) (007) (017)

Domestic sparfloxacin 005 003 005 092 048 051 251 043(002) (001) (001) (017) (005) (006) (016) (010)

Demand patterns in the western region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 427 009dagger 011 286 008 006 011 126(129) (005) (005) (133) (004) (005) (004) (021)

Foreign norfloxacin 714dagger 008 710dagger 594dagger 995 742 770 299dagger

(369) (183) (368) (328) (1015) (297) (296) (178)Foreign ofloxacin 013 012dagger 145 009 012 026 013 119

(007) (006) (037) (029) (005) (033) (005) (021)Domestic ciprofloxacin 019 001 000 174 009 008 011 118

(008) (000) (001) (024) (002) (003) (002) (004)Domestic norfloxacin 004 004 004 067 243 050 047 069

(002) (004) (001) (018) (014) (005) (004) (010)Domestic ofloxacin 003 003 007 048 048 257 048 093

(002) (001) (008) (022) (005) (015) (004) (014)Domestic sparfloxacin 006 003 005 083 046 049 241 046

(002) (001) (001) (014) (004) (005) (012) (009)

Demand patterns in the southern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 560 013dagger 015 402 011 010 017 137(182) (007) (007) (187) (006) (006) (007) (029)

Foreign norfloxacin 533dagger 031 531dagger 431dagger 748 559 584 249dagger

(301) (134) (300) (245) (734) (219) (219) (128)Foreign ofloxacin 014 012dagger 148 011 012 029 016 120

(007) (007) (039) (033) (006) (035) (006) (022)Domestic ciprofloxacin 017 001 000 169 008 007 011 116

(008) (000) (001) (025) (002) (003) (002) (003)Domestic norfloxacin 004 003 003 059 225 043 040 073

(001) (003) (001) (019) (012) (004) (004) (009)Domestic ofloxacin 004dagger 004 009 062 060 296 060 091

(002) (001) (010) (026) (006) (019) (007) (017)Domestic sparfloxacin 008 005 008 138 073 075 323 016

(002) (001) (002) (018) (007) (008) (021) (014)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at 5-percent level anddagger (dagger) denotes significance at the 10-percent level

1499VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

foreign product is in stock) in aggregate datawe would expect to find precisely the substitu-tion patterns that we report in Table 6

Whether the particular explanation we provideabove is the correct one the high degree of sub-stitutability between domestic product groupsturns out to have important implications for thewelfare calculations We discuss these in moredetail below when we present the results of thecounterfactual welfare analysis Another elasticitywith important implications for the counterfactu-als is the price elasticity for the quinolone subseg-ment as a whole which indicates how likelyconsumers are to switch to other antibioticsgroups when faced with a price increase for quin-olones This elasticity is computed on the basis ofthe results in Table A3 and it is at 111 (stan-dard error 024) this is large in magnitudebutmdashas expectedmdashsmaller in absolute value thanthe own-price elasticities of the product groupswithin the quinolone subsegment

The results in Tables 6A and 6B are based onour preferred specification discussed in SectionII In Tables A4 to A6 in the Appendix weexperiment with some alternative specificationsTables A4(a)ndashA4(c) correspond to a specificationthat includes in addition to product-group-specificregional fixed effects product-group-specific (andfor the upper level antibiotics-segment-specific)seasonal effects We distinguish among three sea-sonsmdashthe summer monsoon and wintermdashandreport the unconditional demand elasticities for

the northern region for each of these seasons Asevident from the tables in the Appendix our elas-ticity estimates are robust to the inclusion of sea-sonal effects The demand elasticities in Table A5are based on estimation of the demand system byOLS Compared to the elasticities obtained by IVthe OLS elasticities are smaller in absolute valueimplying that welfare calculations based on theOLS estimates would produce larger welfare lossestimates Nevertheless some of the patterns re-garding the cross-price elasticities discussed ear-lier are also evident in the OLS results inparticular the cross-price elasticities between dif-ferent domestic product groups are all positivelarge and significant and in most instances largerthan the cross-price elasticities between drugs thatcontain the same molecule but are produced byfirms of different domesticforeign status Theclose substitutability of domestic products indi-cated by both the OLS and IV estimates seems tobe one of the most robust findings of the paper

B Cost and Markup Estimates

Table 7 displays the marginal costs markupsand profits implied by the price elasticity esti-mates of Tables 6A and 6B for each of the sevenproduct groups Given that our regional effectsimply different price elasticities for each regionour marginal cost and markup estimates alsodiffer by region Given however that based on

TABLE 7mdashUPPER AND LOWER BOUNDS FOR MARGINAL COST MARKUP AND ANNUAL PROFIT BY PRODUCT GROUPS WITHIN

THE QUINOLONE SUBSEGMENT

Product groupLower boundfor MC (Rs)

Upper boundfor markup

Upper boundfor profit(Rs mill)

Upper boundfor MC (Rs)

Lower boundfor markup

Lower boundfor profit

(Rs)

Foreign ciprofloxacin 83 19 269 103 0 00(123) (012) (1655)

Foreign norfloxacin NA NA NA 53 0 00Foreign ofloxacin 323 70 1061 1085 0 00

(2316) (021) (3185)Domestic ciprofloxacin 47 59 17019 112 0 00

(114) (010) (29858)Domestic norfloxacin 52 43 2807 90 0 00

(020) (002) (1532)Domestic ofloxacin 587 34 1612 901 0 00

(218) (002) (1280)Domestic sparfloxacin 495 37 1985 788 0 00

(157) (002) (1100)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level Estimated lower bound forforeign norfloxacinrsquos marginal cost is negative since the estimated price elasticity is less than one in absolute value

1500 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Tables 6A and 6B the price elasticities do notseem to differ substantially across regions wereport for ease of exposition only the nationalaverages for marginal costs and markups inTable 7

Table 7 has two parts In the left part (firstthree columns) we report the lower bound formarginal cost and the corresponding upperbound for markup and upper bound for totalannual profit for each product group Thesenumbers are based on the price elasticities weobtained from estimating the two-level AIDSdemand system Since we do not have a reliableestimate for the price elasticity of foreign nor-floxacin (the point estimate is negative but lessthan one in absolute value and insignificant)we cannot compute the lower bound for mar-ginal cost in this case It is important to note thatthese estimates do not reflect either the actualmarginal cost or the actual markup for thesedrugs both because the existence of price reg-ulation implies that the unconstrained first-orderconditions are unlikely to hold each period andbecause our aggregation across firms of thesame domesticforeign status supplying thesame molecule makes the interpretation of theseestimates problematic In particular the fact thatwe ignore competition among firms within eachproduct group implies that our estimates willtend to overstate market power These numberswill prove useful however in the counterfac-tual simulations as they can give us a sense ofhow large the maximum profit gains for multi-national firms and the maximum profit lossesfor domestic firms are likely to be under patentenforcement The right part of the table (lastthree columns) reports the upper bound for mar-ginal cost which we obtain by simply taking themarginal cost to be equal to the observed priceWe do not report standard errors in this casesince the numbers are based on actual data Themarkups corresponding to these marginal costupper bounds are of course zero We conductthe counterfactual simulations using both thelower and upper bounds for marginal costs

The (upper bounds for) markups on the leftside of the table are generally plausible Thedomestic ciprofloxacin product group that dom-inates the quinolone subsegment and for whichwe documented high prices a high marketshare and a relatively low elasticity of demandenjoys one of the highest markup upper bounds(60 percent) and accounts for nearly 70 percent

of all profits derived within the subsegmentForeign ciprofloxacin on the other hand whichfaces the stiffest competition from domesticfirms and for which we estimated a highly elas-tic demand has the lowest markup upper bound(19 percent)

C Counterfactual Estimates of the Impact onPrices and Welfare

With estimates of the key demand and costparameters in hand we turn to the counterfac-tuals We consider the five separate scenarioslisted in the previous section All of the scenar-ios involve the withdrawal of one or more of thedomestic product groups from the market Ta-ble 8 displays our estimates of the consumerwelfare losses that result under the differentscenarios The losses are expressed in billion ofrupees per year All numbers presented in Ta-ble 8 and subsequent tables are based on usingthe lower bounds for marginal cost and upperbounds for markup in the simulations As dis-cussed earlier these numbers are the more in-teresting to work with since they give us anupper bound for the changes in the profits ofdomestic and foreign firms that would resultfrom patent enforcement In the Appendix wealso present results based on using the upperbounds for marginal costs in the simulations inwhich case the pre-TRIPS profits of domesticand foreign firms are zero In all cases marginalcosts are assumed to be constant in outputWhile naturally the profit implications differdepending on whether one uses the upper orlower bounds for marginal costs (firm profits arezero if one assumes the upper bound of mar-ginal cost in which case price equals marginalcost) the estimated consumer welfare losses aresimilar in the two cases We discuss these re-sults in more detail at the end of this section

The first column presents our estimates of theconsumer welfare losses attributable to the pureloss of product variety effect where we fix theprices of all remaining products as well as theoverall expenditure on quinolones while with-drawing one or more of the domestic productgroups Note that in our initial specification ofthe demand system had we not allowed for thepossibility that consumers might differentiatebetween domestic and foreign products evenwhen they contain the same molecule this

1501VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

particular component of the loss of consumerwelfare would not have arisen

The estimates reported in the second columnincorporate the expenditure-switching effect ontop of the loss of product variety Here basedupon the price elasticity estimates from thehigher-level AIDS system we adjust (down-ward) the expenditures allocated to the quino-lone subsegment as the composite price ofquinolones effectively increases as a consequenceof the higher virtual prices of the domestic productgroups that are withdrawn from the market Be-cause the estimates in this column are generatedassuming that the prices of the products that re-main in the market are not adjusted upward theyprovide a sense of what consumer losses would beif the introduction of product patents were coupledwith strict price regulation aimed at maintainingprices at prepatent levels Alternatively they canbe thought of as the relevant welfare numbers ifintense competition among firms within the re-maining product groups kept the prices of theproducts that were still offered in the market closeto the firmsrsquo marginal costs

The last column of Table 8 displays the esti-mated consumer welfare losses when bothcross-segment expenditure-switching and with-in-segment upward price adjustments are takeninto account

If we compare the results across the firstsecond and third columns all the counterfac-tual scenarios produce qualitatively similar pat-terns patterns that are consistent with what we

would expect Starting from the initial loss ofwelfare attributable to the loss of product vari-ety the option of switching expenditures outof the quinolone subsegment to other subseg-ments mitigates some of the initial welfareloss But if we then incorporate the upwardprice adjustments that result in response to thereduced competition the welfare losses aremagnified

Of particular interest from a policy perspec-tive are the relative magnitudes of these threeeffects which are similar under all the counter-factual scenarios though absolute levels varyconsiderably First despite the fact that the de-mand for quinolones is quite sensitive to thecomposite price of quinolonesmdashthe upper levelprice elasticity is 111mdashthe cross-subsegmentexpenditure-switching effects are in all thecases small (in absolute value terms) relative tothe other two effects For instance under thescenario where all the domestic quinolone prod-uct groups are withdrawn from the market theoverall consumer welfare loss of Rs 1781 bil-lion per year can be decomposed into an initialloss of Rs 1176 billion (66 percent) attribut-able to the loss of product variety a slightreduction in this initial loss of Rs 041 billion(2 percent) from Rs 1176 billion to Rs1135 billion because of expenditure switchingand a subsequent additional loss of Rs 646billion (36 percent) from Rs 1135 billion toRs 1781 billion because of the reduced com-petition and consequent price increases

TABLE 8mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure loss ofvariety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 732(087) (089) (146)

Only ofloxacin 008 008 023(008) (008) (010)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1253(177) (180) (415)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1058(142) (145) (331)

All four domestic quinolones products 1176dagger 1135dagger 1781(643) (634) (1270)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent significance level and dagger (dagger)denotes significance at the 10-percent level

1502 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 6: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

worldrsquos largest producers of bulk drugs7 Thestructure of the industry has also evolved In1970 the industry was dominated by multina-tional subsidiaries by 2001 Indian-owned firmswere not just the leading players in the industrymany had also become major exporters

The data we use in this paper are from theretail pharmaceutical audits of ORG-MARGIndiarsquos premier market research and consultingfirm The audit provides detailed product-levelinformationmdashestimates of monthly retail salesin each of the four geographic zones of Indiaprice dosage form launch date brand namechemical name therapeutic categorizationmdashon

all pharmaceutical products sold in India byabout 300 of the largest firms representingroughly 90 percent of domestic retail sales ofpharmaceuticals The coverage of the audit isextensive reaching a representative panel ofthousands of retail chemists in over 350 citiesand towns The data collected which providethe only real source of disaggregate informationon the Indian pharmaceutical market are usedby both the government of India in formulatingpricing policy and other decisions and the In-dian pharmaceutical industry in determiningpricing and marketing strategies We have in-formation at a monthly frequency for the periodof January 1999 to December 2000 Tables 1 to3 provide a set of descriptive statistics that areessential for understanding the focus of ouranalysis and interpreting our results

As noted earlier the characteristics of de-mand for pharmaceuticals in India are likely todiffer considerably from those in developedeconomies With a share of 23 percent the

7 Bulk drugs are the therapeutically relevant active phar-maceutical ingredients that are combined with a variety ofinactive ingredients to make the formulations ultimatelyconsumed by patients Firms in the pharmaceutical sectorcan be one of three types bulk drugs producers pureformulators or integrated firms which produce both bulkdrugs and market formulations

TABLE 1mdashTHE QUINOLONES SUBSEGMENT

Molecule

Shares ()of quinolones sales Sales (Rs mill) 2000

Domesticfirms

Foreignsubsidiaries

Domesticfirms

Foreignsubsidiaries

Ciprofloxacin 530 27 3030 156Norfloxacin 112 01 640 3Ofloxacin 116 31 665 177Sparfloxacin 108 01 620 4Lomefloxacin 15 86 Pefloxacin 13 01 72 5Levofloxacin 00 0 Nalidixic acid 13 73

TABLE 2mdashBASIC INFORMATION ABOUT THE FOUR QUINOLONES MOLECULES 2000

Ciprofloxacin Norfloxacin Ofloxacin Sparfloxacin

US or European patent holder Bayer Merck Ortho-McNeil Rhone-PoulencYear of US patent expiry 2003 1998 2003 2010Year of US-FDA approval 1987 1986 1990 1996Year first introduced in India 1989 1988 1990 1996No of domestic Indian firms 75 40 17 25No of foreign subsidiaries 8 2 2 1Sales weighted average per-unit API

price of products produced byDomestic Indian firms 1124 905 9008 7884Foreign subsidiaries 1035 528 10847

API Active pharmaceutical ingredient

1482 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

antiinfectives segment ranks second in Indiawhereas in the world market it is fifth and has ashare of only 90 percent Hence antiinfectivesare important in India not only from a health

and public policy point of view but also as asource of firm revenue

With this in mind we focus on one partic-ular subsegment of antiinfectives namely the

TABLE 3mdashSUMMARY STATISTICS FOR THE QUINOLONES SUBSEGMENT 1999ndash2000

North East West South

Annual quinolones expenditure per household (Rs) 3125 1975 2764 2359(366) (367) (407) (286)

Annual antibiotics expenditure per household (Rs) 11988 8424 11052 9624(1224) (1224) (960) (996)

No of SKUsForeign ciprofloxacin 1238 1129 1308 1246

(150) (190) (102) (106)Foreign norfloxacin 183 171 200 158

(070) (075) (088) (083)Foreign ofloxacin 304 296 296 300

(086) (086) (091) (088)Domestic ciprofloxacin 10621 9763 10342 10550

(599) (434) (722) (451)Domestic norfloxacin 3896 3496 3617 3942

(271) (268) (251) (379)Domestic ofloxacin 1846 1600 1725 1725

(680) (634) (586) (635)Domestic sparfloxacin 2983 2829 3121 2929

(557) (638) (688) (657)Price per-unit API (Rs)

Foreign ciprofloxacin 958 1090 1085 1007(128) (066) (071) (058)

Foreign norfloxacin 563 509 605 435(077) (133) (139) (147)

Foreign ofloxacin 10946 10943 10886 10612(620) (664) (700) (1140)

Domestic ciprofloxacin 1143 1067 1131 1152(016) (015) (017) (013)

Domestic norfloxacin 951 907 888 873(024) (035) (037) (020)

Domestic ofloxacin 9163 8964 8565 9341(1615) (1565) (1422) (1407)

Domestic sparfloxacin 7972 7849 7688 8028(976) (1014) (1185) (1037)

Annual sales (Rs mill)Foreign ciprofloxacin 4179 2431 4520 2947

(1534) (816) (1273) (648)Foreign norfloxacin 128 100 058 073

(101) (082) (044) (057)Foreign ofloxacin 5446 3184 3522 3111

(1399) (933) (906) (703)Domestic ciprofloxacin 96229 58591 67874 70381

(10626) (13026) (12226) (8740)Domestic norfloxacin 22255 11971 14918 15829

(3884) (1945) (2691) (1626)Domestic ofloxacin 12502 9621 14936 11205

(4434) (3011) (5282) (4259)Domestic sparfloxacin 15617 12175 16130 9811

(3141) (2576) (4674) (3420)

Note Standard deviations in parentheses API Active pharmaceutical ingredient

1483VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

quinolone subsegment Quinolones fall into thesystemic antibiotics and antibacterials segmentof the Indian pharmaceuticals market whichgenerates over three-quarters of the revenues inthe antiinfectives segment8 The systemic anti-bacterials segment includes all of the originalmiracle drugs that first sparked the growth ofthe global research-based pharmaceutical indus-try in the postndashWorld War II period as well aslater generations of molecules that have beenintroduced in the last four decades

Among systemic antibacterials quinolonesare the latest generation molecules available inIndia We focus our analysis on quinolones forseveral reasons First quinolones are the drug ofchoice for a large number of bacterial infec-tions some of which are also treated by alter-native drugs (see Table A1 in the Appendixwhich outlines the spectrum of activity for eachmolecule family within the antibacterials seg-ment) Hence if there were one product groupfor which we would expect to have many sub-stitutes readily available this would be quino-lones Second with a share of 20 percent in thesales of systemic antibacterials quinolones rep-resent one of the largest subsegments within thistherapeutic category Finally several moleculeswithin the quinolone subsegment were still un-der patent in the United States at the time of ourinvestigation Table 2 details the basic informa-tion about the four quinolone molecules that arethe focus of our analysis The first row showsthe year of US patent expiry this ranges from1998 for norfloxacin to 2010 for sparfloxacinQuinolones include in principle four moremolecules that are listed at the bottom of Ta-ble 1 the market shares of these molecules arenegligible however so that we exclude thesemolecules from our analysis

Table 2 reveals several other interesting factsabout competition in the quinolone market inIndia First note the large number of firmsoperating in this subsegment The large numberof domestic firms is perhaps not that surprisinggiven that pharmaceutical product patents werenot recognized in India9 What is more surpris-ing is the number of foreign firms selling pat-ented products (eg ciprofloxacin) the fact that

multiple foreign firms sell a patented productindicates that such firms often ldquoinfringerdquo10

patent laws in India while complying with themin developed world countries The last two rowsof Table 2 further indicate that domestic prod-ucts often sell at a premium With the exceptionof ofloxacin the average prices of products of-fered by Indian firms are higher than the pricesof products offered by foreign subsidiaries Thispreliminary evidence suggests that Indian con-sumers do not place a premium on the brandname and reputation of big multinational phar-maceutical concerns Moreover the higherprice of domestic products does not seem toprevent domestic companies from capturing alarge market share This is most evident in thecase of ciprofloxacin where domestic firmshave with 53 percent the largest share in thetotal sales of quinolonesmdashthis despite the factthat the average price of these products is 10percent higher than the price of foreign productscontaining the same molecule

Table 3 provides additional summary statis-tics for our data broken down by region Thefirst two rows of the table report the averageannual household expenditure on quinolonesand antibiotics respectively Note that in bothcases the average expenditure is higher in thenorth and west these regions include states withhigher per capita incomes and tend to be moreindustrialized and urbanized than those in theeast and south Pharmaceutical products areavailable in multiple presentations that is com-binations of dosage forms (eg capsule tabletsyrup) strength (eg 100 milligrams 500 mil-ligrams) and packet sizes (eg 50 capsule bot-tle 100 tablet bottle) The various presentationsin which a product is available are often referredto as stock-keeping units or SKUs11 The num-ber of SKUs for each product group withinquinolones is reported at the top of Table 3 Aswith the more aggregate numbers on firms and

8 In addition to anti-bacterials this segment also containsantivirals

9 Accordingly the common distinction between ldquobrandedrdquoand ldquogenericrdquo products is irrelevant here

10 We emphasize here that the word ldquoinfringerdquo belongsin quotes because patent laws do not currently exist inIndia infringement in the legal sense is not possible It ishowever striking that the same firms that accuse Indianproducers of ldquopiracyrdquo sell in India products that are patentedin the United States and for which the patent is held by adifferent multinational corporation

11 For instance a 100 capsule bottle of 100 milligramcapsules of a particular branded drug and a 50 capsulebottle of 100 milligram capsules of the same branded drugwould be identified as two separate SKUs

1484 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

products reported in Table 2 the differencebetween domestic and foreign products isstriking The number of SKUs offered byIndian firms is consistently larger than thenumber offered by subsidiaries of foreignmultinationals The number of SKUs variesslightly across regions but more importantlyit varies across time as some SKUs disap-pear while new ones get introduced duringour sample period

Many pharmaceutical products in India aresubject to price controls12 While the specificsof the price regulation are too complex for anyeconomic model to capture adequately themain concern for the empirical analysis is thatprice controls may lead to a lack of price vari-ation over time so that the demand functioncannot be identified Prices at the most disag-gregate SKU level are relatively stable overtime there are variations due to occasionalchanges in the estimated cost (due for exampleto changes in exchange rates that affect the costof imported materials or bulk drugs) but suchvariations tend to be infrequent and small inmagnitude The degree of time variation ishowever substantially larger once one aggre-gates to the product level This variability stemsnot only from the fact that the SKUs over whichwe aggregate may experience changes in theirrespective prices at different points in time butalso from the fact that the range of SKUs of-fered in the market does not remain constantover time The entry and exit of presentationswithin the same product group that have differ-

ent prices effectively affects the price that con-sumers face for this drug in each period

The middle portion of Table 3 reports themean price and standard deviation for eachproduct group by region Prices vary by regionthough there is no clear pattern emerging fromthe table with regard to the cross-regional vari-ation (in the sense of some regions being sys-tematically more expensive than others) Toexamine what portion of the total price variationis due to time versus regional variation weconducted an analysis of variance of prices thatwe report in Table 4 The table is based onseparate regressions for each product group(pooling data across groups with big differencesin their average prices is not particularly infor-mative as most of the price variation is ac-counted for by product group dummies) Thelast two columns of the table show the fractionof price variation accounted for by region andtime dummies respectively As evident fromthe table a significant fraction of the total vari-ance in prices can be attributed to time varia-tion In the demand estimation we include a fullset of product groupndashspecific regional dummiesso that the price parameters are identified en-tirely based on this time variation within eachproduct group The time variation of productgroup prices is driven primarily by composi-tional changes within each group the revenueshares of the individual SKUs that compriseeach product group change over time (see therelated discussion in Section IIC) while in eachperiod there is entry and exit of SKUs into thesample To check whether this pattern reflectsgenuine entry and exit as opposed to samplingvariation we examined the revenue shares ofthe SKUs that leave the sample relative to theones that remain during the entire period Theresults are reported in Table 5 While the SKUs

12 The details of the procedures for price fixation canbe found in the official government Web site httpwwwnppaindianicinindex1html under the link ldquoDrug Price Con-trol Order 1995rdquo A new pharmaceutical policy was introducedin 2002 but our data were collected before that year

TABLE 4mdashANALYSIS OF PRODUCT PRICE VARIANCE

Product groupPartial SS

timePartial SS

region Total SSPercentage

explained by timePercentage

explained by region

Foreign ciprofloxacin 0296 0306 1047 283 292Foreign norfloxacin 1036 1002 4179 248 240Foreign ofloxacin 0429 0021 0571 752 36Domestic ciprofloxacin 0005 0088 0104 50 841Domestic norfloxacin 0059 0098 0198 296 494Domestic ofloxacin 2858 0104 3056 935 34Domestic sparfloxacin 1754 0032 1860 943 17

1485VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

that exit tend to be smaller (their average shareis 1 percent as opposed to 34 percent for thoseSKUs that are present during the entire sampleperiod) the shares of the two groups do notseem orders of magnitude apart13 In additionour data cover only the 300 largest firms sellingin the Indian market so that firms with verysmall shares are not included in our sample

II The Analytic Framework and EstimationApproach

Patent enforcement in the Indian pharmaceu-tical market will have the effect of eliminatingdomestic products whose active pharmaceuticalingredients are protected by (foreign) patentsThus assessing the effects of patent enforce-ment is tantamount to assessing the effects ofwithdrawing domestic products from the mar-ket This task is the converse of evaluating newproduct introduction accordingly the concep-tual framework we use to address the questionsof interest is similar to the one developed in theliterature for the valuation of new goods14

We start by estimating demand for quino-lones Given that the market is characterized byimperfect competition the counterfactual anal-ysis requires that we also model the supply sideas removal of one product will affect the pricesof other products especially those that are closecompetitors The existence of price regulationin the Indian pharmaceutical market imposespotential constraints on firmsrsquo maximizationproblem Given these constraints and the com-

plexity of the price regulation process the typ-ical approach of deriving estimates of actualmarginal costs and markups by exploiting thefirst-order conditions of profit maximizing firmsdoes not seem particularly promising Insteadwe use our demand estimates to place upper andlower bounds on marginal costs and markups

With demand elasticities and upper and lowerbounds for marginal costs in place we thenconduct counterfactual simulations We con-sider several alternative scenarios dependingon the number of domestic products affected bypatent enforcement For each scenario we com-pute the counterfactual prices and use them toassess the effects of domestic product with-drawal on consumer welfare (as measured bythe compensating variation) firm profits andsocial welfare As with the valuation of newproducts the big conceptual problem facing thispart of the analysis is that we need to extrapo-late from the region of the data to the point atwhich demand for the products that exit themarket becomes zero This conceptual issue ispresent in any attempt to evaluate a major pol-icy change for which no historical precedentexists like the enforcement of patent laws inIndia One advantage of the present study is thatwe have a limited set of price data for Pakistana country with similar demographics to Indiabut with a market structure that resembles theone that would emerge in India under patentenforcement (monopoly of multinational sub-sidiaries) By comparing the prices of productsoffered by multinationals in Pakistan to thosewe compute in our counterfactual simulationsfor the products that would be offered by mul-tinationals in India if patent laws were enforcedwe can get a sense of how plausible our coun-terfactual estimates are

A Demand

The demand modelling is based on the multi-stage budgeting approach Our primary motiva-tion for adopting this approach is a practical oneIn the multistage budgeting approach the depen-dent variable is defined as a revenue share whichis appealing given that the products we include inthe analysis contain different molecules (ie ac-tive pharmaceutical ingredients or APIs) Eventhough we do have data on the quantity of therelevant API (eg 100 milligrams of ciprofloxa-cin) contained in each product converting the

13 While we cannot completely rule out the possibilitythat some of the exit is due to sampling variation note thatthe latter should be reflected in low precision of the demandparameter estimates The demand parameters however areprecisely estimated

14 See Manuel Trajtenberg (1989) Jerry A Hausman(1994) and Timothy F Bresnahan (1997) for representativeexamples and a discussion of the relevant issues

TABLE 5mdashREVENUE SHARES OF THE EXITING SKUS

Revenue share ofexiting SKUs

Revenue shares ofall SKUs

Full sample 10 34Northern region 08 33Eastern region 12 36Western region 10 34Southern region 12 33

1486 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

revenue shares to physical shares is extremelydifficult if not infeasible in the case of antibioticsBecause such drugs are ldquosystemicrdquo by nature theyare used to treat a large number of infections andthe dosage of each drug depends on the particularinfection it is supposed to address (for example thedosage will differ depending on whether the anti-biotic is used to combat an ear infection or tuber-culosis) This particular feature of antibioticscomplicates the conversion of revenue to physicalshares

The basic idea of the multistage budgetingapproach is to use the therapeutic classificationof a productmdashie the therapeutic segment andsubsegment the product belongs tomdashto orga-nize all products in the systemic antibacterialssegment into a hierarchical taxonomy consist-ing of two levels At the higher level are thevarious subsegments of systemic antibacterialsThe first stage of budgeting corresponds to theallocation of expenditures across the subseg-ments in this upper level of the taxonomy

In the second stage of the budgeting processcorresponding to the lower level of the taxon-omy a flexible functional form is adopted tomodel how the expenditures allocated to eachsubsegment are distributed across the productswithin that subsegment In particular to modeldemand at the second stage we employ theAIDS specification proposed by Deaton andMuellbauer (1980b)15

While the two-stage demand estimation ap-proach offers functional form flexibility its ap-plication to the Indian systemic antibioticsmarket poses a few problems The first is thatdue to entry and exit many SKUs and evenproducts in our sample are not present in everyperiod AIDS does not have a good way ofdealing with a varying number of products as itwas developed with broad commodity catego-ries in mind which are consumed by all con-sumers every period To solve this problem weaggregate within each subsegment (eg quino-lones) SKUs into product groups where withineach product group all SKUs contain the samemolecule and are produced by firms with thesame domesticforeign status Specifically let aSKU k be indexed by its molecule (or API) M its

domesticforeign status DF indicating whether itis produced by a domestic (Indian) or a subsidiaryof a foreign (multinational) firm a particular pre-sentation s and the particular firm f that producesit We aggregate SKUs over presentations andfirms to obtain a newly defined product group iwhich is indexed only by molecule M and domes-ticforeign status DF and has revenue Ri yenfsRk with i (M DF) k (M DF f s) and pricepi yenfs kpk where k denotes the conditional(on M and DF) revenue share of this particularproduct ie

(1) k Rk

Ri

In most cases the resulting product groups arebroad enough to be present every period16

The usual concern with this aggregation pro-cedure is that it may lead us to overstate firmsrsquomarket power as we ignore competition amongfirms with the same domesticforeign statusproducing the same molecule In the presentapplication however this concern is unlikely tobe of great importance as the effect of patentenforcement is to wipe out all domestic compe-tition at once while granting foreign firms mo-nopoly power hence competition among firmsfor patented molecules becomes irrelevant Theaggregation according to the domesticforeignstatus (within a particular molecule) thus corre-sponds to the scope of our analysis and theparticular questions of interest

The second problem is that for our approachto be useful in welfare analysis the allocation oftotal expenditures to group expenditures at thehigher stage has to be modelled in a way con-sistent with utility maximization In general thesolution of this allocation problem requiresknowledge of all individual product prices

15 Representative applications of the multistage budget-ing approach include Sara F Ellison et al (1997) Hausman(1994) and Hausman and Gregory K Leonard (2002)

16 We are missing only four observations (ie monthregion combinations) all for the drug group of ForeignNorfloxacin August 1999 in the south May 2000 in thewest October 2000 in the south and November 2000 in theeast In these cases we set the revenue shares of ForeignNorfloxacin equal to zero In general with 01 percent ofquinolone sales (see Table 1 row 2) Foreign Norfloxacinhas a very small share of the market This probably explainswhy the results pertaining to this drug group are unreliableit is the only drug group for which we do not obtain asignificant price elasticity of demand while its cross-priceelasticities with other foreign drug groups often have thewrong sign

1487VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

From an empirical point of view this is notparticularly useful as it eliminates all compu-tational advantages of the two-stage approachTo address this problem we adopt an approxi-mate solution to model the higher-level expen-diture allocation along the lines suggested byDeaton and Muellbauer (1980b pp 131ndash32)This gives rise to a two-level AIDS specification

Consider the lower-level estimation firstwhich refers to the allocation of a particularsubsegmentrsquos expenditure to the product groupswithin the subsegment In our application therelevant subsegment is quinolones which weindex with Q Let the product groups within thissubsegment be indexed by i 1 N pi be theprice of product group i (where as noted abovei refers to a particular molecule and domesticforeign status combination) and XQ be the totalexpenditure on the quinolone segment The rev-enue share of each product group is given by

(2) i i j

ijln pj ilnXQ

PQ

where i the revenue share of product group iis defined as

(3) i pi qi

yenj

pj qj

xi

XQ with i j Q

XQ is the overall expenditure on the quinolonesubsegment and PQ is a price index given by

(4) ln PQ ap 0 i

iln pi

1

2 i

j

ijln piln pj

With a limited number of product groups and asufficiently large number of time-series observa-tions the flexibility implied by the AIDS modeldoes not impose too many demands on the data Inthe present application however where the num-ber of observations is limited the AIDS model isnot estimable in this general form To reduce thenumber of parameters that need to be estimatedwe impose two sets of restrictions

The first set of restrictions is implied by thetheory of utility maximization Specificallythese restrictions are

Adding-up yenk k 1 yenk k 0 yenk kj 0 j

Homogeneity yenk jk 0 j Symmetry ij 1frasl2 [ij ji] ji This last

restriction by itself reduces the number of parameters to [N(N 1)]2

The second set of restrictions we impose aimsat further reducing the number of parameters tobe estimated by exploiting our knowledge of thisparticular market Specifically for each productgroup i we allow one ij parameter for all productgroups j that have different molecules from prod-uct group i and are produced by foreign firms andone ij for all product groups j with differentmolecules produced by domestic firms We donrsquotimpose any restrictions on the ij parameter whenproduct group j has the same molecule as productgroup i (By construction product groups i and jcontain products produced by firms with differentdomesticforeign status)

To better illustrate the nature of the restric-tions we impose on the patterns of substitutionacross products some additional notation isneeded Let d(i j) be an indicator of the degreeof similarity (or difference) between productgroup i and product group j along the dimen-sions we are able to observe (molecule M anddomesticforeign status DF) For any two prod-uct groups i and j d(i j) can take on one of thefollowing three values17

(5)

di j (1 0) if Mi Mj DFi DFj

(0 1) if Mi Mj DFi DFj

(0 0) if Mi Mj DFi DFj

Let

(6) Diab j di j a b

17 The sequence (1 1) is not possible for two differentproducts in this case the parameter corresponds to theproductrsquos own-price effect that is ii

1488 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The equation at the lower level becomes

(7) i i iiln pi i10ln pj j Di10

j Di

01

i01ln pj

j Di

00

i00ln pj ilnXQ

PQ

Note that

The parameter ii captures a product grouprsquosown-price effect (there will be as many iiparameters as the number of product groups)

The parameter i10 captures the cross-priceeffect of the product group containing prod-ucts with the same molecule but produced byfirms of different nationality

The parameter i01 captures the cross-priceeffects of product groups containing productswith different molecules but produced byfirms with the same nationality

The parameter i00 captures the cross-priceeffects of product groups containing productswith different molecules produced by firms ofdifferent nationality

Before we take the demand equation to thedata we make two modifications The first oneis to let the product-specific effects i vary byregion r The resulting product-specific regionaleffects ir have two interpretations first theycontrol for the ldquoqualityrdquo of each drug withquality differences being allowed to vary acrossregions second they proxy for demographicsand other demand shifters which vary by re-gion and may affect the demand of each prod-uct group differently18 Note that by includingproduct-specific regional effects in the demandspecification we estimate the price parametersbased on the within-product-group variation ofprices in each region In an earlier version of thepaper we also estimated the demand systemwithout regional dummies and obtained similarresults

The second modification that allows us to gofrom a deterministic to a stochastic specificationof the demand equation is to include an additiveerror term in (7) The latter takes into accountthe fact that (7) is not expected to fit the dataexactly The error term irt accounts for mea-surement error (due to the fact that the productgroup prices pjrt we employ in the estimationare not exact price indices but approximationsthereof) and (potentially region-specific) de-mand shocks that may affect the demand for aproduct in a particular period Examples of suchshocks include an advertising campaign for aparticular product which temporarily increasesthe demand for this product or the outbreak ofa (potentially region-specific) epidemic whichcalls for the use of a particular drug We discussthe interpretation and properties of this errorterm in more detail in the next section

The final form of the equation we estimate atthe lower level becomes (with subscript t de-noting month and subscript r denoting region)

(8) irt i ir iiln pirt i10ln pjrt j Di10

j Di

01

i01ln pjrt

j Di

00

i00ln pjrt

ilnXQrt

PQrt irt

The analysis so far has conditioned on the ex-penditure allocated to the quinolone subseg-ment XQ The upper level of the estimationconsiders the problem of allocating total expen-diture across the different systemic antibioticssubsegments one of which is quinolones Theupper-level demand function is given by

(9) G G H

GHln PH GlnX

P

where all variables denoted by capital letters aredefined as before but now refer to subsegments(G H ) rather than product groups within asubsegment and the total expenditures on sys-temic antibiotics X are deflated by the Stone priceindex log P yenH Hlog PH When estimating thesystem above we impose all the restrictions

18 Given the short time span of our sample typicaldemand shifters such as age distribution income distribu-tion and education hardly change over our sample periodSuch shifters are therefore absorbed by the region-specificproduct fixed effects ir

1489VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

implied by utility maximization as we do withthe estimation of the lower-level AIDS We donot however impose any additional restrictionson the substitution patterns at this stage so thatthe cross-price effects across segments remainrelatively unconstrained

Estimation of the higher-level AIDS allowsus to obtain the unconditional own- and cross-price elasticities that are used in the formulationof the supply problem and welfare analysisThese are given by the formula

(10) ij ijXQXQ

ln qi

ln XQ

ln XQ

ln PQ

ln PQ

ln pj

with the conditional cross price elasticitiesgiven by

ijXQXQ ij ij jlnXQ

PQ

i

As in the lower stage we include subsegment-specific regional dummies in the specificationof the upper-stage demand system so that thefinal form of the estimating equation at theupper stage becomes

(11) Grt G Gr H

GHln PHrt

GlnXrt

Prt Grt

In sum the demand system we take to thedata is represented by equations (8) and (11)and the associated parameter restrictions im-plied by economic theory

B Modelling the Supply Side of the Market

Counterfactual simulations concerning theeffects of domestic product withdrawal requireknowledge of the marginal costs of pharmaceu-tical firms operating in the Indian market Theseare unobservable The usual approach in theNew Empirical Industrial Organization litera-ture has been to exploit the firm equilibriumconditions to infer marginal cost For exampleit is often assumed that the marginal cost ci isconstant and that the industry is an oligopoly

engaging in Bertrand competition with differen-tiated products Assuming that firms myopicallymaximize profits each period one can then de-rive the firmsrsquo first-order conditions that corre-spond to the assumptions above about costsmarket structure and firm behavior

We deviate from this procedure as the presenceof price regulation renders the assumption of un-constrained period-by-period maximization un-tenable Ideally one would like to explicitlyincorporate the price and other administrative con-trols into the firmrsquos optimization problem andderive the first-order conditions under the assump-tion of constrained maximization The complexityof the price regulation however makes this ap-proach infeasible Therefore we adopt an alterna-tive approach that does not rely on modelling theprice-setting process but is instead based on de-riving upper and lower bounds for marginal costsand markups

In particular an upper bound for marginalcosts and a lower bound for the markups (zero)can be derived under the assumption of perfectcompetition While this assumption is clearlyunrealistic in the pharmaceuticals market it isuseful in providing an upper bound for costs ci

Uwhich will be given by

(12) ciU pi

On the other hand a lower bound for marginalcosts ci

L (and upper bound for markups) can bederived by assuming that there is perfect collu-sion within each product group i (where i refershere to a moleculedomestic-foreign combina-tion)19 and ignoring price controls so thatprices are determined by the first-order condi-tion of the jointly profit-maximizing firmswithin product group i Solving this first-ordercondition for the (lower bound of) marginal costthen gives

(13) ciL pi 1

1

iipipj

19 The reason that this assumption will lead to an under-statement of marginal costs (and hence overstatement ofmarket power) is that it assumes away competition amongfirms within each drug group

1490 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

where ii(pipj) the own-price elasticity of de-

mand for product i will depend on the productrsquosown-price pi and all other productsrsquo prices pjwith i j

Once we have obtained the demand elastic-ities through estimation of the demand sys-tem we can calculate the upper and lowerbounds for marginal costs and correspondingmarkups according to (12) or (13) These willthen be employed in the counterfactual sim-ulations While our counterfactuals use boththe lower and upper bounds for costs most ofour discussion will be based on using thelower bounds for costs since this is the moreinteresting case in the policy analysis it givesus the largest possible profits for pharmaceuticalfirms and hence corresponds to the worst pos-sible scenario facing Indian firms and the bestpossible scenario facing multinationals underTRIPS

C Identification Assumptions and EstimationApproach

The discussion of the demand system has sofar abstracted from the issue of price endogene-ity The usual premise in the industrial organi-zation literature is that correlation of prices withthe error term in the demand equation arises byvirtue of the first-order conditions of profit-maximizing firms20 As we discuss in this sec-tion this source of simultaneity bias is unlikelyto be of major concern in the present contextbecause of the existence of price regulation Ourprimary concern is instead with the simultane-ity bias that is implied by the particular wayprices are constructed

To understand the sources of the simultaneitybias note first that the price of product group piwe employ in the demand estimation should bethought of as a proxy for an exact price indexthat we do not observe As such pi contains bydefinition measurement error and it will becorrelated with the error term of the demandequation Specifically the demand equation weare interested in estimating can be written insimplified form (suppressing the subscripts rand t and ignoring parameter restrictions forconvenience) as

i i j

ijln pjT ilnXQ

PQ i

where pjT denotes the true (exact) price index for

product group i and i captures unobservedvariables that may affect demand in a particularperiod (eg an advertising campaign or an ep-idemic) The price index pj

T depends on theSKU prices pk Note that in the presence ofSKU entry and exit into the sample the priceindex pj

T will vary over time even if the pricesof the individual SKUs remain stable as theproducts (SKUs) that comprise each productgroup change each period

Ideally we would like to compute the exactprice index pj

T for each group and employ theseindices in the demand estimation Under theassumption of predetermined SKU prices eachperiod the exact price indices would be uncor-related with the error term and the demandparameters could be estimated by simple OLSUnfortunately this strategy is not feasible in thecurrent context

To derive an exact price index for each prod-uct group it is necessary to model consumerrsquoschoice among SKUs conditional on the choiceof the group To this end it is necessary toexplicitly introduce a third stage in the demandspecification and estimate the parameters asso-ciated with the choice at that stage Howeverany specification based on three-stage budget-ing (that would require directly specifying ademand function associated with the SKUchoice at the third stage and imposing additiveseparability at the higher stages) would be in-feasible to estimate for three main reasonsFirst there is a large number of SKUs withinsome groups so that the demand parameters can-not be estimated even if one imposes parameterrestrictions second the SKU prices exhibit littlevariation over time lastly the choice sets at thislowest stage vary over time rendering estimationof the price parameters associated with SKUs thatare not available in all periods infeasible Alterna-tively one could abandon the multistage budget-ing altogether and adopt a discrete choiceapproach which is inherently better suited to deal-ing with varying choice sets and limited pricevariation A discrete choice approach would beparticularly appealing for modeling the SKUchoice conditional on product group choice giventhat within each group all SKUs contain the same

20 See Steven Berry et al (1995) or Aviv Nevo (2001)for a related discussion

1491VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

molecule and are hence comparable As notedearlier however a discrete choice approach at thehigher stages of the demand estimation requiresconverting revenue to physical shares which ischallenging in the case of systemic drugs such asantibiotics for which the appropriate dosage is notwell defined Finally a combination of a multi-stage budgeting approach for the higher stages anda discrete choice approach for the lowest stagewould be feasible to estimate but inconsistentwith a model of utility maximization For thesereasons we have no alternative but to adopt anapproximation for measuring the price index foreach product group

Let us denote this approximation by pjA and let

13j denote the proportional ldquoapproximationrdquo errorassociated with measuring the true price index forgroup j so that ln pj

A ln pjT ln 13j The

estimating demand equation can then be written as

i i j

ijln pjA ilnXQ

PQ

j

ijln 13j i

or

i i j

ijln pjA ilnXQ

PQ i

where the new error term i yenj ijln 13j iis comprised of two components the first onereflects ldquomeasurementrdquo error due to the fact thatthe product group prices we employ are approx-imations and not exact price indices and thesecond one captures conventional demandshocks The demand equation as written abovecorresponds to the equation we take to the data

The particular proxy we use for the exact priceindex for each product group j is the revenue-share-weighted average of the prices of multipleSKUs available within this group that is

pjA

k j

k pk

The notation above illustrates the two source ofprice variation variation in the SKU prices pkand variation in the weights k

We are concerned with potential correlation ofthe so-constructed product group prices with bothcomponents of the product group demand error

term The correlation between pjA and the first

component yenj ijln 13j is independent of theparticular way we construct pj

A and inherent in thefact that we measure the exact price index witherror The potential correlation between pj

A and thesecond component of the product group demanderror the shock i arises however because of thespecific way we construct the proxy pj

A and inparticular because of the presence of the reve-nue share weights k in pj

A since the SKUrevenue share weights k will generally dependon the product group expenditure they arelikely to be correlated with the product groupdemand shock i

To address the simultaneity bias we use in-strumental variables To this end we need vari-ables that are correlated with the proxies pj

A butuncorrelated with the error term i We use thenumber of SKUs within each product group j asan instrument as it does justice to the idea thatvariation in the product group price index stemsin part from variation in the set of SKUs that areavailable each period It is clearly correlatedwith the average group prices pj

A and plausiblyuncorrelated with the demand shock i Theassumption that the number of SKUs is uncor-related with the conventional demand error willhowever be violated if the introduction of anew SKU changes the perceived quality of adrug21 or if it is accompanied by promotionalactivities which will be reflected in i Our hopeis that this is not too often the case Assumingthat the SKU number for each group j is alsouncorrelated with the weighted sum of the mea-surement errors of all product groups yenj ijln 13j itcan be used as an instrument for the product groupprice pj

A22 If one accepts the premise that prices atthe SKU level are exogenous then SKU pricescan also be used as instruments The final list ofinstruments we use hence includes the number ofSKUs in each group the prices of the five largestSKUs for each group and all other exogenousvariables in the demand estimation such as re-gionalproduct dummy interactions and the upper-

21 Note that we already proxy for the quality of each prod-uct group through the product fixed effects i In the actualestimation we actually let these fixed effects be region-specific(ir) so that we allow quality to vary across regions

22 If the measurement error increases in the number ofincluded SKUs this assumption will be violated Unfortu-nately we have no way of checking the validity of thisassumption

1492 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

level total expenditure on antibiotics Regressionsof group prices on the instruments above yieldhigh R-squares with most regressors highly sig-nificant indicating that our instruments are highlycorrelated with prices23

Our sample includes four molecules cipro-floxacin norfloxacin ofloxacin and spar-floxacin Except for sparfloxacin all othermolecules are produced by both foreign anddomestic firms24 So we have seven products(domestic ciprofloxacin foreign ciprofloxa-cin etc) with 96 observations (two years ofmonthly data four geographical regions foreach period) for each product The parametersin the lower-level AIDS demand system asdefined in equation (8) are the product fixedeffects i the productndashregional dummy inter-actions ir the own revenue-share price elas-ticities ii the cross revenue-share priceelasticities i10 i01 i00 and the revenue-share expenditure elasticities i In estimatingthe parameters we first regress prices on allinstrumental variables and then plug the pre-dicted values for prices in the constrainedleast-square regression (For a detailed expla-nation of the constraints see the previoussection)

Given that we impose many cross-equationconstraints and employ instrumental variablesin the estimation it is difficult to derive stan-dard errors for the parameter estimates ana-lytically Our error term interpretation in (8)implies that the error terms for each productgroup are likely to be correlated across re-gions for example if a national advertisingcampaign increases the demand for a rela-tively expensive presentation in one regionsimultaneously increasing the aggregate de-mand and price index for the correspondingproduct group in that region it is likely that thesame effect will be observed in other regions Inprinciple we could exploit the cross-regionalcorrelation in the error terms to estimate thedemand system using seemingly unrelated re-

gression (SUR) This however would requireus to derive the variance-covariance matrix ofthe parameters analytically which as notedabove is cumbersome in the current contextInstead we use the bootstrap method choosingto remain agnostic about the structure of thevariance-covariance matrix The potential dis-advantage of this approach over SUR is a loss inefficiency but as will become apparent in theresults section the model parameters are fairlyprecisely estimated To maintain the marketstructure we randomly sample the periods (withreplacement) and use the same periods for allproducts Regarding the optimal number ofbootstrap repetitions ideally one would followthe three-step method proposed by Donald W KAndrews et al (2000) Empirical evidence sug-gests however that one rarely needs more than200 replications to estimate the standard er-rors25 To be safe we generate 300 bootstrapsamples (with replacement) based on the origi-nal data and estimate the standard errors usingthe standard errors of the bootstrap sampleestimates

The estimation of the top-level AIDS systemis similar The constraints imposed on this top-level demand system are adding-up homogene-ity and symmetry as defined on page 12Again bootstrapping is used to obtain the stan-dard errors of the parameter estimates

D The Counterfactual Scenarios

In assessing the effects of patent enforce-ment we start by focusing on the most extremecase in which compulsory licensing is not anoption and foreign firms are not subject to pricecontrols We use the results from the analysis ofthis case as a benchmark In reality the outcomeof the WTO negotiations is more likely to in-volve some constraints on the monopoly powerof foreign firms selling patented products indeveloping countries such as price caps orcompulsory licensing Our framework can eas-ily accommodate these cases as will becomeapparent in the next subsection

We now focus on the effects of potentialpatent enforcement in the quinolone segmentWe consider several scenarios which vary inthe number and size of domestic products that

23 Specifically the R-squares from the first-stage regres-sions of product group prices on instruments range from057 to 095 except for domestic ciprofloxacin for whichthe R-square is 017

24 Sparfloxacin is actually offered by one foreign subsid-iary in India (see Table 2) However its revenue share isminiscule We therefore treat Sparfloxacin as being pro-duced by domestic firms only 25 See Bradley Efron and Robert J Tibshirani (1993)

1493VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

will be removed from the market In particularwe consider the following five scenarios26

Withdrawal of one large domestic productgroup only domestic ciprofloxacin

Withdrawal of one relatively small domesticproduct group only domestic ofloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand norfloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand sparfloxacin

Withdrawal of all four domestic quinoloneproduct groups

As the list above suggests we proceed fromanalyzing the effects of single-product with-drawal to the analysis of eliminating the en-tire domestic segment This approach wasmotivated by early empirical results that in-dicated that the existence and extent of com-petition from domestic firms has a significantbearing on the predicted effects of patent en-forcement that is our predictions regardingprices and welfare vary substantially depend-ing on how many domestic products are af-fected by patent enforcement and on whethersome domestic competition will remain presentafter TRIPS In addition the size of the affecteddomestic groups is relevant for the welfare pre-dictions accordingly we examine both scenar-ios in which a large domestic product groupsuch as domestic ciprofloxacin is eliminatedfrom the market and scenarios in which the

eliminated domestic group is relatively small(eg domestic ofloxacin) In all scenarios wemaintain the assumption that the set of prod-ucts offered by the remaining firms in themarket does not change in response to patentenforcement

E Computation of Virtual Prices and NewEquilibrium Prices

The first step in the counterfactual analysisis to derive the new equilibrium prices underpatent enforcement27 In this context thereare two sets of prices that are relevant Thefirst set consists of the virtual prices of those(domestic) products that will not be availableonce TRIPS is put in effect To calculate thesevirtual prices we set the expenditures of therelevant products equal to zero The secondset of prices consists of the prices of thoseproducts that remain in the market In deriv-ing these prices we start by assuming profitmaximization without any form of price reg-ulation the firms remaining in the marketreoptimize in response to the policy changeand set new prices taking the prices of allother firms as given28 Of course at the equi-librium all prices change in response to thefact that some domestic products are nolonger present The new equilibrium pricesfor products that remain in the market are thuscomputed by utilizing the first-order condi-tions of profit-maximizing firms into whichthe virtual prices of the eliminated productsare substituted Hence to compute the newequilibrium prices we solve an equation sys-tem of the following form

26 As Table 2 indicates most of the patents for the drugsin the quinolone segment have expired by now so that noneof the scenarios described in this section is going to mate-rialize in practice The reason we consider these alternativescenarios in our counterfactual simulations is to get a senseof how the presence and extent of domestic competitionaffects welfare calculations Such calculations may becomerelevant in the future in other therapeutic classes in whichpatents have not expired yet and where domestic firmscompete with foreign patent holders While the particularexit scenarios will depend on the exact year of patent expiryof drugs in each therapeutic class and the specific way inwhich patents will be enforced (eg whether certain do-mestic drugs will be grandfathered in) the general conclu-sion that emerges from our calculations is that the consumerwelfare loss is substantially smaller in cases where somedomestic competition remains present in the market Hencethe particular way in which patent enforcement is imple-mented is essential for assessing its welfare impact

27 Of course until product patents are in fact introducedthese prices will not be observable Note also that we areassuming here that the range of available products will notchange with the introduction of patents

28 As mentioned above this first set of calculations ab-stracts from the existence of remaining price controls orother government regulations that would impose constraintson the firmsrsquo profit maximization problem Accordingly theresulting numbers should be interpreted as a benchmark ofwhat would happen if markets were completely unregulatedTo examine what the welfare effects of TRIPS would be inthe more realistic scenario of price regulation we subse-quently consider a scenario in which regulation keeps theprices of the products provided by patent-holders at theirpre-TRIPS level For a more detailed discussion see theresults section

1494 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

For products i that are withdrawn from themarket

(14) 0 i ir iiln pirV i10ln pjr j Di

10

j Di

01

i01ln pjr

j Di

00

i00ln pjr

ilnXQr

PQr

For products k that remain in the market

(15) pkr ckr 1 1

kk pkrpjr pirV 1

In the equations above pirV denotes the virtual

prices of the products that are removed fromthe market while pjr denotes the updatedprices of all other products Note that whensolving for the virtual prices we account forthe fact that both the price index for quino-lones PQr and the expenditure allocated tothis subsegment XQr need to be updated toreflect the fact that as a result of the pricechanges there may be substitution away fromthis subsegment To obtain the new quinoloneexpenditure XQr and the new price index PQrwe use the estimates and formulas for thehigher-level AIDS system In equation (15)ck refers to the marginal cost for product k thatwe have obtained from the previous estimationstage As mentioned in the previous subsectionwe conducted the simulations using both theupper and lower bounds for marginal cost ck

u

and ckL respectively The term kk(pkr

pjrpirV) re-

fers to the unconditional own-price elasticity forproduct k which is a function of the eliminatedproductsrsquo virtual prices and the remaining prod-uctsrsquo new equilibrium prices We conduct thecounterfactual simulations and welfare analysisat the regional level because the presence ofregion-specific product effects in the demandestimation implies region-specific demand elas-ticities and hence region-specific marginalcosts and markups The presentation and discus-sion of our results focus on national averages ofthe relevant variables that we construct by com-

puting weighted averages across regions usingthe population of each region as a weight

F Welfare Assessment

The simulation of the new equilibrium underpatent protection can provide important insightsinto how consumers and firms will respond tothe removal of domestic products in the market(for example toward which products consumerswill substitute or which prices will increase themost) To get a more precise idea of how peo-plersquos well-being will ultimately be affected byTRIPS we compute as a last step in our anal-ysis the welfare effects of the policy changeSocial welfare is defined as consumer welfareand the sum of domestic firm profits Thechange in domestic profits can easily be calcu-lated by comparing the domestic firm (variable)profits at the pre-TRIPS prices to the profitsthese firms will realize at the new simulatedprices Although foreign firm profits do notcount in domestic welfare calculations we alsocompute the effects of patent enforcement onforeign firm profits to get an idea of how largethe expected benefits of TRIPS are for thesefirms This provides in some sense an indirectway of assessing whether the claims that patentenforcement in countries like India will lead tomore research on developing-country-specificdiseases (such as malaria) have any validity iffor example we find that the effect of patentenforcement on the foreign firm profits realizedin India is small in magnitude it is unlikely thatforeign firms will engage in more developing-country-specific research in response to TRIPSIt is important to note that in all these calcula-tions we work with the lower-bound estimatesfor marginal costs since these give rise to thehighest possible markups Hence our estimateof profit loss for domestic producers most likelyoverstates this loss

On the consumer side we measure changes inconsumer welfare by the compensating variation(CV) defined as the additional expenditure thatconsumers need in order to achieve the sameutility level as before patent enforcement at thenew prices Specifically let P0 denote the pricevector before patent enforcement P the simulatedprice vector post-TRIPS (that we obtained usingthe methods described in the previous subsection)u0 the utility attained by consumers before TRIPS

1495VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

and E(u P) the higher-level expenditure functionThen the compensating variation is given by

(16) CV Eu0 P Eu0 P0

Note that the CV as computed in (16) representsthe combination of three effects

The pure product-variety effect that is theeffect that arises because one or more prod-ucts are not available to consumers anymoreholding the prices of all other remainingproducts and the total expenditure on thequinolone subsegment XQ constant

The expenditure-switching effect that is theeffect arising from substitution away fromquinolones and toward other subsegments ofthe antibiotics market again holding the pricesof all other remaining products constant

The reduced-competition effect that is theeffect that arises because the firms remainingin the market adjust (increase) their prices inresponse to the removal of domestic products

From both an analytical and a policy point ofview it is desirable to assess how large each ofthese effects is Accordingly we decompose thetotal effect on consumer welfare (the CV as givenby equation (16)) using the following procedure

To get the pure product variety effect wecompute virtual prices for the products that areremoved from the market holding the quino-lone expenditure XQ and the prices of all otherproducts fixed Let us call the resulting pricevector P1 Then the pure product variety effectis represented by E(u0 P1) E(u0 P0)

To compute the expenditure-switching effectwe compute another set of virtual prices againholding the prices of all remaining products fixedbut letting quinolone expenditure adjust in re-sponse to the new price index for the quinolonesegment (Given that the prices of the remainingproducts remain fixed the change in the priceindex arises only because of the removal of one ormore domestic products) Note that this scenariomost closely resembles the case in which patentlaws are enforced but strict price regulation keepsthe prices of the products offered in the market attheir pre-TRIP level Let us label the so-computedprice vector P2 The expenditure-switching effectis then E(u0 P2) E(u0 P1)

Finally the reduced competition effect aris-ing from higher prices for the remaining prod-

ucts is computed as the residual change in thecompensating variation once the product-varietyand expenditure-switching effects have been ac-counted for that is E(u0 P) E(u0 P2)where the price vector P is computed accordingto the formulas (14) and (15) to reflect theadjustment of prices to the new regime

To compute the standard errors associatedwith the counterfactual simulations (that is thestandard errors for the counterfactual prices andwelfare estimates) we again use bootstrappingWe first bootstrap the original sample (sales andprices of all the drugs) 300 times Next we esti-mate the AIDS model for each of these samplesand compute the counterfactual equilibrium pricesand welfare losses corresponding to each policyscenario for each of the 300 simulations In thefinal step we compute standard errors29

One limitation of our framework is that itdoes not allow for a heterogeneous response ofconsumers to the policy change Accordinglyour framework is not suited to addressing thequestion of how different groups in the popula-tion will be affected by patent law enforce-ment30 Along the same lines our frameworkdoes not accommodate the possibility of pricediscrimination which might lead to differentresults regarding welfare losses and profit gainsrelative to uniform pricing31 Still we believethat the results of the counterfactual simulationscan provide important insights into the likelyaggregate response to patent enforcement andthe factors that drive this response

III Results

A The Structure of Demand

Tables A2 and A3 in the Appendix display theresults from estimation of the lower- and upper-level AIDS systems respectively For ease of in-terpretation rather than discussing the coefficientestimates we focus our discussion on the implied

29 In cases some of the simulations do not converge wecompute the standard errors using only those simulationsthat converged

30 Addressing this question would require at a mini-mum micro data on consumer purchases as in Goldberg(1995) To our knowledge such data do not exist for theIndian pharmaceutical market

31 See Ernst R Berndt (1994) for an extensive discussionof uniform pricing

1496 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

unconditional price and expenditure elasticities re-ported in Table 6 Given that the region-specificproduct group effects ir reported in the last fourcolumns of Tables A2 and A3 imply region-spe-cific demand elasticities we report separate elas-ticities for each region in Tables 6A and 6B Asevident from the comparison of these tables theelasticities are very similar across regions so thatwe can focus the remaining discussion on theelasticities of one region only the northern regiondisplayed in Table 6A

The diagonal terms of Table 6A report theown-price elasticities which are in all but onecase negative and highly significant The oneexception is the foreign norfloxacin productgroupmdashwhose share of quinolone sales is 007percentmdashfor which we estimate a negative butinsignificant own-price elasticity For the remain-ing product groups demand appears to be highlyelastic with the estimated elasticities lower than2 in four out of the six cases The magnitude ofthe own-price elasticities matches the features ofthe Indian pharmaceutical market mentioned ear-lier which would suggest that Indian consumersare likely to be quite price-sensitive32 The elas-ticities appear especially large if one takes intoaccount that they refer to product groups (such asdomestic ciprofloxacin) and not individual drugsoffered by particular firms Their relative magni-tudes are also intuitive the drug with the largestmarket share and a relatively high price (domesticciprofloxacin) appears to be one of the least elas-tic In contrast foreign ciprofloxacin is highlyprice elastic this is plausible as ciprofloxacindrugs offered by subsidiaries of multinationalsface the stiffest competition in this market seg-ment from approximately 75 Indian firms offeringthe same molecule

The estimated expenditure elasticities appear inthe last column These are all positive indicatingthat the demand for all product groups is normalThe remaining cells display the estimated cross-price elasticities As one might perhaps expect forproducts within a therapeutic subsegment theseare mostly positive Out of a total of 49 priceelasticities we estimate there are six that do notconform to expectations these are the cross-price elasticities between different foreign prod-

uct groups which are estimated to be negativeand significant33 Fortunately these elasticitieshave negligible impact on the welfare analysisgiven that our counterfactuals focus on theeffect of withdrawing one or more domesticproducts from the market the most relevantelasticities are the ones that capture the responseof various product group shares to a change in theprice of one or more domestic groups these elas-ticities are the cross-price elasticities between var-ious domestic groups and the ones betweendomestic and foreign groups which are plausibleand precisely estimated34

Regarding these elasticities a striking aspectof our estimates is how large positive andsignificant the cross-price elasticities betweendifferent domestic product groups aremdashin factfor norfloxacin and ofloxacin we estimate thatdomestic product groups containing differentmolecules are closer substitutes for one anotherthan product groups that contain the same mol-ecule but are produced by foreign firms Incontrast for ciprofloxacin (the molecule withthe largest revenue share) we estimate a largepositive cross-price elasticity between the do-mestic and foreign versions

The fact that domestic products appear to beclose substitutes for other domestic productsthat contain different molecules truly representsan ldquoempiricalrdquo finding in the sense that we do

32 In developed economies elasticities of this magnitude havetypically been found only for generic drugs (and even then onlyrarely) or among consumers who lack health insurance

33 While these elasticities are clearly counterintuitive theyare not inconsistent with the underlying demand system whichimposes no restrictions on the sign of the cross-price elastici-ties We do not have a good explanation of why these elastic-ities are estimated to be negative A potential explanation isthat the shares of the foreign products are very small giventhis we observe very few consumers switching from one (verysmall) foreign group to another (very small) foreign groupwhen the price of the first foreign group goes up and hence theinference is not very reliable in this case

34 The cross-price elasticities between foreign drug groupscontaining different molecules will also have an effect on thewelfare estimates given that the withdrawal of domestic prod-ucts will generally lead to changes in the prices of all foreignproducts and market shares will be reallocated from someforeign products to other foreign products based on the newprices However this reallocation from ldquoforeign to foreignrdquo istruly second order in our case compared to the reallocationfrom ldquodomestic to domesticrdquo and ldquodomestic to foreignrdquo prod-ucts In scenarios in which we consider patent enforcementaccompanied by strict price regulation the cross-price elastic-ities between foreign product groups are in fact completelyirrelevant as the prices of foreign products are not allowed toincrease in these cases Still our welfare loss estimates remainsubstantial drivenmdashas beforemdashby the loss of product variety

1497VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

not impose it through any of our assumptionsregarding the demand function The questionthat naturally arises then is what might explainthis finding While we cannot formally addressthis question anecdotal accounts in various in-dustry studies suggest that the explanation maylie in the differences between domestic andforeign firms in the structure and coverage ofretail distribution networks

Distribution networks for pharmaceuticals inIndia are typically organized in a hierarchicalfashion Pharmaceutical companies deal mainlywith carrying and forwarding (CampF) agents inmany instances regionally based who each sup-ply a network of stockists (wholesalers) Thesestockists in turn deal with the retail pharma-cists through whom retail sales ultimately oc-cur35 The market share enjoyed by a particularpharmaceutical product therefore depends inpart on the number of retail pharmacists who

stock the product And it is here that thereappears to be a distinction between domesticfirms and multinational subsidiaries In particu-lar the retail reach of domestic firms as agroup tends to be much more comprehensivethan that of multinational subsidiaries (IndianCredit Rating Agency (ICRA) 1999)36

There appear to be two reasons for this Thefirst is that many of the larger Indian firmsbecause they have a much larger portfolio ofproducts over which to spread the associatedfixed costs typically have more extensive net-works of medical representatives The second issimply that there are many more domestic firms(and products) on the market At the retail levelthis would imply that local pharmacists mightbe more likely to stock domestic products con-taining two different molecules say ciprofloxa-cin and norfloxacin than they would domesticand foreign versions of the same molecule Tothe extent that patients (or their doctors) arewilling to substitute across molecules in order tosave on transport or search costs (eg going toanother pharmacy to check whether a particular

35 There are estimated to be some 300000 retail pharma-cists in India On average stockists deal with about 75 retailers(ICRA 1999) There are naturally variations in this structureand a host of specific exclusive dealing and other arrangementsexists in practice Pharmaceutical firms also maintain networksof medical representatives whose main function is to marketthe companyrsquos products to doctors who do the actual prescrib-ing of drugs In some instances firms do sell directly to thedoctors who then become the ldquoretailerrdquo as far as patients areconcerned but these are relatively rare

36 These differences were also highlighted in conversa-tions one of the authors had with CEOs and managingdirectors of several pharmaceutical firms as part of a sep-arate study

TABLE 6AmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN THE NORTHERN REGION

Product group

Elasticity with respect to

Prices of foreign productgroups Prices of domestic product groups Overall

quinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 557 013dagger 015 401 011dagger 011dagger 016 137(179) (007) (007) (184) (006) (006) (006) (029)

Foreign norfloxacin 427dagger 045 427dagger 350dagger 602 451 465 220(242) (112) (242) (210) (623) (184) (183) (105)

Foreign ofloxacin 011 010dagger 138 009 009dagger 023 011 116(005) (005) (031) (027) (005) (028) (004) (017)

Domestic ciprofloxacin 018 001 001 168 008 008 010 117(008) (000) (001) (023) (002) (002) (002) (003)

Domestic norfloxacin 004 003 004 058 223 042 040 073(001) (003) (001) (017) (011) (004) (003) (009)

Domestic ofloxacin 005 005 011 077 074 342 074 089(002) (002) (013) (028) (008) (025) (008) (021)

Domestic sparfloxacin 007 004 007 115 063 063 288 028(002) (001) (002) (015) (006) (006) (017) (012)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent significance level and dagger (dagger) denotes significance at the 10-percent level

1498 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE 6BmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN OTHER REGIONS

Demand patterns in the eastern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 594 014dagger 016 431 013 011 017 140(195) (008) (008) (201) (006) (007) (007) (031)

Foreign norfloxacin 329dagger 058 329dagger 264dagger 460 346 359 192(180) (083) (179) (153) (451) (138) (139) (082)

Foreign ofloxacin 012 010dagger 140 010 010 024 013 117(006) (006) (034) (031) (005) (030) (005) (019)

Domestic ciprofloxacin 018 001 001 172 009 008 011 117(008) (000) (002) (027) (002) (003) (002) (003)

Domestic norfloxacin 004 004 004 068 242 049 046 070(002) (004) (002) (020) (013) (005) (004) (010)

Domestic ofloxacin 004dagger 004 009 061 060 295 060 092(002) (001) (010) (028) (007) (022) (007) (017)

Domestic sparfloxacin 005 003 005 092 048 051 251 043(002) (001) (001) (017) (005) (006) (016) (010)

Demand patterns in the western region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 427 009dagger 011 286 008 006 011 126(129) (005) (005) (133) (004) (005) (004) (021)

Foreign norfloxacin 714dagger 008 710dagger 594dagger 995 742 770 299dagger

(369) (183) (368) (328) (1015) (297) (296) (178)Foreign ofloxacin 013 012dagger 145 009 012 026 013 119

(007) (006) (037) (029) (005) (033) (005) (021)Domestic ciprofloxacin 019 001 000 174 009 008 011 118

(008) (000) (001) (024) (002) (003) (002) (004)Domestic norfloxacin 004 004 004 067 243 050 047 069

(002) (004) (001) (018) (014) (005) (004) (010)Domestic ofloxacin 003 003 007 048 048 257 048 093

(002) (001) (008) (022) (005) (015) (004) (014)Domestic sparfloxacin 006 003 005 083 046 049 241 046

(002) (001) (001) (014) (004) (005) (012) (009)

Demand patterns in the southern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 560 013dagger 015 402 011 010 017 137(182) (007) (007) (187) (006) (006) (007) (029)

Foreign norfloxacin 533dagger 031 531dagger 431dagger 748 559 584 249dagger

(301) (134) (300) (245) (734) (219) (219) (128)Foreign ofloxacin 014 012dagger 148 011 012 029 016 120

(007) (007) (039) (033) (006) (035) (006) (022)Domestic ciprofloxacin 017 001 000 169 008 007 011 116

(008) (000) (001) (025) (002) (003) (002) (003)Domestic norfloxacin 004 003 003 059 225 043 040 073

(001) (003) (001) (019) (012) (004) (004) (009)Domestic ofloxacin 004dagger 004 009 062 060 296 060 091

(002) (001) (010) (026) (006) (019) (007) (017)Domestic sparfloxacin 008 005 008 138 073 075 323 016

(002) (001) (002) (018) (007) (008) (021) (014)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at 5-percent level anddagger (dagger) denotes significance at the 10-percent level

1499VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

foreign product is in stock) in aggregate datawe would expect to find precisely the substitu-tion patterns that we report in Table 6

Whether the particular explanation we provideabove is the correct one the high degree of sub-stitutability between domestic product groupsturns out to have important implications for thewelfare calculations We discuss these in moredetail below when we present the results of thecounterfactual welfare analysis Another elasticitywith important implications for the counterfactu-als is the price elasticity for the quinolone subseg-ment as a whole which indicates how likelyconsumers are to switch to other antibioticsgroups when faced with a price increase for quin-olones This elasticity is computed on the basis ofthe results in Table A3 and it is at 111 (stan-dard error 024) this is large in magnitudebutmdashas expectedmdashsmaller in absolute value thanthe own-price elasticities of the product groupswithin the quinolone subsegment

The results in Tables 6A and 6B are based onour preferred specification discussed in SectionII In Tables A4 to A6 in the Appendix weexperiment with some alternative specificationsTables A4(a)ndashA4(c) correspond to a specificationthat includes in addition to product-group-specificregional fixed effects product-group-specific (andfor the upper level antibiotics-segment-specific)seasonal effects We distinguish among three sea-sonsmdashthe summer monsoon and wintermdashandreport the unconditional demand elasticities for

the northern region for each of these seasons Asevident from the tables in the Appendix our elas-ticity estimates are robust to the inclusion of sea-sonal effects The demand elasticities in Table A5are based on estimation of the demand system byOLS Compared to the elasticities obtained by IVthe OLS elasticities are smaller in absolute valueimplying that welfare calculations based on theOLS estimates would produce larger welfare lossestimates Nevertheless some of the patterns re-garding the cross-price elasticities discussed ear-lier are also evident in the OLS results inparticular the cross-price elasticities between dif-ferent domestic product groups are all positivelarge and significant and in most instances largerthan the cross-price elasticities between drugs thatcontain the same molecule but are produced byfirms of different domesticforeign status Theclose substitutability of domestic products indi-cated by both the OLS and IV estimates seems tobe one of the most robust findings of the paper

B Cost and Markup Estimates

Table 7 displays the marginal costs markupsand profits implied by the price elasticity esti-mates of Tables 6A and 6B for each of the sevenproduct groups Given that our regional effectsimply different price elasticities for each regionour marginal cost and markup estimates alsodiffer by region Given however that based on

TABLE 7mdashUPPER AND LOWER BOUNDS FOR MARGINAL COST MARKUP AND ANNUAL PROFIT BY PRODUCT GROUPS WITHIN

THE QUINOLONE SUBSEGMENT

Product groupLower boundfor MC (Rs)

Upper boundfor markup

Upper boundfor profit(Rs mill)

Upper boundfor MC (Rs)

Lower boundfor markup

Lower boundfor profit

(Rs)

Foreign ciprofloxacin 83 19 269 103 0 00(123) (012) (1655)

Foreign norfloxacin NA NA NA 53 0 00Foreign ofloxacin 323 70 1061 1085 0 00

(2316) (021) (3185)Domestic ciprofloxacin 47 59 17019 112 0 00

(114) (010) (29858)Domestic norfloxacin 52 43 2807 90 0 00

(020) (002) (1532)Domestic ofloxacin 587 34 1612 901 0 00

(218) (002) (1280)Domestic sparfloxacin 495 37 1985 788 0 00

(157) (002) (1100)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level Estimated lower bound forforeign norfloxacinrsquos marginal cost is negative since the estimated price elasticity is less than one in absolute value

1500 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Tables 6A and 6B the price elasticities do notseem to differ substantially across regions wereport for ease of exposition only the nationalaverages for marginal costs and markups inTable 7

Table 7 has two parts In the left part (firstthree columns) we report the lower bound formarginal cost and the corresponding upperbound for markup and upper bound for totalannual profit for each product group Thesenumbers are based on the price elasticities weobtained from estimating the two-level AIDSdemand system Since we do not have a reliableestimate for the price elasticity of foreign nor-floxacin (the point estimate is negative but lessthan one in absolute value and insignificant)we cannot compute the lower bound for mar-ginal cost in this case It is important to note thatthese estimates do not reflect either the actualmarginal cost or the actual markup for thesedrugs both because the existence of price reg-ulation implies that the unconstrained first-orderconditions are unlikely to hold each period andbecause our aggregation across firms of thesame domesticforeign status supplying thesame molecule makes the interpretation of theseestimates problematic In particular the fact thatwe ignore competition among firms within eachproduct group implies that our estimates willtend to overstate market power These numberswill prove useful however in the counterfac-tual simulations as they can give us a sense ofhow large the maximum profit gains for multi-national firms and the maximum profit lossesfor domestic firms are likely to be under patentenforcement The right part of the table (lastthree columns) reports the upper bound for mar-ginal cost which we obtain by simply taking themarginal cost to be equal to the observed priceWe do not report standard errors in this casesince the numbers are based on actual data Themarkups corresponding to these marginal costupper bounds are of course zero We conductthe counterfactual simulations using both thelower and upper bounds for marginal costs

The (upper bounds for) markups on the leftside of the table are generally plausible Thedomestic ciprofloxacin product group that dom-inates the quinolone subsegment and for whichwe documented high prices a high marketshare and a relatively low elasticity of demandenjoys one of the highest markup upper bounds(60 percent) and accounts for nearly 70 percent

of all profits derived within the subsegmentForeign ciprofloxacin on the other hand whichfaces the stiffest competition from domesticfirms and for which we estimated a highly elas-tic demand has the lowest markup upper bound(19 percent)

C Counterfactual Estimates of the Impact onPrices and Welfare

With estimates of the key demand and costparameters in hand we turn to the counterfac-tuals We consider the five separate scenarioslisted in the previous section All of the scenar-ios involve the withdrawal of one or more of thedomestic product groups from the market Ta-ble 8 displays our estimates of the consumerwelfare losses that result under the differentscenarios The losses are expressed in billion ofrupees per year All numbers presented in Ta-ble 8 and subsequent tables are based on usingthe lower bounds for marginal cost and upperbounds for markup in the simulations As dis-cussed earlier these numbers are the more in-teresting to work with since they give us anupper bound for the changes in the profits ofdomestic and foreign firms that would resultfrom patent enforcement In the Appendix wealso present results based on using the upperbounds for marginal costs in the simulations inwhich case the pre-TRIPS profits of domesticand foreign firms are zero In all cases marginalcosts are assumed to be constant in outputWhile naturally the profit implications differdepending on whether one uses the upper orlower bounds for marginal costs (firm profits arezero if one assumes the upper bound of mar-ginal cost in which case price equals marginalcost) the estimated consumer welfare losses aresimilar in the two cases We discuss these re-sults in more detail at the end of this section

The first column presents our estimates of theconsumer welfare losses attributable to the pureloss of product variety effect where we fix theprices of all remaining products as well as theoverall expenditure on quinolones while with-drawing one or more of the domestic productgroups Note that in our initial specification ofthe demand system had we not allowed for thepossibility that consumers might differentiatebetween domestic and foreign products evenwhen they contain the same molecule this

1501VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

particular component of the loss of consumerwelfare would not have arisen

The estimates reported in the second columnincorporate the expenditure-switching effect ontop of the loss of product variety Here basedupon the price elasticity estimates from thehigher-level AIDS system we adjust (down-ward) the expenditures allocated to the quino-lone subsegment as the composite price ofquinolones effectively increases as a consequenceof the higher virtual prices of the domestic productgroups that are withdrawn from the market Be-cause the estimates in this column are generatedassuming that the prices of the products that re-main in the market are not adjusted upward theyprovide a sense of what consumer losses would beif the introduction of product patents were coupledwith strict price regulation aimed at maintainingprices at prepatent levels Alternatively they canbe thought of as the relevant welfare numbers ifintense competition among firms within the re-maining product groups kept the prices of theproducts that were still offered in the market closeto the firmsrsquo marginal costs

The last column of Table 8 displays the esti-mated consumer welfare losses when bothcross-segment expenditure-switching and with-in-segment upward price adjustments are takeninto account

If we compare the results across the firstsecond and third columns all the counterfac-tual scenarios produce qualitatively similar pat-terns patterns that are consistent with what we

would expect Starting from the initial loss ofwelfare attributable to the loss of product vari-ety the option of switching expenditures outof the quinolone subsegment to other subseg-ments mitigates some of the initial welfareloss But if we then incorporate the upwardprice adjustments that result in response to thereduced competition the welfare losses aremagnified

Of particular interest from a policy perspec-tive are the relative magnitudes of these threeeffects which are similar under all the counter-factual scenarios though absolute levels varyconsiderably First despite the fact that the de-mand for quinolones is quite sensitive to thecomposite price of quinolonesmdashthe upper levelprice elasticity is 111mdashthe cross-subsegmentexpenditure-switching effects are in all thecases small (in absolute value terms) relative tothe other two effects For instance under thescenario where all the domestic quinolone prod-uct groups are withdrawn from the market theoverall consumer welfare loss of Rs 1781 bil-lion per year can be decomposed into an initialloss of Rs 1176 billion (66 percent) attribut-able to the loss of product variety a slightreduction in this initial loss of Rs 041 billion(2 percent) from Rs 1176 billion to Rs1135 billion because of expenditure switchingand a subsequent additional loss of Rs 646billion (36 percent) from Rs 1135 billion toRs 1781 billion because of the reduced com-petition and consequent price increases

TABLE 8mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure loss ofvariety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 732(087) (089) (146)

Only ofloxacin 008 008 023(008) (008) (010)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1253(177) (180) (415)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1058(142) (145) (331)

All four domestic quinolones products 1176dagger 1135dagger 1781(643) (634) (1270)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent significance level and dagger (dagger)denotes significance at the 10-percent level

1502 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 7: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

antiinfectives segment ranks second in Indiawhereas in the world market it is fifth and has ashare of only 90 percent Hence antiinfectivesare important in India not only from a health

and public policy point of view but also as asource of firm revenue

With this in mind we focus on one partic-ular subsegment of antiinfectives namely the

TABLE 3mdashSUMMARY STATISTICS FOR THE QUINOLONES SUBSEGMENT 1999ndash2000

North East West South

Annual quinolones expenditure per household (Rs) 3125 1975 2764 2359(366) (367) (407) (286)

Annual antibiotics expenditure per household (Rs) 11988 8424 11052 9624(1224) (1224) (960) (996)

No of SKUsForeign ciprofloxacin 1238 1129 1308 1246

(150) (190) (102) (106)Foreign norfloxacin 183 171 200 158

(070) (075) (088) (083)Foreign ofloxacin 304 296 296 300

(086) (086) (091) (088)Domestic ciprofloxacin 10621 9763 10342 10550

(599) (434) (722) (451)Domestic norfloxacin 3896 3496 3617 3942

(271) (268) (251) (379)Domestic ofloxacin 1846 1600 1725 1725

(680) (634) (586) (635)Domestic sparfloxacin 2983 2829 3121 2929

(557) (638) (688) (657)Price per-unit API (Rs)

Foreign ciprofloxacin 958 1090 1085 1007(128) (066) (071) (058)

Foreign norfloxacin 563 509 605 435(077) (133) (139) (147)

Foreign ofloxacin 10946 10943 10886 10612(620) (664) (700) (1140)

Domestic ciprofloxacin 1143 1067 1131 1152(016) (015) (017) (013)

Domestic norfloxacin 951 907 888 873(024) (035) (037) (020)

Domestic ofloxacin 9163 8964 8565 9341(1615) (1565) (1422) (1407)

Domestic sparfloxacin 7972 7849 7688 8028(976) (1014) (1185) (1037)

Annual sales (Rs mill)Foreign ciprofloxacin 4179 2431 4520 2947

(1534) (816) (1273) (648)Foreign norfloxacin 128 100 058 073

(101) (082) (044) (057)Foreign ofloxacin 5446 3184 3522 3111

(1399) (933) (906) (703)Domestic ciprofloxacin 96229 58591 67874 70381

(10626) (13026) (12226) (8740)Domestic norfloxacin 22255 11971 14918 15829

(3884) (1945) (2691) (1626)Domestic ofloxacin 12502 9621 14936 11205

(4434) (3011) (5282) (4259)Domestic sparfloxacin 15617 12175 16130 9811

(3141) (2576) (4674) (3420)

Note Standard deviations in parentheses API Active pharmaceutical ingredient

1483VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

quinolone subsegment Quinolones fall into thesystemic antibiotics and antibacterials segmentof the Indian pharmaceuticals market whichgenerates over three-quarters of the revenues inthe antiinfectives segment8 The systemic anti-bacterials segment includes all of the originalmiracle drugs that first sparked the growth ofthe global research-based pharmaceutical indus-try in the postndashWorld War II period as well aslater generations of molecules that have beenintroduced in the last four decades

Among systemic antibacterials quinolonesare the latest generation molecules available inIndia We focus our analysis on quinolones forseveral reasons First quinolones are the drug ofchoice for a large number of bacterial infec-tions some of which are also treated by alter-native drugs (see Table A1 in the Appendixwhich outlines the spectrum of activity for eachmolecule family within the antibacterials seg-ment) Hence if there were one product groupfor which we would expect to have many sub-stitutes readily available this would be quino-lones Second with a share of 20 percent in thesales of systemic antibacterials quinolones rep-resent one of the largest subsegments within thistherapeutic category Finally several moleculeswithin the quinolone subsegment were still un-der patent in the United States at the time of ourinvestigation Table 2 details the basic informa-tion about the four quinolone molecules that arethe focus of our analysis The first row showsthe year of US patent expiry this ranges from1998 for norfloxacin to 2010 for sparfloxacinQuinolones include in principle four moremolecules that are listed at the bottom of Ta-ble 1 the market shares of these molecules arenegligible however so that we exclude thesemolecules from our analysis

Table 2 reveals several other interesting factsabout competition in the quinolone market inIndia First note the large number of firmsoperating in this subsegment The large numberof domestic firms is perhaps not that surprisinggiven that pharmaceutical product patents werenot recognized in India9 What is more surpris-ing is the number of foreign firms selling pat-ented products (eg ciprofloxacin) the fact that

multiple foreign firms sell a patented productindicates that such firms often ldquoinfringerdquo10

patent laws in India while complying with themin developed world countries The last two rowsof Table 2 further indicate that domestic prod-ucts often sell at a premium With the exceptionof ofloxacin the average prices of products of-fered by Indian firms are higher than the pricesof products offered by foreign subsidiaries Thispreliminary evidence suggests that Indian con-sumers do not place a premium on the brandname and reputation of big multinational phar-maceutical concerns Moreover the higherprice of domestic products does not seem toprevent domestic companies from capturing alarge market share This is most evident in thecase of ciprofloxacin where domestic firmshave with 53 percent the largest share in thetotal sales of quinolonesmdashthis despite the factthat the average price of these products is 10percent higher than the price of foreign productscontaining the same molecule

Table 3 provides additional summary statis-tics for our data broken down by region Thefirst two rows of the table report the averageannual household expenditure on quinolonesand antibiotics respectively Note that in bothcases the average expenditure is higher in thenorth and west these regions include states withhigher per capita incomes and tend to be moreindustrialized and urbanized than those in theeast and south Pharmaceutical products areavailable in multiple presentations that is com-binations of dosage forms (eg capsule tabletsyrup) strength (eg 100 milligrams 500 mil-ligrams) and packet sizes (eg 50 capsule bot-tle 100 tablet bottle) The various presentationsin which a product is available are often referredto as stock-keeping units or SKUs11 The num-ber of SKUs for each product group withinquinolones is reported at the top of Table 3 Aswith the more aggregate numbers on firms and

8 In addition to anti-bacterials this segment also containsantivirals

9 Accordingly the common distinction between ldquobrandedrdquoand ldquogenericrdquo products is irrelevant here

10 We emphasize here that the word ldquoinfringerdquo belongsin quotes because patent laws do not currently exist inIndia infringement in the legal sense is not possible It ishowever striking that the same firms that accuse Indianproducers of ldquopiracyrdquo sell in India products that are patentedin the United States and for which the patent is held by adifferent multinational corporation

11 For instance a 100 capsule bottle of 100 milligramcapsules of a particular branded drug and a 50 capsulebottle of 100 milligram capsules of the same branded drugwould be identified as two separate SKUs

1484 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

products reported in Table 2 the differencebetween domestic and foreign products isstriking The number of SKUs offered byIndian firms is consistently larger than thenumber offered by subsidiaries of foreignmultinationals The number of SKUs variesslightly across regions but more importantlyit varies across time as some SKUs disap-pear while new ones get introduced duringour sample period

Many pharmaceutical products in India aresubject to price controls12 While the specificsof the price regulation are too complex for anyeconomic model to capture adequately themain concern for the empirical analysis is thatprice controls may lead to a lack of price vari-ation over time so that the demand functioncannot be identified Prices at the most disag-gregate SKU level are relatively stable overtime there are variations due to occasionalchanges in the estimated cost (due for exampleto changes in exchange rates that affect the costof imported materials or bulk drugs) but suchvariations tend to be infrequent and small inmagnitude The degree of time variation ishowever substantially larger once one aggre-gates to the product level This variability stemsnot only from the fact that the SKUs over whichwe aggregate may experience changes in theirrespective prices at different points in time butalso from the fact that the range of SKUs of-fered in the market does not remain constantover time The entry and exit of presentationswithin the same product group that have differ-

ent prices effectively affects the price that con-sumers face for this drug in each period

The middle portion of Table 3 reports themean price and standard deviation for eachproduct group by region Prices vary by regionthough there is no clear pattern emerging fromthe table with regard to the cross-regional vari-ation (in the sense of some regions being sys-tematically more expensive than others) Toexamine what portion of the total price variationis due to time versus regional variation weconducted an analysis of variance of prices thatwe report in Table 4 The table is based onseparate regressions for each product group(pooling data across groups with big differencesin their average prices is not particularly infor-mative as most of the price variation is ac-counted for by product group dummies) Thelast two columns of the table show the fractionof price variation accounted for by region andtime dummies respectively As evident fromthe table a significant fraction of the total vari-ance in prices can be attributed to time varia-tion In the demand estimation we include a fullset of product groupndashspecific regional dummiesso that the price parameters are identified en-tirely based on this time variation within eachproduct group The time variation of productgroup prices is driven primarily by composi-tional changes within each group the revenueshares of the individual SKUs that compriseeach product group change over time (see therelated discussion in Section IIC) while in eachperiod there is entry and exit of SKUs into thesample To check whether this pattern reflectsgenuine entry and exit as opposed to samplingvariation we examined the revenue shares ofthe SKUs that leave the sample relative to theones that remain during the entire period Theresults are reported in Table 5 While the SKUs

12 The details of the procedures for price fixation canbe found in the official government Web site httpwwwnppaindianicinindex1html under the link ldquoDrug Price Con-trol Order 1995rdquo A new pharmaceutical policy was introducedin 2002 but our data were collected before that year

TABLE 4mdashANALYSIS OF PRODUCT PRICE VARIANCE

Product groupPartial SS

timePartial SS

region Total SSPercentage

explained by timePercentage

explained by region

Foreign ciprofloxacin 0296 0306 1047 283 292Foreign norfloxacin 1036 1002 4179 248 240Foreign ofloxacin 0429 0021 0571 752 36Domestic ciprofloxacin 0005 0088 0104 50 841Domestic norfloxacin 0059 0098 0198 296 494Domestic ofloxacin 2858 0104 3056 935 34Domestic sparfloxacin 1754 0032 1860 943 17

1485VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

that exit tend to be smaller (their average shareis 1 percent as opposed to 34 percent for thoseSKUs that are present during the entire sampleperiod) the shares of the two groups do notseem orders of magnitude apart13 In additionour data cover only the 300 largest firms sellingin the Indian market so that firms with verysmall shares are not included in our sample

II The Analytic Framework and EstimationApproach

Patent enforcement in the Indian pharmaceu-tical market will have the effect of eliminatingdomestic products whose active pharmaceuticalingredients are protected by (foreign) patentsThus assessing the effects of patent enforce-ment is tantamount to assessing the effects ofwithdrawing domestic products from the mar-ket This task is the converse of evaluating newproduct introduction accordingly the concep-tual framework we use to address the questionsof interest is similar to the one developed in theliterature for the valuation of new goods14

We start by estimating demand for quino-lones Given that the market is characterized byimperfect competition the counterfactual anal-ysis requires that we also model the supply sideas removal of one product will affect the pricesof other products especially those that are closecompetitors The existence of price regulationin the Indian pharmaceutical market imposespotential constraints on firmsrsquo maximizationproblem Given these constraints and the com-

plexity of the price regulation process the typ-ical approach of deriving estimates of actualmarginal costs and markups by exploiting thefirst-order conditions of profit maximizing firmsdoes not seem particularly promising Insteadwe use our demand estimates to place upper andlower bounds on marginal costs and markups

With demand elasticities and upper and lowerbounds for marginal costs in place we thenconduct counterfactual simulations We con-sider several alternative scenarios dependingon the number of domestic products affected bypatent enforcement For each scenario we com-pute the counterfactual prices and use them toassess the effects of domestic product with-drawal on consumer welfare (as measured bythe compensating variation) firm profits andsocial welfare As with the valuation of newproducts the big conceptual problem facing thispart of the analysis is that we need to extrapo-late from the region of the data to the point atwhich demand for the products that exit themarket becomes zero This conceptual issue ispresent in any attempt to evaluate a major pol-icy change for which no historical precedentexists like the enforcement of patent laws inIndia One advantage of the present study is thatwe have a limited set of price data for Pakistana country with similar demographics to Indiabut with a market structure that resembles theone that would emerge in India under patentenforcement (monopoly of multinational sub-sidiaries) By comparing the prices of productsoffered by multinationals in Pakistan to thosewe compute in our counterfactual simulationsfor the products that would be offered by mul-tinationals in India if patent laws were enforcedwe can get a sense of how plausible our coun-terfactual estimates are

A Demand

The demand modelling is based on the multi-stage budgeting approach Our primary motiva-tion for adopting this approach is a practical oneIn the multistage budgeting approach the depen-dent variable is defined as a revenue share whichis appealing given that the products we include inthe analysis contain different molecules (ie ac-tive pharmaceutical ingredients or APIs) Eventhough we do have data on the quantity of therelevant API (eg 100 milligrams of ciprofloxa-cin) contained in each product converting the

13 While we cannot completely rule out the possibilitythat some of the exit is due to sampling variation note thatthe latter should be reflected in low precision of the demandparameter estimates The demand parameters however areprecisely estimated

14 See Manuel Trajtenberg (1989) Jerry A Hausman(1994) and Timothy F Bresnahan (1997) for representativeexamples and a discussion of the relevant issues

TABLE 5mdashREVENUE SHARES OF THE EXITING SKUS

Revenue share ofexiting SKUs

Revenue shares ofall SKUs

Full sample 10 34Northern region 08 33Eastern region 12 36Western region 10 34Southern region 12 33

1486 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

revenue shares to physical shares is extremelydifficult if not infeasible in the case of antibioticsBecause such drugs are ldquosystemicrdquo by nature theyare used to treat a large number of infections andthe dosage of each drug depends on the particularinfection it is supposed to address (for example thedosage will differ depending on whether the anti-biotic is used to combat an ear infection or tuber-culosis) This particular feature of antibioticscomplicates the conversion of revenue to physicalshares

The basic idea of the multistage budgetingapproach is to use the therapeutic classificationof a productmdashie the therapeutic segment andsubsegment the product belongs tomdashto orga-nize all products in the systemic antibacterialssegment into a hierarchical taxonomy consist-ing of two levels At the higher level are thevarious subsegments of systemic antibacterialsThe first stage of budgeting corresponds to theallocation of expenditures across the subseg-ments in this upper level of the taxonomy

In the second stage of the budgeting processcorresponding to the lower level of the taxon-omy a flexible functional form is adopted tomodel how the expenditures allocated to eachsubsegment are distributed across the productswithin that subsegment In particular to modeldemand at the second stage we employ theAIDS specification proposed by Deaton andMuellbauer (1980b)15

While the two-stage demand estimation ap-proach offers functional form flexibility its ap-plication to the Indian systemic antibioticsmarket poses a few problems The first is thatdue to entry and exit many SKUs and evenproducts in our sample are not present in everyperiod AIDS does not have a good way ofdealing with a varying number of products as itwas developed with broad commodity catego-ries in mind which are consumed by all con-sumers every period To solve this problem weaggregate within each subsegment (eg quino-lones) SKUs into product groups where withineach product group all SKUs contain the samemolecule and are produced by firms with thesame domesticforeign status Specifically let aSKU k be indexed by its molecule (or API) M its

domesticforeign status DF indicating whether itis produced by a domestic (Indian) or a subsidiaryof a foreign (multinational) firm a particular pre-sentation s and the particular firm f that producesit We aggregate SKUs over presentations andfirms to obtain a newly defined product group iwhich is indexed only by molecule M and domes-ticforeign status DF and has revenue Ri yenfsRk with i (M DF) k (M DF f s) and pricepi yenfs kpk where k denotes the conditional(on M and DF) revenue share of this particularproduct ie

(1) k Rk

Ri

In most cases the resulting product groups arebroad enough to be present every period16

The usual concern with this aggregation pro-cedure is that it may lead us to overstate firmsrsquomarket power as we ignore competition amongfirms with the same domesticforeign statusproducing the same molecule In the presentapplication however this concern is unlikely tobe of great importance as the effect of patentenforcement is to wipe out all domestic compe-tition at once while granting foreign firms mo-nopoly power hence competition among firmsfor patented molecules becomes irrelevant Theaggregation according to the domesticforeignstatus (within a particular molecule) thus corre-sponds to the scope of our analysis and theparticular questions of interest

The second problem is that for our approachto be useful in welfare analysis the allocation oftotal expenditures to group expenditures at thehigher stage has to be modelled in a way con-sistent with utility maximization In general thesolution of this allocation problem requiresknowledge of all individual product prices

15 Representative applications of the multistage budget-ing approach include Sara F Ellison et al (1997) Hausman(1994) and Hausman and Gregory K Leonard (2002)

16 We are missing only four observations (ie monthregion combinations) all for the drug group of ForeignNorfloxacin August 1999 in the south May 2000 in thewest October 2000 in the south and November 2000 in theeast In these cases we set the revenue shares of ForeignNorfloxacin equal to zero In general with 01 percent ofquinolone sales (see Table 1 row 2) Foreign Norfloxacinhas a very small share of the market This probably explainswhy the results pertaining to this drug group are unreliableit is the only drug group for which we do not obtain asignificant price elasticity of demand while its cross-priceelasticities with other foreign drug groups often have thewrong sign

1487VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

From an empirical point of view this is notparticularly useful as it eliminates all compu-tational advantages of the two-stage approachTo address this problem we adopt an approxi-mate solution to model the higher-level expen-diture allocation along the lines suggested byDeaton and Muellbauer (1980b pp 131ndash32)This gives rise to a two-level AIDS specification

Consider the lower-level estimation firstwhich refers to the allocation of a particularsubsegmentrsquos expenditure to the product groupswithin the subsegment In our application therelevant subsegment is quinolones which weindex with Q Let the product groups within thissubsegment be indexed by i 1 N pi be theprice of product group i (where as noted abovei refers to a particular molecule and domesticforeign status combination) and XQ be the totalexpenditure on the quinolone segment The rev-enue share of each product group is given by

(2) i i j

ijln pj ilnXQ

PQ

where i the revenue share of product group iis defined as

(3) i pi qi

yenj

pj qj

xi

XQ with i j Q

XQ is the overall expenditure on the quinolonesubsegment and PQ is a price index given by

(4) ln PQ ap 0 i

iln pi

1

2 i

j

ijln piln pj

With a limited number of product groups and asufficiently large number of time-series observa-tions the flexibility implied by the AIDS modeldoes not impose too many demands on the data Inthe present application however where the num-ber of observations is limited the AIDS model isnot estimable in this general form To reduce thenumber of parameters that need to be estimatedwe impose two sets of restrictions

The first set of restrictions is implied by thetheory of utility maximization Specificallythese restrictions are

Adding-up yenk k 1 yenk k 0 yenk kj 0 j

Homogeneity yenk jk 0 j Symmetry ij 1frasl2 [ij ji] ji This last

restriction by itself reduces the number of parameters to [N(N 1)]2

The second set of restrictions we impose aimsat further reducing the number of parameters tobe estimated by exploiting our knowledge of thisparticular market Specifically for each productgroup i we allow one ij parameter for all productgroups j that have different molecules from prod-uct group i and are produced by foreign firms andone ij for all product groups j with differentmolecules produced by domestic firms We donrsquotimpose any restrictions on the ij parameter whenproduct group j has the same molecule as productgroup i (By construction product groups i and jcontain products produced by firms with differentdomesticforeign status)

To better illustrate the nature of the restric-tions we impose on the patterns of substitutionacross products some additional notation isneeded Let d(i j) be an indicator of the degreeof similarity (or difference) between productgroup i and product group j along the dimen-sions we are able to observe (molecule M anddomesticforeign status DF) For any two prod-uct groups i and j d(i j) can take on one of thefollowing three values17

(5)

di j (1 0) if Mi Mj DFi DFj

(0 1) if Mi Mj DFi DFj

(0 0) if Mi Mj DFi DFj

Let

(6) Diab j di j a b

17 The sequence (1 1) is not possible for two differentproducts in this case the parameter corresponds to theproductrsquos own-price effect that is ii

1488 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The equation at the lower level becomes

(7) i i iiln pi i10ln pj j Di10

j Di

01

i01ln pj

j Di

00

i00ln pj ilnXQ

PQ

Note that

The parameter ii captures a product grouprsquosown-price effect (there will be as many iiparameters as the number of product groups)

The parameter i10 captures the cross-priceeffect of the product group containing prod-ucts with the same molecule but produced byfirms of different nationality

The parameter i01 captures the cross-priceeffects of product groups containing productswith different molecules but produced byfirms with the same nationality

The parameter i00 captures the cross-priceeffects of product groups containing productswith different molecules produced by firms ofdifferent nationality

Before we take the demand equation to thedata we make two modifications The first oneis to let the product-specific effects i vary byregion r The resulting product-specific regionaleffects ir have two interpretations first theycontrol for the ldquoqualityrdquo of each drug withquality differences being allowed to vary acrossregions second they proxy for demographicsand other demand shifters which vary by re-gion and may affect the demand of each prod-uct group differently18 Note that by includingproduct-specific regional effects in the demandspecification we estimate the price parametersbased on the within-product-group variation ofprices in each region In an earlier version of thepaper we also estimated the demand systemwithout regional dummies and obtained similarresults

The second modification that allows us to gofrom a deterministic to a stochastic specificationof the demand equation is to include an additiveerror term in (7) The latter takes into accountthe fact that (7) is not expected to fit the dataexactly The error term irt accounts for mea-surement error (due to the fact that the productgroup prices pjrt we employ in the estimationare not exact price indices but approximationsthereof) and (potentially region-specific) de-mand shocks that may affect the demand for aproduct in a particular period Examples of suchshocks include an advertising campaign for aparticular product which temporarily increasesthe demand for this product or the outbreak ofa (potentially region-specific) epidemic whichcalls for the use of a particular drug We discussthe interpretation and properties of this errorterm in more detail in the next section

The final form of the equation we estimate atthe lower level becomes (with subscript t de-noting month and subscript r denoting region)

(8) irt i ir iiln pirt i10ln pjrt j Di10

j Di

01

i01ln pjrt

j Di

00

i00ln pjrt

ilnXQrt

PQrt irt

The analysis so far has conditioned on the ex-penditure allocated to the quinolone subseg-ment XQ The upper level of the estimationconsiders the problem of allocating total expen-diture across the different systemic antibioticssubsegments one of which is quinolones Theupper-level demand function is given by

(9) G G H

GHln PH GlnX

P

where all variables denoted by capital letters aredefined as before but now refer to subsegments(G H ) rather than product groups within asubsegment and the total expenditures on sys-temic antibiotics X are deflated by the Stone priceindex log P yenH Hlog PH When estimating thesystem above we impose all the restrictions

18 Given the short time span of our sample typicaldemand shifters such as age distribution income distribu-tion and education hardly change over our sample periodSuch shifters are therefore absorbed by the region-specificproduct fixed effects ir

1489VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

implied by utility maximization as we do withthe estimation of the lower-level AIDS We donot however impose any additional restrictionson the substitution patterns at this stage so thatthe cross-price effects across segments remainrelatively unconstrained

Estimation of the higher-level AIDS allowsus to obtain the unconditional own- and cross-price elasticities that are used in the formulationof the supply problem and welfare analysisThese are given by the formula

(10) ij ijXQXQ

ln qi

ln XQ

ln XQ

ln PQ

ln PQ

ln pj

with the conditional cross price elasticitiesgiven by

ijXQXQ ij ij jlnXQ

PQ

i

As in the lower stage we include subsegment-specific regional dummies in the specificationof the upper-stage demand system so that thefinal form of the estimating equation at theupper stage becomes

(11) Grt G Gr H

GHln PHrt

GlnXrt

Prt Grt

In sum the demand system we take to thedata is represented by equations (8) and (11)and the associated parameter restrictions im-plied by economic theory

B Modelling the Supply Side of the Market

Counterfactual simulations concerning theeffects of domestic product withdrawal requireknowledge of the marginal costs of pharmaceu-tical firms operating in the Indian market Theseare unobservable The usual approach in theNew Empirical Industrial Organization litera-ture has been to exploit the firm equilibriumconditions to infer marginal cost For exampleit is often assumed that the marginal cost ci isconstant and that the industry is an oligopoly

engaging in Bertrand competition with differen-tiated products Assuming that firms myopicallymaximize profits each period one can then de-rive the firmsrsquo first-order conditions that corre-spond to the assumptions above about costsmarket structure and firm behavior

We deviate from this procedure as the presenceof price regulation renders the assumption of un-constrained period-by-period maximization un-tenable Ideally one would like to explicitlyincorporate the price and other administrative con-trols into the firmrsquos optimization problem andderive the first-order conditions under the assump-tion of constrained maximization The complexityof the price regulation however makes this ap-proach infeasible Therefore we adopt an alterna-tive approach that does not rely on modelling theprice-setting process but is instead based on de-riving upper and lower bounds for marginal costsand markups

In particular an upper bound for marginalcosts and a lower bound for the markups (zero)can be derived under the assumption of perfectcompetition While this assumption is clearlyunrealistic in the pharmaceuticals market it isuseful in providing an upper bound for costs ci

Uwhich will be given by

(12) ciU pi

On the other hand a lower bound for marginalcosts ci

L (and upper bound for markups) can bederived by assuming that there is perfect collu-sion within each product group i (where i refershere to a moleculedomestic-foreign combina-tion)19 and ignoring price controls so thatprices are determined by the first-order condi-tion of the jointly profit-maximizing firmswithin product group i Solving this first-ordercondition for the (lower bound of) marginal costthen gives

(13) ciL pi 1

1

iipipj

19 The reason that this assumption will lead to an under-statement of marginal costs (and hence overstatement ofmarket power) is that it assumes away competition amongfirms within each drug group

1490 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

where ii(pipj) the own-price elasticity of de-

mand for product i will depend on the productrsquosown-price pi and all other productsrsquo prices pjwith i j

Once we have obtained the demand elastic-ities through estimation of the demand sys-tem we can calculate the upper and lowerbounds for marginal costs and correspondingmarkups according to (12) or (13) These willthen be employed in the counterfactual sim-ulations While our counterfactuals use boththe lower and upper bounds for costs most ofour discussion will be based on using thelower bounds for costs since this is the moreinteresting case in the policy analysis it givesus the largest possible profits for pharmaceuticalfirms and hence corresponds to the worst pos-sible scenario facing Indian firms and the bestpossible scenario facing multinationals underTRIPS

C Identification Assumptions and EstimationApproach

The discussion of the demand system has sofar abstracted from the issue of price endogene-ity The usual premise in the industrial organi-zation literature is that correlation of prices withthe error term in the demand equation arises byvirtue of the first-order conditions of profit-maximizing firms20 As we discuss in this sec-tion this source of simultaneity bias is unlikelyto be of major concern in the present contextbecause of the existence of price regulation Ourprimary concern is instead with the simultane-ity bias that is implied by the particular wayprices are constructed

To understand the sources of the simultaneitybias note first that the price of product group piwe employ in the demand estimation should bethought of as a proxy for an exact price indexthat we do not observe As such pi contains bydefinition measurement error and it will becorrelated with the error term of the demandequation Specifically the demand equation weare interested in estimating can be written insimplified form (suppressing the subscripts rand t and ignoring parameter restrictions forconvenience) as

i i j

ijln pjT ilnXQ

PQ i

where pjT denotes the true (exact) price index for

product group i and i captures unobservedvariables that may affect demand in a particularperiod (eg an advertising campaign or an ep-idemic) The price index pj

T depends on theSKU prices pk Note that in the presence ofSKU entry and exit into the sample the priceindex pj

T will vary over time even if the pricesof the individual SKUs remain stable as theproducts (SKUs) that comprise each productgroup change each period

Ideally we would like to compute the exactprice index pj

T for each group and employ theseindices in the demand estimation Under theassumption of predetermined SKU prices eachperiod the exact price indices would be uncor-related with the error term and the demandparameters could be estimated by simple OLSUnfortunately this strategy is not feasible in thecurrent context

To derive an exact price index for each prod-uct group it is necessary to model consumerrsquoschoice among SKUs conditional on the choiceof the group To this end it is necessary toexplicitly introduce a third stage in the demandspecification and estimate the parameters asso-ciated with the choice at that stage Howeverany specification based on three-stage budget-ing (that would require directly specifying ademand function associated with the SKUchoice at the third stage and imposing additiveseparability at the higher stages) would be in-feasible to estimate for three main reasonsFirst there is a large number of SKUs withinsome groups so that the demand parameters can-not be estimated even if one imposes parameterrestrictions second the SKU prices exhibit littlevariation over time lastly the choice sets at thislowest stage vary over time rendering estimationof the price parameters associated with SKUs thatare not available in all periods infeasible Alterna-tively one could abandon the multistage budget-ing altogether and adopt a discrete choiceapproach which is inherently better suited to deal-ing with varying choice sets and limited pricevariation A discrete choice approach would beparticularly appealing for modeling the SKUchoice conditional on product group choice giventhat within each group all SKUs contain the same

20 See Steven Berry et al (1995) or Aviv Nevo (2001)for a related discussion

1491VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

molecule and are hence comparable As notedearlier however a discrete choice approach at thehigher stages of the demand estimation requiresconverting revenue to physical shares which ischallenging in the case of systemic drugs such asantibiotics for which the appropriate dosage is notwell defined Finally a combination of a multi-stage budgeting approach for the higher stages anda discrete choice approach for the lowest stagewould be feasible to estimate but inconsistentwith a model of utility maximization For thesereasons we have no alternative but to adopt anapproximation for measuring the price index foreach product group

Let us denote this approximation by pjA and let

13j denote the proportional ldquoapproximationrdquo errorassociated with measuring the true price index forgroup j so that ln pj

A ln pjT ln 13j The

estimating demand equation can then be written as

i i j

ijln pjA ilnXQ

PQ

j

ijln 13j i

or

i i j

ijln pjA ilnXQ

PQ i

where the new error term i yenj ijln 13j iis comprised of two components the first onereflects ldquomeasurementrdquo error due to the fact thatthe product group prices we employ are approx-imations and not exact price indices and thesecond one captures conventional demandshocks The demand equation as written abovecorresponds to the equation we take to the data

The particular proxy we use for the exact priceindex for each product group j is the revenue-share-weighted average of the prices of multipleSKUs available within this group that is

pjA

k j

k pk

The notation above illustrates the two source ofprice variation variation in the SKU prices pkand variation in the weights k

We are concerned with potential correlation ofthe so-constructed product group prices with bothcomponents of the product group demand error

term The correlation between pjA and the first

component yenj ijln 13j is independent of theparticular way we construct pj

A and inherent in thefact that we measure the exact price index witherror The potential correlation between pj

A and thesecond component of the product group demanderror the shock i arises however because of thespecific way we construct the proxy pj

A and inparticular because of the presence of the reve-nue share weights k in pj

A since the SKUrevenue share weights k will generally dependon the product group expenditure they arelikely to be correlated with the product groupdemand shock i

To address the simultaneity bias we use in-strumental variables To this end we need vari-ables that are correlated with the proxies pj

A butuncorrelated with the error term i We use thenumber of SKUs within each product group j asan instrument as it does justice to the idea thatvariation in the product group price index stemsin part from variation in the set of SKUs that areavailable each period It is clearly correlatedwith the average group prices pj

A and plausiblyuncorrelated with the demand shock i Theassumption that the number of SKUs is uncor-related with the conventional demand error willhowever be violated if the introduction of anew SKU changes the perceived quality of adrug21 or if it is accompanied by promotionalactivities which will be reflected in i Our hopeis that this is not too often the case Assumingthat the SKU number for each group j is alsouncorrelated with the weighted sum of the mea-surement errors of all product groups yenj ijln 13j itcan be used as an instrument for the product groupprice pj

A22 If one accepts the premise that prices atthe SKU level are exogenous then SKU pricescan also be used as instruments The final list ofinstruments we use hence includes the number ofSKUs in each group the prices of the five largestSKUs for each group and all other exogenousvariables in the demand estimation such as re-gionalproduct dummy interactions and the upper-

21 Note that we already proxy for the quality of each prod-uct group through the product fixed effects i In the actualestimation we actually let these fixed effects be region-specific(ir) so that we allow quality to vary across regions

22 If the measurement error increases in the number ofincluded SKUs this assumption will be violated Unfortu-nately we have no way of checking the validity of thisassumption

1492 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

level total expenditure on antibiotics Regressionsof group prices on the instruments above yieldhigh R-squares with most regressors highly sig-nificant indicating that our instruments are highlycorrelated with prices23

Our sample includes four molecules cipro-floxacin norfloxacin ofloxacin and spar-floxacin Except for sparfloxacin all othermolecules are produced by both foreign anddomestic firms24 So we have seven products(domestic ciprofloxacin foreign ciprofloxa-cin etc) with 96 observations (two years ofmonthly data four geographical regions foreach period) for each product The parametersin the lower-level AIDS demand system asdefined in equation (8) are the product fixedeffects i the productndashregional dummy inter-actions ir the own revenue-share price elas-ticities ii the cross revenue-share priceelasticities i10 i01 i00 and the revenue-share expenditure elasticities i In estimatingthe parameters we first regress prices on allinstrumental variables and then plug the pre-dicted values for prices in the constrainedleast-square regression (For a detailed expla-nation of the constraints see the previoussection)

Given that we impose many cross-equationconstraints and employ instrumental variablesin the estimation it is difficult to derive stan-dard errors for the parameter estimates ana-lytically Our error term interpretation in (8)implies that the error terms for each productgroup are likely to be correlated across re-gions for example if a national advertisingcampaign increases the demand for a rela-tively expensive presentation in one regionsimultaneously increasing the aggregate de-mand and price index for the correspondingproduct group in that region it is likely that thesame effect will be observed in other regions Inprinciple we could exploit the cross-regionalcorrelation in the error terms to estimate thedemand system using seemingly unrelated re-

gression (SUR) This however would requireus to derive the variance-covariance matrix ofthe parameters analytically which as notedabove is cumbersome in the current contextInstead we use the bootstrap method choosingto remain agnostic about the structure of thevariance-covariance matrix The potential dis-advantage of this approach over SUR is a loss inefficiency but as will become apparent in theresults section the model parameters are fairlyprecisely estimated To maintain the marketstructure we randomly sample the periods (withreplacement) and use the same periods for allproducts Regarding the optimal number ofbootstrap repetitions ideally one would followthe three-step method proposed by Donald W KAndrews et al (2000) Empirical evidence sug-gests however that one rarely needs more than200 replications to estimate the standard er-rors25 To be safe we generate 300 bootstrapsamples (with replacement) based on the origi-nal data and estimate the standard errors usingthe standard errors of the bootstrap sampleestimates

The estimation of the top-level AIDS systemis similar The constraints imposed on this top-level demand system are adding-up homogene-ity and symmetry as defined on page 12Again bootstrapping is used to obtain the stan-dard errors of the parameter estimates

D The Counterfactual Scenarios

In assessing the effects of patent enforce-ment we start by focusing on the most extremecase in which compulsory licensing is not anoption and foreign firms are not subject to pricecontrols We use the results from the analysis ofthis case as a benchmark In reality the outcomeof the WTO negotiations is more likely to in-volve some constraints on the monopoly powerof foreign firms selling patented products indeveloping countries such as price caps orcompulsory licensing Our framework can eas-ily accommodate these cases as will becomeapparent in the next subsection

We now focus on the effects of potentialpatent enforcement in the quinolone segmentWe consider several scenarios which vary inthe number and size of domestic products that

23 Specifically the R-squares from the first-stage regres-sions of product group prices on instruments range from057 to 095 except for domestic ciprofloxacin for whichthe R-square is 017

24 Sparfloxacin is actually offered by one foreign subsid-iary in India (see Table 2) However its revenue share isminiscule We therefore treat Sparfloxacin as being pro-duced by domestic firms only 25 See Bradley Efron and Robert J Tibshirani (1993)

1493VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

will be removed from the market In particularwe consider the following five scenarios26

Withdrawal of one large domestic productgroup only domestic ciprofloxacin

Withdrawal of one relatively small domesticproduct group only domestic ofloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand norfloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand sparfloxacin

Withdrawal of all four domestic quinoloneproduct groups

As the list above suggests we proceed fromanalyzing the effects of single-product with-drawal to the analysis of eliminating the en-tire domestic segment This approach wasmotivated by early empirical results that in-dicated that the existence and extent of com-petition from domestic firms has a significantbearing on the predicted effects of patent en-forcement that is our predictions regardingprices and welfare vary substantially depend-ing on how many domestic products are af-fected by patent enforcement and on whethersome domestic competition will remain presentafter TRIPS In addition the size of the affecteddomestic groups is relevant for the welfare pre-dictions accordingly we examine both scenar-ios in which a large domestic product groupsuch as domestic ciprofloxacin is eliminatedfrom the market and scenarios in which the

eliminated domestic group is relatively small(eg domestic ofloxacin) In all scenarios wemaintain the assumption that the set of prod-ucts offered by the remaining firms in themarket does not change in response to patentenforcement

E Computation of Virtual Prices and NewEquilibrium Prices

The first step in the counterfactual analysisis to derive the new equilibrium prices underpatent enforcement27 In this context thereare two sets of prices that are relevant Thefirst set consists of the virtual prices of those(domestic) products that will not be availableonce TRIPS is put in effect To calculate thesevirtual prices we set the expenditures of therelevant products equal to zero The secondset of prices consists of the prices of thoseproducts that remain in the market In deriv-ing these prices we start by assuming profitmaximization without any form of price reg-ulation the firms remaining in the marketreoptimize in response to the policy changeand set new prices taking the prices of allother firms as given28 Of course at the equi-librium all prices change in response to thefact that some domestic products are nolonger present The new equilibrium pricesfor products that remain in the market are thuscomputed by utilizing the first-order condi-tions of profit-maximizing firms into whichthe virtual prices of the eliminated productsare substituted Hence to compute the newequilibrium prices we solve an equation sys-tem of the following form

26 As Table 2 indicates most of the patents for the drugsin the quinolone segment have expired by now so that noneof the scenarios described in this section is going to mate-rialize in practice The reason we consider these alternativescenarios in our counterfactual simulations is to get a senseof how the presence and extent of domestic competitionaffects welfare calculations Such calculations may becomerelevant in the future in other therapeutic classes in whichpatents have not expired yet and where domestic firmscompete with foreign patent holders While the particularexit scenarios will depend on the exact year of patent expiryof drugs in each therapeutic class and the specific way inwhich patents will be enforced (eg whether certain do-mestic drugs will be grandfathered in) the general conclu-sion that emerges from our calculations is that the consumerwelfare loss is substantially smaller in cases where somedomestic competition remains present in the market Hencethe particular way in which patent enforcement is imple-mented is essential for assessing its welfare impact

27 Of course until product patents are in fact introducedthese prices will not be observable Note also that we areassuming here that the range of available products will notchange with the introduction of patents

28 As mentioned above this first set of calculations ab-stracts from the existence of remaining price controls orother government regulations that would impose constraintson the firmsrsquo profit maximization problem Accordingly theresulting numbers should be interpreted as a benchmark ofwhat would happen if markets were completely unregulatedTo examine what the welfare effects of TRIPS would be inthe more realistic scenario of price regulation we subse-quently consider a scenario in which regulation keeps theprices of the products provided by patent-holders at theirpre-TRIPS level For a more detailed discussion see theresults section

1494 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

For products i that are withdrawn from themarket

(14) 0 i ir iiln pirV i10ln pjr j Di

10

j Di

01

i01ln pjr

j Di

00

i00ln pjr

ilnXQr

PQr

For products k that remain in the market

(15) pkr ckr 1 1

kk pkrpjr pirV 1

In the equations above pirV denotes the virtual

prices of the products that are removed fromthe market while pjr denotes the updatedprices of all other products Note that whensolving for the virtual prices we account forthe fact that both the price index for quino-lones PQr and the expenditure allocated tothis subsegment XQr need to be updated toreflect the fact that as a result of the pricechanges there may be substitution away fromthis subsegment To obtain the new quinoloneexpenditure XQr and the new price index PQrwe use the estimates and formulas for thehigher-level AIDS system In equation (15)ck refers to the marginal cost for product k thatwe have obtained from the previous estimationstage As mentioned in the previous subsectionwe conducted the simulations using both theupper and lower bounds for marginal cost ck

u

and ckL respectively The term kk(pkr

pjrpirV) re-

fers to the unconditional own-price elasticity forproduct k which is a function of the eliminatedproductsrsquo virtual prices and the remaining prod-uctsrsquo new equilibrium prices We conduct thecounterfactual simulations and welfare analysisat the regional level because the presence ofregion-specific product effects in the demandestimation implies region-specific demand elas-ticities and hence region-specific marginalcosts and markups The presentation and discus-sion of our results focus on national averages ofthe relevant variables that we construct by com-

puting weighted averages across regions usingthe population of each region as a weight

F Welfare Assessment

The simulation of the new equilibrium underpatent protection can provide important insightsinto how consumers and firms will respond tothe removal of domestic products in the market(for example toward which products consumerswill substitute or which prices will increase themost) To get a more precise idea of how peo-plersquos well-being will ultimately be affected byTRIPS we compute as a last step in our anal-ysis the welfare effects of the policy changeSocial welfare is defined as consumer welfareand the sum of domestic firm profits Thechange in domestic profits can easily be calcu-lated by comparing the domestic firm (variable)profits at the pre-TRIPS prices to the profitsthese firms will realize at the new simulatedprices Although foreign firm profits do notcount in domestic welfare calculations we alsocompute the effects of patent enforcement onforeign firm profits to get an idea of how largethe expected benefits of TRIPS are for thesefirms This provides in some sense an indirectway of assessing whether the claims that patentenforcement in countries like India will lead tomore research on developing-country-specificdiseases (such as malaria) have any validity iffor example we find that the effect of patentenforcement on the foreign firm profits realizedin India is small in magnitude it is unlikely thatforeign firms will engage in more developing-country-specific research in response to TRIPSIt is important to note that in all these calcula-tions we work with the lower-bound estimatesfor marginal costs since these give rise to thehighest possible markups Hence our estimateof profit loss for domestic producers most likelyoverstates this loss

On the consumer side we measure changes inconsumer welfare by the compensating variation(CV) defined as the additional expenditure thatconsumers need in order to achieve the sameutility level as before patent enforcement at thenew prices Specifically let P0 denote the pricevector before patent enforcement P the simulatedprice vector post-TRIPS (that we obtained usingthe methods described in the previous subsection)u0 the utility attained by consumers before TRIPS

1495VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

and E(u P) the higher-level expenditure functionThen the compensating variation is given by

(16) CV Eu0 P Eu0 P0

Note that the CV as computed in (16) representsthe combination of three effects

The pure product-variety effect that is theeffect that arises because one or more prod-ucts are not available to consumers anymoreholding the prices of all other remainingproducts and the total expenditure on thequinolone subsegment XQ constant

The expenditure-switching effect that is theeffect arising from substitution away fromquinolones and toward other subsegments ofthe antibiotics market again holding the pricesof all other remaining products constant

The reduced-competition effect that is theeffect that arises because the firms remainingin the market adjust (increase) their prices inresponse to the removal of domestic products

From both an analytical and a policy point ofview it is desirable to assess how large each ofthese effects is Accordingly we decompose thetotal effect on consumer welfare (the CV as givenby equation (16)) using the following procedure

To get the pure product variety effect wecompute virtual prices for the products that areremoved from the market holding the quino-lone expenditure XQ and the prices of all otherproducts fixed Let us call the resulting pricevector P1 Then the pure product variety effectis represented by E(u0 P1) E(u0 P0)

To compute the expenditure-switching effectwe compute another set of virtual prices againholding the prices of all remaining products fixedbut letting quinolone expenditure adjust in re-sponse to the new price index for the quinolonesegment (Given that the prices of the remainingproducts remain fixed the change in the priceindex arises only because of the removal of one ormore domestic products) Note that this scenariomost closely resembles the case in which patentlaws are enforced but strict price regulation keepsthe prices of the products offered in the market attheir pre-TRIP level Let us label the so-computedprice vector P2 The expenditure-switching effectis then E(u0 P2) E(u0 P1)

Finally the reduced competition effect aris-ing from higher prices for the remaining prod-

ucts is computed as the residual change in thecompensating variation once the product-varietyand expenditure-switching effects have been ac-counted for that is E(u0 P) E(u0 P2)where the price vector P is computed accordingto the formulas (14) and (15) to reflect theadjustment of prices to the new regime

To compute the standard errors associatedwith the counterfactual simulations (that is thestandard errors for the counterfactual prices andwelfare estimates) we again use bootstrappingWe first bootstrap the original sample (sales andprices of all the drugs) 300 times Next we esti-mate the AIDS model for each of these samplesand compute the counterfactual equilibrium pricesand welfare losses corresponding to each policyscenario for each of the 300 simulations In thefinal step we compute standard errors29

One limitation of our framework is that itdoes not allow for a heterogeneous response ofconsumers to the policy change Accordinglyour framework is not suited to addressing thequestion of how different groups in the popula-tion will be affected by patent law enforce-ment30 Along the same lines our frameworkdoes not accommodate the possibility of pricediscrimination which might lead to differentresults regarding welfare losses and profit gainsrelative to uniform pricing31 Still we believethat the results of the counterfactual simulationscan provide important insights into the likelyaggregate response to patent enforcement andthe factors that drive this response

III Results

A The Structure of Demand

Tables A2 and A3 in the Appendix display theresults from estimation of the lower- and upper-level AIDS systems respectively For ease of in-terpretation rather than discussing the coefficientestimates we focus our discussion on the implied

29 In cases some of the simulations do not converge wecompute the standard errors using only those simulationsthat converged

30 Addressing this question would require at a mini-mum micro data on consumer purchases as in Goldberg(1995) To our knowledge such data do not exist for theIndian pharmaceutical market

31 See Ernst R Berndt (1994) for an extensive discussionof uniform pricing

1496 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

unconditional price and expenditure elasticities re-ported in Table 6 Given that the region-specificproduct group effects ir reported in the last fourcolumns of Tables A2 and A3 imply region-spe-cific demand elasticities we report separate elas-ticities for each region in Tables 6A and 6B Asevident from the comparison of these tables theelasticities are very similar across regions so thatwe can focus the remaining discussion on theelasticities of one region only the northern regiondisplayed in Table 6A

The diagonal terms of Table 6A report theown-price elasticities which are in all but onecase negative and highly significant The oneexception is the foreign norfloxacin productgroupmdashwhose share of quinolone sales is 007percentmdashfor which we estimate a negative butinsignificant own-price elasticity For the remain-ing product groups demand appears to be highlyelastic with the estimated elasticities lower than2 in four out of the six cases The magnitude ofthe own-price elasticities matches the features ofthe Indian pharmaceutical market mentioned ear-lier which would suggest that Indian consumersare likely to be quite price-sensitive32 The elas-ticities appear especially large if one takes intoaccount that they refer to product groups (such asdomestic ciprofloxacin) and not individual drugsoffered by particular firms Their relative magni-tudes are also intuitive the drug with the largestmarket share and a relatively high price (domesticciprofloxacin) appears to be one of the least elas-tic In contrast foreign ciprofloxacin is highlyprice elastic this is plausible as ciprofloxacindrugs offered by subsidiaries of multinationalsface the stiffest competition in this market seg-ment from approximately 75 Indian firms offeringthe same molecule

The estimated expenditure elasticities appear inthe last column These are all positive indicatingthat the demand for all product groups is normalThe remaining cells display the estimated cross-price elasticities As one might perhaps expect forproducts within a therapeutic subsegment theseare mostly positive Out of a total of 49 priceelasticities we estimate there are six that do notconform to expectations these are the cross-price elasticities between different foreign prod-

uct groups which are estimated to be negativeand significant33 Fortunately these elasticitieshave negligible impact on the welfare analysisgiven that our counterfactuals focus on theeffect of withdrawing one or more domesticproducts from the market the most relevantelasticities are the ones that capture the responseof various product group shares to a change in theprice of one or more domestic groups these elas-ticities are the cross-price elasticities between var-ious domestic groups and the ones betweendomestic and foreign groups which are plausibleand precisely estimated34

Regarding these elasticities a striking aspectof our estimates is how large positive andsignificant the cross-price elasticities betweendifferent domestic product groups aremdashin factfor norfloxacin and ofloxacin we estimate thatdomestic product groups containing differentmolecules are closer substitutes for one anotherthan product groups that contain the same mol-ecule but are produced by foreign firms Incontrast for ciprofloxacin (the molecule withthe largest revenue share) we estimate a largepositive cross-price elasticity between the do-mestic and foreign versions

The fact that domestic products appear to beclose substitutes for other domestic productsthat contain different molecules truly representsan ldquoempiricalrdquo finding in the sense that we do

32 In developed economies elasticities of this magnitude havetypically been found only for generic drugs (and even then onlyrarely) or among consumers who lack health insurance

33 While these elasticities are clearly counterintuitive theyare not inconsistent with the underlying demand system whichimposes no restrictions on the sign of the cross-price elastici-ties We do not have a good explanation of why these elastic-ities are estimated to be negative A potential explanation isthat the shares of the foreign products are very small giventhis we observe very few consumers switching from one (verysmall) foreign group to another (very small) foreign groupwhen the price of the first foreign group goes up and hence theinference is not very reliable in this case

34 The cross-price elasticities between foreign drug groupscontaining different molecules will also have an effect on thewelfare estimates given that the withdrawal of domestic prod-ucts will generally lead to changes in the prices of all foreignproducts and market shares will be reallocated from someforeign products to other foreign products based on the newprices However this reallocation from ldquoforeign to foreignrdquo istruly second order in our case compared to the reallocationfrom ldquodomestic to domesticrdquo and ldquodomestic to foreignrdquo prod-ucts In scenarios in which we consider patent enforcementaccompanied by strict price regulation the cross-price elastic-ities between foreign product groups are in fact completelyirrelevant as the prices of foreign products are not allowed toincrease in these cases Still our welfare loss estimates remainsubstantial drivenmdashas beforemdashby the loss of product variety

1497VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

not impose it through any of our assumptionsregarding the demand function The questionthat naturally arises then is what might explainthis finding While we cannot formally addressthis question anecdotal accounts in various in-dustry studies suggest that the explanation maylie in the differences between domestic andforeign firms in the structure and coverage ofretail distribution networks

Distribution networks for pharmaceuticals inIndia are typically organized in a hierarchicalfashion Pharmaceutical companies deal mainlywith carrying and forwarding (CampF) agents inmany instances regionally based who each sup-ply a network of stockists (wholesalers) Thesestockists in turn deal with the retail pharma-cists through whom retail sales ultimately oc-cur35 The market share enjoyed by a particularpharmaceutical product therefore depends inpart on the number of retail pharmacists who

stock the product And it is here that thereappears to be a distinction between domesticfirms and multinational subsidiaries In particu-lar the retail reach of domestic firms as agroup tends to be much more comprehensivethan that of multinational subsidiaries (IndianCredit Rating Agency (ICRA) 1999)36

There appear to be two reasons for this Thefirst is that many of the larger Indian firmsbecause they have a much larger portfolio ofproducts over which to spread the associatedfixed costs typically have more extensive net-works of medical representatives The second issimply that there are many more domestic firms(and products) on the market At the retail levelthis would imply that local pharmacists mightbe more likely to stock domestic products con-taining two different molecules say ciprofloxa-cin and norfloxacin than they would domesticand foreign versions of the same molecule Tothe extent that patients (or their doctors) arewilling to substitute across molecules in order tosave on transport or search costs (eg going toanother pharmacy to check whether a particular

35 There are estimated to be some 300000 retail pharma-cists in India On average stockists deal with about 75 retailers(ICRA 1999) There are naturally variations in this structureand a host of specific exclusive dealing and other arrangementsexists in practice Pharmaceutical firms also maintain networksof medical representatives whose main function is to marketthe companyrsquos products to doctors who do the actual prescrib-ing of drugs In some instances firms do sell directly to thedoctors who then become the ldquoretailerrdquo as far as patients areconcerned but these are relatively rare

36 These differences were also highlighted in conversa-tions one of the authors had with CEOs and managingdirectors of several pharmaceutical firms as part of a sep-arate study

TABLE 6AmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN THE NORTHERN REGION

Product group

Elasticity with respect to

Prices of foreign productgroups Prices of domestic product groups Overall

quinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 557 013dagger 015 401 011dagger 011dagger 016 137(179) (007) (007) (184) (006) (006) (006) (029)

Foreign norfloxacin 427dagger 045 427dagger 350dagger 602 451 465 220(242) (112) (242) (210) (623) (184) (183) (105)

Foreign ofloxacin 011 010dagger 138 009 009dagger 023 011 116(005) (005) (031) (027) (005) (028) (004) (017)

Domestic ciprofloxacin 018 001 001 168 008 008 010 117(008) (000) (001) (023) (002) (002) (002) (003)

Domestic norfloxacin 004 003 004 058 223 042 040 073(001) (003) (001) (017) (011) (004) (003) (009)

Domestic ofloxacin 005 005 011 077 074 342 074 089(002) (002) (013) (028) (008) (025) (008) (021)

Domestic sparfloxacin 007 004 007 115 063 063 288 028(002) (001) (002) (015) (006) (006) (017) (012)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent significance level and dagger (dagger) denotes significance at the 10-percent level

1498 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE 6BmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN OTHER REGIONS

Demand patterns in the eastern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 594 014dagger 016 431 013 011 017 140(195) (008) (008) (201) (006) (007) (007) (031)

Foreign norfloxacin 329dagger 058 329dagger 264dagger 460 346 359 192(180) (083) (179) (153) (451) (138) (139) (082)

Foreign ofloxacin 012 010dagger 140 010 010 024 013 117(006) (006) (034) (031) (005) (030) (005) (019)

Domestic ciprofloxacin 018 001 001 172 009 008 011 117(008) (000) (002) (027) (002) (003) (002) (003)

Domestic norfloxacin 004 004 004 068 242 049 046 070(002) (004) (002) (020) (013) (005) (004) (010)

Domestic ofloxacin 004dagger 004 009 061 060 295 060 092(002) (001) (010) (028) (007) (022) (007) (017)

Domestic sparfloxacin 005 003 005 092 048 051 251 043(002) (001) (001) (017) (005) (006) (016) (010)

Demand patterns in the western region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 427 009dagger 011 286 008 006 011 126(129) (005) (005) (133) (004) (005) (004) (021)

Foreign norfloxacin 714dagger 008 710dagger 594dagger 995 742 770 299dagger

(369) (183) (368) (328) (1015) (297) (296) (178)Foreign ofloxacin 013 012dagger 145 009 012 026 013 119

(007) (006) (037) (029) (005) (033) (005) (021)Domestic ciprofloxacin 019 001 000 174 009 008 011 118

(008) (000) (001) (024) (002) (003) (002) (004)Domestic norfloxacin 004 004 004 067 243 050 047 069

(002) (004) (001) (018) (014) (005) (004) (010)Domestic ofloxacin 003 003 007 048 048 257 048 093

(002) (001) (008) (022) (005) (015) (004) (014)Domestic sparfloxacin 006 003 005 083 046 049 241 046

(002) (001) (001) (014) (004) (005) (012) (009)

Demand patterns in the southern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 560 013dagger 015 402 011 010 017 137(182) (007) (007) (187) (006) (006) (007) (029)

Foreign norfloxacin 533dagger 031 531dagger 431dagger 748 559 584 249dagger

(301) (134) (300) (245) (734) (219) (219) (128)Foreign ofloxacin 014 012dagger 148 011 012 029 016 120

(007) (007) (039) (033) (006) (035) (006) (022)Domestic ciprofloxacin 017 001 000 169 008 007 011 116

(008) (000) (001) (025) (002) (003) (002) (003)Domestic norfloxacin 004 003 003 059 225 043 040 073

(001) (003) (001) (019) (012) (004) (004) (009)Domestic ofloxacin 004dagger 004 009 062 060 296 060 091

(002) (001) (010) (026) (006) (019) (007) (017)Domestic sparfloxacin 008 005 008 138 073 075 323 016

(002) (001) (002) (018) (007) (008) (021) (014)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at 5-percent level anddagger (dagger) denotes significance at the 10-percent level

1499VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

foreign product is in stock) in aggregate datawe would expect to find precisely the substitu-tion patterns that we report in Table 6

Whether the particular explanation we provideabove is the correct one the high degree of sub-stitutability between domestic product groupsturns out to have important implications for thewelfare calculations We discuss these in moredetail below when we present the results of thecounterfactual welfare analysis Another elasticitywith important implications for the counterfactu-als is the price elasticity for the quinolone subseg-ment as a whole which indicates how likelyconsumers are to switch to other antibioticsgroups when faced with a price increase for quin-olones This elasticity is computed on the basis ofthe results in Table A3 and it is at 111 (stan-dard error 024) this is large in magnitudebutmdashas expectedmdashsmaller in absolute value thanthe own-price elasticities of the product groupswithin the quinolone subsegment

The results in Tables 6A and 6B are based onour preferred specification discussed in SectionII In Tables A4 to A6 in the Appendix weexperiment with some alternative specificationsTables A4(a)ndashA4(c) correspond to a specificationthat includes in addition to product-group-specificregional fixed effects product-group-specific (andfor the upper level antibiotics-segment-specific)seasonal effects We distinguish among three sea-sonsmdashthe summer monsoon and wintermdashandreport the unconditional demand elasticities for

the northern region for each of these seasons Asevident from the tables in the Appendix our elas-ticity estimates are robust to the inclusion of sea-sonal effects The demand elasticities in Table A5are based on estimation of the demand system byOLS Compared to the elasticities obtained by IVthe OLS elasticities are smaller in absolute valueimplying that welfare calculations based on theOLS estimates would produce larger welfare lossestimates Nevertheless some of the patterns re-garding the cross-price elasticities discussed ear-lier are also evident in the OLS results inparticular the cross-price elasticities between dif-ferent domestic product groups are all positivelarge and significant and in most instances largerthan the cross-price elasticities between drugs thatcontain the same molecule but are produced byfirms of different domesticforeign status Theclose substitutability of domestic products indi-cated by both the OLS and IV estimates seems tobe one of the most robust findings of the paper

B Cost and Markup Estimates

Table 7 displays the marginal costs markupsand profits implied by the price elasticity esti-mates of Tables 6A and 6B for each of the sevenproduct groups Given that our regional effectsimply different price elasticities for each regionour marginal cost and markup estimates alsodiffer by region Given however that based on

TABLE 7mdashUPPER AND LOWER BOUNDS FOR MARGINAL COST MARKUP AND ANNUAL PROFIT BY PRODUCT GROUPS WITHIN

THE QUINOLONE SUBSEGMENT

Product groupLower boundfor MC (Rs)

Upper boundfor markup

Upper boundfor profit(Rs mill)

Upper boundfor MC (Rs)

Lower boundfor markup

Lower boundfor profit

(Rs)

Foreign ciprofloxacin 83 19 269 103 0 00(123) (012) (1655)

Foreign norfloxacin NA NA NA 53 0 00Foreign ofloxacin 323 70 1061 1085 0 00

(2316) (021) (3185)Domestic ciprofloxacin 47 59 17019 112 0 00

(114) (010) (29858)Domestic norfloxacin 52 43 2807 90 0 00

(020) (002) (1532)Domestic ofloxacin 587 34 1612 901 0 00

(218) (002) (1280)Domestic sparfloxacin 495 37 1985 788 0 00

(157) (002) (1100)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level Estimated lower bound forforeign norfloxacinrsquos marginal cost is negative since the estimated price elasticity is less than one in absolute value

1500 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Tables 6A and 6B the price elasticities do notseem to differ substantially across regions wereport for ease of exposition only the nationalaverages for marginal costs and markups inTable 7

Table 7 has two parts In the left part (firstthree columns) we report the lower bound formarginal cost and the corresponding upperbound for markup and upper bound for totalannual profit for each product group Thesenumbers are based on the price elasticities weobtained from estimating the two-level AIDSdemand system Since we do not have a reliableestimate for the price elasticity of foreign nor-floxacin (the point estimate is negative but lessthan one in absolute value and insignificant)we cannot compute the lower bound for mar-ginal cost in this case It is important to note thatthese estimates do not reflect either the actualmarginal cost or the actual markup for thesedrugs both because the existence of price reg-ulation implies that the unconstrained first-orderconditions are unlikely to hold each period andbecause our aggregation across firms of thesame domesticforeign status supplying thesame molecule makes the interpretation of theseestimates problematic In particular the fact thatwe ignore competition among firms within eachproduct group implies that our estimates willtend to overstate market power These numberswill prove useful however in the counterfac-tual simulations as they can give us a sense ofhow large the maximum profit gains for multi-national firms and the maximum profit lossesfor domestic firms are likely to be under patentenforcement The right part of the table (lastthree columns) reports the upper bound for mar-ginal cost which we obtain by simply taking themarginal cost to be equal to the observed priceWe do not report standard errors in this casesince the numbers are based on actual data Themarkups corresponding to these marginal costupper bounds are of course zero We conductthe counterfactual simulations using both thelower and upper bounds for marginal costs

The (upper bounds for) markups on the leftside of the table are generally plausible Thedomestic ciprofloxacin product group that dom-inates the quinolone subsegment and for whichwe documented high prices a high marketshare and a relatively low elasticity of demandenjoys one of the highest markup upper bounds(60 percent) and accounts for nearly 70 percent

of all profits derived within the subsegmentForeign ciprofloxacin on the other hand whichfaces the stiffest competition from domesticfirms and for which we estimated a highly elas-tic demand has the lowest markup upper bound(19 percent)

C Counterfactual Estimates of the Impact onPrices and Welfare

With estimates of the key demand and costparameters in hand we turn to the counterfac-tuals We consider the five separate scenarioslisted in the previous section All of the scenar-ios involve the withdrawal of one or more of thedomestic product groups from the market Ta-ble 8 displays our estimates of the consumerwelfare losses that result under the differentscenarios The losses are expressed in billion ofrupees per year All numbers presented in Ta-ble 8 and subsequent tables are based on usingthe lower bounds for marginal cost and upperbounds for markup in the simulations As dis-cussed earlier these numbers are the more in-teresting to work with since they give us anupper bound for the changes in the profits ofdomestic and foreign firms that would resultfrom patent enforcement In the Appendix wealso present results based on using the upperbounds for marginal costs in the simulations inwhich case the pre-TRIPS profits of domesticand foreign firms are zero In all cases marginalcosts are assumed to be constant in outputWhile naturally the profit implications differdepending on whether one uses the upper orlower bounds for marginal costs (firm profits arezero if one assumes the upper bound of mar-ginal cost in which case price equals marginalcost) the estimated consumer welfare losses aresimilar in the two cases We discuss these re-sults in more detail at the end of this section

The first column presents our estimates of theconsumer welfare losses attributable to the pureloss of product variety effect where we fix theprices of all remaining products as well as theoverall expenditure on quinolones while with-drawing one or more of the domestic productgroups Note that in our initial specification ofthe demand system had we not allowed for thepossibility that consumers might differentiatebetween domestic and foreign products evenwhen they contain the same molecule this

1501VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

particular component of the loss of consumerwelfare would not have arisen

The estimates reported in the second columnincorporate the expenditure-switching effect ontop of the loss of product variety Here basedupon the price elasticity estimates from thehigher-level AIDS system we adjust (down-ward) the expenditures allocated to the quino-lone subsegment as the composite price ofquinolones effectively increases as a consequenceof the higher virtual prices of the domestic productgroups that are withdrawn from the market Be-cause the estimates in this column are generatedassuming that the prices of the products that re-main in the market are not adjusted upward theyprovide a sense of what consumer losses would beif the introduction of product patents were coupledwith strict price regulation aimed at maintainingprices at prepatent levels Alternatively they canbe thought of as the relevant welfare numbers ifintense competition among firms within the re-maining product groups kept the prices of theproducts that were still offered in the market closeto the firmsrsquo marginal costs

The last column of Table 8 displays the esti-mated consumer welfare losses when bothcross-segment expenditure-switching and with-in-segment upward price adjustments are takeninto account

If we compare the results across the firstsecond and third columns all the counterfac-tual scenarios produce qualitatively similar pat-terns patterns that are consistent with what we

would expect Starting from the initial loss ofwelfare attributable to the loss of product vari-ety the option of switching expenditures outof the quinolone subsegment to other subseg-ments mitigates some of the initial welfareloss But if we then incorporate the upwardprice adjustments that result in response to thereduced competition the welfare losses aremagnified

Of particular interest from a policy perspec-tive are the relative magnitudes of these threeeffects which are similar under all the counter-factual scenarios though absolute levels varyconsiderably First despite the fact that the de-mand for quinolones is quite sensitive to thecomposite price of quinolonesmdashthe upper levelprice elasticity is 111mdashthe cross-subsegmentexpenditure-switching effects are in all thecases small (in absolute value terms) relative tothe other two effects For instance under thescenario where all the domestic quinolone prod-uct groups are withdrawn from the market theoverall consumer welfare loss of Rs 1781 bil-lion per year can be decomposed into an initialloss of Rs 1176 billion (66 percent) attribut-able to the loss of product variety a slightreduction in this initial loss of Rs 041 billion(2 percent) from Rs 1176 billion to Rs1135 billion because of expenditure switchingand a subsequent additional loss of Rs 646billion (36 percent) from Rs 1135 billion toRs 1781 billion because of the reduced com-petition and consequent price increases

TABLE 8mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure loss ofvariety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 732(087) (089) (146)

Only ofloxacin 008 008 023(008) (008) (010)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1253(177) (180) (415)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1058(142) (145) (331)

All four domestic quinolones products 1176dagger 1135dagger 1781(643) (634) (1270)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent significance level and dagger (dagger)denotes significance at the 10-percent level

1502 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 8: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

quinolone subsegment Quinolones fall into thesystemic antibiotics and antibacterials segmentof the Indian pharmaceuticals market whichgenerates over three-quarters of the revenues inthe antiinfectives segment8 The systemic anti-bacterials segment includes all of the originalmiracle drugs that first sparked the growth ofthe global research-based pharmaceutical indus-try in the postndashWorld War II period as well aslater generations of molecules that have beenintroduced in the last four decades

Among systemic antibacterials quinolonesare the latest generation molecules available inIndia We focus our analysis on quinolones forseveral reasons First quinolones are the drug ofchoice for a large number of bacterial infec-tions some of which are also treated by alter-native drugs (see Table A1 in the Appendixwhich outlines the spectrum of activity for eachmolecule family within the antibacterials seg-ment) Hence if there were one product groupfor which we would expect to have many sub-stitutes readily available this would be quino-lones Second with a share of 20 percent in thesales of systemic antibacterials quinolones rep-resent one of the largest subsegments within thistherapeutic category Finally several moleculeswithin the quinolone subsegment were still un-der patent in the United States at the time of ourinvestigation Table 2 details the basic informa-tion about the four quinolone molecules that arethe focus of our analysis The first row showsthe year of US patent expiry this ranges from1998 for norfloxacin to 2010 for sparfloxacinQuinolones include in principle four moremolecules that are listed at the bottom of Ta-ble 1 the market shares of these molecules arenegligible however so that we exclude thesemolecules from our analysis

Table 2 reveals several other interesting factsabout competition in the quinolone market inIndia First note the large number of firmsoperating in this subsegment The large numberof domestic firms is perhaps not that surprisinggiven that pharmaceutical product patents werenot recognized in India9 What is more surpris-ing is the number of foreign firms selling pat-ented products (eg ciprofloxacin) the fact that

multiple foreign firms sell a patented productindicates that such firms often ldquoinfringerdquo10

patent laws in India while complying with themin developed world countries The last two rowsof Table 2 further indicate that domestic prod-ucts often sell at a premium With the exceptionof ofloxacin the average prices of products of-fered by Indian firms are higher than the pricesof products offered by foreign subsidiaries Thispreliminary evidence suggests that Indian con-sumers do not place a premium on the brandname and reputation of big multinational phar-maceutical concerns Moreover the higherprice of domestic products does not seem toprevent domestic companies from capturing alarge market share This is most evident in thecase of ciprofloxacin where domestic firmshave with 53 percent the largest share in thetotal sales of quinolonesmdashthis despite the factthat the average price of these products is 10percent higher than the price of foreign productscontaining the same molecule

Table 3 provides additional summary statis-tics for our data broken down by region Thefirst two rows of the table report the averageannual household expenditure on quinolonesand antibiotics respectively Note that in bothcases the average expenditure is higher in thenorth and west these regions include states withhigher per capita incomes and tend to be moreindustrialized and urbanized than those in theeast and south Pharmaceutical products areavailable in multiple presentations that is com-binations of dosage forms (eg capsule tabletsyrup) strength (eg 100 milligrams 500 mil-ligrams) and packet sizes (eg 50 capsule bot-tle 100 tablet bottle) The various presentationsin which a product is available are often referredto as stock-keeping units or SKUs11 The num-ber of SKUs for each product group withinquinolones is reported at the top of Table 3 Aswith the more aggregate numbers on firms and

8 In addition to anti-bacterials this segment also containsantivirals

9 Accordingly the common distinction between ldquobrandedrdquoand ldquogenericrdquo products is irrelevant here

10 We emphasize here that the word ldquoinfringerdquo belongsin quotes because patent laws do not currently exist inIndia infringement in the legal sense is not possible It ishowever striking that the same firms that accuse Indianproducers of ldquopiracyrdquo sell in India products that are patentedin the United States and for which the patent is held by adifferent multinational corporation

11 For instance a 100 capsule bottle of 100 milligramcapsules of a particular branded drug and a 50 capsulebottle of 100 milligram capsules of the same branded drugwould be identified as two separate SKUs

1484 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

products reported in Table 2 the differencebetween domestic and foreign products isstriking The number of SKUs offered byIndian firms is consistently larger than thenumber offered by subsidiaries of foreignmultinationals The number of SKUs variesslightly across regions but more importantlyit varies across time as some SKUs disap-pear while new ones get introduced duringour sample period

Many pharmaceutical products in India aresubject to price controls12 While the specificsof the price regulation are too complex for anyeconomic model to capture adequately themain concern for the empirical analysis is thatprice controls may lead to a lack of price vari-ation over time so that the demand functioncannot be identified Prices at the most disag-gregate SKU level are relatively stable overtime there are variations due to occasionalchanges in the estimated cost (due for exampleto changes in exchange rates that affect the costof imported materials or bulk drugs) but suchvariations tend to be infrequent and small inmagnitude The degree of time variation ishowever substantially larger once one aggre-gates to the product level This variability stemsnot only from the fact that the SKUs over whichwe aggregate may experience changes in theirrespective prices at different points in time butalso from the fact that the range of SKUs of-fered in the market does not remain constantover time The entry and exit of presentationswithin the same product group that have differ-

ent prices effectively affects the price that con-sumers face for this drug in each period

The middle portion of Table 3 reports themean price and standard deviation for eachproduct group by region Prices vary by regionthough there is no clear pattern emerging fromthe table with regard to the cross-regional vari-ation (in the sense of some regions being sys-tematically more expensive than others) Toexamine what portion of the total price variationis due to time versus regional variation weconducted an analysis of variance of prices thatwe report in Table 4 The table is based onseparate regressions for each product group(pooling data across groups with big differencesin their average prices is not particularly infor-mative as most of the price variation is ac-counted for by product group dummies) Thelast two columns of the table show the fractionof price variation accounted for by region andtime dummies respectively As evident fromthe table a significant fraction of the total vari-ance in prices can be attributed to time varia-tion In the demand estimation we include a fullset of product groupndashspecific regional dummiesso that the price parameters are identified en-tirely based on this time variation within eachproduct group The time variation of productgroup prices is driven primarily by composi-tional changes within each group the revenueshares of the individual SKUs that compriseeach product group change over time (see therelated discussion in Section IIC) while in eachperiod there is entry and exit of SKUs into thesample To check whether this pattern reflectsgenuine entry and exit as opposed to samplingvariation we examined the revenue shares ofthe SKUs that leave the sample relative to theones that remain during the entire period Theresults are reported in Table 5 While the SKUs

12 The details of the procedures for price fixation canbe found in the official government Web site httpwwwnppaindianicinindex1html under the link ldquoDrug Price Con-trol Order 1995rdquo A new pharmaceutical policy was introducedin 2002 but our data were collected before that year

TABLE 4mdashANALYSIS OF PRODUCT PRICE VARIANCE

Product groupPartial SS

timePartial SS

region Total SSPercentage

explained by timePercentage

explained by region

Foreign ciprofloxacin 0296 0306 1047 283 292Foreign norfloxacin 1036 1002 4179 248 240Foreign ofloxacin 0429 0021 0571 752 36Domestic ciprofloxacin 0005 0088 0104 50 841Domestic norfloxacin 0059 0098 0198 296 494Domestic ofloxacin 2858 0104 3056 935 34Domestic sparfloxacin 1754 0032 1860 943 17

1485VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

that exit tend to be smaller (their average shareis 1 percent as opposed to 34 percent for thoseSKUs that are present during the entire sampleperiod) the shares of the two groups do notseem orders of magnitude apart13 In additionour data cover only the 300 largest firms sellingin the Indian market so that firms with verysmall shares are not included in our sample

II The Analytic Framework and EstimationApproach

Patent enforcement in the Indian pharmaceu-tical market will have the effect of eliminatingdomestic products whose active pharmaceuticalingredients are protected by (foreign) patentsThus assessing the effects of patent enforce-ment is tantamount to assessing the effects ofwithdrawing domestic products from the mar-ket This task is the converse of evaluating newproduct introduction accordingly the concep-tual framework we use to address the questionsof interest is similar to the one developed in theliterature for the valuation of new goods14

We start by estimating demand for quino-lones Given that the market is characterized byimperfect competition the counterfactual anal-ysis requires that we also model the supply sideas removal of one product will affect the pricesof other products especially those that are closecompetitors The existence of price regulationin the Indian pharmaceutical market imposespotential constraints on firmsrsquo maximizationproblem Given these constraints and the com-

plexity of the price regulation process the typ-ical approach of deriving estimates of actualmarginal costs and markups by exploiting thefirst-order conditions of profit maximizing firmsdoes not seem particularly promising Insteadwe use our demand estimates to place upper andlower bounds on marginal costs and markups

With demand elasticities and upper and lowerbounds for marginal costs in place we thenconduct counterfactual simulations We con-sider several alternative scenarios dependingon the number of domestic products affected bypatent enforcement For each scenario we com-pute the counterfactual prices and use them toassess the effects of domestic product with-drawal on consumer welfare (as measured bythe compensating variation) firm profits andsocial welfare As with the valuation of newproducts the big conceptual problem facing thispart of the analysis is that we need to extrapo-late from the region of the data to the point atwhich demand for the products that exit themarket becomes zero This conceptual issue ispresent in any attempt to evaluate a major pol-icy change for which no historical precedentexists like the enforcement of patent laws inIndia One advantage of the present study is thatwe have a limited set of price data for Pakistana country with similar demographics to Indiabut with a market structure that resembles theone that would emerge in India under patentenforcement (monopoly of multinational sub-sidiaries) By comparing the prices of productsoffered by multinationals in Pakistan to thosewe compute in our counterfactual simulationsfor the products that would be offered by mul-tinationals in India if patent laws were enforcedwe can get a sense of how plausible our coun-terfactual estimates are

A Demand

The demand modelling is based on the multi-stage budgeting approach Our primary motiva-tion for adopting this approach is a practical oneIn the multistage budgeting approach the depen-dent variable is defined as a revenue share whichis appealing given that the products we include inthe analysis contain different molecules (ie ac-tive pharmaceutical ingredients or APIs) Eventhough we do have data on the quantity of therelevant API (eg 100 milligrams of ciprofloxa-cin) contained in each product converting the

13 While we cannot completely rule out the possibilitythat some of the exit is due to sampling variation note thatthe latter should be reflected in low precision of the demandparameter estimates The demand parameters however areprecisely estimated

14 See Manuel Trajtenberg (1989) Jerry A Hausman(1994) and Timothy F Bresnahan (1997) for representativeexamples and a discussion of the relevant issues

TABLE 5mdashREVENUE SHARES OF THE EXITING SKUS

Revenue share ofexiting SKUs

Revenue shares ofall SKUs

Full sample 10 34Northern region 08 33Eastern region 12 36Western region 10 34Southern region 12 33

1486 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

revenue shares to physical shares is extremelydifficult if not infeasible in the case of antibioticsBecause such drugs are ldquosystemicrdquo by nature theyare used to treat a large number of infections andthe dosage of each drug depends on the particularinfection it is supposed to address (for example thedosage will differ depending on whether the anti-biotic is used to combat an ear infection or tuber-culosis) This particular feature of antibioticscomplicates the conversion of revenue to physicalshares

The basic idea of the multistage budgetingapproach is to use the therapeutic classificationof a productmdashie the therapeutic segment andsubsegment the product belongs tomdashto orga-nize all products in the systemic antibacterialssegment into a hierarchical taxonomy consist-ing of two levels At the higher level are thevarious subsegments of systemic antibacterialsThe first stage of budgeting corresponds to theallocation of expenditures across the subseg-ments in this upper level of the taxonomy

In the second stage of the budgeting processcorresponding to the lower level of the taxon-omy a flexible functional form is adopted tomodel how the expenditures allocated to eachsubsegment are distributed across the productswithin that subsegment In particular to modeldemand at the second stage we employ theAIDS specification proposed by Deaton andMuellbauer (1980b)15

While the two-stage demand estimation ap-proach offers functional form flexibility its ap-plication to the Indian systemic antibioticsmarket poses a few problems The first is thatdue to entry and exit many SKUs and evenproducts in our sample are not present in everyperiod AIDS does not have a good way ofdealing with a varying number of products as itwas developed with broad commodity catego-ries in mind which are consumed by all con-sumers every period To solve this problem weaggregate within each subsegment (eg quino-lones) SKUs into product groups where withineach product group all SKUs contain the samemolecule and are produced by firms with thesame domesticforeign status Specifically let aSKU k be indexed by its molecule (or API) M its

domesticforeign status DF indicating whether itis produced by a domestic (Indian) or a subsidiaryof a foreign (multinational) firm a particular pre-sentation s and the particular firm f that producesit We aggregate SKUs over presentations andfirms to obtain a newly defined product group iwhich is indexed only by molecule M and domes-ticforeign status DF and has revenue Ri yenfsRk with i (M DF) k (M DF f s) and pricepi yenfs kpk where k denotes the conditional(on M and DF) revenue share of this particularproduct ie

(1) k Rk

Ri

In most cases the resulting product groups arebroad enough to be present every period16

The usual concern with this aggregation pro-cedure is that it may lead us to overstate firmsrsquomarket power as we ignore competition amongfirms with the same domesticforeign statusproducing the same molecule In the presentapplication however this concern is unlikely tobe of great importance as the effect of patentenforcement is to wipe out all domestic compe-tition at once while granting foreign firms mo-nopoly power hence competition among firmsfor patented molecules becomes irrelevant Theaggregation according to the domesticforeignstatus (within a particular molecule) thus corre-sponds to the scope of our analysis and theparticular questions of interest

The second problem is that for our approachto be useful in welfare analysis the allocation oftotal expenditures to group expenditures at thehigher stage has to be modelled in a way con-sistent with utility maximization In general thesolution of this allocation problem requiresknowledge of all individual product prices

15 Representative applications of the multistage budget-ing approach include Sara F Ellison et al (1997) Hausman(1994) and Hausman and Gregory K Leonard (2002)

16 We are missing only four observations (ie monthregion combinations) all for the drug group of ForeignNorfloxacin August 1999 in the south May 2000 in thewest October 2000 in the south and November 2000 in theeast In these cases we set the revenue shares of ForeignNorfloxacin equal to zero In general with 01 percent ofquinolone sales (see Table 1 row 2) Foreign Norfloxacinhas a very small share of the market This probably explainswhy the results pertaining to this drug group are unreliableit is the only drug group for which we do not obtain asignificant price elasticity of demand while its cross-priceelasticities with other foreign drug groups often have thewrong sign

1487VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

From an empirical point of view this is notparticularly useful as it eliminates all compu-tational advantages of the two-stage approachTo address this problem we adopt an approxi-mate solution to model the higher-level expen-diture allocation along the lines suggested byDeaton and Muellbauer (1980b pp 131ndash32)This gives rise to a two-level AIDS specification

Consider the lower-level estimation firstwhich refers to the allocation of a particularsubsegmentrsquos expenditure to the product groupswithin the subsegment In our application therelevant subsegment is quinolones which weindex with Q Let the product groups within thissubsegment be indexed by i 1 N pi be theprice of product group i (where as noted abovei refers to a particular molecule and domesticforeign status combination) and XQ be the totalexpenditure on the quinolone segment The rev-enue share of each product group is given by

(2) i i j

ijln pj ilnXQ

PQ

where i the revenue share of product group iis defined as

(3) i pi qi

yenj

pj qj

xi

XQ with i j Q

XQ is the overall expenditure on the quinolonesubsegment and PQ is a price index given by

(4) ln PQ ap 0 i

iln pi

1

2 i

j

ijln piln pj

With a limited number of product groups and asufficiently large number of time-series observa-tions the flexibility implied by the AIDS modeldoes not impose too many demands on the data Inthe present application however where the num-ber of observations is limited the AIDS model isnot estimable in this general form To reduce thenumber of parameters that need to be estimatedwe impose two sets of restrictions

The first set of restrictions is implied by thetheory of utility maximization Specificallythese restrictions are

Adding-up yenk k 1 yenk k 0 yenk kj 0 j

Homogeneity yenk jk 0 j Symmetry ij 1frasl2 [ij ji] ji This last

restriction by itself reduces the number of parameters to [N(N 1)]2

The second set of restrictions we impose aimsat further reducing the number of parameters tobe estimated by exploiting our knowledge of thisparticular market Specifically for each productgroup i we allow one ij parameter for all productgroups j that have different molecules from prod-uct group i and are produced by foreign firms andone ij for all product groups j with differentmolecules produced by domestic firms We donrsquotimpose any restrictions on the ij parameter whenproduct group j has the same molecule as productgroup i (By construction product groups i and jcontain products produced by firms with differentdomesticforeign status)

To better illustrate the nature of the restric-tions we impose on the patterns of substitutionacross products some additional notation isneeded Let d(i j) be an indicator of the degreeof similarity (or difference) between productgroup i and product group j along the dimen-sions we are able to observe (molecule M anddomesticforeign status DF) For any two prod-uct groups i and j d(i j) can take on one of thefollowing three values17

(5)

di j (1 0) if Mi Mj DFi DFj

(0 1) if Mi Mj DFi DFj

(0 0) if Mi Mj DFi DFj

Let

(6) Diab j di j a b

17 The sequence (1 1) is not possible for two differentproducts in this case the parameter corresponds to theproductrsquos own-price effect that is ii

1488 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The equation at the lower level becomes

(7) i i iiln pi i10ln pj j Di10

j Di

01

i01ln pj

j Di

00

i00ln pj ilnXQ

PQ

Note that

The parameter ii captures a product grouprsquosown-price effect (there will be as many iiparameters as the number of product groups)

The parameter i10 captures the cross-priceeffect of the product group containing prod-ucts with the same molecule but produced byfirms of different nationality

The parameter i01 captures the cross-priceeffects of product groups containing productswith different molecules but produced byfirms with the same nationality

The parameter i00 captures the cross-priceeffects of product groups containing productswith different molecules produced by firms ofdifferent nationality

Before we take the demand equation to thedata we make two modifications The first oneis to let the product-specific effects i vary byregion r The resulting product-specific regionaleffects ir have two interpretations first theycontrol for the ldquoqualityrdquo of each drug withquality differences being allowed to vary acrossregions second they proxy for demographicsand other demand shifters which vary by re-gion and may affect the demand of each prod-uct group differently18 Note that by includingproduct-specific regional effects in the demandspecification we estimate the price parametersbased on the within-product-group variation ofprices in each region In an earlier version of thepaper we also estimated the demand systemwithout regional dummies and obtained similarresults

The second modification that allows us to gofrom a deterministic to a stochastic specificationof the demand equation is to include an additiveerror term in (7) The latter takes into accountthe fact that (7) is not expected to fit the dataexactly The error term irt accounts for mea-surement error (due to the fact that the productgroup prices pjrt we employ in the estimationare not exact price indices but approximationsthereof) and (potentially region-specific) de-mand shocks that may affect the demand for aproduct in a particular period Examples of suchshocks include an advertising campaign for aparticular product which temporarily increasesthe demand for this product or the outbreak ofa (potentially region-specific) epidemic whichcalls for the use of a particular drug We discussthe interpretation and properties of this errorterm in more detail in the next section

The final form of the equation we estimate atthe lower level becomes (with subscript t de-noting month and subscript r denoting region)

(8) irt i ir iiln pirt i10ln pjrt j Di10

j Di

01

i01ln pjrt

j Di

00

i00ln pjrt

ilnXQrt

PQrt irt

The analysis so far has conditioned on the ex-penditure allocated to the quinolone subseg-ment XQ The upper level of the estimationconsiders the problem of allocating total expen-diture across the different systemic antibioticssubsegments one of which is quinolones Theupper-level demand function is given by

(9) G G H

GHln PH GlnX

P

where all variables denoted by capital letters aredefined as before but now refer to subsegments(G H ) rather than product groups within asubsegment and the total expenditures on sys-temic antibiotics X are deflated by the Stone priceindex log P yenH Hlog PH When estimating thesystem above we impose all the restrictions

18 Given the short time span of our sample typicaldemand shifters such as age distribution income distribu-tion and education hardly change over our sample periodSuch shifters are therefore absorbed by the region-specificproduct fixed effects ir

1489VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

implied by utility maximization as we do withthe estimation of the lower-level AIDS We donot however impose any additional restrictionson the substitution patterns at this stage so thatthe cross-price effects across segments remainrelatively unconstrained

Estimation of the higher-level AIDS allowsus to obtain the unconditional own- and cross-price elasticities that are used in the formulationof the supply problem and welfare analysisThese are given by the formula

(10) ij ijXQXQ

ln qi

ln XQ

ln XQ

ln PQ

ln PQ

ln pj

with the conditional cross price elasticitiesgiven by

ijXQXQ ij ij jlnXQ

PQ

i

As in the lower stage we include subsegment-specific regional dummies in the specificationof the upper-stage demand system so that thefinal form of the estimating equation at theupper stage becomes

(11) Grt G Gr H

GHln PHrt

GlnXrt

Prt Grt

In sum the demand system we take to thedata is represented by equations (8) and (11)and the associated parameter restrictions im-plied by economic theory

B Modelling the Supply Side of the Market

Counterfactual simulations concerning theeffects of domestic product withdrawal requireknowledge of the marginal costs of pharmaceu-tical firms operating in the Indian market Theseare unobservable The usual approach in theNew Empirical Industrial Organization litera-ture has been to exploit the firm equilibriumconditions to infer marginal cost For exampleit is often assumed that the marginal cost ci isconstant and that the industry is an oligopoly

engaging in Bertrand competition with differen-tiated products Assuming that firms myopicallymaximize profits each period one can then de-rive the firmsrsquo first-order conditions that corre-spond to the assumptions above about costsmarket structure and firm behavior

We deviate from this procedure as the presenceof price regulation renders the assumption of un-constrained period-by-period maximization un-tenable Ideally one would like to explicitlyincorporate the price and other administrative con-trols into the firmrsquos optimization problem andderive the first-order conditions under the assump-tion of constrained maximization The complexityof the price regulation however makes this ap-proach infeasible Therefore we adopt an alterna-tive approach that does not rely on modelling theprice-setting process but is instead based on de-riving upper and lower bounds for marginal costsand markups

In particular an upper bound for marginalcosts and a lower bound for the markups (zero)can be derived under the assumption of perfectcompetition While this assumption is clearlyunrealistic in the pharmaceuticals market it isuseful in providing an upper bound for costs ci

Uwhich will be given by

(12) ciU pi

On the other hand a lower bound for marginalcosts ci

L (and upper bound for markups) can bederived by assuming that there is perfect collu-sion within each product group i (where i refershere to a moleculedomestic-foreign combina-tion)19 and ignoring price controls so thatprices are determined by the first-order condi-tion of the jointly profit-maximizing firmswithin product group i Solving this first-ordercondition for the (lower bound of) marginal costthen gives

(13) ciL pi 1

1

iipipj

19 The reason that this assumption will lead to an under-statement of marginal costs (and hence overstatement ofmarket power) is that it assumes away competition amongfirms within each drug group

1490 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

where ii(pipj) the own-price elasticity of de-

mand for product i will depend on the productrsquosown-price pi and all other productsrsquo prices pjwith i j

Once we have obtained the demand elastic-ities through estimation of the demand sys-tem we can calculate the upper and lowerbounds for marginal costs and correspondingmarkups according to (12) or (13) These willthen be employed in the counterfactual sim-ulations While our counterfactuals use boththe lower and upper bounds for costs most ofour discussion will be based on using thelower bounds for costs since this is the moreinteresting case in the policy analysis it givesus the largest possible profits for pharmaceuticalfirms and hence corresponds to the worst pos-sible scenario facing Indian firms and the bestpossible scenario facing multinationals underTRIPS

C Identification Assumptions and EstimationApproach

The discussion of the demand system has sofar abstracted from the issue of price endogene-ity The usual premise in the industrial organi-zation literature is that correlation of prices withthe error term in the demand equation arises byvirtue of the first-order conditions of profit-maximizing firms20 As we discuss in this sec-tion this source of simultaneity bias is unlikelyto be of major concern in the present contextbecause of the existence of price regulation Ourprimary concern is instead with the simultane-ity bias that is implied by the particular wayprices are constructed

To understand the sources of the simultaneitybias note first that the price of product group piwe employ in the demand estimation should bethought of as a proxy for an exact price indexthat we do not observe As such pi contains bydefinition measurement error and it will becorrelated with the error term of the demandequation Specifically the demand equation weare interested in estimating can be written insimplified form (suppressing the subscripts rand t and ignoring parameter restrictions forconvenience) as

i i j

ijln pjT ilnXQ

PQ i

where pjT denotes the true (exact) price index for

product group i and i captures unobservedvariables that may affect demand in a particularperiod (eg an advertising campaign or an ep-idemic) The price index pj

T depends on theSKU prices pk Note that in the presence ofSKU entry and exit into the sample the priceindex pj

T will vary over time even if the pricesof the individual SKUs remain stable as theproducts (SKUs) that comprise each productgroup change each period

Ideally we would like to compute the exactprice index pj

T for each group and employ theseindices in the demand estimation Under theassumption of predetermined SKU prices eachperiod the exact price indices would be uncor-related with the error term and the demandparameters could be estimated by simple OLSUnfortunately this strategy is not feasible in thecurrent context

To derive an exact price index for each prod-uct group it is necessary to model consumerrsquoschoice among SKUs conditional on the choiceof the group To this end it is necessary toexplicitly introduce a third stage in the demandspecification and estimate the parameters asso-ciated with the choice at that stage Howeverany specification based on three-stage budget-ing (that would require directly specifying ademand function associated with the SKUchoice at the third stage and imposing additiveseparability at the higher stages) would be in-feasible to estimate for three main reasonsFirst there is a large number of SKUs withinsome groups so that the demand parameters can-not be estimated even if one imposes parameterrestrictions second the SKU prices exhibit littlevariation over time lastly the choice sets at thislowest stage vary over time rendering estimationof the price parameters associated with SKUs thatare not available in all periods infeasible Alterna-tively one could abandon the multistage budget-ing altogether and adopt a discrete choiceapproach which is inherently better suited to deal-ing with varying choice sets and limited pricevariation A discrete choice approach would beparticularly appealing for modeling the SKUchoice conditional on product group choice giventhat within each group all SKUs contain the same

20 See Steven Berry et al (1995) or Aviv Nevo (2001)for a related discussion

1491VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

molecule and are hence comparable As notedearlier however a discrete choice approach at thehigher stages of the demand estimation requiresconverting revenue to physical shares which ischallenging in the case of systemic drugs such asantibiotics for which the appropriate dosage is notwell defined Finally a combination of a multi-stage budgeting approach for the higher stages anda discrete choice approach for the lowest stagewould be feasible to estimate but inconsistentwith a model of utility maximization For thesereasons we have no alternative but to adopt anapproximation for measuring the price index foreach product group

Let us denote this approximation by pjA and let

13j denote the proportional ldquoapproximationrdquo errorassociated with measuring the true price index forgroup j so that ln pj

A ln pjT ln 13j The

estimating demand equation can then be written as

i i j

ijln pjA ilnXQ

PQ

j

ijln 13j i

or

i i j

ijln pjA ilnXQ

PQ i

where the new error term i yenj ijln 13j iis comprised of two components the first onereflects ldquomeasurementrdquo error due to the fact thatthe product group prices we employ are approx-imations and not exact price indices and thesecond one captures conventional demandshocks The demand equation as written abovecorresponds to the equation we take to the data

The particular proxy we use for the exact priceindex for each product group j is the revenue-share-weighted average of the prices of multipleSKUs available within this group that is

pjA

k j

k pk

The notation above illustrates the two source ofprice variation variation in the SKU prices pkand variation in the weights k

We are concerned with potential correlation ofthe so-constructed product group prices with bothcomponents of the product group demand error

term The correlation between pjA and the first

component yenj ijln 13j is independent of theparticular way we construct pj

A and inherent in thefact that we measure the exact price index witherror The potential correlation between pj

A and thesecond component of the product group demanderror the shock i arises however because of thespecific way we construct the proxy pj

A and inparticular because of the presence of the reve-nue share weights k in pj

A since the SKUrevenue share weights k will generally dependon the product group expenditure they arelikely to be correlated with the product groupdemand shock i

To address the simultaneity bias we use in-strumental variables To this end we need vari-ables that are correlated with the proxies pj

A butuncorrelated with the error term i We use thenumber of SKUs within each product group j asan instrument as it does justice to the idea thatvariation in the product group price index stemsin part from variation in the set of SKUs that areavailable each period It is clearly correlatedwith the average group prices pj

A and plausiblyuncorrelated with the demand shock i Theassumption that the number of SKUs is uncor-related with the conventional demand error willhowever be violated if the introduction of anew SKU changes the perceived quality of adrug21 or if it is accompanied by promotionalactivities which will be reflected in i Our hopeis that this is not too often the case Assumingthat the SKU number for each group j is alsouncorrelated with the weighted sum of the mea-surement errors of all product groups yenj ijln 13j itcan be used as an instrument for the product groupprice pj

A22 If one accepts the premise that prices atthe SKU level are exogenous then SKU pricescan also be used as instruments The final list ofinstruments we use hence includes the number ofSKUs in each group the prices of the five largestSKUs for each group and all other exogenousvariables in the demand estimation such as re-gionalproduct dummy interactions and the upper-

21 Note that we already proxy for the quality of each prod-uct group through the product fixed effects i In the actualestimation we actually let these fixed effects be region-specific(ir) so that we allow quality to vary across regions

22 If the measurement error increases in the number ofincluded SKUs this assumption will be violated Unfortu-nately we have no way of checking the validity of thisassumption

1492 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

level total expenditure on antibiotics Regressionsof group prices on the instruments above yieldhigh R-squares with most regressors highly sig-nificant indicating that our instruments are highlycorrelated with prices23

Our sample includes four molecules cipro-floxacin norfloxacin ofloxacin and spar-floxacin Except for sparfloxacin all othermolecules are produced by both foreign anddomestic firms24 So we have seven products(domestic ciprofloxacin foreign ciprofloxa-cin etc) with 96 observations (two years ofmonthly data four geographical regions foreach period) for each product The parametersin the lower-level AIDS demand system asdefined in equation (8) are the product fixedeffects i the productndashregional dummy inter-actions ir the own revenue-share price elas-ticities ii the cross revenue-share priceelasticities i10 i01 i00 and the revenue-share expenditure elasticities i In estimatingthe parameters we first regress prices on allinstrumental variables and then plug the pre-dicted values for prices in the constrainedleast-square regression (For a detailed expla-nation of the constraints see the previoussection)

Given that we impose many cross-equationconstraints and employ instrumental variablesin the estimation it is difficult to derive stan-dard errors for the parameter estimates ana-lytically Our error term interpretation in (8)implies that the error terms for each productgroup are likely to be correlated across re-gions for example if a national advertisingcampaign increases the demand for a rela-tively expensive presentation in one regionsimultaneously increasing the aggregate de-mand and price index for the correspondingproduct group in that region it is likely that thesame effect will be observed in other regions Inprinciple we could exploit the cross-regionalcorrelation in the error terms to estimate thedemand system using seemingly unrelated re-

gression (SUR) This however would requireus to derive the variance-covariance matrix ofthe parameters analytically which as notedabove is cumbersome in the current contextInstead we use the bootstrap method choosingto remain agnostic about the structure of thevariance-covariance matrix The potential dis-advantage of this approach over SUR is a loss inefficiency but as will become apparent in theresults section the model parameters are fairlyprecisely estimated To maintain the marketstructure we randomly sample the periods (withreplacement) and use the same periods for allproducts Regarding the optimal number ofbootstrap repetitions ideally one would followthe three-step method proposed by Donald W KAndrews et al (2000) Empirical evidence sug-gests however that one rarely needs more than200 replications to estimate the standard er-rors25 To be safe we generate 300 bootstrapsamples (with replacement) based on the origi-nal data and estimate the standard errors usingthe standard errors of the bootstrap sampleestimates

The estimation of the top-level AIDS systemis similar The constraints imposed on this top-level demand system are adding-up homogene-ity and symmetry as defined on page 12Again bootstrapping is used to obtain the stan-dard errors of the parameter estimates

D The Counterfactual Scenarios

In assessing the effects of patent enforce-ment we start by focusing on the most extremecase in which compulsory licensing is not anoption and foreign firms are not subject to pricecontrols We use the results from the analysis ofthis case as a benchmark In reality the outcomeof the WTO negotiations is more likely to in-volve some constraints on the monopoly powerof foreign firms selling patented products indeveloping countries such as price caps orcompulsory licensing Our framework can eas-ily accommodate these cases as will becomeapparent in the next subsection

We now focus on the effects of potentialpatent enforcement in the quinolone segmentWe consider several scenarios which vary inthe number and size of domestic products that

23 Specifically the R-squares from the first-stage regres-sions of product group prices on instruments range from057 to 095 except for domestic ciprofloxacin for whichthe R-square is 017

24 Sparfloxacin is actually offered by one foreign subsid-iary in India (see Table 2) However its revenue share isminiscule We therefore treat Sparfloxacin as being pro-duced by domestic firms only 25 See Bradley Efron and Robert J Tibshirani (1993)

1493VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

will be removed from the market In particularwe consider the following five scenarios26

Withdrawal of one large domestic productgroup only domestic ciprofloxacin

Withdrawal of one relatively small domesticproduct group only domestic ofloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand norfloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand sparfloxacin

Withdrawal of all four domestic quinoloneproduct groups

As the list above suggests we proceed fromanalyzing the effects of single-product with-drawal to the analysis of eliminating the en-tire domestic segment This approach wasmotivated by early empirical results that in-dicated that the existence and extent of com-petition from domestic firms has a significantbearing on the predicted effects of patent en-forcement that is our predictions regardingprices and welfare vary substantially depend-ing on how many domestic products are af-fected by patent enforcement and on whethersome domestic competition will remain presentafter TRIPS In addition the size of the affecteddomestic groups is relevant for the welfare pre-dictions accordingly we examine both scenar-ios in which a large domestic product groupsuch as domestic ciprofloxacin is eliminatedfrom the market and scenarios in which the

eliminated domestic group is relatively small(eg domestic ofloxacin) In all scenarios wemaintain the assumption that the set of prod-ucts offered by the remaining firms in themarket does not change in response to patentenforcement

E Computation of Virtual Prices and NewEquilibrium Prices

The first step in the counterfactual analysisis to derive the new equilibrium prices underpatent enforcement27 In this context thereare two sets of prices that are relevant Thefirst set consists of the virtual prices of those(domestic) products that will not be availableonce TRIPS is put in effect To calculate thesevirtual prices we set the expenditures of therelevant products equal to zero The secondset of prices consists of the prices of thoseproducts that remain in the market In deriv-ing these prices we start by assuming profitmaximization without any form of price reg-ulation the firms remaining in the marketreoptimize in response to the policy changeand set new prices taking the prices of allother firms as given28 Of course at the equi-librium all prices change in response to thefact that some domestic products are nolonger present The new equilibrium pricesfor products that remain in the market are thuscomputed by utilizing the first-order condi-tions of profit-maximizing firms into whichthe virtual prices of the eliminated productsare substituted Hence to compute the newequilibrium prices we solve an equation sys-tem of the following form

26 As Table 2 indicates most of the patents for the drugsin the quinolone segment have expired by now so that noneof the scenarios described in this section is going to mate-rialize in practice The reason we consider these alternativescenarios in our counterfactual simulations is to get a senseof how the presence and extent of domestic competitionaffects welfare calculations Such calculations may becomerelevant in the future in other therapeutic classes in whichpatents have not expired yet and where domestic firmscompete with foreign patent holders While the particularexit scenarios will depend on the exact year of patent expiryof drugs in each therapeutic class and the specific way inwhich patents will be enforced (eg whether certain do-mestic drugs will be grandfathered in) the general conclu-sion that emerges from our calculations is that the consumerwelfare loss is substantially smaller in cases where somedomestic competition remains present in the market Hencethe particular way in which patent enforcement is imple-mented is essential for assessing its welfare impact

27 Of course until product patents are in fact introducedthese prices will not be observable Note also that we areassuming here that the range of available products will notchange with the introduction of patents

28 As mentioned above this first set of calculations ab-stracts from the existence of remaining price controls orother government regulations that would impose constraintson the firmsrsquo profit maximization problem Accordingly theresulting numbers should be interpreted as a benchmark ofwhat would happen if markets were completely unregulatedTo examine what the welfare effects of TRIPS would be inthe more realistic scenario of price regulation we subse-quently consider a scenario in which regulation keeps theprices of the products provided by patent-holders at theirpre-TRIPS level For a more detailed discussion see theresults section

1494 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

For products i that are withdrawn from themarket

(14) 0 i ir iiln pirV i10ln pjr j Di

10

j Di

01

i01ln pjr

j Di

00

i00ln pjr

ilnXQr

PQr

For products k that remain in the market

(15) pkr ckr 1 1

kk pkrpjr pirV 1

In the equations above pirV denotes the virtual

prices of the products that are removed fromthe market while pjr denotes the updatedprices of all other products Note that whensolving for the virtual prices we account forthe fact that both the price index for quino-lones PQr and the expenditure allocated tothis subsegment XQr need to be updated toreflect the fact that as a result of the pricechanges there may be substitution away fromthis subsegment To obtain the new quinoloneexpenditure XQr and the new price index PQrwe use the estimates and formulas for thehigher-level AIDS system In equation (15)ck refers to the marginal cost for product k thatwe have obtained from the previous estimationstage As mentioned in the previous subsectionwe conducted the simulations using both theupper and lower bounds for marginal cost ck

u

and ckL respectively The term kk(pkr

pjrpirV) re-

fers to the unconditional own-price elasticity forproduct k which is a function of the eliminatedproductsrsquo virtual prices and the remaining prod-uctsrsquo new equilibrium prices We conduct thecounterfactual simulations and welfare analysisat the regional level because the presence ofregion-specific product effects in the demandestimation implies region-specific demand elas-ticities and hence region-specific marginalcosts and markups The presentation and discus-sion of our results focus on national averages ofthe relevant variables that we construct by com-

puting weighted averages across regions usingthe population of each region as a weight

F Welfare Assessment

The simulation of the new equilibrium underpatent protection can provide important insightsinto how consumers and firms will respond tothe removal of domestic products in the market(for example toward which products consumerswill substitute or which prices will increase themost) To get a more precise idea of how peo-plersquos well-being will ultimately be affected byTRIPS we compute as a last step in our anal-ysis the welfare effects of the policy changeSocial welfare is defined as consumer welfareand the sum of domestic firm profits Thechange in domestic profits can easily be calcu-lated by comparing the domestic firm (variable)profits at the pre-TRIPS prices to the profitsthese firms will realize at the new simulatedprices Although foreign firm profits do notcount in domestic welfare calculations we alsocompute the effects of patent enforcement onforeign firm profits to get an idea of how largethe expected benefits of TRIPS are for thesefirms This provides in some sense an indirectway of assessing whether the claims that patentenforcement in countries like India will lead tomore research on developing-country-specificdiseases (such as malaria) have any validity iffor example we find that the effect of patentenforcement on the foreign firm profits realizedin India is small in magnitude it is unlikely thatforeign firms will engage in more developing-country-specific research in response to TRIPSIt is important to note that in all these calcula-tions we work with the lower-bound estimatesfor marginal costs since these give rise to thehighest possible markups Hence our estimateof profit loss for domestic producers most likelyoverstates this loss

On the consumer side we measure changes inconsumer welfare by the compensating variation(CV) defined as the additional expenditure thatconsumers need in order to achieve the sameutility level as before patent enforcement at thenew prices Specifically let P0 denote the pricevector before patent enforcement P the simulatedprice vector post-TRIPS (that we obtained usingthe methods described in the previous subsection)u0 the utility attained by consumers before TRIPS

1495VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

and E(u P) the higher-level expenditure functionThen the compensating variation is given by

(16) CV Eu0 P Eu0 P0

Note that the CV as computed in (16) representsthe combination of three effects

The pure product-variety effect that is theeffect that arises because one or more prod-ucts are not available to consumers anymoreholding the prices of all other remainingproducts and the total expenditure on thequinolone subsegment XQ constant

The expenditure-switching effect that is theeffect arising from substitution away fromquinolones and toward other subsegments ofthe antibiotics market again holding the pricesof all other remaining products constant

The reduced-competition effect that is theeffect that arises because the firms remainingin the market adjust (increase) their prices inresponse to the removal of domestic products

From both an analytical and a policy point ofview it is desirable to assess how large each ofthese effects is Accordingly we decompose thetotal effect on consumer welfare (the CV as givenby equation (16)) using the following procedure

To get the pure product variety effect wecompute virtual prices for the products that areremoved from the market holding the quino-lone expenditure XQ and the prices of all otherproducts fixed Let us call the resulting pricevector P1 Then the pure product variety effectis represented by E(u0 P1) E(u0 P0)

To compute the expenditure-switching effectwe compute another set of virtual prices againholding the prices of all remaining products fixedbut letting quinolone expenditure adjust in re-sponse to the new price index for the quinolonesegment (Given that the prices of the remainingproducts remain fixed the change in the priceindex arises only because of the removal of one ormore domestic products) Note that this scenariomost closely resembles the case in which patentlaws are enforced but strict price regulation keepsthe prices of the products offered in the market attheir pre-TRIP level Let us label the so-computedprice vector P2 The expenditure-switching effectis then E(u0 P2) E(u0 P1)

Finally the reduced competition effect aris-ing from higher prices for the remaining prod-

ucts is computed as the residual change in thecompensating variation once the product-varietyand expenditure-switching effects have been ac-counted for that is E(u0 P) E(u0 P2)where the price vector P is computed accordingto the formulas (14) and (15) to reflect theadjustment of prices to the new regime

To compute the standard errors associatedwith the counterfactual simulations (that is thestandard errors for the counterfactual prices andwelfare estimates) we again use bootstrappingWe first bootstrap the original sample (sales andprices of all the drugs) 300 times Next we esti-mate the AIDS model for each of these samplesand compute the counterfactual equilibrium pricesand welfare losses corresponding to each policyscenario for each of the 300 simulations In thefinal step we compute standard errors29

One limitation of our framework is that itdoes not allow for a heterogeneous response ofconsumers to the policy change Accordinglyour framework is not suited to addressing thequestion of how different groups in the popula-tion will be affected by patent law enforce-ment30 Along the same lines our frameworkdoes not accommodate the possibility of pricediscrimination which might lead to differentresults regarding welfare losses and profit gainsrelative to uniform pricing31 Still we believethat the results of the counterfactual simulationscan provide important insights into the likelyaggregate response to patent enforcement andthe factors that drive this response

III Results

A The Structure of Demand

Tables A2 and A3 in the Appendix display theresults from estimation of the lower- and upper-level AIDS systems respectively For ease of in-terpretation rather than discussing the coefficientestimates we focus our discussion on the implied

29 In cases some of the simulations do not converge wecompute the standard errors using only those simulationsthat converged

30 Addressing this question would require at a mini-mum micro data on consumer purchases as in Goldberg(1995) To our knowledge such data do not exist for theIndian pharmaceutical market

31 See Ernst R Berndt (1994) for an extensive discussionof uniform pricing

1496 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

unconditional price and expenditure elasticities re-ported in Table 6 Given that the region-specificproduct group effects ir reported in the last fourcolumns of Tables A2 and A3 imply region-spe-cific demand elasticities we report separate elas-ticities for each region in Tables 6A and 6B Asevident from the comparison of these tables theelasticities are very similar across regions so thatwe can focus the remaining discussion on theelasticities of one region only the northern regiondisplayed in Table 6A

The diagonal terms of Table 6A report theown-price elasticities which are in all but onecase negative and highly significant The oneexception is the foreign norfloxacin productgroupmdashwhose share of quinolone sales is 007percentmdashfor which we estimate a negative butinsignificant own-price elasticity For the remain-ing product groups demand appears to be highlyelastic with the estimated elasticities lower than2 in four out of the six cases The magnitude ofthe own-price elasticities matches the features ofthe Indian pharmaceutical market mentioned ear-lier which would suggest that Indian consumersare likely to be quite price-sensitive32 The elas-ticities appear especially large if one takes intoaccount that they refer to product groups (such asdomestic ciprofloxacin) and not individual drugsoffered by particular firms Their relative magni-tudes are also intuitive the drug with the largestmarket share and a relatively high price (domesticciprofloxacin) appears to be one of the least elas-tic In contrast foreign ciprofloxacin is highlyprice elastic this is plausible as ciprofloxacindrugs offered by subsidiaries of multinationalsface the stiffest competition in this market seg-ment from approximately 75 Indian firms offeringthe same molecule

The estimated expenditure elasticities appear inthe last column These are all positive indicatingthat the demand for all product groups is normalThe remaining cells display the estimated cross-price elasticities As one might perhaps expect forproducts within a therapeutic subsegment theseare mostly positive Out of a total of 49 priceelasticities we estimate there are six that do notconform to expectations these are the cross-price elasticities between different foreign prod-

uct groups which are estimated to be negativeand significant33 Fortunately these elasticitieshave negligible impact on the welfare analysisgiven that our counterfactuals focus on theeffect of withdrawing one or more domesticproducts from the market the most relevantelasticities are the ones that capture the responseof various product group shares to a change in theprice of one or more domestic groups these elas-ticities are the cross-price elasticities between var-ious domestic groups and the ones betweendomestic and foreign groups which are plausibleand precisely estimated34

Regarding these elasticities a striking aspectof our estimates is how large positive andsignificant the cross-price elasticities betweendifferent domestic product groups aremdashin factfor norfloxacin and ofloxacin we estimate thatdomestic product groups containing differentmolecules are closer substitutes for one anotherthan product groups that contain the same mol-ecule but are produced by foreign firms Incontrast for ciprofloxacin (the molecule withthe largest revenue share) we estimate a largepositive cross-price elasticity between the do-mestic and foreign versions

The fact that domestic products appear to beclose substitutes for other domestic productsthat contain different molecules truly representsan ldquoempiricalrdquo finding in the sense that we do

32 In developed economies elasticities of this magnitude havetypically been found only for generic drugs (and even then onlyrarely) or among consumers who lack health insurance

33 While these elasticities are clearly counterintuitive theyare not inconsistent with the underlying demand system whichimposes no restrictions on the sign of the cross-price elastici-ties We do not have a good explanation of why these elastic-ities are estimated to be negative A potential explanation isthat the shares of the foreign products are very small giventhis we observe very few consumers switching from one (verysmall) foreign group to another (very small) foreign groupwhen the price of the first foreign group goes up and hence theinference is not very reliable in this case

34 The cross-price elasticities between foreign drug groupscontaining different molecules will also have an effect on thewelfare estimates given that the withdrawal of domestic prod-ucts will generally lead to changes in the prices of all foreignproducts and market shares will be reallocated from someforeign products to other foreign products based on the newprices However this reallocation from ldquoforeign to foreignrdquo istruly second order in our case compared to the reallocationfrom ldquodomestic to domesticrdquo and ldquodomestic to foreignrdquo prod-ucts In scenarios in which we consider patent enforcementaccompanied by strict price regulation the cross-price elastic-ities between foreign product groups are in fact completelyirrelevant as the prices of foreign products are not allowed toincrease in these cases Still our welfare loss estimates remainsubstantial drivenmdashas beforemdashby the loss of product variety

1497VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

not impose it through any of our assumptionsregarding the demand function The questionthat naturally arises then is what might explainthis finding While we cannot formally addressthis question anecdotal accounts in various in-dustry studies suggest that the explanation maylie in the differences between domestic andforeign firms in the structure and coverage ofretail distribution networks

Distribution networks for pharmaceuticals inIndia are typically organized in a hierarchicalfashion Pharmaceutical companies deal mainlywith carrying and forwarding (CampF) agents inmany instances regionally based who each sup-ply a network of stockists (wholesalers) Thesestockists in turn deal with the retail pharma-cists through whom retail sales ultimately oc-cur35 The market share enjoyed by a particularpharmaceutical product therefore depends inpart on the number of retail pharmacists who

stock the product And it is here that thereappears to be a distinction between domesticfirms and multinational subsidiaries In particu-lar the retail reach of domestic firms as agroup tends to be much more comprehensivethan that of multinational subsidiaries (IndianCredit Rating Agency (ICRA) 1999)36

There appear to be two reasons for this Thefirst is that many of the larger Indian firmsbecause they have a much larger portfolio ofproducts over which to spread the associatedfixed costs typically have more extensive net-works of medical representatives The second issimply that there are many more domestic firms(and products) on the market At the retail levelthis would imply that local pharmacists mightbe more likely to stock domestic products con-taining two different molecules say ciprofloxa-cin and norfloxacin than they would domesticand foreign versions of the same molecule Tothe extent that patients (or their doctors) arewilling to substitute across molecules in order tosave on transport or search costs (eg going toanother pharmacy to check whether a particular

35 There are estimated to be some 300000 retail pharma-cists in India On average stockists deal with about 75 retailers(ICRA 1999) There are naturally variations in this structureand a host of specific exclusive dealing and other arrangementsexists in practice Pharmaceutical firms also maintain networksof medical representatives whose main function is to marketthe companyrsquos products to doctors who do the actual prescrib-ing of drugs In some instances firms do sell directly to thedoctors who then become the ldquoretailerrdquo as far as patients areconcerned but these are relatively rare

36 These differences were also highlighted in conversa-tions one of the authors had with CEOs and managingdirectors of several pharmaceutical firms as part of a sep-arate study

TABLE 6AmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN THE NORTHERN REGION

Product group

Elasticity with respect to

Prices of foreign productgroups Prices of domestic product groups Overall

quinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 557 013dagger 015 401 011dagger 011dagger 016 137(179) (007) (007) (184) (006) (006) (006) (029)

Foreign norfloxacin 427dagger 045 427dagger 350dagger 602 451 465 220(242) (112) (242) (210) (623) (184) (183) (105)

Foreign ofloxacin 011 010dagger 138 009 009dagger 023 011 116(005) (005) (031) (027) (005) (028) (004) (017)

Domestic ciprofloxacin 018 001 001 168 008 008 010 117(008) (000) (001) (023) (002) (002) (002) (003)

Domestic norfloxacin 004 003 004 058 223 042 040 073(001) (003) (001) (017) (011) (004) (003) (009)

Domestic ofloxacin 005 005 011 077 074 342 074 089(002) (002) (013) (028) (008) (025) (008) (021)

Domestic sparfloxacin 007 004 007 115 063 063 288 028(002) (001) (002) (015) (006) (006) (017) (012)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent significance level and dagger (dagger) denotes significance at the 10-percent level

1498 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE 6BmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN OTHER REGIONS

Demand patterns in the eastern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 594 014dagger 016 431 013 011 017 140(195) (008) (008) (201) (006) (007) (007) (031)

Foreign norfloxacin 329dagger 058 329dagger 264dagger 460 346 359 192(180) (083) (179) (153) (451) (138) (139) (082)

Foreign ofloxacin 012 010dagger 140 010 010 024 013 117(006) (006) (034) (031) (005) (030) (005) (019)

Domestic ciprofloxacin 018 001 001 172 009 008 011 117(008) (000) (002) (027) (002) (003) (002) (003)

Domestic norfloxacin 004 004 004 068 242 049 046 070(002) (004) (002) (020) (013) (005) (004) (010)

Domestic ofloxacin 004dagger 004 009 061 060 295 060 092(002) (001) (010) (028) (007) (022) (007) (017)

Domestic sparfloxacin 005 003 005 092 048 051 251 043(002) (001) (001) (017) (005) (006) (016) (010)

Demand patterns in the western region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 427 009dagger 011 286 008 006 011 126(129) (005) (005) (133) (004) (005) (004) (021)

Foreign norfloxacin 714dagger 008 710dagger 594dagger 995 742 770 299dagger

(369) (183) (368) (328) (1015) (297) (296) (178)Foreign ofloxacin 013 012dagger 145 009 012 026 013 119

(007) (006) (037) (029) (005) (033) (005) (021)Domestic ciprofloxacin 019 001 000 174 009 008 011 118

(008) (000) (001) (024) (002) (003) (002) (004)Domestic norfloxacin 004 004 004 067 243 050 047 069

(002) (004) (001) (018) (014) (005) (004) (010)Domestic ofloxacin 003 003 007 048 048 257 048 093

(002) (001) (008) (022) (005) (015) (004) (014)Domestic sparfloxacin 006 003 005 083 046 049 241 046

(002) (001) (001) (014) (004) (005) (012) (009)

Demand patterns in the southern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 560 013dagger 015 402 011 010 017 137(182) (007) (007) (187) (006) (006) (007) (029)

Foreign norfloxacin 533dagger 031 531dagger 431dagger 748 559 584 249dagger

(301) (134) (300) (245) (734) (219) (219) (128)Foreign ofloxacin 014 012dagger 148 011 012 029 016 120

(007) (007) (039) (033) (006) (035) (006) (022)Domestic ciprofloxacin 017 001 000 169 008 007 011 116

(008) (000) (001) (025) (002) (003) (002) (003)Domestic norfloxacin 004 003 003 059 225 043 040 073

(001) (003) (001) (019) (012) (004) (004) (009)Domestic ofloxacin 004dagger 004 009 062 060 296 060 091

(002) (001) (010) (026) (006) (019) (007) (017)Domestic sparfloxacin 008 005 008 138 073 075 323 016

(002) (001) (002) (018) (007) (008) (021) (014)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at 5-percent level anddagger (dagger) denotes significance at the 10-percent level

1499VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

foreign product is in stock) in aggregate datawe would expect to find precisely the substitu-tion patterns that we report in Table 6

Whether the particular explanation we provideabove is the correct one the high degree of sub-stitutability between domestic product groupsturns out to have important implications for thewelfare calculations We discuss these in moredetail below when we present the results of thecounterfactual welfare analysis Another elasticitywith important implications for the counterfactu-als is the price elasticity for the quinolone subseg-ment as a whole which indicates how likelyconsumers are to switch to other antibioticsgroups when faced with a price increase for quin-olones This elasticity is computed on the basis ofthe results in Table A3 and it is at 111 (stan-dard error 024) this is large in magnitudebutmdashas expectedmdashsmaller in absolute value thanthe own-price elasticities of the product groupswithin the quinolone subsegment

The results in Tables 6A and 6B are based onour preferred specification discussed in SectionII In Tables A4 to A6 in the Appendix weexperiment with some alternative specificationsTables A4(a)ndashA4(c) correspond to a specificationthat includes in addition to product-group-specificregional fixed effects product-group-specific (andfor the upper level antibiotics-segment-specific)seasonal effects We distinguish among three sea-sonsmdashthe summer monsoon and wintermdashandreport the unconditional demand elasticities for

the northern region for each of these seasons Asevident from the tables in the Appendix our elas-ticity estimates are robust to the inclusion of sea-sonal effects The demand elasticities in Table A5are based on estimation of the demand system byOLS Compared to the elasticities obtained by IVthe OLS elasticities are smaller in absolute valueimplying that welfare calculations based on theOLS estimates would produce larger welfare lossestimates Nevertheless some of the patterns re-garding the cross-price elasticities discussed ear-lier are also evident in the OLS results inparticular the cross-price elasticities between dif-ferent domestic product groups are all positivelarge and significant and in most instances largerthan the cross-price elasticities between drugs thatcontain the same molecule but are produced byfirms of different domesticforeign status Theclose substitutability of domestic products indi-cated by both the OLS and IV estimates seems tobe one of the most robust findings of the paper

B Cost and Markup Estimates

Table 7 displays the marginal costs markupsand profits implied by the price elasticity esti-mates of Tables 6A and 6B for each of the sevenproduct groups Given that our regional effectsimply different price elasticities for each regionour marginal cost and markup estimates alsodiffer by region Given however that based on

TABLE 7mdashUPPER AND LOWER BOUNDS FOR MARGINAL COST MARKUP AND ANNUAL PROFIT BY PRODUCT GROUPS WITHIN

THE QUINOLONE SUBSEGMENT

Product groupLower boundfor MC (Rs)

Upper boundfor markup

Upper boundfor profit(Rs mill)

Upper boundfor MC (Rs)

Lower boundfor markup

Lower boundfor profit

(Rs)

Foreign ciprofloxacin 83 19 269 103 0 00(123) (012) (1655)

Foreign norfloxacin NA NA NA 53 0 00Foreign ofloxacin 323 70 1061 1085 0 00

(2316) (021) (3185)Domestic ciprofloxacin 47 59 17019 112 0 00

(114) (010) (29858)Domestic norfloxacin 52 43 2807 90 0 00

(020) (002) (1532)Domestic ofloxacin 587 34 1612 901 0 00

(218) (002) (1280)Domestic sparfloxacin 495 37 1985 788 0 00

(157) (002) (1100)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level Estimated lower bound forforeign norfloxacinrsquos marginal cost is negative since the estimated price elasticity is less than one in absolute value

1500 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Tables 6A and 6B the price elasticities do notseem to differ substantially across regions wereport for ease of exposition only the nationalaverages for marginal costs and markups inTable 7

Table 7 has two parts In the left part (firstthree columns) we report the lower bound formarginal cost and the corresponding upperbound for markup and upper bound for totalannual profit for each product group Thesenumbers are based on the price elasticities weobtained from estimating the two-level AIDSdemand system Since we do not have a reliableestimate for the price elasticity of foreign nor-floxacin (the point estimate is negative but lessthan one in absolute value and insignificant)we cannot compute the lower bound for mar-ginal cost in this case It is important to note thatthese estimates do not reflect either the actualmarginal cost or the actual markup for thesedrugs both because the existence of price reg-ulation implies that the unconstrained first-orderconditions are unlikely to hold each period andbecause our aggregation across firms of thesame domesticforeign status supplying thesame molecule makes the interpretation of theseestimates problematic In particular the fact thatwe ignore competition among firms within eachproduct group implies that our estimates willtend to overstate market power These numberswill prove useful however in the counterfac-tual simulations as they can give us a sense ofhow large the maximum profit gains for multi-national firms and the maximum profit lossesfor domestic firms are likely to be under patentenforcement The right part of the table (lastthree columns) reports the upper bound for mar-ginal cost which we obtain by simply taking themarginal cost to be equal to the observed priceWe do not report standard errors in this casesince the numbers are based on actual data Themarkups corresponding to these marginal costupper bounds are of course zero We conductthe counterfactual simulations using both thelower and upper bounds for marginal costs

The (upper bounds for) markups on the leftside of the table are generally plausible Thedomestic ciprofloxacin product group that dom-inates the quinolone subsegment and for whichwe documented high prices a high marketshare and a relatively low elasticity of demandenjoys one of the highest markup upper bounds(60 percent) and accounts for nearly 70 percent

of all profits derived within the subsegmentForeign ciprofloxacin on the other hand whichfaces the stiffest competition from domesticfirms and for which we estimated a highly elas-tic demand has the lowest markup upper bound(19 percent)

C Counterfactual Estimates of the Impact onPrices and Welfare

With estimates of the key demand and costparameters in hand we turn to the counterfac-tuals We consider the five separate scenarioslisted in the previous section All of the scenar-ios involve the withdrawal of one or more of thedomestic product groups from the market Ta-ble 8 displays our estimates of the consumerwelfare losses that result under the differentscenarios The losses are expressed in billion ofrupees per year All numbers presented in Ta-ble 8 and subsequent tables are based on usingthe lower bounds for marginal cost and upperbounds for markup in the simulations As dis-cussed earlier these numbers are the more in-teresting to work with since they give us anupper bound for the changes in the profits ofdomestic and foreign firms that would resultfrom patent enforcement In the Appendix wealso present results based on using the upperbounds for marginal costs in the simulations inwhich case the pre-TRIPS profits of domesticand foreign firms are zero In all cases marginalcosts are assumed to be constant in outputWhile naturally the profit implications differdepending on whether one uses the upper orlower bounds for marginal costs (firm profits arezero if one assumes the upper bound of mar-ginal cost in which case price equals marginalcost) the estimated consumer welfare losses aresimilar in the two cases We discuss these re-sults in more detail at the end of this section

The first column presents our estimates of theconsumer welfare losses attributable to the pureloss of product variety effect where we fix theprices of all remaining products as well as theoverall expenditure on quinolones while with-drawing one or more of the domestic productgroups Note that in our initial specification ofthe demand system had we not allowed for thepossibility that consumers might differentiatebetween domestic and foreign products evenwhen they contain the same molecule this

1501VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

particular component of the loss of consumerwelfare would not have arisen

The estimates reported in the second columnincorporate the expenditure-switching effect ontop of the loss of product variety Here basedupon the price elasticity estimates from thehigher-level AIDS system we adjust (down-ward) the expenditures allocated to the quino-lone subsegment as the composite price ofquinolones effectively increases as a consequenceof the higher virtual prices of the domestic productgroups that are withdrawn from the market Be-cause the estimates in this column are generatedassuming that the prices of the products that re-main in the market are not adjusted upward theyprovide a sense of what consumer losses would beif the introduction of product patents were coupledwith strict price regulation aimed at maintainingprices at prepatent levels Alternatively they canbe thought of as the relevant welfare numbers ifintense competition among firms within the re-maining product groups kept the prices of theproducts that were still offered in the market closeto the firmsrsquo marginal costs

The last column of Table 8 displays the esti-mated consumer welfare losses when bothcross-segment expenditure-switching and with-in-segment upward price adjustments are takeninto account

If we compare the results across the firstsecond and third columns all the counterfac-tual scenarios produce qualitatively similar pat-terns patterns that are consistent with what we

would expect Starting from the initial loss ofwelfare attributable to the loss of product vari-ety the option of switching expenditures outof the quinolone subsegment to other subseg-ments mitigates some of the initial welfareloss But if we then incorporate the upwardprice adjustments that result in response to thereduced competition the welfare losses aremagnified

Of particular interest from a policy perspec-tive are the relative magnitudes of these threeeffects which are similar under all the counter-factual scenarios though absolute levels varyconsiderably First despite the fact that the de-mand for quinolones is quite sensitive to thecomposite price of quinolonesmdashthe upper levelprice elasticity is 111mdashthe cross-subsegmentexpenditure-switching effects are in all thecases small (in absolute value terms) relative tothe other two effects For instance under thescenario where all the domestic quinolone prod-uct groups are withdrawn from the market theoverall consumer welfare loss of Rs 1781 bil-lion per year can be decomposed into an initialloss of Rs 1176 billion (66 percent) attribut-able to the loss of product variety a slightreduction in this initial loss of Rs 041 billion(2 percent) from Rs 1176 billion to Rs1135 billion because of expenditure switchingand a subsequent additional loss of Rs 646billion (36 percent) from Rs 1135 billion toRs 1781 billion because of the reduced com-petition and consequent price increases

TABLE 8mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure loss ofvariety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 732(087) (089) (146)

Only ofloxacin 008 008 023(008) (008) (010)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1253(177) (180) (415)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1058(142) (145) (331)

All four domestic quinolones products 1176dagger 1135dagger 1781(643) (634) (1270)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent significance level and dagger (dagger)denotes significance at the 10-percent level

1502 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 9: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

products reported in Table 2 the differencebetween domestic and foreign products isstriking The number of SKUs offered byIndian firms is consistently larger than thenumber offered by subsidiaries of foreignmultinationals The number of SKUs variesslightly across regions but more importantlyit varies across time as some SKUs disap-pear while new ones get introduced duringour sample period

Many pharmaceutical products in India aresubject to price controls12 While the specificsof the price regulation are too complex for anyeconomic model to capture adequately themain concern for the empirical analysis is thatprice controls may lead to a lack of price vari-ation over time so that the demand functioncannot be identified Prices at the most disag-gregate SKU level are relatively stable overtime there are variations due to occasionalchanges in the estimated cost (due for exampleto changes in exchange rates that affect the costof imported materials or bulk drugs) but suchvariations tend to be infrequent and small inmagnitude The degree of time variation ishowever substantially larger once one aggre-gates to the product level This variability stemsnot only from the fact that the SKUs over whichwe aggregate may experience changes in theirrespective prices at different points in time butalso from the fact that the range of SKUs of-fered in the market does not remain constantover time The entry and exit of presentationswithin the same product group that have differ-

ent prices effectively affects the price that con-sumers face for this drug in each period

The middle portion of Table 3 reports themean price and standard deviation for eachproduct group by region Prices vary by regionthough there is no clear pattern emerging fromthe table with regard to the cross-regional vari-ation (in the sense of some regions being sys-tematically more expensive than others) Toexamine what portion of the total price variationis due to time versus regional variation weconducted an analysis of variance of prices thatwe report in Table 4 The table is based onseparate regressions for each product group(pooling data across groups with big differencesin their average prices is not particularly infor-mative as most of the price variation is ac-counted for by product group dummies) Thelast two columns of the table show the fractionof price variation accounted for by region andtime dummies respectively As evident fromthe table a significant fraction of the total vari-ance in prices can be attributed to time varia-tion In the demand estimation we include a fullset of product groupndashspecific regional dummiesso that the price parameters are identified en-tirely based on this time variation within eachproduct group The time variation of productgroup prices is driven primarily by composi-tional changes within each group the revenueshares of the individual SKUs that compriseeach product group change over time (see therelated discussion in Section IIC) while in eachperiod there is entry and exit of SKUs into thesample To check whether this pattern reflectsgenuine entry and exit as opposed to samplingvariation we examined the revenue shares ofthe SKUs that leave the sample relative to theones that remain during the entire period Theresults are reported in Table 5 While the SKUs

12 The details of the procedures for price fixation canbe found in the official government Web site httpwwwnppaindianicinindex1html under the link ldquoDrug Price Con-trol Order 1995rdquo A new pharmaceutical policy was introducedin 2002 but our data were collected before that year

TABLE 4mdashANALYSIS OF PRODUCT PRICE VARIANCE

Product groupPartial SS

timePartial SS

region Total SSPercentage

explained by timePercentage

explained by region

Foreign ciprofloxacin 0296 0306 1047 283 292Foreign norfloxacin 1036 1002 4179 248 240Foreign ofloxacin 0429 0021 0571 752 36Domestic ciprofloxacin 0005 0088 0104 50 841Domestic norfloxacin 0059 0098 0198 296 494Domestic ofloxacin 2858 0104 3056 935 34Domestic sparfloxacin 1754 0032 1860 943 17

1485VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

that exit tend to be smaller (their average shareis 1 percent as opposed to 34 percent for thoseSKUs that are present during the entire sampleperiod) the shares of the two groups do notseem orders of magnitude apart13 In additionour data cover only the 300 largest firms sellingin the Indian market so that firms with verysmall shares are not included in our sample

II The Analytic Framework and EstimationApproach

Patent enforcement in the Indian pharmaceu-tical market will have the effect of eliminatingdomestic products whose active pharmaceuticalingredients are protected by (foreign) patentsThus assessing the effects of patent enforce-ment is tantamount to assessing the effects ofwithdrawing domestic products from the mar-ket This task is the converse of evaluating newproduct introduction accordingly the concep-tual framework we use to address the questionsof interest is similar to the one developed in theliterature for the valuation of new goods14

We start by estimating demand for quino-lones Given that the market is characterized byimperfect competition the counterfactual anal-ysis requires that we also model the supply sideas removal of one product will affect the pricesof other products especially those that are closecompetitors The existence of price regulationin the Indian pharmaceutical market imposespotential constraints on firmsrsquo maximizationproblem Given these constraints and the com-

plexity of the price regulation process the typ-ical approach of deriving estimates of actualmarginal costs and markups by exploiting thefirst-order conditions of profit maximizing firmsdoes not seem particularly promising Insteadwe use our demand estimates to place upper andlower bounds on marginal costs and markups

With demand elasticities and upper and lowerbounds for marginal costs in place we thenconduct counterfactual simulations We con-sider several alternative scenarios dependingon the number of domestic products affected bypatent enforcement For each scenario we com-pute the counterfactual prices and use them toassess the effects of domestic product with-drawal on consumer welfare (as measured bythe compensating variation) firm profits andsocial welfare As with the valuation of newproducts the big conceptual problem facing thispart of the analysis is that we need to extrapo-late from the region of the data to the point atwhich demand for the products that exit themarket becomes zero This conceptual issue ispresent in any attempt to evaluate a major pol-icy change for which no historical precedentexists like the enforcement of patent laws inIndia One advantage of the present study is thatwe have a limited set of price data for Pakistana country with similar demographics to Indiabut with a market structure that resembles theone that would emerge in India under patentenforcement (monopoly of multinational sub-sidiaries) By comparing the prices of productsoffered by multinationals in Pakistan to thosewe compute in our counterfactual simulationsfor the products that would be offered by mul-tinationals in India if patent laws were enforcedwe can get a sense of how plausible our coun-terfactual estimates are

A Demand

The demand modelling is based on the multi-stage budgeting approach Our primary motiva-tion for adopting this approach is a practical oneIn the multistage budgeting approach the depen-dent variable is defined as a revenue share whichis appealing given that the products we include inthe analysis contain different molecules (ie ac-tive pharmaceutical ingredients or APIs) Eventhough we do have data on the quantity of therelevant API (eg 100 milligrams of ciprofloxa-cin) contained in each product converting the

13 While we cannot completely rule out the possibilitythat some of the exit is due to sampling variation note thatthe latter should be reflected in low precision of the demandparameter estimates The demand parameters however areprecisely estimated

14 See Manuel Trajtenberg (1989) Jerry A Hausman(1994) and Timothy F Bresnahan (1997) for representativeexamples and a discussion of the relevant issues

TABLE 5mdashREVENUE SHARES OF THE EXITING SKUS

Revenue share ofexiting SKUs

Revenue shares ofall SKUs

Full sample 10 34Northern region 08 33Eastern region 12 36Western region 10 34Southern region 12 33

1486 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

revenue shares to physical shares is extremelydifficult if not infeasible in the case of antibioticsBecause such drugs are ldquosystemicrdquo by nature theyare used to treat a large number of infections andthe dosage of each drug depends on the particularinfection it is supposed to address (for example thedosage will differ depending on whether the anti-biotic is used to combat an ear infection or tuber-culosis) This particular feature of antibioticscomplicates the conversion of revenue to physicalshares

The basic idea of the multistage budgetingapproach is to use the therapeutic classificationof a productmdashie the therapeutic segment andsubsegment the product belongs tomdashto orga-nize all products in the systemic antibacterialssegment into a hierarchical taxonomy consist-ing of two levels At the higher level are thevarious subsegments of systemic antibacterialsThe first stage of budgeting corresponds to theallocation of expenditures across the subseg-ments in this upper level of the taxonomy

In the second stage of the budgeting processcorresponding to the lower level of the taxon-omy a flexible functional form is adopted tomodel how the expenditures allocated to eachsubsegment are distributed across the productswithin that subsegment In particular to modeldemand at the second stage we employ theAIDS specification proposed by Deaton andMuellbauer (1980b)15

While the two-stage demand estimation ap-proach offers functional form flexibility its ap-plication to the Indian systemic antibioticsmarket poses a few problems The first is thatdue to entry and exit many SKUs and evenproducts in our sample are not present in everyperiod AIDS does not have a good way ofdealing with a varying number of products as itwas developed with broad commodity catego-ries in mind which are consumed by all con-sumers every period To solve this problem weaggregate within each subsegment (eg quino-lones) SKUs into product groups where withineach product group all SKUs contain the samemolecule and are produced by firms with thesame domesticforeign status Specifically let aSKU k be indexed by its molecule (or API) M its

domesticforeign status DF indicating whether itis produced by a domestic (Indian) or a subsidiaryof a foreign (multinational) firm a particular pre-sentation s and the particular firm f that producesit We aggregate SKUs over presentations andfirms to obtain a newly defined product group iwhich is indexed only by molecule M and domes-ticforeign status DF and has revenue Ri yenfsRk with i (M DF) k (M DF f s) and pricepi yenfs kpk where k denotes the conditional(on M and DF) revenue share of this particularproduct ie

(1) k Rk

Ri

In most cases the resulting product groups arebroad enough to be present every period16

The usual concern with this aggregation pro-cedure is that it may lead us to overstate firmsrsquomarket power as we ignore competition amongfirms with the same domesticforeign statusproducing the same molecule In the presentapplication however this concern is unlikely tobe of great importance as the effect of patentenforcement is to wipe out all domestic compe-tition at once while granting foreign firms mo-nopoly power hence competition among firmsfor patented molecules becomes irrelevant Theaggregation according to the domesticforeignstatus (within a particular molecule) thus corre-sponds to the scope of our analysis and theparticular questions of interest

The second problem is that for our approachto be useful in welfare analysis the allocation oftotal expenditures to group expenditures at thehigher stage has to be modelled in a way con-sistent with utility maximization In general thesolution of this allocation problem requiresknowledge of all individual product prices

15 Representative applications of the multistage budget-ing approach include Sara F Ellison et al (1997) Hausman(1994) and Hausman and Gregory K Leonard (2002)

16 We are missing only four observations (ie monthregion combinations) all for the drug group of ForeignNorfloxacin August 1999 in the south May 2000 in thewest October 2000 in the south and November 2000 in theeast In these cases we set the revenue shares of ForeignNorfloxacin equal to zero In general with 01 percent ofquinolone sales (see Table 1 row 2) Foreign Norfloxacinhas a very small share of the market This probably explainswhy the results pertaining to this drug group are unreliableit is the only drug group for which we do not obtain asignificant price elasticity of demand while its cross-priceelasticities with other foreign drug groups often have thewrong sign

1487VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

From an empirical point of view this is notparticularly useful as it eliminates all compu-tational advantages of the two-stage approachTo address this problem we adopt an approxi-mate solution to model the higher-level expen-diture allocation along the lines suggested byDeaton and Muellbauer (1980b pp 131ndash32)This gives rise to a two-level AIDS specification

Consider the lower-level estimation firstwhich refers to the allocation of a particularsubsegmentrsquos expenditure to the product groupswithin the subsegment In our application therelevant subsegment is quinolones which weindex with Q Let the product groups within thissubsegment be indexed by i 1 N pi be theprice of product group i (where as noted abovei refers to a particular molecule and domesticforeign status combination) and XQ be the totalexpenditure on the quinolone segment The rev-enue share of each product group is given by

(2) i i j

ijln pj ilnXQ

PQ

where i the revenue share of product group iis defined as

(3) i pi qi

yenj

pj qj

xi

XQ with i j Q

XQ is the overall expenditure on the quinolonesubsegment and PQ is a price index given by

(4) ln PQ ap 0 i

iln pi

1

2 i

j

ijln piln pj

With a limited number of product groups and asufficiently large number of time-series observa-tions the flexibility implied by the AIDS modeldoes not impose too many demands on the data Inthe present application however where the num-ber of observations is limited the AIDS model isnot estimable in this general form To reduce thenumber of parameters that need to be estimatedwe impose two sets of restrictions

The first set of restrictions is implied by thetheory of utility maximization Specificallythese restrictions are

Adding-up yenk k 1 yenk k 0 yenk kj 0 j

Homogeneity yenk jk 0 j Symmetry ij 1frasl2 [ij ji] ji This last

restriction by itself reduces the number of parameters to [N(N 1)]2

The second set of restrictions we impose aimsat further reducing the number of parameters tobe estimated by exploiting our knowledge of thisparticular market Specifically for each productgroup i we allow one ij parameter for all productgroups j that have different molecules from prod-uct group i and are produced by foreign firms andone ij for all product groups j with differentmolecules produced by domestic firms We donrsquotimpose any restrictions on the ij parameter whenproduct group j has the same molecule as productgroup i (By construction product groups i and jcontain products produced by firms with differentdomesticforeign status)

To better illustrate the nature of the restric-tions we impose on the patterns of substitutionacross products some additional notation isneeded Let d(i j) be an indicator of the degreeof similarity (or difference) between productgroup i and product group j along the dimen-sions we are able to observe (molecule M anddomesticforeign status DF) For any two prod-uct groups i and j d(i j) can take on one of thefollowing three values17

(5)

di j (1 0) if Mi Mj DFi DFj

(0 1) if Mi Mj DFi DFj

(0 0) if Mi Mj DFi DFj

Let

(6) Diab j di j a b

17 The sequence (1 1) is not possible for two differentproducts in this case the parameter corresponds to theproductrsquos own-price effect that is ii

1488 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The equation at the lower level becomes

(7) i i iiln pi i10ln pj j Di10

j Di

01

i01ln pj

j Di

00

i00ln pj ilnXQ

PQ

Note that

The parameter ii captures a product grouprsquosown-price effect (there will be as many iiparameters as the number of product groups)

The parameter i10 captures the cross-priceeffect of the product group containing prod-ucts with the same molecule but produced byfirms of different nationality

The parameter i01 captures the cross-priceeffects of product groups containing productswith different molecules but produced byfirms with the same nationality

The parameter i00 captures the cross-priceeffects of product groups containing productswith different molecules produced by firms ofdifferent nationality

Before we take the demand equation to thedata we make two modifications The first oneis to let the product-specific effects i vary byregion r The resulting product-specific regionaleffects ir have two interpretations first theycontrol for the ldquoqualityrdquo of each drug withquality differences being allowed to vary acrossregions second they proxy for demographicsand other demand shifters which vary by re-gion and may affect the demand of each prod-uct group differently18 Note that by includingproduct-specific regional effects in the demandspecification we estimate the price parametersbased on the within-product-group variation ofprices in each region In an earlier version of thepaper we also estimated the demand systemwithout regional dummies and obtained similarresults

The second modification that allows us to gofrom a deterministic to a stochastic specificationof the demand equation is to include an additiveerror term in (7) The latter takes into accountthe fact that (7) is not expected to fit the dataexactly The error term irt accounts for mea-surement error (due to the fact that the productgroup prices pjrt we employ in the estimationare not exact price indices but approximationsthereof) and (potentially region-specific) de-mand shocks that may affect the demand for aproduct in a particular period Examples of suchshocks include an advertising campaign for aparticular product which temporarily increasesthe demand for this product or the outbreak ofa (potentially region-specific) epidemic whichcalls for the use of a particular drug We discussthe interpretation and properties of this errorterm in more detail in the next section

The final form of the equation we estimate atthe lower level becomes (with subscript t de-noting month and subscript r denoting region)

(8) irt i ir iiln pirt i10ln pjrt j Di10

j Di

01

i01ln pjrt

j Di

00

i00ln pjrt

ilnXQrt

PQrt irt

The analysis so far has conditioned on the ex-penditure allocated to the quinolone subseg-ment XQ The upper level of the estimationconsiders the problem of allocating total expen-diture across the different systemic antibioticssubsegments one of which is quinolones Theupper-level demand function is given by

(9) G G H

GHln PH GlnX

P

where all variables denoted by capital letters aredefined as before but now refer to subsegments(G H ) rather than product groups within asubsegment and the total expenditures on sys-temic antibiotics X are deflated by the Stone priceindex log P yenH Hlog PH When estimating thesystem above we impose all the restrictions

18 Given the short time span of our sample typicaldemand shifters such as age distribution income distribu-tion and education hardly change over our sample periodSuch shifters are therefore absorbed by the region-specificproduct fixed effects ir

1489VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

implied by utility maximization as we do withthe estimation of the lower-level AIDS We donot however impose any additional restrictionson the substitution patterns at this stage so thatthe cross-price effects across segments remainrelatively unconstrained

Estimation of the higher-level AIDS allowsus to obtain the unconditional own- and cross-price elasticities that are used in the formulationof the supply problem and welfare analysisThese are given by the formula

(10) ij ijXQXQ

ln qi

ln XQ

ln XQ

ln PQ

ln PQ

ln pj

with the conditional cross price elasticitiesgiven by

ijXQXQ ij ij jlnXQ

PQ

i

As in the lower stage we include subsegment-specific regional dummies in the specificationof the upper-stage demand system so that thefinal form of the estimating equation at theupper stage becomes

(11) Grt G Gr H

GHln PHrt

GlnXrt

Prt Grt

In sum the demand system we take to thedata is represented by equations (8) and (11)and the associated parameter restrictions im-plied by economic theory

B Modelling the Supply Side of the Market

Counterfactual simulations concerning theeffects of domestic product withdrawal requireknowledge of the marginal costs of pharmaceu-tical firms operating in the Indian market Theseare unobservable The usual approach in theNew Empirical Industrial Organization litera-ture has been to exploit the firm equilibriumconditions to infer marginal cost For exampleit is often assumed that the marginal cost ci isconstant and that the industry is an oligopoly

engaging in Bertrand competition with differen-tiated products Assuming that firms myopicallymaximize profits each period one can then de-rive the firmsrsquo first-order conditions that corre-spond to the assumptions above about costsmarket structure and firm behavior

We deviate from this procedure as the presenceof price regulation renders the assumption of un-constrained period-by-period maximization un-tenable Ideally one would like to explicitlyincorporate the price and other administrative con-trols into the firmrsquos optimization problem andderive the first-order conditions under the assump-tion of constrained maximization The complexityof the price regulation however makes this ap-proach infeasible Therefore we adopt an alterna-tive approach that does not rely on modelling theprice-setting process but is instead based on de-riving upper and lower bounds for marginal costsand markups

In particular an upper bound for marginalcosts and a lower bound for the markups (zero)can be derived under the assumption of perfectcompetition While this assumption is clearlyunrealistic in the pharmaceuticals market it isuseful in providing an upper bound for costs ci

Uwhich will be given by

(12) ciU pi

On the other hand a lower bound for marginalcosts ci

L (and upper bound for markups) can bederived by assuming that there is perfect collu-sion within each product group i (where i refershere to a moleculedomestic-foreign combina-tion)19 and ignoring price controls so thatprices are determined by the first-order condi-tion of the jointly profit-maximizing firmswithin product group i Solving this first-ordercondition for the (lower bound of) marginal costthen gives

(13) ciL pi 1

1

iipipj

19 The reason that this assumption will lead to an under-statement of marginal costs (and hence overstatement ofmarket power) is that it assumes away competition amongfirms within each drug group

1490 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

where ii(pipj) the own-price elasticity of de-

mand for product i will depend on the productrsquosown-price pi and all other productsrsquo prices pjwith i j

Once we have obtained the demand elastic-ities through estimation of the demand sys-tem we can calculate the upper and lowerbounds for marginal costs and correspondingmarkups according to (12) or (13) These willthen be employed in the counterfactual sim-ulations While our counterfactuals use boththe lower and upper bounds for costs most ofour discussion will be based on using thelower bounds for costs since this is the moreinteresting case in the policy analysis it givesus the largest possible profits for pharmaceuticalfirms and hence corresponds to the worst pos-sible scenario facing Indian firms and the bestpossible scenario facing multinationals underTRIPS

C Identification Assumptions and EstimationApproach

The discussion of the demand system has sofar abstracted from the issue of price endogene-ity The usual premise in the industrial organi-zation literature is that correlation of prices withthe error term in the demand equation arises byvirtue of the first-order conditions of profit-maximizing firms20 As we discuss in this sec-tion this source of simultaneity bias is unlikelyto be of major concern in the present contextbecause of the existence of price regulation Ourprimary concern is instead with the simultane-ity bias that is implied by the particular wayprices are constructed

To understand the sources of the simultaneitybias note first that the price of product group piwe employ in the demand estimation should bethought of as a proxy for an exact price indexthat we do not observe As such pi contains bydefinition measurement error and it will becorrelated with the error term of the demandequation Specifically the demand equation weare interested in estimating can be written insimplified form (suppressing the subscripts rand t and ignoring parameter restrictions forconvenience) as

i i j

ijln pjT ilnXQ

PQ i

where pjT denotes the true (exact) price index for

product group i and i captures unobservedvariables that may affect demand in a particularperiod (eg an advertising campaign or an ep-idemic) The price index pj

T depends on theSKU prices pk Note that in the presence ofSKU entry and exit into the sample the priceindex pj

T will vary over time even if the pricesof the individual SKUs remain stable as theproducts (SKUs) that comprise each productgroup change each period

Ideally we would like to compute the exactprice index pj

T for each group and employ theseindices in the demand estimation Under theassumption of predetermined SKU prices eachperiod the exact price indices would be uncor-related with the error term and the demandparameters could be estimated by simple OLSUnfortunately this strategy is not feasible in thecurrent context

To derive an exact price index for each prod-uct group it is necessary to model consumerrsquoschoice among SKUs conditional on the choiceof the group To this end it is necessary toexplicitly introduce a third stage in the demandspecification and estimate the parameters asso-ciated with the choice at that stage Howeverany specification based on three-stage budget-ing (that would require directly specifying ademand function associated with the SKUchoice at the third stage and imposing additiveseparability at the higher stages) would be in-feasible to estimate for three main reasonsFirst there is a large number of SKUs withinsome groups so that the demand parameters can-not be estimated even if one imposes parameterrestrictions second the SKU prices exhibit littlevariation over time lastly the choice sets at thislowest stage vary over time rendering estimationof the price parameters associated with SKUs thatare not available in all periods infeasible Alterna-tively one could abandon the multistage budget-ing altogether and adopt a discrete choiceapproach which is inherently better suited to deal-ing with varying choice sets and limited pricevariation A discrete choice approach would beparticularly appealing for modeling the SKUchoice conditional on product group choice giventhat within each group all SKUs contain the same

20 See Steven Berry et al (1995) or Aviv Nevo (2001)for a related discussion

1491VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

molecule and are hence comparable As notedearlier however a discrete choice approach at thehigher stages of the demand estimation requiresconverting revenue to physical shares which ischallenging in the case of systemic drugs such asantibiotics for which the appropriate dosage is notwell defined Finally a combination of a multi-stage budgeting approach for the higher stages anda discrete choice approach for the lowest stagewould be feasible to estimate but inconsistentwith a model of utility maximization For thesereasons we have no alternative but to adopt anapproximation for measuring the price index foreach product group

Let us denote this approximation by pjA and let

13j denote the proportional ldquoapproximationrdquo errorassociated with measuring the true price index forgroup j so that ln pj

A ln pjT ln 13j The

estimating demand equation can then be written as

i i j

ijln pjA ilnXQ

PQ

j

ijln 13j i

or

i i j

ijln pjA ilnXQ

PQ i

where the new error term i yenj ijln 13j iis comprised of two components the first onereflects ldquomeasurementrdquo error due to the fact thatthe product group prices we employ are approx-imations and not exact price indices and thesecond one captures conventional demandshocks The demand equation as written abovecorresponds to the equation we take to the data

The particular proxy we use for the exact priceindex for each product group j is the revenue-share-weighted average of the prices of multipleSKUs available within this group that is

pjA

k j

k pk

The notation above illustrates the two source ofprice variation variation in the SKU prices pkand variation in the weights k

We are concerned with potential correlation ofthe so-constructed product group prices with bothcomponents of the product group demand error

term The correlation between pjA and the first

component yenj ijln 13j is independent of theparticular way we construct pj

A and inherent in thefact that we measure the exact price index witherror The potential correlation between pj

A and thesecond component of the product group demanderror the shock i arises however because of thespecific way we construct the proxy pj

A and inparticular because of the presence of the reve-nue share weights k in pj

A since the SKUrevenue share weights k will generally dependon the product group expenditure they arelikely to be correlated with the product groupdemand shock i

To address the simultaneity bias we use in-strumental variables To this end we need vari-ables that are correlated with the proxies pj

A butuncorrelated with the error term i We use thenumber of SKUs within each product group j asan instrument as it does justice to the idea thatvariation in the product group price index stemsin part from variation in the set of SKUs that areavailable each period It is clearly correlatedwith the average group prices pj

A and plausiblyuncorrelated with the demand shock i Theassumption that the number of SKUs is uncor-related with the conventional demand error willhowever be violated if the introduction of anew SKU changes the perceived quality of adrug21 or if it is accompanied by promotionalactivities which will be reflected in i Our hopeis that this is not too often the case Assumingthat the SKU number for each group j is alsouncorrelated with the weighted sum of the mea-surement errors of all product groups yenj ijln 13j itcan be used as an instrument for the product groupprice pj

A22 If one accepts the premise that prices atthe SKU level are exogenous then SKU pricescan also be used as instruments The final list ofinstruments we use hence includes the number ofSKUs in each group the prices of the five largestSKUs for each group and all other exogenousvariables in the demand estimation such as re-gionalproduct dummy interactions and the upper-

21 Note that we already proxy for the quality of each prod-uct group through the product fixed effects i In the actualestimation we actually let these fixed effects be region-specific(ir) so that we allow quality to vary across regions

22 If the measurement error increases in the number ofincluded SKUs this assumption will be violated Unfortu-nately we have no way of checking the validity of thisassumption

1492 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

level total expenditure on antibiotics Regressionsof group prices on the instruments above yieldhigh R-squares with most regressors highly sig-nificant indicating that our instruments are highlycorrelated with prices23

Our sample includes four molecules cipro-floxacin norfloxacin ofloxacin and spar-floxacin Except for sparfloxacin all othermolecules are produced by both foreign anddomestic firms24 So we have seven products(domestic ciprofloxacin foreign ciprofloxa-cin etc) with 96 observations (two years ofmonthly data four geographical regions foreach period) for each product The parametersin the lower-level AIDS demand system asdefined in equation (8) are the product fixedeffects i the productndashregional dummy inter-actions ir the own revenue-share price elas-ticities ii the cross revenue-share priceelasticities i10 i01 i00 and the revenue-share expenditure elasticities i In estimatingthe parameters we first regress prices on allinstrumental variables and then plug the pre-dicted values for prices in the constrainedleast-square regression (For a detailed expla-nation of the constraints see the previoussection)

Given that we impose many cross-equationconstraints and employ instrumental variablesin the estimation it is difficult to derive stan-dard errors for the parameter estimates ana-lytically Our error term interpretation in (8)implies that the error terms for each productgroup are likely to be correlated across re-gions for example if a national advertisingcampaign increases the demand for a rela-tively expensive presentation in one regionsimultaneously increasing the aggregate de-mand and price index for the correspondingproduct group in that region it is likely that thesame effect will be observed in other regions Inprinciple we could exploit the cross-regionalcorrelation in the error terms to estimate thedemand system using seemingly unrelated re-

gression (SUR) This however would requireus to derive the variance-covariance matrix ofthe parameters analytically which as notedabove is cumbersome in the current contextInstead we use the bootstrap method choosingto remain agnostic about the structure of thevariance-covariance matrix The potential dis-advantage of this approach over SUR is a loss inefficiency but as will become apparent in theresults section the model parameters are fairlyprecisely estimated To maintain the marketstructure we randomly sample the periods (withreplacement) and use the same periods for allproducts Regarding the optimal number ofbootstrap repetitions ideally one would followthe three-step method proposed by Donald W KAndrews et al (2000) Empirical evidence sug-gests however that one rarely needs more than200 replications to estimate the standard er-rors25 To be safe we generate 300 bootstrapsamples (with replacement) based on the origi-nal data and estimate the standard errors usingthe standard errors of the bootstrap sampleestimates

The estimation of the top-level AIDS systemis similar The constraints imposed on this top-level demand system are adding-up homogene-ity and symmetry as defined on page 12Again bootstrapping is used to obtain the stan-dard errors of the parameter estimates

D The Counterfactual Scenarios

In assessing the effects of patent enforce-ment we start by focusing on the most extremecase in which compulsory licensing is not anoption and foreign firms are not subject to pricecontrols We use the results from the analysis ofthis case as a benchmark In reality the outcomeof the WTO negotiations is more likely to in-volve some constraints on the monopoly powerof foreign firms selling patented products indeveloping countries such as price caps orcompulsory licensing Our framework can eas-ily accommodate these cases as will becomeapparent in the next subsection

We now focus on the effects of potentialpatent enforcement in the quinolone segmentWe consider several scenarios which vary inthe number and size of domestic products that

23 Specifically the R-squares from the first-stage regres-sions of product group prices on instruments range from057 to 095 except for domestic ciprofloxacin for whichthe R-square is 017

24 Sparfloxacin is actually offered by one foreign subsid-iary in India (see Table 2) However its revenue share isminiscule We therefore treat Sparfloxacin as being pro-duced by domestic firms only 25 See Bradley Efron and Robert J Tibshirani (1993)

1493VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

will be removed from the market In particularwe consider the following five scenarios26

Withdrawal of one large domestic productgroup only domestic ciprofloxacin

Withdrawal of one relatively small domesticproduct group only domestic ofloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand norfloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand sparfloxacin

Withdrawal of all four domestic quinoloneproduct groups

As the list above suggests we proceed fromanalyzing the effects of single-product with-drawal to the analysis of eliminating the en-tire domestic segment This approach wasmotivated by early empirical results that in-dicated that the existence and extent of com-petition from domestic firms has a significantbearing on the predicted effects of patent en-forcement that is our predictions regardingprices and welfare vary substantially depend-ing on how many domestic products are af-fected by patent enforcement and on whethersome domestic competition will remain presentafter TRIPS In addition the size of the affecteddomestic groups is relevant for the welfare pre-dictions accordingly we examine both scenar-ios in which a large domestic product groupsuch as domestic ciprofloxacin is eliminatedfrom the market and scenarios in which the

eliminated domestic group is relatively small(eg domestic ofloxacin) In all scenarios wemaintain the assumption that the set of prod-ucts offered by the remaining firms in themarket does not change in response to patentenforcement

E Computation of Virtual Prices and NewEquilibrium Prices

The first step in the counterfactual analysisis to derive the new equilibrium prices underpatent enforcement27 In this context thereare two sets of prices that are relevant Thefirst set consists of the virtual prices of those(domestic) products that will not be availableonce TRIPS is put in effect To calculate thesevirtual prices we set the expenditures of therelevant products equal to zero The secondset of prices consists of the prices of thoseproducts that remain in the market In deriv-ing these prices we start by assuming profitmaximization without any form of price reg-ulation the firms remaining in the marketreoptimize in response to the policy changeand set new prices taking the prices of allother firms as given28 Of course at the equi-librium all prices change in response to thefact that some domestic products are nolonger present The new equilibrium pricesfor products that remain in the market are thuscomputed by utilizing the first-order condi-tions of profit-maximizing firms into whichthe virtual prices of the eliminated productsare substituted Hence to compute the newequilibrium prices we solve an equation sys-tem of the following form

26 As Table 2 indicates most of the patents for the drugsin the quinolone segment have expired by now so that noneof the scenarios described in this section is going to mate-rialize in practice The reason we consider these alternativescenarios in our counterfactual simulations is to get a senseof how the presence and extent of domestic competitionaffects welfare calculations Such calculations may becomerelevant in the future in other therapeutic classes in whichpatents have not expired yet and where domestic firmscompete with foreign patent holders While the particularexit scenarios will depend on the exact year of patent expiryof drugs in each therapeutic class and the specific way inwhich patents will be enforced (eg whether certain do-mestic drugs will be grandfathered in) the general conclu-sion that emerges from our calculations is that the consumerwelfare loss is substantially smaller in cases where somedomestic competition remains present in the market Hencethe particular way in which patent enforcement is imple-mented is essential for assessing its welfare impact

27 Of course until product patents are in fact introducedthese prices will not be observable Note also that we areassuming here that the range of available products will notchange with the introduction of patents

28 As mentioned above this first set of calculations ab-stracts from the existence of remaining price controls orother government regulations that would impose constraintson the firmsrsquo profit maximization problem Accordingly theresulting numbers should be interpreted as a benchmark ofwhat would happen if markets were completely unregulatedTo examine what the welfare effects of TRIPS would be inthe more realistic scenario of price regulation we subse-quently consider a scenario in which regulation keeps theprices of the products provided by patent-holders at theirpre-TRIPS level For a more detailed discussion see theresults section

1494 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

For products i that are withdrawn from themarket

(14) 0 i ir iiln pirV i10ln pjr j Di

10

j Di

01

i01ln pjr

j Di

00

i00ln pjr

ilnXQr

PQr

For products k that remain in the market

(15) pkr ckr 1 1

kk pkrpjr pirV 1

In the equations above pirV denotes the virtual

prices of the products that are removed fromthe market while pjr denotes the updatedprices of all other products Note that whensolving for the virtual prices we account forthe fact that both the price index for quino-lones PQr and the expenditure allocated tothis subsegment XQr need to be updated toreflect the fact that as a result of the pricechanges there may be substitution away fromthis subsegment To obtain the new quinoloneexpenditure XQr and the new price index PQrwe use the estimates and formulas for thehigher-level AIDS system In equation (15)ck refers to the marginal cost for product k thatwe have obtained from the previous estimationstage As mentioned in the previous subsectionwe conducted the simulations using both theupper and lower bounds for marginal cost ck

u

and ckL respectively The term kk(pkr

pjrpirV) re-

fers to the unconditional own-price elasticity forproduct k which is a function of the eliminatedproductsrsquo virtual prices and the remaining prod-uctsrsquo new equilibrium prices We conduct thecounterfactual simulations and welfare analysisat the regional level because the presence ofregion-specific product effects in the demandestimation implies region-specific demand elas-ticities and hence region-specific marginalcosts and markups The presentation and discus-sion of our results focus on national averages ofthe relevant variables that we construct by com-

puting weighted averages across regions usingthe population of each region as a weight

F Welfare Assessment

The simulation of the new equilibrium underpatent protection can provide important insightsinto how consumers and firms will respond tothe removal of domestic products in the market(for example toward which products consumerswill substitute or which prices will increase themost) To get a more precise idea of how peo-plersquos well-being will ultimately be affected byTRIPS we compute as a last step in our anal-ysis the welfare effects of the policy changeSocial welfare is defined as consumer welfareand the sum of domestic firm profits Thechange in domestic profits can easily be calcu-lated by comparing the domestic firm (variable)profits at the pre-TRIPS prices to the profitsthese firms will realize at the new simulatedprices Although foreign firm profits do notcount in domestic welfare calculations we alsocompute the effects of patent enforcement onforeign firm profits to get an idea of how largethe expected benefits of TRIPS are for thesefirms This provides in some sense an indirectway of assessing whether the claims that patentenforcement in countries like India will lead tomore research on developing-country-specificdiseases (such as malaria) have any validity iffor example we find that the effect of patentenforcement on the foreign firm profits realizedin India is small in magnitude it is unlikely thatforeign firms will engage in more developing-country-specific research in response to TRIPSIt is important to note that in all these calcula-tions we work with the lower-bound estimatesfor marginal costs since these give rise to thehighest possible markups Hence our estimateof profit loss for domestic producers most likelyoverstates this loss

On the consumer side we measure changes inconsumer welfare by the compensating variation(CV) defined as the additional expenditure thatconsumers need in order to achieve the sameutility level as before patent enforcement at thenew prices Specifically let P0 denote the pricevector before patent enforcement P the simulatedprice vector post-TRIPS (that we obtained usingthe methods described in the previous subsection)u0 the utility attained by consumers before TRIPS

1495VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

and E(u P) the higher-level expenditure functionThen the compensating variation is given by

(16) CV Eu0 P Eu0 P0

Note that the CV as computed in (16) representsthe combination of three effects

The pure product-variety effect that is theeffect that arises because one or more prod-ucts are not available to consumers anymoreholding the prices of all other remainingproducts and the total expenditure on thequinolone subsegment XQ constant

The expenditure-switching effect that is theeffect arising from substitution away fromquinolones and toward other subsegments ofthe antibiotics market again holding the pricesof all other remaining products constant

The reduced-competition effect that is theeffect that arises because the firms remainingin the market adjust (increase) their prices inresponse to the removal of domestic products

From both an analytical and a policy point ofview it is desirable to assess how large each ofthese effects is Accordingly we decompose thetotal effect on consumer welfare (the CV as givenby equation (16)) using the following procedure

To get the pure product variety effect wecompute virtual prices for the products that areremoved from the market holding the quino-lone expenditure XQ and the prices of all otherproducts fixed Let us call the resulting pricevector P1 Then the pure product variety effectis represented by E(u0 P1) E(u0 P0)

To compute the expenditure-switching effectwe compute another set of virtual prices againholding the prices of all remaining products fixedbut letting quinolone expenditure adjust in re-sponse to the new price index for the quinolonesegment (Given that the prices of the remainingproducts remain fixed the change in the priceindex arises only because of the removal of one ormore domestic products) Note that this scenariomost closely resembles the case in which patentlaws are enforced but strict price regulation keepsthe prices of the products offered in the market attheir pre-TRIP level Let us label the so-computedprice vector P2 The expenditure-switching effectis then E(u0 P2) E(u0 P1)

Finally the reduced competition effect aris-ing from higher prices for the remaining prod-

ucts is computed as the residual change in thecompensating variation once the product-varietyand expenditure-switching effects have been ac-counted for that is E(u0 P) E(u0 P2)where the price vector P is computed accordingto the formulas (14) and (15) to reflect theadjustment of prices to the new regime

To compute the standard errors associatedwith the counterfactual simulations (that is thestandard errors for the counterfactual prices andwelfare estimates) we again use bootstrappingWe first bootstrap the original sample (sales andprices of all the drugs) 300 times Next we esti-mate the AIDS model for each of these samplesand compute the counterfactual equilibrium pricesand welfare losses corresponding to each policyscenario for each of the 300 simulations In thefinal step we compute standard errors29

One limitation of our framework is that itdoes not allow for a heterogeneous response ofconsumers to the policy change Accordinglyour framework is not suited to addressing thequestion of how different groups in the popula-tion will be affected by patent law enforce-ment30 Along the same lines our frameworkdoes not accommodate the possibility of pricediscrimination which might lead to differentresults regarding welfare losses and profit gainsrelative to uniform pricing31 Still we believethat the results of the counterfactual simulationscan provide important insights into the likelyaggregate response to patent enforcement andthe factors that drive this response

III Results

A The Structure of Demand

Tables A2 and A3 in the Appendix display theresults from estimation of the lower- and upper-level AIDS systems respectively For ease of in-terpretation rather than discussing the coefficientestimates we focus our discussion on the implied

29 In cases some of the simulations do not converge wecompute the standard errors using only those simulationsthat converged

30 Addressing this question would require at a mini-mum micro data on consumer purchases as in Goldberg(1995) To our knowledge such data do not exist for theIndian pharmaceutical market

31 See Ernst R Berndt (1994) for an extensive discussionof uniform pricing

1496 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

unconditional price and expenditure elasticities re-ported in Table 6 Given that the region-specificproduct group effects ir reported in the last fourcolumns of Tables A2 and A3 imply region-spe-cific demand elasticities we report separate elas-ticities for each region in Tables 6A and 6B Asevident from the comparison of these tables theelasticities are very similar across regions so thatwe can focus the remaining discussion on theelasticities of one region only the northern regiondisplayed in Table 6A

The diagonal terms of Table 6A report theown-price elasticities which are in all but onecase negative and highly significant The oneexception is the foreign norfloxacin productgroupmdashwhose share of quinolone sales is 007percentmdashfor which we estimate a negative butinsignificant own-price elasticity For the remain-ing product groups demand appears to be highlyelastic with the estimated elasticities lower than2 in four out of the six cases The magnitude ofthe own-price elasticities matches the features ofthe Indian pharmaceutical market mentioned ear-lier which would suggest that Indian consumersare likely to be quite price-sensitive32 The elas-ticities appear especially large if one takes intoaccount that they refer to product groups (such asdomestic ciprofloxacin) and not individual drugsoffered by particular firms Their relative magni-tudes are also intuitive the drug with the largestmarket share and a relatively high price (domesticciprofloxacin) appears to be one of the least elas-tic In contrast foreign ciprofloxacin is highlyprice elastic this is plausible as ciprofloxacindrugs offered by subsidiaries of multinationalsface the stiffest competition in this market seg-ment from approximately 75 Indian firms offeringthe same molecule

The estimated expenditure elasticities appear inthe last column These are all positive indicatingthat the demand for all product groups is normalThe remaining cells display the estimated cross-price elasticities As one might perhaps expect forproducts within a therapeutic subsegment theseare mostly positive Out of a total of 49 priceelasticities we estimate there are six that do notconform to expectations these are the cross-price elasticities between different foreign prod-

uct groups which are estimated to be negativeand significant33 Fortunately these elasticitieshave negligible impact on the welfare analysisgiven that our counterfactuals focus on theeffect of withdrawing one or more domesticproducts from the market the most relevantelasticities are the ones that capture the responseof various product group shares to a change in theprice of one or more domestic groups these elas-ticities are the cross-price elasticities between var-ious domestic groups and the ones betweendomestic and foreign groups which are plausibleand precisely estimated34

Regarding these elasticities a striking aspectof our estimates is how large positive andsignificant the cross-price elasticities betweendifferent domestic product groups aremdashin factfor norfloxacin and ofloxacin we estimate thatdomestic product groups containing differentmolecules are closer substitutes for one anotherthan product groups that contain the same mol-ecule but are produced by foreign firms Incontrast for ciprofloxacin (the molecule withthe largest revenue share) we estimate a largepositive cross-price elasticity between the do-mestic and foreign versions

The fact that domestic products appear to beclose substitutes for other domestic productsthat contain different molecules truly representsan ldquoempiricalrdquo finding in the sense that we do

32 In developed economies elasticities of this magnitude havetypically been found only for generic drugs (and even then onlyrarely) or among consumers who lack health insurance

33 While these elasticities are clearly counterintuitive theyare not inconsistent with the underlying demand system whichimposes no restrictions on the sign of the cross-price elastici-ties We do not have a good explanation of why these elastic-ities are estimated to be negative A potential explanation isthat the shares of the foreign products are very small giventhis we observe very few consumers switching from one (verysmall) foreign group to another (very small) foreign groupwhen the price of the first foreign group goes up and hence theinference is not very reliable in this case

34 The cross-price elasticities between foreign drug groupscontaining different molecules will also have an effect on thewelfare estimates given that the withdrawal of domestic prod-ucts will generally lead to changes in the prices of all foreignproducts and market shares will be reallocated from someforeign products to other foreign products based on the newprices However this reallocation from ldquoforeign to foreignrdquo istruly second order in our case compared to the reallocationfrom ldquodomestic to domesticrdquo and ldquodomestic to foreignrdquo prod-ucts In scenarios in which we consider patent enforcementaccompanied by strict price regulation the cross-price elastic-ities between foreign product groups are in fact completelyirrelevant as the prices of foreign products are not allowed toincrease in these cases Still our welfare loss estimates remainsubstantial drivenmdashas beforemdashby the loss of product variety

1497VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

not impose it through any of our assumptionsregarding the demand function The questionthat naturally arises then is what might explainthis finding While we cannot formally addressthis question anecdotal accounts in various in-dustry studies suggest that the explanation maylie in the differences between domestic andforeign firms in the structure and coverage ofretail distribution networks

Distribution networks for pharmaceuticals inIndia are typically organized in a hierarchicalfashion Pharmaceutical companies deal mainlywith carrying and forwarding (CampF) agents inmany instances regionally based who each sup-ply a network of stockists (wholesalers) Thesestockists in turn deal with the retail pharma-cists through whom retail sales ultimately oc-cur35 The market share enjoyed by a particularpharmaceutical product therefore depends inpart on the number of retail pharmacists who

stock the product And it is here that thereappears to be a distinction between domesticfirms and multinational subsidiaries In particu-lar the retail reach of domestic firms as agroup tends to be much more comprehensivethan that of multinational subsidiaries (IndianCredit Rating Agency (ICRA) 1999)36

There appear to be two reasons for this Thefirst is that many of the larger Indian firmsbecause they have a much larger portfolio ofproducts over which to spread the associatedfixed costs typically have more extensive net-works of medical representatives The second issimply that there are many more domestic firms(and products) on the market At the retail levelthis would imply that local pharmacists mightbe more likely to stock domestic products con-taining two different molecules say ciprofloxa-cin and norfloxacin than they would domesticand foreign versions of the same molecule Tothe extent that patients (or their doctors) arewilling to substitute across molecules in order tosave on transport or search costs (eg going toanother pharmacy to check whether a particular

35 There are estimated to be some 300000 retail pharma-cists in India On average stockists deal with about 75 retailers(ICRA 1999) There are naturally variations in this structureand a host of specific exclusive dealing and other arrangementsexists in practice Pharmaceutical firms also maintain networksof medical representatives whose main function is to marketthe companyrsquos products to doctors who do the actual prescrib-ing of drugs In some instances firms do sell directly to thedoctors who then become the ldquoretailerrdquo as far as patients areconcerned but these are relatively rare

36 These differences were also highlighted in conversa-tions one of the authors had with CEOs and managingdirectors of several pharmaceutical firms as part of a sep-arate study

TABLE 6AmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN THE NORTHERN REGION

Product group

Elasticity with respect to

Prices of foreign productgroups Prices of domestic product groups Overall

quinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 557 013dagger 015 401 011dagger 011dagger 016 137(179) (007) (007) (184) (006) (006) (006) (029)

Foreign norfloxacin 427dagger 045 427dagger 350dagger 602 451 465 220(242) (112) (242) (210) (623) (184) (183) (105)

Foreign ofloxacin 011 010dagger 138 009 009dagger 023 011 116(005) (005) (031) (027) (005) (028) (004) (017)

Domestic ciprofloxacin 018 001 001 168 008 008 010 117(008) (000) (001) (023) (002) (002) (002) (003)

Domestic norfloxacin 004 003 004 058 223 042 040 073(001) (003) (001) (017) (011) (004) (003) (009)

Domestic ofloxacin 005 005 011 077 074 342 074 089(002) (002) (013) (028) (008) (025) (008) (021)

Domestic sparfloxacin 007 004 007 115 063 063 288 028(002) (001) (002) (015) (006) (006) (017) (012)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent significance level and dagger (dagger) denotes significance at the 10-percent level

1498 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE 6BmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN OTHER REGIONS

Demand patterns in the eastern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 594 014dagger 016 431 013 011 017 140(195) (008) (008) (201) (006) (007) (007) (031)

Foreign norfloxacin 329dagger 058 329dagger 264dagger 460 346 359 192(180) (083) (179) (153) (451) (138) (139) (082)

Foreign ofloxacin 012 010dagger 140 010 010 024 013 117(006) (006) (034) (031) (005) (030) (005) (019)

Domestic ciprofloxacin 018 001 001 172 009 008 011 117(008) (000) (002) (027) (002) (003) (002) (003)

Domestic norfloxacin 004 004 004 068 242 049 046 070(002) (004) (002) (020) (013) (005) (004) (010)

Domestic ofloxacin 004dagger 004 009 061 060 295 060 092(002) (001) (010) (028) (007) (022) (007) (017)

Domestic sparfloxacin 005 003 005 092 048 051 251 043(002) (001) (001) (017) (005) (006) (016) (010)

Demand patterns in the western region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 427 009dagger 011 286 008 006 011 126(129) (005) (005) (133) (004) (005) (004) (021)

Foreign norfloxacin 714dagger 008 710dagger 594dagger 995 742 770 299dagger

(369) (183) (368) (328) (1015) (297) (296) (178)Foreign ofloxacin 013 012dagger 145 009 012 026 013 119

(007) (006) (037) (029) (005) (033) (005) (021)Domestic ciprofloxacin 019 001 000 174 009 008 011 118

(008) (000) (001) (024) (002) (003) (002) (004)Domestic norfloxacin 004 004 004 067 243 050 047 069

(002) (004) (001) (018) (014) (005) (004) (010)Domestic ofloxacin 003 003 007 048 048 257 048 093

(002) (001) (008) (022) (005) (015) (004) (014)Domestic sparfloxacin 006 003 005 083 046 049 241 046

(002) (001) (001) (014) (004) (005) (012) (009)

Demand patterns in the southern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 560 013dagger 015 402 011 010 017 137(182) (007) (007) (187) (006) (006) (007) (029)

Foreign norfloxacin 533dagger 031 531dagger 431dagger 748 559 584 249dagger

(301) (134) (300) (245) (734) (219) (219) (128)Foreign ofloxacin 014 012dagger 148 011 012 029 016 120

(007) (007) (039) (033) (006) (035) (006) (022)Domestic ciprofloxacin 017 001 000 169 008 007 011 116

(008) (000) (001) (025) (002) (003) (002) (003)Domestic norfloxacin 004 003 003 059 225 043 040 073

(001) (003) (001) (019) (012) (004) (004) (009)Domestic ofloxacin 004dagger 004 009 062 060 296 060 091

(002) (001) (010) (026) (006) (019) (007) (017)Domestic sparfloxacin 008 005 008 138 073 075 323 016

(002) (001) (002) (018) (007) (008) (021) (014)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at 5-percent level anddagger (dagger) denotes significance at the 10-percent level

1499VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

foreign product is in stock) in aggregate datawe would expect to find precisely the substitu-tion patterns that we report in Table 6

Whether the particular explanation we provideabove is the correct one the high degree of sub-stitutability between domestic product groupsturns out to have important implications for thewelfare calculations We discuss these in moredetail below when we present the results of thecounterfactual welfare analysis Another elasticitywith important implications for the counterfactu-als is the price elasticity for the quinolone subseg-ment as a whole which indicates how likelyconsumers are to switch to other antibioticsgroups when faced with a price increase for quin-olones This elasticity is computed on the basis ofthe results in Table A3 and it is at 111 (stan-dard error 024) this is large in magnitudebutmdashas expectedmdashsmaller in absolute value thanthe own-price elasticities of the product groupswithin the quinolone subsegment

The results in Tables 6A and 6B are based onour preferred specification discussed in SectionII In Tables A4 to A6 in the Appendix weexperiment with some alternative specificationsTables A4(a)ndashA4(c) correspond to a specificationthat includes in addition to product-group-specificregional fixed effects product-group-specific (andfor the upper level antibiotics-segment-specific)seasonal effects We distinguish among three sea-sonsmdashthe summer monsoon and wintermdashandreport the unconditional demand elasticities for

the northern region for each of these seasons Asevident from the tables in the Appendix our elas-ticity estimates are robust to the inclusion of sea-sonal effects The demand elasticities in Table A5are based on estimation of the demand system byOLS Compared to the elasticities obtained by IVthe OLS elasticities are smaller in absolute valueimplying that welfare calculations based on theOLS estimates would produce larger welfare lossestimates Nevertheless some of the patterns re-garding the cross-price elasticities discussed ear-lier are also evident in the OLS results inparticular the cross-price elasticities between dif-ferent domestic product groups are all positivelarge and significant and in most instances largerthan the cross-price elasticities between drugs thatcontain the same molecule but are produced byfirms of different domesticforeign status Theclose substitutability of domestic products indi-cated by both the OLS and IV estimates seems tobe one of the most robust findings of the paper

B Cost and Markup Estimates

Table 7 displays the marginal costs markupsand profits implied by the price elasticity esti-mates of Tables 6A and 6B for each of the sevenproduct groups Given that our regional effectsimply different price elasticities for each regionour marginal cost and markup estimates alsodiffer by region Given however that based on

TABLE 7mdashUPPER AND LOWER BOUNDS FOR MARGINAL COST MARKUP AND ANNUAL PROFIT BY PRODUCT GROUPS WITHIN

THE QUINOLONE SUBSEGMENT

Product groupLower boundfor MC (Rs)

Upper boundfor markup

Upper boundfor profit(Rs mill)

Upper boundfor MC (Rs)

Lower boundfor markup

Lower boundfor profit

(Rs)

Foreign ciprofloxacin 83 19 269 103 0 00(123) (012) (1655)

Foreign norfloxacin NA NA NA 53 0 00Foreign ofloxacin 323 70 1061 1085 0 00

(2316) (021) (3185)Domestic ciprofloxacin 47 59 17019 112 0 00

(114) (010) (29858)Domestic norfloxacin 52 43 2807 90 0 00

(020) (002) (1532)Domestic ofloxacin 587 34 1612 901 0 00

(218) (002) (1280)Domestic sparfloxacin 495 37 1985 788 0 00

(157) (002) (1100)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level Estimated lower bound forforeign norfloxacinrsquos marginal cost is negative since the estimated price elasticity is less than one in absolute value

1500 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Tables 6A and 6B the price elasticities do notseem to differ substantially across regions wereport for ease of exposition only the nationalaverages for marginal costs and markups inTable 7

Table 7 has two parts In the left part (firstthree columns) we report the lower bound formarginal cost and the corresponding upperbound for markup and upper bound for totalannual profit for each product group Thesenumbers are based on the price elasticities weobtained from estimating the two-level AIDSdemand system Since we do not have a reliableestimate for the price elasticity of foreign nor-floxacin (the point estimate is negative but lessthan one in absolute value and insignificant)we cannot compute the lower bound for mar-ginal cost in this case It is important to note thatthese estimates do not reflect either the actualmarginal cost or the actual markup for thesedrugs both because the existence of price reg-ulation implies that the unconstrained first-orderconditions are unlikely to hold each period andbecause our aggregation across firms of thesame domesticforeign status supplying thesame molecule makes the interpretation of theseestimates problematic In particular the fact thatwe ignore competition among firms within eachproduct group implies that our estimates willtend to overstate market power These numberswill prove useful however in the counterfac-tual simulations as they can give us a sense ofhow large the maximum profit gains for multi-national firms and the maximum profit lossesfor domestic firms are likely to be under patentenforcement The right part of the table (lastthree columns) reports the upper bound for mar-ginal cost which we obtain by simply taking themarginal cost to be equal to the observed priceWe do not report standard errors in this casesince the numbers are based on actual data Themarkups corresponding to these marginal costupper bounds are of course zero We conductthe counterfactual simulations using both thelower and upper bounds for marginal costs

The (upper bounds for) markups on the leftside of the table are generally plausible Thedomestic ciprofloxacin product group that dom-inates the quinolone subsegment and for whichwe documented high prices a high marketshare and a relatively low elasticity of demandenjoys one of the highest markup upper bounds(60 percent) and accounts for nearly 70 percent

of all profits derived within the subsegmentForeign ciprofloxacin on the other hand whichfaces the stiffest competition from domesticfirms and for which we estimated a highly elas-tic demand has the lowest markup upper bound(19 percent)

C Counterfactual Estimates of the Impact onPrices and Welfare

With estimates of the key demand and costparameters in hand we turn to the counterfac-tuals We consider the five separate scenarioslisted in the previous section All of the scenar-ios involve the withdrawal of one or more of thedomestic product groups from the market Ta-ble 8 displays our estimates of the consumerwelfare losses that result under the differentscenarios The losses are expressed in billion ofrupees per year All numbers presented in Ta-ble 8 and subsequent tables are based on usingthe lower bounds for marginal cost and upperbounds for markup in the simulations As dis-cussed earlier these numbers are the more in-teresting to work with since they give us anupper bound for the changes in the profits ofdomestic and foreign firms that would resultfrom patent enforcement In the Appendix wealso present results based on using the upperbounds for marginal costs in the simulations inwhich case the pre-TRIPS profits of domesticand foreign firms are zero In all cases marginalcosts are assumed to be constant in outputWhile naturally the profit implications differdepending on whether one uses the upper orlower bounds for marginal costs (firm profits arezero if one assumes the upper bound of mar-ginal cost in which case price equals marginalcost) the estimated consumer welfare losses aresimilar in the two cases We discuss these re-sults in more detail at the end of this section

The first column presents our estimates of theconsumer welfare losses attributable to the pureloss of product variety effect where we fix theprices of all remaining products as well as theoverall expenditure on quinolones while with-drawing one or more of the domestic productgroups Note that in our initial specification ofthe demand system had we not allowed for thepossibility that consumers might differentiatebetween domestic and foreign products evenwhen they contain the same molecule this

1501VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

particular component of the loss of consumerwelfare would not have arisen

The estimates reported in the second columnincorporate the expenditure-switching effect ontop of the loss of product variety Here basedupon the price elasticity estimates from thehigher-level AIDS system we adjust (down-ward) the expenditures allocated to the quino-lone subsegment as the composite price ofquinolones effectively increases as a consequenceof the higher virtual prices of the domestic productgroups that are withdrawn from the market Be-cause the estimates in this column are generatedassuming that the prices of the products that re-main in the market are not adjusted upward theyprovide a sense of what consumer losses would beif the introduction of product patents were coupledwith strict price regulation aimed at maintainingprices at prepatent levels Alternatively they canbe thought of as the relevant welfare numbers ifintense competition among firms within the re-maining product groups kept the prices of theproducts that were still offered in the market closeto the firmsrsquo marginal costs

The last column of Table 8 displays the esti-mated consumer welfare losses when bothcross-segment expenditure-switching and with-in-segment upward price adjustments are takeninto account

If we compare the results across the firstsecond and third columns all the counterfac-tual scenarios produce qualitatively similar pat-terns patterns that are consistent with what we

would expect Starting from the initial loss ofwelfare attributable to the loss of product vari-ety the option of switching expenditures outof the quinolone subsegment to other subseg-ments mitigates some of the initial welfareloss But if we then incorporate the upwardprice adjustments that result in response to thereduced competition the welfare losses aremagnified

Of particular interest from a policy perspec-tive are the relative magnitudes of these threeeffects which are similar under all the counter-factual scenarios though absolute levels varyconsiderably First despite the fact that the de-mand for quinolones is quite sensitive to thecomposite price of quinolonesmdashthe upper levelprice elasticity is 111mdashthe cross-subsegmentexpenditure-switching effects are in all thecases small (in absolute value terms) relative tothe other two effects For instance under thescenario where all the domestic quinolone prod-uct groups are withdrawn from the market theoverall consumer welfare loss of Rs 1781 bil-lion per year can be decomposed into an initialloss of Rs 1176 billion (66 percent) attribut-able to the loss of product variety a slightreduction in this initial loss of Rs 041 billion(2 percent) from Rs 1176 billion to Rs1135 billion because of expenditure switchingand a subsequent additional loss of Rs 646billion (36 percent) from Rs 1135 billion toRs 1781 billion because of the reduced com-petition and consequent price increases

TABLE 8mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure loss ofvariety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 732(087) (089) (146)

Only ofloxacin 008 008 023(008) (008) (010)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1253(177) (180) (415)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1058(142) (145) (331)

All four domestic quinolones products 1176dagger 1135dagger 1781(643) (634) (1270)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent significance level and dagger (dagger)denotes significance at the 10-percent level

1502 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 10: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

that exit tend to be smaller (their average shareis 1 percent as opposed to 34 percent for thoseSKUs that are present during the entire sampleperiod) the shares of the two groups do notseem orders of magnitude apart13 In additionour data cover only the 300 largest firms sellingin the Indian market so that firms with verysmall shares are not included in our sample

II The Analytic Framework and EstimationApproach

Patent enforcement in the Indian pharmaceu-tical market will have the effect of eliminatingdomestic products whose active pharmaceuticalingredients are protected by (foreign) patentsThus assessing the effects of patent enforce-ment is tantamount to assessing the effects ofwithdrawing domestic products from the mar-ket This task is the converse of evaluating newproduct introduction accordingly the concep-tual framework we use to address the questionsof interest is similar to the one developed in theliterature for the valuation of new goods14

We start by estimating demand for quino-lones Given that the market is characterized byimperfect competition the counterfactual anal-ysis requires that we also model the supply sideas removal of one product will affect the pricesof other products especially those that are closecompetitors The existence of price regulationin the Indian pharmaceutical market imposespotential constraints on firmsrsquo maximizationproblem Given these constraints and the com-

plexity of the price regulation process the typ-ical approach of deriving estimates of actualmarginal costs and markups by exploiting thefirst-order conditions of profit maximizing firmsdoes not seem particularly promising Insteadwe use our demand estimates to place upper andlower bounds on marginal costs and markups

With demand elasticities and upper and lowerbounds for marginal costs in place we thenconduct counterfactual simulations We con-sider several alternative scenarios dependingon the number of domestic products affected bypatent enforcement For each scenario we com-pute the counterfactual prices and use them toassess the effects of domestic product with-drawal on consumer welfare (as measured bythe compensating variation) firm profits andsocial welfare As with the valuation of newproducts the big conceptual problem facing thispart of the analysis is that we need to extrapo-late from the region of the data to the point atwhich demand for the products that exit themarket becomes zero This conceptual issue ispresent in any attempt to evaluate a major pol-icy change for which no historical precedentexists like the enforcement of patent laws inIndia One advantage of the present study is thatwe have a limited set of price data for Pakistana country with similar demographics to Indiabut with a market structure that resembles theone that would emerge in India under patentenforcement (monopoly of multinational sub-sidiaries) By comparing the prices of productsoffered by multinationals in Pakistan to thosewe compute in our counterfactual simulationsfor the products that would be offered by mul-tinationals in India if patent laws were enforcedwe can get a sense of how plausible our coun-terfactual estimates are

A Demand

The demand modelling is based on the multi-stage budgeting approach Our primary motiva-tion for adopting this approach is a practical oneIn the multistage budgeting approach the depen-dent variable is defined as a revenue share whichis appealing given that the products we include inthe analysis contain different molecules (ie ac-tive pharmaceutical ingredients or APIs) Eventhough we do have data on the quantity of therelevant API (eg 100 milligrams of ciprofloxa-cin) contained in each product converting the

13 While we cannot completely rule out the possibilitythat some of the exit is due to sampling variation note thatthe latter should be reflected in low precision of the demandparameter estimates The demand parameters however areprecisely estimated

14 See Manuel Trajtenberg (1989) Jerry A Hausman(1994) and Timothy F Bresnahan (1997) for representativeexamples and a discussion of the relevant issues

TABLE 5mdashREVENUE SHARES OF THE EXITING SKUS

Revenue share ofexiting SKUs

Revenue shares ofall SKUs

Full sample 10 34Northern region 08 33Eastern region 12 36Western region 10 34Southern region 12 33

1486 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

revenue shares to physical shares is extremelydifficult if not infeasible in the case of antibioticsBecause such drugs are ldquosystemicrdquo by nature theyare used to treat a large number of infections andthe dosage of each drug depends on the particularinfection it is supposed to address (for example thedosage will differ depending on whether the anti-biotic is used to combat an ear infection or tuber-culosis) This particular feature of antibioticscomplicates the conversion of revenue to physicalshares

The basic idea of the multistage budgetingapproach is to use the therapeutic classificationof a productmdashie the therapeutic segment andsubsegment the product belongs tomdashto orga-nize all products in the systemic antibacterialssegment into a hierarchical taxonomy consist-ing of two levels At the higher level are thevarious subsegments of systemic antibacterialsThe first stage of budgeting corresponds to theallocation of expenditures across the subseg-ments in this upper level of the taxonomy

In the second stage of the budgeting processcorresponding to the lower level of the taxon-omy a flexible functional form is adopted tomodel how the expenditures allocated to eachsubsegment are distributed across the productswithin that subsegment In particular to modeldemand at the second stage we employ theAIDS specification proposed by Deaton andMuellbauer (1980b)15

While the two-stage demand estimation ap-proach offers functional form flexibility its ap-plication to the Indian systemic antibioticsmarket poses a few problems The first is thatdue to entry and exit many SKUs and evenproducts in our sample are not present in everyperiod AIDS does not have a good way ofdealing with a varying number of products as itwas developed with broad commodity catego-ries in mind which are consumed by all con-sumers every period To solve this problem weaggregate within each subsegment (eg quino-lones) SKUs into product groups where withineach product group all SKUs contain the samemolecule and are produced by firms with thesame domesticforeign status Specifically let aSKU k be indexed by its molecule (or API) M its

domesticforeign status DF indicating whether itis produced by a domestic (Indian) or a subsidiaryof a foreign (multinational) firm a particular pre-sentation s and the particular firm f that producesit We aggregate SKUs over presentations andfirms to obtain a newly defined product group iwhich is indexed only by molecule M and domes-ticforeign status DF and has revenue Ri yenfsRk with i (M DF) k (M DF f s) and pricepi yenfs kpk where k denotes the conditional(on M and DF) revenue share of this particularproduct ie

(1) k Rk

Ri

In most cases the resulting product groups arebroad enough to be present every period16

The usual concern with this aggregation pro-cedure is that it may lead us to overstate firmsrsquomarket power as we ignore competition amongfirms with the same domesticforeign statusproducing the same molecule In the presentapplication however this concern is unlikely tobe of great importance as the effect of patentenforcement is to wipe out all domestic compe-tition at once while granting foreign firms mo-nopoly power hence competition among firmsfor patented molecules becomes irrelevant Theaggregation according to the domesticforeignstatus (within a particular molecule) thus corre-sponds to the scope of our analysis and theparticular questions of interest

The second problem is that for our approachto be useful in welfare analysis the allocation oftotal expenditures to group expenditures at thehigher stage has to be modelled in a way con-sistent with utility maximization In general thesolution of this allocation problem requiresknowledge of all individual product prices

15 Representative applications of the multistage budget-ing approach include Sara F Ellison et al (1997) Hausman(1994) and Hausman and Gregory K Leonard (2002)

16 We are missing only four observations (ie monthregion combinations) all for the drug group of ForeignNorfloxacin August 1999 in the south May 2000 in thewest October 2000 in the south and November 2000 in theeast In these cases we set the revenue shares of ForeignNorfloxacin equal to zero In general with 01 percent ofquinolone sales (see Table 1 row 2) Foreign Norfloxacinhas a very small share of the market This probably explainswhy the results pertaining to this drug group are unreliableit is the only drug group for which we do not obtain asignificant price elasticity of demand while its cross-priceelasticities with other foreign drug groups often have thewrong sign

1487VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

From an empirical point of view this is notparticularly useful as it eliminates all compu-tational advantages of the two-stage approachTo address this problem we adopt an approxi-mate solution to model the higher-level expen-diture allocation along the lines suggested byDeaton and Muellbauer (1980b pp 131ndash32)This gives rise to a two-level AIDS specification

Consider the lower-level estimation firstwhich refers to the allocation of a particularsubsegmentrsquos expenditure to the product groupswithin the subsegment In our application therelevant subsegment is quinolones which weindex with Q Let the product groups within thissubsegment be indexed by i 1 N pi be theprice of product group i (where as noted abovei refers to a particular molecule and domesticforeign status combination) and XQ be the totalexpenditure on the quinolone segment The rev-enue share of each product group is given by

(2) i i j

ijln pj ilnXQ

PQ

where i the revenue share of product group iis defined as

(3) i pi qi

yenj

pj qj

xi

XQ with i j Q

XQ is the overall expenditure on the quinolonesubsegment and PQ is a price index given by

(4) ln PQ ap 0 i

iln pi

1

2 i

j

ijln piln pj

With a limited number of product groups and asufficiently large number of time-series observa-tions the flexibility implied by the AIDS modeldoes not impose too many demands on the data Inthe present application however where the num-ber of observations is limited the AIDS model isnot estimable in this general form To reduce thenumber of parameters that need to be estimatedwe impose two sets of restrictions

The first set of restrictions is implied by thetheory of utility maximization Specificallythese restrictions are

Adding-up yenk k 1 yenk k 0 yenk kj 0 j

Homogeneity yenk jk 0 j Symmetry ij 1frasl2 [ij ji] ji This last

restriction by itself reduces the number of parameters to [N(N 1)]2

The second set of restrictions we impose aimsat further reducing the number of parameters tobe estimated by exploiting our knowledge of thisparticular market Specifically for each productgroup i we allow one ij parameter for all productgroups j that have different molecules from prod-uct group i and are produced by foreign firms andone ij for all product groups j with differentmolecules produced by domestic firms We donrsquotimpose any restrictions on the ij parameter whenproduct group j has the same molecule as productgroup i (By construction product groups i and jcontain products produced by firms with differentdomesticforeign status)

To better illustrate the nature of the restric-tions we impose on the patterns of substitutionacross products some additional notation isneeded Let d(i j) be an indicator of the degreeof similarity (or difference) between productgroup i and product group j along the dimen-sions we are able to observe (molecule M anddomesticforeign status DF) For any two prod-uct groups i and j d(i j) can take on one of thefollowing three values17

(5)

di j (1 0) if Mi Mj DFi DFj

(0 1) if Mi Mj DFi DFj

(0 0) if Mi Mj DFi DFj

Let

(6) Diab j di j a b

17 The sequence (1 1) is not possible for two differentproducts in this case the parameter corresponds to theproductrsquos own-price effect that is ii

1488 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The equation at the lower level becomes

(7) i i iiln pi i10ln pj j Di10

j Di

01

i01ln pj

j Di

00

i00ln pj ilnXQ

PQ

Note that

The parameter ii captures a product grouprsquosown-price effect (there will be as many iiparameters as the number of product groups)

The parameter i10 captures the cross-priceeffect of the product group containing prod-ucts with the same molecule but produced byfirms of different nationality

The parameter i01 captures the cross-priceeffects of product groups containing productswith different molecules but produced byfirms with the same nationality

The parameter i00 captures the cross-priceeffects of product groups containing productswith different molecules produced by firms ofdifferent nationality

Before we take the demand equation to thedata we make two modifications The first oneis to let the product-specific effects i vary byregion r The resulting product-specific regionaleffects ir have two interpretations first theycontrol for the ldquoqualityrdquo of each drug withquality differences being allowed to vary acrossregions second they proxy for demographicsand other demand shifters which vary by re-gion and may affect the demand of each prod-uct group differently18 Note that by includingproduct-specific regional effects in the demandspecification we estimate the price parametersbased on the within-product-group variation ofprices in each region In an earlier version of thepaper we also estimated the demand systemwithout regional dummies and obtained similarresults

The second modification that allows us to gofrom a deterministic to a stochastic specificationof the demand equation is to include an additiveerror term in (7) The latter takes into accountthe fact that (7) is not expected to fit the dataexactly The error term irt accounts for mea-surement error (due to the fact that the productgroup prices pjrt we employ in the estimationare not exact price indices but approximationsthereof) and (potentially region-specific) de-mand shocks that may affect the demand for aproduct in a particular period Examples of suchshocks include an advertising campaign for aparticular product which temporarily increasesthe demand for this product or the outbreak ofa (potentially region-specific) epidemic whichcalls for the use of a particular drug We discussthe interpretation and properties of this errorterm in more detail in the next section

The final form of the equation we estimate atthe lower level becomes (with subscript t de-noting month and subscript r denoting region)

(8) irt i ir iiln pirt i10ln pjrt j Di10

j Di

01

i01ln pjrt

j Di

00

i00ln pjrt

ilnXQrt

PQrt irt

The analysis so far has conditioned on the ex-penditure allocated to the quinolone subseg-ment XQ The upper level of the estimationconsiders the problem of allocating total expen-diture across the different systemic antibioticssubsegments one of which is quinolones Theupper-level demand function is given by

(9) G G H

GHln PH GlnX

P

where all variables denoted by capital letters aredefined as before but now refer to subsegments(G H ) rather than product groups within asubsegment and the total expenditures on sys-temic antibiotics X are deflated by the Stone priceindex log P yenH Hlog PH When estimating thesystem above we impose all the restrictions

18 Given the short time span of our sample typicaldemand shifters such as age distribution income distribu-tion and education hardly change over our sample periodSuch shifters are therefore absorbed by the region-specificproduct fixed effects ir

1489VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

implied by utility maximization as we do withthe estimation of the lower-level AIDS We donot however impose any additional restrictionson the substitution patterns at this stage so thatthe cross-price effects across segments remainrelatively unconstrained

Estimation of the higher-level AIDS allowsus to obtain the unconditional own- and cross-price elasticities that are used in the formulationof the supply problem and welfare analysisThese are given by the formula

(10) ij ijXQXQ

ln qi

ln XQ

ln XQ

ln PQ

ln PQ

ln pj

with the conditional cross price elasticitiesgiven by

ijXQXQ ij ij jlnXQ

PQ

i

As in the lower stage we include subsegment-specific regional dummies in the specificationof the upper-stage demand system so that thefinal form of the estimating equation at theupper stage becomes

(11) Grt G Gr H

GHln PHrt

GlnXrt

Prt Grt

In sum the demand system we take to thedata is represented by equations (8) and (11)and the associated parameter restrictions im-plied by economic theory

B Modelling the Supply Side of the Market

Counterfactual simulations concerning theeffects of domestic product withdrawal requireknowledge of the marginal costs of pharmaceu-tical firms operating in the Indian market Theseare unobservable The usual approach in theNew Empirical Industrial Organization litera-ture has been to exploit the firm equilibriumconditions to infer marginal cost For exampleit is often assumed that the marginal cost ci isconstant and that the industry is an oligopoly

engaging in Bertrand competition with differen-tiated products Assuming that firms myopicallymaximize profits each period one can then de-rive the firmsrsquo first-order conditions that corre-spond to the assumptions above about costsmarket structure and firm behavior

We deviate from this procedure as the presenceof price regulation renders the assumption of un-constrained period-by-period maximization un-tenable Ideally one would like to explicitlyincorporate the price and other administrative con-trols into the firmrsquos optimization problem andderive the first-order conditions under the assump-tion of constrained maximization The complexityof the price regulation however makes this ap-proach infeasible Therefore we adopt an alterna-tive approach that does not rely on modelling theprice-setting process but is instead based on de-riving upper and lower bounds for marginal costsand markups

In particular an upper bound for marginalcosts and a lower bound for the markups (zero)can be derived under the assumption of perfectcompetition While this assumption is clearlyunrealistic in the pharmaceuticals market it isuseful in providing an upper bound for costs ci

Uwhich will be given by

(12) ciU pi

On the other hand a lower bound for marginalcosts ci

L (and upper bound for markups) can bederived by assuming that there is perfect collu-sion within each product group i (where i refershere to a moleculedomestic-foreign combina-tion)19 and ignoring price controls so thatprices are determined by the first-order condi-tion of the jointly profit-maximizing firmswithin product group i Solving this first-ordercondition for the (lower bound of) marginal costthen gives

(13) ciL pi 1

1

iipipj

19 The reason that this assumption will lead to an under-statement of marginal costs (and hence overstatement ofmarket power) is that it assumes away competition amongfirms within each drug group

1490 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

where ii(pipj) the own-price elasticity of de-

mand for product i will depend on the productrsquosown-price pi and all other productsrsquo prices pjwith i j

Once we have obtained the demand elastic-ities through estimation of the demand sys-tem we can calculate the upper and lowerbounds for marginal costs and correspondingmarkups according to (12) or (13) These willthen be employed in the counterfactual sim-ulations While our counterfactuals use boththe lower and upper bounds for costs most ofour discussion will be based on using thelower bounds for costs since this is the moreinteresting case in the policy analysis it givesus the largest possible profits for pharmaceuticalfirms and hence corresponds to the worst pos-sible scenario facing Indian firms and the bestpossible scenario facing multinationals underTRIPS

C Identification Assumptions and EstimationApproach

The discussion of the demand system has sofar abstracted from the issue of price endogene-ity The usual premise in the industrial organi-zation literature is that correlation of prices withthe error term in the demand equation arises byvirtue of the first-order conditions of profit-maximizing firms20 As we discuss in this sec-tion this source of simultaneity bias is unlikelyto be of major concern in the present contextbecause of the existence of price regulation Ourprimary concern is instead with the simultane-ity bias that is implied by the particular wayprices are constructed

To understand the sources of the simultaneitybias note first that the price of product group piwe employ in the demand estimation should bethought of as a proxy for an exact price indexthat we do not observe As such pi contains bydefinition measurement error and it will becorrelated with the error term of the demandequation Specifically the demand equation weare interested in estimating can be written insimplified form (suppressing the subscripts rand t and ignoring parameter restrictions forconvenience) as

i i j

ijln pjT ilnXQ

PQ i

where pjT denotes the true (exact) price index for

product group i and i captures unobservedvariables that may affect demand in a particularperiod (eg an advertising campaign or an ep-idemic) The price index pj

T depends on theSKU prices pk Note that in the presence ofSKU entry and exit into the sample the priceindex pj

T will vary over time even if the pricesof the individual SKUs remain stable as theproducts (SKUs) that comprise each productgroup change each period

Ideally we would like to compute the exactprice index pj

T for each group and employ theseindices in the demand estimation Under theassumption of predetermined SKU prices eachperiod the exact price indices would be uncor-related with the error term and the demandparameters could be estimated by simple OLSUnfortunately this strategy is not feasible in thecurrent context

To derive an exact price index for each prod-uct group it is necessary to model consumerrsquoschoice among SKUs conditional on the choiceof the group To this end it is necessary toexplicitly introduce a third stage in the demandspecification and estimate the parameters asso-ciated with the choice at that stage Howeverany specification based on three-stage budget-ing (that would require directly specifying ademand function associated with the SKUchoice at the third stage and imposing additiveseparability at the higher stages) would be in-feasible to estimate for three main reasonsFirst there is a large number of SKUs withinsome groups so that the demand parameters can-not be estimated even if one imposes parameterrestrictions second the SKU prices exhibit littlevariation over time lastly the choice sets at thislowest stage vary over time rendering estimationof the price parameters associated with SKUs thatare not available in all periods infeasible Alterna-tively one could abandon the multistage budget-ing altogether and adopt a discrete choiceapproach which is inherently better suited to deal-ing with varying choice sets and limited pricevariation A discrete choice approach would beparticularly appealing for modeling the SKUchoice conditional on product group choice giventhat within each group all SKUs contain the same

20 See Steven Berry et al (1995) or Aviv Nevo (2001)for a related discussion

1491VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

molecule and are hence comparable As notedearlier however a discrete choice approach at thehigher stages of the demand estimation requiresconverting revenue to physical shares which ischallenging in the case of systemic drugs such asantibiotics for which the appropriate dosage is notwell defined Finally a combination of a multi-stage budgeting approach for the higher stages anda discrete choice approach for the lowest stagewould be feasible to estimate but inconsistentwith a model of utility maximization For thesereasons we have no alternative but to adopt anapproximation for measuring the price index foreach product group

Let us denote this approximation by pjA and let

13j denote the proportional ldquoapproximationrdquo errorassociated with measuring the true price index forgroup j so that ln pj

A ln pjT ln 13j The

estimating demand equation can then be written as

i i j

ijln pjA ilnXQ

PQ

j

ijln 13j i

or

i i j

ijln pjA ilnXQ

PQ i

where the new error term i yenj ijln 13j iis comprised of two components the first onereflects ldquomeasurementrdquo error due to the fact thatthe product group prices we employ are approx-imations and not exact price indices and thesecond one captures conventional demandshocks The demand equation as written abovecorresponds to the equation we take to the data

The particular proxy we use for the exact priceindex for each product group j is the revenue-share-weighted average of the prices of multipleSKUs available within this group that is

pjA

k j

k pk

The notation above illustrates the two source ofprice variation variation in the SKU prices pkand variation in the weights k

We are concerned with potential correlation ofthe so-constructed product group prices with bothcomponents of the product group demand error

term The correlation between pjA and the first

component yenj ijln 13j is independent of theparticular way we construct pj

A and inherent in thefact that we measure the exact price index witherror The potential correlation between pj

A and thesecond component of the product group demanderror the shock i arises however because of thespecific way we construct the proxy pj

A and inparticular because of the presence of the reve-nue share weights k in pj

A since the SKUrevenue share weights k will generally dependon the product group expenditure they arelikely to be correlated with the product groupdemand shock i

To address the simultaneity bias we use in-strumental variables To this end we need vari-ables that are correlated with the proxies pj

A butuncorrelated with the error term i We use thenumber of SKUs within each product group j asan instrument as it does justice to the idea thatvariation in the product group price index stemsin part from variation in the set of SKUs that areavailable each period It is clearly correlatedwith the average group prices pj

A and plausiblyuncorrelated with the demand shock i Theassumption that the number of SKUs is uncor-related with the conventional demand error willhowever be violated if the introduction of anew SKU changes the perceived quality of adrug21 or if it is accompanied by promotionalactivities which will be reflected in i Our hopeis that this is not too often the case Assumingthat the SKU number for each group j is alsouncorrelated with the weighted sum of the mea-surement errors of all product groups yenj ijln 13j itcan be used as an instrument for the product groupprice pj

A22 If one accepts the premise that prices atthe SKU level are exogenous then SKU pricescan also be used as instruments The final list ofinstruments we use hence includes the number ofSKUs in each group the prices of the five largestSKUs for each group and all other exogenousvariables in the demand estimation such as re-gionalproduct dummy interactions and the upper-

21 Note that we already proxy for the quality of each prod-uct group through the product fixed effects i In the actualestimation we actually let these fixed effects be region-specific(ir) so that we allow quality to vary across regions

22 If the measurement error increases in the number ofincluded SKUs this assumption will be violated Unfortu-nately we have no way of checking the validity of thisassumption

1492 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

level total expenditure on antibiotics Regressionsof group prices on the instruments above yieldhigh R-squares with most regressors highly sig-nificant indicating that our instruments are highlycorrelated with prices23

Our sample includes four molecules cipro-floxacin norfloxacin ofloxacin and spar-floxacin Except for sparfloxacin all othermolecules are produced by both foreign anddomestic firms24 So we have seven products(domestic ciprofloxacin foreign ciprofloxa-cin etc) with 96 observations (two years ofmonthly data four geographical regions foreach period) for each product The parametersin the lower-level AIDS demand system asdefined in equation (8) are the product fixedeffects i the productndashregional dummy inter-actions ir the own revenue-share price elas-ticities ii the cross revenue-share priceelasticities i10 i01 i00 and the revenue-share expenditure elasticities i In estimatingthe parameters we first regress prices on allinstrumental variables and then plug the pre-dicted values for prices in the constrainedleast-square regression (For a detailed expla-nation of the constraints see the previoussection)

Given that we impose many cross-equationconstraints and employ instrumental variablesin the estimation it is difficult to derive stan-dard errors for the parameter estimates ana-lytically Our error term interpretation in (8)implies that the error terms for each productgroup are likely to be correlated across re-gions for example if a national advertisingcampaign increases the demand for a rela-tively expensive presentation in one regionsimultaneously increasing the aggregate de-mand and price index for the correspondingproduct group in that region it is likely that thesame effect will be observed in other regions Inprinciple we could exploit the cross-regionalcorrelation in the error terms to estimate thedemand system using seemingly unrelated re-

gression (SUR) This however would requireus to derive the variance-covariance matrix ofthe parameters analytically which as notedabove is cumbersome in the current contextInstead we use the bootstrap method choosingto remain agnostic about the structure of thevariance-covariance matrix The potential dis-advantage of this approach over SUR is a loss inefficiency but as will become apparent in theresults section the model parameters are fairlyprecisely estimated To maintain the marketstructure we randomly sample the periods (withreplacement) and use the same periods for allproducts Regarding the optimal number ofbootstrap repetitions ideally one would followthe three-step method proposed by Donald W KAndrews et al (2000) Empirical evidence sug-gests however that one rarely needs more than200 replications to estimate the standard er-rors25 To be safe we generate 300 bootstrapsamples (with replacement) based on the origi-nal data and estimate the standard errors usingthe standard errors of the bootstrap sampleestimates

The estimation of the top-level AIDS systemis similar The constraints imposed on this top-level demand system are adding-up homogene-ity and symmetry as defined on page 12Again bootstrapping is used to obtain the stan-dard errors of the parameter estimates

D The Counterfactual Scenarios

In assessing the effects of patent enforce-ment we start by focusing on the most extremecase in which compulsory licensing is not anoption and foreign firms are not subject to pricecontrols We use the results from the analysis ofthis case as a benchmark In reality the outcomeof the WTO negotiations is more likely to in-volve some constraints on the monopoly powerof foreign firms selling patented products indeveloping countries such as price caps orcompulsory licensing Our framework can eas-ily accommodate these cases as will becomeapparent in the next subsection

We now focus on the effects of potentialpatent enforcement in the quinolone segmentWe consider several scenarios which vary inthe number and size of domestic products that

23 Specifically the R-squares from the first-stage regres-sions of product group prices on instruments range from057 to 095 except for domestic ciprofloxacin for whichthe R-square is 017

24 Sparfloxacin is actually offered by one foreign subsid-iary in India (see Table 2) However its revenue share isminiscule We therefore treat Sparfloxacin as being pro-duced by domestic firms only 25 See Bradley Efron and Robert J Tibshirani (1993)

1493VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

will be removed from the market In particularwe consider the following five scenarios26

Withdrawal of one large domestic productgroup only domestic ciprofloxacin

Withdrawal of one relatively small domesticproduct group only domestic ofloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand norfloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand sparfloxacin

Withdrawal of all four domestic quinoloneproduct groups

As the list above suggests we proceed fromanalyzing the effects of single-product with-drawal to the analysis of eliminating the en-tire domestic segment This approach wasmotivated by early empirical results that in-dicated that the existence and extent of com-petition from domestic firms has a significantbearing on the predicted effects of patent en-forcement that is our predictions regardingprices and welfare vary substantially depend-ing on how many domestic products are af-fected by patent enforcement and on whethersome domestic competition will remain presentafter TRIPS In addition the size of the affecteddomestic groups is relevant for the welfare pre-dictions accordingly we examine both scenar-ios in which a large domestic product groupsuch as domestic ciprofloxacin is eliminatedfrom the market and scenarios in which the

eliminated domestic group is relatively small(eg domestic ofloxacin) In all scenarios wemaintain the assumption that the set of prod-ucts offered by the remaining firms in themarket does not change in response to patentenforcement

E Computation of Virtual Prices and NewEquilibrium Prices

The first step in the counterfactual analysisis to derive the new equilibrium prices underpatent enforcement27 In this context thereare two sets of prices that are relevant Thefirst set consists of the virtual prices of those(domestic) products that will not be availableonce TRIPS is put in effect To calculate thesevirtual prices we set the expenditures of therelevant products equal to zero The secondset of prices consists of the prices of thoseproducts that remain in the market In deriv-ing these prices we start by assuming profitmaximization without any form of price reg-ulation the firms remaining in the marketreoptimize in response to the policy changeand set new prices taking the prices of allother firms as given28 Of course at the equi-librium all prices change in response to thefact that some domestic products are nolonger present The new equilibrium pricesfor products that remain in the market are thuscomputed by utilizing the first-order condi-tions of profit-maximizing firms into whichthe virtual prices of the eliminated productsare substituted Hence to compute the newequilibrium prices we solve an equation sys-tem of the following form

26 As Table 2 indicates most of the patents for the drugsin the quinolone segment have expired by now so that noneof the scenarios described in this section is going to mate-rialize in practice The reason we consider these alternativescenarios in our counterfactual simulations is to get a senseof how the presence and extent of domestic competitionaffects welfare calculations Such calculations may becomerelevant in the future in other therapeutic classes in whichpatents have not expired yet and where domestic firmscompete with foreign patent holders While the particularexit scenarios will depend on the exact year of patent expiryof drugs in each therapeutic class and the specific way inwhich patents will be enforced (eg whether certain do-mestic drugs will be grandfathered in) the general conclu-sion that emerges from our calculations is that the consumerwelfare loss is substantially smaller in cases where somedomestic competition remains present in the market Hencethe particular way in which patent enforcement is imple-mented is essential for assessing its welfare impact

27 Of course until product patents are in fact introducedthese prices will not be observable Note also that we areassuming here that the range of available products will notchange with the introduction of patents

28 As mentioned above this first set of calculations ab-stracts from the existence of remaining price controls orother government regulations that would impose constraintson the firmsrsquo profit maximization problem Accordingly theresulting numbers should be interpreted as a benchmark ofwhat would happen if markets were completely unregulatedTo examine what the welfare effects of TRIPS would be inthe more realistic scenario of price regulation we subse-quently consider a scenario in which regulation keeps theprices of the products provided by patent-holders at theirpre-TRIPS level For a more detailed discussion see theresults section

1494 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

For products i that are withdrawn from themarket

(14) 0 i ir iiln pirV i10ln pjr j Di

10

j Di

01

i01ln pjr

j Di

00

i00ln pjr

ilnXQr

PQr

For products k that remain in the market

(15) pkr ckr 1 1

kk pkrpjr pirV 1

In the equations above pirV denotes the virtual

prices of the products that are removed fromthe market while pjr denotes the updatedprices of all other products Note that whensolving for the virtual prices we account forthe fact that both the price index for quino-lones PQr and the expenditure allocated tothis subsegment XQr need to be updated toreflect the fact that as a result of the pricechanges there may be substitution away fromthis subsegment To obtain the new quinoloneexpenditure XQr and the new price index PQrwe use the estimates and formulas for thehigher-level AIDS system In equation (15)ck refers to the marginal cost for product k thatwe have obtained from the previous estimationstage As mentioned in the previous subsectionwe conducted the simulations using both theupper and lower bounds for marginal cost ck

u

and ckL respectively The term kk(pkr

pjrpirV) re-

fers to the unconditional own-price elasticity forproduct k which is a function of the eliminatedproductsrsquo virtual prices and the remaining prod-uctsrsquo new equilibrium prices We conduct thecounterfactual simulations and welfare analysisat the regional level because the presence ofregion-specific product effects in the demandestimation implies region-specific demand elas-ticities and hence region-specific marginalcosts and markups The presentation and discus-sion of our results focus on national averages ofthe relevant variables that we construct by com-

puting weighted averages across regions usingthe population of each region as a weight

F Welfare Assessment

The simulation of the new equilibrium underpatent protection can provide important insightsinto how consumers and firms will respond tothe removal of domestic products in the market(for example toward which products consumerswill substitute or which prices will increase themost) To get a more precise idea of how peo-plersquos well-being will ultimately be affected byTRIPS we compute as a last step in our anal-ysis the welfare effects of the policy changeSocial welfare is defined as consumer welfareand the sum of domestic firm profits Thechange in domestic profits can easily be calcu-lated by comparing the domestic firm (variable)profits at the pre-TRIPS prices to the profitsthese firms will realize at the new simulatedprices Although foreign firm profits do notcount in domestic welfare calculations we alsocompute the effects of patent enforcement onforeign firm profits to get an idea of how largethe expected benefits of TRIPS are for thesefirms This provides in some sense an indirectway of assessing whether the claims that patentenforcement in countries like India will lead tomore research on developing-country-specificdiseases (such as malaria) have any validity iffor example we find that the effect of patentenforcement on the foreign firm profits realizedin India is small in magnitude it is unlikely thatforeign firms will engage in more developing-country-specific research in response to TRIPSIt is important to note that in all these calcula-tions we work with the lower-bound estimatesfor marginal costs since these give rise to thehighest possible markups Hence our estimateof profit loss for domestic producers most likelyoverstates this loss

On the consumer side we measure changes inconsumer welfare by the compensating variation(CV) defined as the additional expenditure thatconsumers need in order to achieve the sameutility level as before patent enforcement at thenew prices Specifically let P0 denote the pricevector before patent enforcement P the simulatedprice vector post-TRIPS (that we obtained usingthe methods described in the previous subsection)u0 the utility attained by consumers before TRIPS

1495VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

and E(u P) the higher-level expenditure functionThen the compensating variation is given by

(16) CV Eu0 P Eu0 P0

Note that the CV as computed in (16) representsthe combination of three effects

The pure product-variety effect that is theeffect that arises because one or more prod-ucts are not available to consumers anymoreholding the prices of all other remainingproducts and the total expenditure on thequinolone subsegment XQ constant

The expenditure-switching effect that is theeffect arising from substitution away fromquinolones and toward other subsegments ofthe antibiotics market again holding the pricesof all other remaining products constant

The reduced-competition effect that is theeffect that arises because the firms remainingin the market adjust (increase) their prices inresponse to the removal of domestic products

From both an analytical and a policy point ofview it is desirable to assess how large each ofthese effects is Accordingly we decompose thetotal effect on consumer welfare (the CV as givenby equation (16)) using the following procedure

To get the pure product variety effect wecompute virtual prices for the products that areremoved from the market holding the quino-lone expenditure XQ and the prices of all otherproducts fixed Let us call the resulting pricevector P1 Then the pure product variety effectis represented by E(u0 P1) E(u0 P0)

To compute the expenditure-switching effectwe compute another set of virtual prices againholding the prices of all remaining products fixedbut letting quinolone expenditure adjust in re-sponse to the new price index for the quinolonesegment (Given that the prices of the remainingproducts remain fixed the change in the priceindex arises only because of the removal of one ormore domestic products) Note that this scenariomost closely resembles the case in which patentlaws are enforced but strict price regulation keepsthe prices of the products offered in the market attheir pre-TRIP level Let us label the so-computedprice vector P2 The expenditure-switching effectis then E(u0 P2) E(u0 P1)

Finally the reduced competition effect aris-ing from higher prices for the remaining prod-

ucts is computed as the residual change in thecompensating variation once the product-varietyand expenditure-switching effects have been ac-counted for that is E(u0 P) E(u0 P2)where the price vector P is computed accordingto the formulas (14) and (15) to reflect theadjustment of prices to the new regime

To compute the standard errors associatedwith the counterfactual simulations (that is thestandard errors for the counterfactual prices andwelfare estimates) we again use bootstrappingWe first bootstrap the original sample (sales andprices of all the drugs) 300 times Next we esti-mate the AIDS model for each of these samplesand compute the counterfactual equilibrium pricesand welfare losses corresponding to each policyscenario for each of the 300 simulations In thefinal step we compute standard errors29

One limitation of our framework is that itdoes not allow for a heterogeneous response ofconsumers to the policy change Accordinglyour framework is not suited to addressing thequestion of how different groups in the popula-tion will be affected by patent law enforce-ment30 Along the same lines our frameworkdoes not accommodate the possibility of pricediscrimination which might lead to differentresults regarding welfare losses and profit gainsrelative to uniform pricing31 Still we believethat the results of the counterfactual simulationscan provide important insights into the likelyaggregate response to patent enforcement andthe factors that drive this response

III Results

A The Structure of Demand

Tables A2 and A3 in the Appendix display theresults from estimation of the lower- and upper-level AIDS systems respectively For ease of in-terpretation rather than discussing the coefficientestimates we focus our discussion on the implied

29 In cases some of the simulations do not converge wecompute the standard errors using only those simulationsthat converged

30 Addressing this question would require at a mini-mum micro data on consumer purchases as in Goldberg(1995) To our knowledge such data do not exist for theIndian pharmaceutical market

31 See Ernst R Berndt (1994) for an extensive discussionof uniform pricing

1496 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

unconditional price and expenditure elasticities re-ported in Table 6 Given that the region-specificproduct group effects ir reported in the last fourcolumns of Tables A2 and A3 imply region-spe-cific demand elasticities we report separate elas-ticities for each region in Tables 6A and 6B Asevident from the comparison of these tables theelasticities are very similar across regions so thatwe can focus the remaining discussion on theelasticities of one region only the northern regiondisplayed in Table 6A

The diagonal terms of Table 6A report theown-price elasticities which are in all but onecase negative and highly significant The oneexception is the foreign norfloxacin productgroupmdashwhose share of quinolone sales is 007percentmdashfor which we estimate a negative butinsignificant own-price elasticity For the remain-ing product groups demand appears to be highlyelastic with the estimated elasticities lower than2 in four out of the six cases The magnitude ofthe own-price elasticities matches the features ofthe Indian pharmaceutical market mentioned ear-lier which would suggest that Indian consumersare likely to be quite price-sensitive32 The elas-ticities appear especially large if one takes intoaccount that they refer to product groups (such asdomestic ciprofloxacin) and not individual drugsoffered by particular firms Their relative magni-tudes are also intuitive the drug with the largestmarket share and a relatively high price (domesticciprofloxacin) appears to be one of the least elas-tic In contrast foreign ciprofloxacin is highlyprice elastic this is plausible as ciprofloxacindrugs offered by subsidiaries of multinationalsface the stiffest competition in this market seg-ment from approximately 75 Indian firms offeringthe same molecule

The estimated expenditure elasticities appear inthe last column These are all positive indicatingthat the demand for all product groups is normalThe remaining cells display the estimated cross-price elasticities As one might perhaps expect forproducts within a therapeutic subsegment theseare mostly positive Out of a total of 49 priceelasticities we estimate there are six that do notconform to expectations these are the cross-price elasticities between different foreign prod-

uct groups which are estimated to be negativeand significant33 Fortunately these elasticitieshave negligible impact on the welfare analysisgiven that our counterfactuals focus on theeffect of withdrawing one or more domesticproducts from the market the most relevantelasticities are the ones that capture the responseof various product group shares to a change in theprice of one or more domestic groups these elas-ticities are the cross-price elasticities between var-ious domestic groups and the ones betweendomestic and foreign groups which are plausibleand precisely estimated34

Regarding these elasticities a striking aspectof our estimates is how large positive andsignificant the cross-price elasticities betweendifferent domestic product groups aremdashin factfor norfloxacin and ofloxacin we estimate thatdomestic product groups containing differentmolecules are closer substitutes for one anotherthan product groups that contain the same mol-ecule but are produced by foreign firms Incontrast for ciprofloxacin (the molecule withthe largest revenue share) we estimate a largepositive cross-price elasticity between the do-mestic and foreign versions

The fact that domestic products appear to beclose substitutes for other domestic productsthat contain different molecules truly representsan ldquoempiricalrdquo finding in the sense that we do

32 In developed economies elasticities of this magnitude havetypically been found only for generic drugs (and even then onlyrarely) or among consumers who lack health insurance

33 While these elasticities are clearly counterintuitive theyare not inconsistent with the underlying demand system whichimposes no restrictions on the sign of the cross-price elastici-ties We do not have a good explanation of why these elastic-ities are estimated to be negative A potential explanation isthat the shares of the foreign products are very small giventhis we observe very few consumers switching from one (verysmall) foreign group to another (very small) foreign groupwhen the price of the first foreign group goes up and hence theinference is not very reliable in this case

34 The cross-price elasticities between foreign drug groupscontaining different molecules will also have an effect on thewelfare estimates given that the withdrawal of domestic prod-ucts will generally lead to changes in the prices of all foreignproducts and market shares will be reallocated from someforeign products to other foreign products based on the newprices However this reallocation from ldquoforeign to foreignrdquo istruly second order in our case compared to the reallocationfrom ldquodomestic to domesticrdquo and ldquodomestic to foreignrdquo prod-ucts In scenarios in which we consider patent enforcementaccompanied by strict price regulation the cross-price elastic-ities between foreign product groups are in fact completelyirrelevant as the prices of foreign products are not allowed toincrease in these cases Still our welfare loss estimates remainsubstantial drivenmdashas beforemdashby the loss of product variety

1497VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

not impose it through any of our assumptionsregarding the demand function The questionthat naturally arises then is what might explainthis finding While we cannot formally addressthis question anecdotal accounts in various in-dustry studies suggest that the explanation maylie in the differences between domestic andforeign firms in the structure and coverage ofretail distribution networks

Distribution networks for pharmaceuticals inIndia are typically organized in a hierarchicalfashion Pharmaceutical companies deal mainlywith carrying and forwarding (CampF) agents inmany instances regionally based who each sup-ply a network of stockists (wholesalers) Thesestockists in turn deal with the retail pharma-cists through whom retail sales ultimately oc-cur35 The market share enjoyed by a particularpharmaceutical product therefore depends inpart on the number of retail pharmacists who

stock the product And it is here that thereappears to be a distinction between domesticfirms and multinational subsidiaries In particu-lar the retail reach of domestic firms as agroup tends to be much more comprehensivethan that of multinational subsidiaries (IndianCredit Rating Agency (ICRA) 1999)36

There appear to be two reasons for this Thefirst is that many of the larger Indian firmsbecause they have a much larger portfolio ofproducts over which to spread the associatedfixed costs typically have more extensive net-works of medical representatives The second issimply that there are many more domestic firms(and products) on the market At the retail levelthis would imply that local pharmacists mightbe more likely to stock domestic products con-taining two different molecules say ciprofloxa-cin and norfloxacin than they would domesticand foreign versions of the same molecule Tothe extent that patients (or their doctors) arewilling to substitute across molecules in order tosave on transport or search costs (eg going toanother pharmacy to check whether a particular

35 There are estimated to be some 300000 retail pharma-cists in India On average stockists deal with about 75 retailers(ICRA 1999) There are naturally variations in this structureand a host of specific exclusive dealing and other arrangementsexists in practice Pharmaceutical firms also maintain networksof medical representatives whose main function is to marketthe companyrsquos products to doctors who do the actual prescrib-ing of drugs In some instances firms do sell directly to thedoctors who then become the ldquoretailerrdquo as far as patients areconcerned but these are relatively rare

36 These differences were also highlighted in conversa-tions one of the authors had with CEOs and managingdirectors of several pharmaceutical firms as part of a sep-arate study

TABLE 6AmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN THE NORTHERN REGION

Product group

Elasticity with respect to

Prices of foreign productgroups Prices of domestic product groups Overall

quinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 557 013dagger 015 401 011dagger 011dagger 016 137(179) (007) (007) (184) (006) (006) (006) (029)

Foreign norfloxacin 427dagger 045 427dagger 350dagger 602 451 465 220(242) (112) (242) (210) (623) (184) (183) (105)

Foreign ofloxacin 011 010dagger 138 009 009dagger 023 011 116(005) (005) (031) (027) (005) (028) (004) (017)

Domestic ciprofloxacin 018 001 001 168 008 008 010 117(008) (000) (001) (023) (002) (002) (002) (003)

Domestic norfloxacin 004 003 004 058 223 042 040 073(001) (003) (001) (017) (011) (004) (003) (009)

Domestic ofloxacin 005 005 011 077 074 342 074 089(002) (002) (013) (028) (008) (025) (008) (021)

Domestic sparfloxacin 007 004 007 115 063 063 288 028(002) (001) (002) (015) (006) (006) (017) (012)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent significance level and dagger (dagger) denotes significance at the 10-percent level

1498 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE 6BmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN OTHER REGIONS

Demand patterns in the eastern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 594 014dagger 016 431 013 011 017 140(195) (008) (008) (201) (006) (007) (007) (031)

Foreign norfloxacin 329dagger 058 329dagger 264dagger 460 346 359 192(180) (083) (179) (153) (451) (138) (139) (082)

Foreign ofloxacin 012 010dagger 140 010 010 024 013 117(006) (006) (034) (031) (005) (030) (005) (019)

Domestic ciprofloxacin 018 001 001 172 009 008 011 117(008) (000) (002) (027) (002) (003) (002) (003)

Domestic norfloxacin 004 004 004 068 242 049 046 070(002) (004) (002) (020) (013) (005) (004) (010)

Domestic ofloxacin 004dagger 004 009 061 060 295 060 092(002) (001) (010) (028) (007) (022) (007) (017)

Domestic sparfloxacin 005 003 005 092 048 051 251 043(002) (001) (001) (017) (005) (006) (016) (010)

Demand patterns in the western region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 427 009dagger 011 286 008 006 011 126(129) (005) (005) (133) (004) (005) (004) (021)

Foreign norfloxacin 714dagger 008 710dagger 594dagger 995 742 770 299dagger

(369) (183) (368) (328) (1015) (297) (296) (178)Foreign ofloxacin 013 012dagger 145 009 012 026 013 119

(007) (006) (037) (029) (005) (033) (005) (021)Domestic ciprofloxacin 019 001 000 174 009 008 011 118

(008) (000) (001) (024) (002) (003) (002) (004)Domestic norfloxacin 004 004 004 067 243 050 047 069

(002) (004) (001) (018) (014) (005) (004) (010)Domestic ofloxacin 003 003 007 048 048 257 048 093

(002) (001) (008) (022) (005) (015) (004) (014)Domestic sparfloxacin 006 003 005 083 046 049 241 046

(002) (001) (001) (014) (004) (005) (012) (009)

Demand patterns in the southern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 560 013dagger 015 402 011 010 017 137(182) (007) (007) (187) (006) (006) (007) (029)

Foreign norfloxacin 533dagger 031 531dagger 431dagger 748 559 584 249dagger

(301) (134) (300) (245) (734) (219) (219) (128)Foreign ofloxacin 014 012dagger 148 011 012 029 016 120

(007) (007) (039) (033) (006) (035) (006) (022)Domestic ciprofloxacin 017 001 000 169 008 007 011 116

(008) (000) (001) (025) (002) (003) (002) (003)Domestic norfloxacin 004 003 003 059 225 043 040 073

(001) (003) (001) (019) (012) (004) (004) (009)Domestic ofloxacin 004dagger 004 009 062 060 296 060 091

(002) (001) (010) (026) (006) (019) (007) (017)Domestic sparfloxacin 008 005 008 138 073 075 323 016

(002) (001) (002) (018) (007) (008) (021) (014)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at 5-percent level anddagger (dagger) denotes significance at the 10-percent level

1499VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

foreign product is in stock) in aggregate datawe would expect to find precisely the substitu-tion patterns that we report in Table 6

Whether the particular explanation we provideabove is the correct one the high degree of sub-stitutability between domestic product groupsturns out to have important implications for thewelfare calculations We discuss these in moredetail below when we present the results of thecounterfactual welfare analysis Another elasticitywith important implications for the counterfactu-als is the price elasticity for the quinolone subseg-ment as a whole which indicates how likelyconsumers are to switch to other antibioticsgroups when faced with a price increase for quin-olones This elasticity is computed on the basis ofthe results in Table A3 and it is at 111 (stan-dard error 024) this is large in magnitudebutmdashas expectedmdashsmaller in absolute value thanthe own-price elasticities of the product groupswithin the quinolone subsegment

The results in Tables 6A and 6B are based onour preferred specification discussed in SectionII In Tables A4 to A6 in the Appendix weexperiment with some alternative specificationsTables A4(a)ndashA4(c) correspond to a specificationthat includes in addition to product-group-specificregional fixed effects product-group-specific (andfor the upper level antibiotics-segment-specific)seasonal effects We distinguish among three sea-sonsmdashthe summer monsoon and wintermdashandreport the unconditional demand elasticities for

the northern region for each of these seasons Asevident from the tables in the Appendix our elas-ticity estimates are robust to the inclusion of sea-sonal effects The demand elasticities in Table A5are based on estimation of the demand system byOLS Compared to the elasticities obtained by IVthe OLS elasticities are smaller in absolute valueimplying that welfare calculations based on theOLS estimates would produce larger welfare lossestimates Nevertheless some of the patterns re-garding the cross-price elasticities discussed ear-lier are also evident in the OLS results inparticular the cross-price elasticities between dif-ferent domestic product groups are all positivelarge and significant and in most instances largerthan the cross-price elasticities between drugs thatcontain the same molecule but are produced byfirms of different domesticforeign status Theclose substitutability of domestic products indi-cated by both the OLS and IV estimates seems tobe one of the most robust findings of the paper

B Cost and Markup Estimates

Table 7 displays the marginal costs markupsand profits implied by the price elasticity esti-mates of Tables 6A and 6B for each of the sevenproduct groups Given that our regional effectsimply different price elasticities for each regionour marginal cost and markup estimates alsodiffer by region Given however that based on

TABLE 7mdashUPPER AND LOWER BOUNDS FOR MARGINAL COST MARKUP AND ANNUAL PROFIT BY PRODUCT GROUPS WITHIN

THE QUINOLONE SUBSEGMENT

Product groupLower boundfor MC (Rs)

Upper boundfor markup

Upper boundfor profit(Rs mill)

Upper boundfor MC (Rs)

Lower boundfor markup

Lower boundfor profit

(Rs)

Foreign ciprofloxacin 83 19 269 103 0 00(123) (012) (1655)

Foreign norfloxacin NA NA NA 53 0 00Foreign ofloxacin 323 70 1061 1085 0 00

(2316) (021) (3185)Domestic ciprofloxacin 47 59 17019 112 0 00

(114) (010) (29858)Domestic norfloxacin 52 43 2807 90 0 00

(020) (002) (1532)Domestic ofloxacin 587 34 1612 901 0 00

(218) (002) (1280)Domestic sparfloxacin 495 37 1985 788 0 00

(157) (002) (1100)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level Estimated lower bound forforeign norfloxacinrsquos marginal cost is negative since the estimated price elasticity is less than one in absolute value

1500 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Tables 6A and 6B the price elasticities do notseem to differ substantially across regions wereport for ease of exposition only the nationalaverages for marginal costs and markups inTable 7

Table 7 has two parts In the left part (firstthree columns) we report the lower bound formarginal cost and the corresponding upperbound for markup and upper bound for totalannual profit for each product group Thesenumbers are based on the price elasticities weobtained from estimating the two-level AIDSdemand system Since we do not have a reliableestimate for the price elasticity of foreign nor-floxacin (the point estimate is negative but lessthan one in absolute value and insignificant)we cannot compute the lower bound for mar-ginal cost in this case It is important to note thatthese estimates do not reflect either the actualmarginal cost or the actual markup for thesedrugs both because the existence of price reg-ulation implies that the unconstrained first-orderconditions are unlikely to hold each period andbecause our aggregation across firms of thesame domesticforeign status supplying thesame molecule makes the interpretation of theseestimates problematic In particular the fact thatwe ignore competition among firms within eachproduct group implies that our estimates willtend to overstate market power These numberswill prove useful however in the counterfac-tual simulations as they can give us a sense ofhow large the maximum profit gains for multi-national firms and the maximum profit lossesfor domestic firms are likely to be under patentenforcement The right part of the table (lastthree columns) reports the upper bound for mar-ginal cost which we obtain by simply taking themarginal cost to be equal to the observed priceWe do not report standard errors in this casesince the numbers are based on actual data Themarkups corresponding to these marginal costupper bounds are of course zero We conductthe counterfactual simulations using both thelower and upper bounds for marginal costs

The (upper bounds for) markups on the leftside of the table are generally plausible Thedomestic ciprofloxacin product group that dom-inates the quinolone subsegment and for whichwe documented high prices a high marketshare and a relatively low elasticity of demandenjoys one of the highest markup upper bounds(60 percent) and accounts for nearly 70 percent

of all profits derived within the subsegmentForeign ciprofloxacin on the other hand whichfaces the stiffest competition from domesticfirms and for which we estimated a highly elas-tic demand has the lowest markup upper bound(19 percent)

C Counterfactual Estimates of the Impact onPrices and Welfare

With estimates of the key demand and costparameters in hand we turn to the counterfac-tuals We consider the five separate scenarioslisted in the previous section All of the scenar-ios involve the withdrawal of one or more of thedomestic product groups from the market Ta-ble 8 displays our estimates of the consumerwelfare losses that result under the differentscenarios The losses are expressed in billion ofrupees per year All numbers presented in Ta-ble 8 and subsequent tables are based on usingthe lower bounds for marginal cost and upperbounds for markup in the simulations As dis-cussed earlier these numbers are the more in-teresting to work with since they give us anupper bound for the changes in the profits ofdomestic and foreign firms that would resultfrom patent enforcement In the Appendix wealso present results based on using the upperbounds for marginal costs in the simulations inwhich case the pre-TRIPS profits of domesticand foreign firms are zero In all cases marginalcosts are assumed to be constant in outputWhile naturally the profit implications differdepending on whether one uses the upper orlower bounds for marginal costs (firm profits arezero if one assumes the upper bound of mar-ginal cost in which case price equals marginalcost) the estimated consumer welfare losses aresimilar in the two cases We discuss these re-sults in more detail at the end of this section

The first column presents our estimates of theconsumer welfare losses attributable to the pureloss of product variety effect where we fix theprices of all remaining products as well as theoverall expenditure on quinolones while with-drawing one or more of the domestic productgroups Note that in our initial specification ofthe demand system had we not allowed for thepossibility that consumers might differentiatebetween domestic and foreign products evenwhen they contain the same molecule this

1501VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

particular component of the loss of consumerwelfare would not have arisen

The estimates reported in the second columnincorporate the expenditure-switching effect ontop of the loss of product variety Here basedupon the price elasticity estimates from thehigher-level AIDS system we adjust (down-ward) the expenditures allocated to the quino-lone subsegment as the composite price ofquinolones effectively increases as a consequenceof the higher virtual prices of the domestic productgroups that are withdrawn from the market Be-cause the estimates in this column are generatedassuming that the prices of the products that re-main in the market are not adjusted upward theyprovide a sense of what consumer losses would beif the introduction of product patents were coupledwith strict price regulation aimed at maintainingprices at prepatent levels Alternatively they canbe thought of as the relevant welfare numbers ifintense competition among firms within the re-maining product groups kept the prices of theproducts that were still offered in the market closeto the firmsrsquo marginal costs

The last column of Table 8 displays the esti-mated consumer welfare losses when bothcross-segment expenditure-switching and with-in-segment upward price adjustments are takeninto account

If we compare the results across the firstsecond and third columns all the counterfac-tual scenarios produce qualitatively similar pat-terns patterns that are consistent with what we

would expect Starting from the initial loss ofwelfare attributable to the loss of product vari-ety the option of switching expenditures outof the quinolone subsegment to other subseg-ments mitigates some of the initial welfareloss But if we then incorporate the upwardprice adjustments that result in response to thereduced competition the welfare losses aremagnified

Of particular interest from a policy perspec-tive are the relative magnitudes of these threeeffects which are similar under all the counter-factual scenarios though absolute levels varyconsiderably First despite the fact that the de-mand for quinolones is quite sensitive to thecomposite price of quinolonesmdashthe upper levelprice elasticity is 111mdashthe cross-subsegmentexpenditure-switching effects are in all thecases small (in absolute value terms) relative tothe other two effects For instance under thescenario where all the domestic quinolone prod-uct groups are withdrawn from the market theoverall consumer welfare loss of Rs 1781 bil-lion per year can be decomposed into an initialloss of Rs 1176 billion (66 percent) attribut-able to the loss of product variety a slightreduction in this initial loss of Rs 041 billion(2 percent) from Rs 1176 billion to Rs1135 billion because of expenditure switchingand a subsequent additional loss of Rs 646billion (36 percent) from Rs 1135 billion toRs 1781 billion because of the reduced com-petition and consequent price increases

TABLE 8mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure loss ofvariety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 732(087) (089) (146)

Only ofloxacin 008 008 023(008) (008) (010)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1253(177) (180) (415)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1058(142) (145) (331)

All four domestic quinolones products 1176dagger 1135dagger 1781(643) (634) (1270)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent significance level and dagger (dagger)denotes significance at the 10-percent level

1502 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 11: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

revenue shares to physical shares is extremelydifficult if not infeasible in the case of antibioticsBecause such drugs are ldquosystemicrdquo by nature theyare used to treat a large number of infections andthe dosage of each drug depends on the particularinfection it is supposed to address (for example thedosage will differ depending on whether the anti-biotic is used to combat an ear infection or tuber-culosis) This particular feature of antibioticscomplicates the conversion of revenue to physicalshares

The basic idea of the multistage budgetingapproach is to use the therapeutic classificationof a productmdashie the therapeutic segment andsubsegment the product belongs tomdashto orga-nize all products in the systemic antibacterialssegment into a hierarchical taxonomy consist-ing of two levels At the higher level are thevarious subsegments of systemic antibacterialsThe first stage of budgeting corresponds to theallocation of expenditures across the subseg-ments in this upper level of the taxonomy

In the second stage of the budgeting processcorresponding to the lower level of the taxon-omy a flexible functional form is adopted tomodel how the expenditures allocated to eachsubsegment are distributed across the productswithin that subsegment In particular to modeldemand at the second stage we employ theAIDS specification proposed by Deaton andMuellbauer (1980b)15

While the two-stage demand estimation ap-proach offers functional form flexibility its ap-plication to the Indian systemic antibioticsmarket poses a few problems The first is thatdue to entry and exit many SKUs and evenproducts in our sample are not present in everyperiod AIDS does not have a good way ofdealing with a varying number of products as itwas developed with broad commodity catego-ries in mind which are consumed by all con-sumers every period To solve this problem weaggregate within each subsegment (eg quino-lones) SKUs into product groups where withineach product group all SKUs contain the samemolecule and are produced by firms with thesame domesticforeign status Specifically let aSKU k be indexed by its molecule (or API) M its

domesticforeign status DF indicating whether itis produced by a domestic (Indian) or a subsidiaryof a foreign (multinational) firm a particular pre-sentation s and the particular firm f that producesit We aggregate SKUs over presentations andfirms to obtain a newly defined product group iwhich is indexed only by molecule M and domes-ticforeign status DF and has revenue Ri yenfsRk with i (M DF) k (M DF f s) and pricepi yenfs kpk where k denotes the conditional(on M and DF) revenue share of this particularproduct ie

(1) k Rk

Ri

In most cases the resulting product groups arebroad enough to be present every period16

The usual concern with this aggregation pro-cedure is that it may lead us to overstate firmsrsquomarket power as we ignore competition amongfirms with the same domesticforeign statusproducing the same molecule In the presentapplication however this concern is unlikely tobe of great importance as the effect of patentenforcement is to wipe out all domestic compe-tition at once while granting foreign firms mo-nopoly power hence competition among firmsfor patented molecules becomes irrelevant Theaggregation according to the domesticforeignstatus (within a particular molecule) thus corre-sponds to the scope of our analysis and theparticular questions of interest

The second problem is that for our approachto be useful in welfare analysis the allocation oftotal expenditures to group expenditures at thehigher stage has to be modelled in a way con-sistent with utility maximization In general thesolution of this allocation problem requiresknowledge of all individual product prices

15 Representative applications of the multistage budget-ing approach include Sara F Ellison et al (1997) Hausman(1994) and Hausman and Gregory K Leonard (2002)

16 We are missing only four observations (ie monthregion combinations) all for the drug group of ForeignNorfloxacin August 1999 in the south May 2000 in thewest October 2000 in the south and November 2000 in theeast In these cases we set the revenue shares of ForeignNorfloxacin equal to zero In general with 01 percent ofquinolone sales (see Table 1 row 2) Foreign Norfloxacinhas a very small share of the market This probably explainswhy the results pertaining to this drug group are unreliableit is the only drug group for which we do not obtain asignificant price elasticity of demand while its cross-priceelasticities with other foreign drug groups often have thewrong sign

1487VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

From an empirical point of view this is notparticularly useful as it eliminates all compu-tational advantages of the two-stage approachTo address this problem we adopt an approxi-mate solution to model the higher-level expen-diture allocation along the lines suggested byDeaton and Muellbauer (1980b pp 131ndash32)This gives rise to a two-level AIDS specification

Consider the lower-level estimation firstwhich refers to the allocation of a particularsubsegmentrsquos expenditure to the product groupswithin the subsegment In our application therelevant subsegment is quinolones which weindex with Q Let the product groups within thissubsegment be indexed by i 1 N pi be theprice of product group i (where as noted abovei refers to a particular molecule and domesticforeign status combination) and XQ be the totalexpenditure on the quinolone segment The rev-enue share of each product group is given by

(2) i i j

ijln pj ilnXQ

PQ

where i the revenue share of product group iis defined as

(3) i pi qi

yenj

pj qj

xi

XQ with i j Q

XQ is the overall expenditure on the quinolonesubsegment and PQ is a price index given by

(4) ln PQ ap 0 i

iln pi

1

2 i

j

ijln piln pj

With a limited number of product groups and asufficiently large number of time-series observa-tions the flexibility implied by the AIDS modeldoes not impose too many demands on the data Inthe present application however where the num-ber of observations is limited the AIDS model isnot estimable in this general form To reduce thenumber of parameters that need to be estimatedwe impose two sets of restrictions

The first set of restrictions is implied by thetheory of utility maximization Specificallythese restrictions are

Adding-up yenk k 1 yenk k 0 yenk kj 0 j

Homogeneity yenk jk 0 j Symmetry ij 1frasl2 [ij ji] ji This last

restriction by itself reduces the number of parameters to [N(N 1)]2

The second set of restrictions we impose aimsat further reducing the number of parameters tobe estimated by exploiting our knowledge of thisparticular market Specifically for each productgroup i we allow one ij parameter for all productgroups j that have different molecules from prod-uct group i and are produced by foreign firms andone ij for all product groups j with differentmolecules produced by domestic firms We donrsquotimpose any restrictions on the ij parameter whenproduct group j has the same molecule as productgroup i (By construction product groups i and jcontain products produced by firms with differentdomesticforeign status)

To better illustrate the nature of the restric-tions we impose on the patterns of substitutionacross products some additional notation isneeded Let d(i j) be an indicator of the degreeof similarity (or difference) between productgroup i and product group j along the dimen-sions we are able to observe (molecule M anddomesticforeign status DF) For any two prod-uct groups i and j d(i j) can take on one of thefollowing three values17

(5)

di j (1 0) if Mi Mj DFi DFj

(0 1) if Mi Mj DFi DFj

(0 0) if Mi Mj DFi DFj

Let

(6) Diab j di j a b

17 The sequence (1 1) is not possible for two differentproducts in this case the parameter corresponds to theproductrsquos own-price effect that is ii

1488 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The equation at the lower level becomes

(7) i i iiln pi i10ln pj j Di10

j Di

01

i01ln pj

j Di

00

i00ln pj ilnXQ

PQ

Note that

The parameter ii captures a product grouprsquosown-price effect (there will be as many iiparameters as the number of product groups)

The parameter i10 captures the cross-priceeffect of the product group containing prod-ucts with the same molecule but produced byfirms of different nationality

The parameter i01 captures the cross-priceeffects of product groups containing productswith different molecules but produced byfirms with the same nationality

The parameter i00 captures the cross-priceeffects of product groups containing productswith different molecules produced by firms ofdifferent nationality

Before we take the demand equation to thedata we make two modifications The first oneis to let the product-specific effects i vary byregion r The resulting product-specific regionaleffects ir have two interpretations first theycontrol for the ldquoqualityrdquo of each drug withquality differences being allowed to vary acrossregions second they proxy for demographicsand other demand shifters which vary by re-gion and may affect the demand of each prod-uct group differently18 Note that by includingproduct-specific regional effects in the demandspecification we estimate the price parametersbased on the within-product-group variation ofprices in each region In an earlier version of thepaper we also estimated the demand systemwithout regional dummies and obtained similarresults

The second modification that allows us to gofrom a deterministic to a stochastic specificationof the demand equation is to include an additiveerror term in (7) The latter takes into accountthe fact that (7) is not expected to fit the dataexactly The error term irt accounts for mea-surement error (due to the fact that the productgroup prices pjrt we employ in the estimationare not exact price indices but approximationsthereof) and (potentially region-specific) de-mand shocks that may affect the demand for aproduct in a particular period Examples of suchshocks include an advertising campaign for aparticular product which temporarily increasesthe demand for this product or the outbreak ofa (potentially region-specific) epidemic whichcalls for the use of a particular drug We discussthe interpretation and properties of this errorterm in more detail in the next section

The final form of the equation we estimate atthe lower level becomes (with subscript t de-noting month and subscript r denoting region)

(8) irt i ir iiln pirt i10ln pjrt j Di10

j Di

01

i01ln pjrt

j Di

00

i00ln pjrt

ilnXQrt

PQrt irt

The analysis so far has conditioned on the ex-penditure allocated to the quinolone subseg-ment XQ The upper level of the estimationconsiders the problem of allocating total expen-diture across the different systemic antibioticssubsegments one of which is quinolones Theupper-level demand function is given by

(9) G G H

GHln PH GlnX

P

where all variables denoted by capital letters aredefined as before but now refer to subsegments(G H ) rather than product groups within asubsegment and the total expenditures on sys-temic antibiotics X are deflated by the Stone priceindex log P yenH Hlog PH When estimating thesystem above we impose all the restrictions

18 Given the short time span of our sample typicaldemand shifters such as age distribution income distribu-tion and education hardly change over our sample periodSuch shifters are therefore absorbed by the region-specificproduct fixed effects ir

1489VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

implied by utility maximization as we do withthe estimation of the lower-level AIDS We donot however impose any additional restrictionson the substitution patterns at this stage so thatthe cross-price effects across segments remainrelatively unconstrained

Estimation of the higher-level AIDS allowsus to obtain the unconditional own- and cross-price elasticities that are used in the formulationof the supply problem and welfare analysisThese are given by the formula

(10) ij ijXQXQ

ln qi

ln XQ

ln XQ

ln PQ

ln PQ

ln pj

with the conditional cross price elasticitiesgiven by

ijXQXQ ij ij jlnXQ

PQ

i

As in the lower stage we include subsegment-specific regional dummies in the specificationof the upper-stage demand system so that thefinal form of the estimating equation at theupper stage becomes

(11) Grt G Gr H

GHln PHrt

GlnXrt

Prt Grt

In sum the demand system we take to thedata is represented by equations (8) and (11)and the associated parameter restrictions im-plied by economic theory

B Modelling the Supply Side of the Market

Counterfactual simulations concerning theeffects of domestic product withdrawal requireknowledge of the marginal costs of pharmaceu-tical firms operating in the Indian market Theseare unobservable The usual approach in theNew Empirical Industrial Organization litera-ture has been to exploit the firm equilibriumconditions to infer marginal cost For exampleit is often assumed that the marginal cost ci isconstant and that the industry is an oligopoly

engaging in Bertrand competition with differen-tiated products Assuming that firms myopicallymaximize profits each period one can then de-rive the firmsrsquo first-order conditions that corre-spond to the assumptions above about costsmarket structure and firm behavior

We deviate from this procedure as the presenceof price regulation renders the assumption of un-constrained period-by-period maximization un-tenable Ideally one would like to explicitlyincorporate the price and other administrative con-trols into the firmrsquos optimization problem andderive the first-order conditions under the assump-tion of constrained maximization The complexityof the price regulation however makes this ap-proach infeasible Therefore we adopt an alterna-tive approach that does not rely on modelling theprice-setting process but is instead based on de-riving upper and lower bounds for marginal costsand markups

In particular an upper bound for marginalcosts and a lower bound for the markups (zero)can be derived under the assumption of perfectcompetition While this assumption is clearlyunrealistic in the pharmaceuticals market it isuseful in providing an upper bound for costs ci

Uwhich will be given by

(12) ciU pi

On the other hand a lower bound for marginalcosts ci

L (and upper bound for markups) can bederived by assuming that there is perfect collu-sion within each product group i (where i refershere to a moleculedomestic-foreign combina-tion)19 and ignoring price controls so thatprices are determined by the first-order condi-tion of the jointly profit-maximizing firmswithin product group i Solving this first-ordercondition for the (lower bound of) marginal costthen gives

(13) ciL pi 1

1

iipipj

19 The reason that this assumption will lead to an under-statement of marginal costs (and hence overstatement ofmarket power) is that it assumes away competition amongfirms within each drug group

1490 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

where ii(pipj) the own-price elasticity of de-

mand for product i will depend on the productrsquosown-price pi and all other productsrsquo prices pjwith i j

Once we have obtained the demand elastic-ities through estimation of the demand sys-tem we can calculate the upper and lowerbounds for marginal costs and correspondingmarkups according to (12) or (13) These willthen be employed in the counterfactual sim-ulations While our counterfactuals use boththe lower and upper bounds for costs most ofour discussion will be based on using thelower bounds for costs since this is the moreinteresting case in the policy analysis it givesus the largest possible profits for pharmaceuticalfirms and hence corresponds to the worst pos-sible scenario facing Indian firms and the bestpossible scenario facing multinationals underTRIPS

C Identification Assumptions and EstimationApproach

The discussion of the demand system has sofar abstracted from the issue of price endogene-ity The usual premise in the industrial organi-zation literature is that correlation of prices withthe error term in the demand equation arises byvirtue of the first-order conditions of profit-maximizing firms20 As we discuss in this sec-tion this source of simultaneity bias is unlikelyto be of major concern in the present contextbecause of the existence of price regulation Ourprimary concern is instead with the simultane-ity bias that is implied by the particular wayprices are constructed

To understand the sources of the simultaneitybias note first that the price of product group piwe employ in the demand estimation should bethought of as a proxy for an exact price indexthat we do not observe As such pi contains bydefinition measurement error and it will becorrelated with the error term of the demandequation Specifically the demand equation weare interested in estimating can be written insimplified form (suppressing the subscripts rand t and ignoring parameter restrictions forconvenience) as

i i j

ijln pjT ilnXQ

PQ i

where pjT denotes the true (exact) price index for

product group i and i captures unobservedvariables that may affect demand in a particularperiod (eg an advertising campaign or an ep-idemic) The price index pj

T depends on theSKU prices pk Note that in the presence ofSKU entry and exit into the sample the priceindex pj

T will vary over time even if the pricesof the individual SKUs remain stable as theproducts (SKUs) that comprise each productgroup change each period

Ideally we would like to compute the exactprice index pj

T for each group and employ theseindices in the demand estimation Under theassumption of predetermined SKU prices eachperiod the exact price indices would be uncor-related with the error term and the demandparameters could be estimated by simple OLSUnfortunately this strategy is not feasible in thecurrent context

To derive an exact price index for each prod-uct group it is necessary to model consumerrsquoschoice among SKUs conditional on the choiceof the group To this end it is necessary toexplicitly introduce a third stage in the demandspecification and estimate the parameters asso-ciated with the choice at that stage Howeverany specification based on three-stage budget-ing (that would require directly specifying ademand function associated with the SKUchoice at the third stage and imposing additiveseparability at the higher stages) would be in-feasible to estimate for three main reasonsFirst there is a large number of SKUs withinsome groups so that the demand parameters can-not be estimated even if one imposes parameterrestrictions second the SKU prices exhibit littlevariation over time lastly the choice sets at thislowest stage vary over time rendering estimationof the price parameters associated with SKUs thatare not available in all periods infeasible Alterna-tively one could abandon the multistage budget-ing altogether and adopt a discrete choiceapproach which is inherently better suited to deal-ing with varying choice sets and limited pricevariation A discrete choice approach would beparticularly appealing for modeling the SKUchoice conditional on product group choice giventhat within each group all SKUs contain the same

20 See Steven Berry et al (1995) or Aviv Nevo (2001)for a related discussion

1491VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

molecule and are hence comparable As notedearlier however a discrete choice approach at thehigher stages of the demand estimation requiresconverting revenue to physical shares which ischallenging in the case of systemic drugs such asantibiotics for which the appropriate dosage is notwell defined Finally a combination of a multi-stage budgeting approach for the higher stages anda discrete choice approach for the lowest stagewould be feasible to estimate but inconsistentwith a model of utility maximization For thesereasons we have no alternative but to adopt anapproximation for measuring the price index foreach product group

Let us denote this approximation by pjA and let

13j denote the proportional ldquoapproximationrdquo errorassociated with measuring the true price index forgroup j so that ln pj

A ln pjT ln 13j The

estimating demand equation can then be written as

i i j

ijln pjA ilnXQ

PQ

j

ijln 13j i

or

i i j

ijln pjA ilnXQ

PQ i

where the new error term i yenj ijln 13j iis comprised of two components the first onereflects ldquomeasurementrdquo error due to the fact thatthe product group prices we employ are approx-imations and not exact price indices and thesecond one captures conventional demandshocks The demand equation as written abovecorresponds to the equation we take to the data

The particular proxy we use for the exact priceindex for each product group j is the revenue-share-weighted average of the prices of multipleSKUs available within this group that is

pjA

k j

k pk

The notation above illustrates the two source ofprice variation variation in the SKU prices pkand variation in the weights k

We are concerned with potential correlation ofthe so-constructed product group prices with bothcomponents of the product group demand error

term The correlation between pjA and the first

component yenj ijln 13j is independent of theparticular way we construct pj

A and inherent in thefact that we measure the exact price index witherror The potential correlation between pj

A and thesecond component of the product group demanderror the shock i arises however because of thespecific way we construct the proxy pj

A and inparticular because of the presence of the reve-nue share weights k in pj

A since the SKUrevenue share weights k will generally dependon the product group expenditure they arelikely to be correlated with the product groupdemand shock i

To address the simultaneity bias we use in-strumental variables To this end we need vari-ables that are correlated with the proxies pj

A butuncorrelated with the error term i We use thenumber of SKUs within each product group j asan instrument as it does justice to the idea thatvariation in the product group price index stemsin part from variation in the set of SKUs that areavailable each period It is clearly correlatedwith the average group prices pj

A and plausiblyuncorrelated with the demand shock i Theassumption that the number of SKUs is uncor-related with the conventional demand error willhowever be violated if the introduction of anew SKU changes the perceived quality of adrug21 or if it is accompanied by promotionalactivities which will be reflected in i Our hopeis that this is not too often the case Assumingthat the SKU number for each group j is alsouncorrelated with the weighted sum of the mea-surement errors of all product groups yenj ijln 13j itcan be used as an instrument for the product groupprice pj

A22 If one accepts the premise that prices atthe SKU level are exogenous then SKU pricescan also be used as instruments The final list ofinstruments we use hence includes the number ofSKUs in each group the prices of the five largestSKUs for each group and all other exogenousvariables in the demand estimation such as re-gionalproduct dummy interactions and the upper-

21 Note that we already proxy for the quality of each prod-uct group through the product fixed effects i In the actualestimation we actually let these fixed effects be region-specific(ir) so that we allow quality to vary across regions

22 If the measurement error increases in the number ofincluded SKUs this assumption will be violated Unfortu-nately we have no way of checking the validity of thisassumption

1492 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

level total expenditure on antibiotics Regressionsof group prices on the instruments above yieldhigh R-squares with most regressors highly sig-nificant indicating that our instruments are highlycorrelated with prices23

Our sample includes four molecules cipro-floxacin norfloxacin ofloxacin and spar-floxacin Except for sparfloxacin all othermolecules are produced by both foreign anddomestic firms24 So we have seven products(domestic ciprofloxacin foreign ciprofloxa-cin etc) with 96 observations (two years ofmonthly data four geographical regions foreach period) for each product The parametersin the lower-level AIDS demand system asdefined in equation (8) are the product fixedeffects i the productndashregional dummy inter-actions ir the own revenue-share price elas-ticities ii the cross revenue-share priceelasticities i10 i01 i00 and the revenue-share expenditure elasticities i In estimatingthe parameters we first regress prices on allinstrumental variables and then plug the pre-dicted values for prices in the constrainedleast-square regression (For a detailed expla-nation of the constraints see the previoussection)

Given that we impose many cross-equationconstraints and employ instrumental variablesin the estimation it is difficult to derive stan-dard errors for the parameter estimates ana-lytically Our error term interpretation in (8)implies that the error terms for each productgroup are likely to be correlated across re-gions for example if a national advertisingcampaign increases the demand for a rela-tively expensive presentation in one regionsimultaneously increasing the aggregate de-mand and price index for the correspondingproduct group in that region it is likely that thesame effect will be observed in other regions Inprinciple we could exploit the cross-regionalcorrelation in the error terms to estimate thedemand system using seemingly unrelated re-

gression (SUR) This however would requireus to derive the variance-covariance matrix ofthe parameters analytically which as notedabove is cumbersome in the current contextInstead we use the bootstrap method choosingto remain agnostic about the structure of thevariance-covariance matrix The potential dis-advantage of this approach over SUR is a loss inefficiency but as will become apparent in theresults section the model parameters are fairlyprecisely estimated To maintain the marketstructure we randomly sample the periods (withreplacement) and use the same periods for allproducts Regarding the optimal number ofbootstrap repetitions ideally one would followthe three-step method proposed by Donald W KAndrews et al (2000) Empirical evidence sug-gests however that one rarely needs more than200 replications to estimate the standard er-rors25 To be safe we generate 300 bootstrapsamples (with replacement) based on the origi-nal data and estimate the standard errors usingthe standard errors of the bootstrap sampleestimates

The estimation of the top-level AIDS systemis similar The constraints imposed on this top-level demand system are adding-up homogene-ity and symmetry as defined on page 12Again bootstrapping is used to obtain the stan-dard errors of the parameter estimates

D The Counterfactual Scenarios

In assessing the effects of patent enforce-ment we start by focusing on the most extremecase in which compulsory licensing is not anoption and foreign firms are not subject to pricecontrols We use the results from the analysis ofthis case as a benchmark In reality the outcomeof the WTO negotiations is more likely to in-volve some constraints on the monopoly powerof foreign firms selling patented products indeveloping countries such as price caps orcompulsory licensing Our framework can eas-ily accommodate these cases as will becomeapparent in the next subsection

We now focus on the effects of potentialpatent enforcement in the quinolone segmentWe consider several scenarios which vary inthe number and size of domestic products that

23 Specifically the R-squares from the first-stage regres-sions of product group prices on instruments range from057 to 095 except for domestic ciprofloxacin for whichthe R-square is 017

24 Sparfloxacin is actually offered by one foreign subsid-iary in India (see Table 2) However its revenue share isminiscule We therefore treat Sparfloxacin as being pro-duced by domestic firms only 25 See Bradley Efron and Robert J Tibshirani (1993)

1493VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

will be removed from the market In particularwe consider the following five scenarios26

Withdrawal of one large domestic productgroup only domestic ciprofloxacin

Withdrawal of one relatively small domesticproduct group only domestic ofloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand norfloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand sparfloxacin

Withdrawal of all four domestic quinoloneproduct groups

As the list above suggests we proceed fromanalyzing the effects of single-product with-drawal to the analysis of eliminating the en-tire domestic segment This approach wasmotivated by early empirical results that in-dicated that the existence and extent of com-petition from domestic firms has a significantbearing on the predicted effects of patent en-forcement that is our predictions regardingprices and welfare vary substantially depend-ing on how many domestic products are af-fected by patent enforcement and on whethersome domestic competition will remain presentafter TRIPS In addition the size of the affecteddomestic groups is relevant for the welfare pre-dictions accordingly we examine both scenar-ios in which a large domestic product groupsuch as domestic ciprofloxacin is eliminatedfrom the market and scenarios in which the

eliminated domestic group is relatively small(eg domestic ofloxacin) In all scenarios wemaintain the assumption that the set of prod-ucts offered by the remaining firms in themarket does not change in response to patentenforcement

E Computation of Virtual Prices and NewEquilibrium Prices

The first step in the counterfactual analysisis to derive the new equilibrium prices underpatent enforcement27 In this context thereare two sets of prices that are relevant Thefirst set consists of the virtual prices of those(domestic) products that will not be availableonce TRIPS is put in effect To calculate thesevirtual prices we set the expenditures of therelevant products equal to zero The secondset of prices consists of the prices of thoseproducts that remain in the market In deriv-ing these prices we start by assuming profitmaximization without any form of price reg-ulation the firms remaining in the marketreoptimize in response to the policy changeand set new prices taking the prices of allother firms as given28 Of course at the equi-librium all prices change in response to thefact that some domestic products are nolonger present The new equilibrium pricesfor products that remain in the market are thuscomputed by utilizing the first-order condi-tions of profit-maximizing firms into whichthe virtual prices of the eliminated productsare substituted Hence to compute the newequilibrium prices we solve an equation sys-tem of the following form

26 As Table 2 indicates most of the patents for the drugsin the quinolone segment have expired by now so that noneof the scenarios described in this section is going to mate-rialize in practice The reason we consider these alternativescenarios in our counterfactual simulations is to get a senseof how the presence and extent of domestic competitionaffects welfare calculations Such calculations may becomerelevant in the future in other therapeutic classes in whichpatents have not expired yet and where domestic firmscompete with foreign patent holders While the particularexit scenarios will depend on the exact year of patent expiryof drugs in each therapeutic class and the specific way inwhich patents will be enforced (eg whether certain do-mestic drugs will be grandfathered in) the general conclu-sion that emerges from our calculations is that the consumerwelfare loss is substantially smaller in cases where somedomestic competition remains present in the market Hencethe particular way in which patent enforcement is imple-mented is essential for assessing its welfare impact

27 Of course until product patents are in fact introducedthese prices will not be observable Note also that we areassuming here that the range of available products will notchange with the introduction of patents

28 As mentioned above this first set of calculations ab-stracts from the existence of remaining price controls orother government regulations that would impose constraintson the firmsrsquo profit maximization problem Accordingly theresulting numbers should be interpreted as a benchmark ofwhat would happen if markets were completely unregulatedTo examine what the welfare effects of TRIPS would be inthe more realistic scenario of price regulation we subse-quently consider a scenario in which regulation keeps theprices of the products provided by patent-holders at theirpre-TRIPS level For a more detailed discussion see theresults section

1494 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

For products i that are withdrawn from themarket

(14) 0 i ir iiln pirV i10ln pjr j Di

10

j Di

01

i01ln pjr

j Di

00

i00ln pjr

ilnXQr

PQr

For products k that remain in the market

(15) pkr ckr 1 1

kk pkrpjr pirV 1

In the equations above pirV denotes the virtual

prices of the products that are removed fromthe market while pjr denotes the updatedprices of all other products Note that whensolving for the virtual prices we account forthe fact that both the price index for quino-lones PQr and the expenditure allocated tothis subsegment XQr need to be updated toreflect the fact that as a result of the pricechanges there may be substitution away fromthis subsegment To obtain the new quinoloneexpenditure XQr and the new price index PQrwe use the estimates and formulas for thehigher-level AIDS system In equation (15)ck refers to the marginal cost for product k thatwe have obtained from the previous estimationstage As mentioned in the previous subsectionwe conducted the simulations using both theupper and lower bounds for marginal cost ck

u

and ckL respectively The term kk(pkr

pjrpirV) re-

fers to the unconditional own-price elasticity forproduct k which is a function of the eliminatedproductsrsquo virtual prices and the remaining prod-uctsrsquo new equilibrium prices We conduct thecounterfactual simulations and welfare analysisat the regional level because the presence ofregion-specific product effects in the demandestimation implies region-specific demand elas-ticities and hence region-specific marginalcosts and markups The presentation and discus-sion of our results focus on national averages ofthe relevant variables that we construct by com-

puting weighted averages across regions usingthe population of each region as a weight

F Welfare Assessment

The simulation of the new equilibrium underpatent protection can provide important insightsinto how consumers and firms will respond tothe removal of domestic products in the market(for example toward which products consumerswill substitute or which prices will increase themost) To get a more precise idea of how peo-plersquos well-being will ultimately be affected byTRIPS we compute as a last step in our anal-ysis the welfare effects of the policy changeSocial welfare is defined as consumer welfareand the sum of domestic firm profits Thechange in domestic profits can easily be calcu-lated by comparing the domestic firm (variable)profits at the pre-TRIPS prices to the profitsthese firms will realize at the new simulatedprices Although foreign firm profits do notcount in domestic welfare calculations we alsocompute the effects of patent enforcement onforeign firm profits to get an idea of how largethe expected benefits of TRIPS are for thesefirms This provides in some sense an indirectway of assessing whether the claims that patentenforcement in countries like India will lead tomore research on developing-country-specificdiseases (such as malaria) have any validity iffor example we find that the effect of patentenforcement on the foreign firm profits realizedin India is small in magnitude it is unlikely thatforeign firms will engage in more developing-country-specific research in response to TRIPSIt is important to note that in all these calcula-tions we work with the lower-bound estimatesfor marginal costs since these give rise to thehighest possible markups Hence our estimateof profit loss for domestic producers most likelyoverstates this loss

On the consumer side we measure changes inconsumer welfare by the compensating variation(CV) defined as the additional expenditure thatconsumers need in order to achieve the sameutility level as before patent enforcement at thenew prices Specifically let P0 denote the pricevector before patent enforcement P the simulatedprice vector post-TRIPS (that we obtained usingthe methods described in the previous subsection)u0 the utility attained by consumers before TRIPS

1495VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

and E(u P) the higher-level expenditure functionThen the compensating variation is given by

(16) CV Eu0 P Eu0 P0

Note that the CV as computed in (16) representsthe combination of three effects

The pure product-variety effect that is theeffect that arises because one or more prod-ucts are not available to consumers anymoreholding the prices of all other remainingproducts and the total expenditure on thequinolone subsegment XQ constant

The expenditure-switching effect that is theeffect arising from substitution away fromquinolones and toward other subsegments ofthe antibiotics market again holding the pricesof all other remaining products constant

The reduced-competition effect that is theeffect that arises because the firms remainingin the market adjust (increase) their prices inresponse to the removal of domestic products

From both an analytical and a policy point ofview it is desirable to assess how large each ofthese effects is Accordingly we decompose thetotal effect on consumer welfare (the CV as givenby equation (16)) using the following procedure

To get the pure product variety effect wecompute virtual prices for the products that areremoved from the market holding the quino-lone expenditure XQ and the prices of all otherproducts fixed Let us call the resulting pricevector P1 Then the pure product variety effectis represented by E(u0 P1) E(u0 P0)

To compute the expenditure-switching effectwe compute another set of virtual prices againholding the prices of all remaining products fixedbut letting quinolone expenditure adjust in re-sponse to the new price index for the quinolonesegment (Given that the prices of the remainingproducts remain fixed the change in the priceindex arises only because of the removal of one ormore domestic products) Note that this scenariomost closely resembles the case in which patentlaws are enforced but strict price regulation keepsthe prices of the products offered in the market attheir pre-TRIP level Let us label the so-computedprice vector P2 The expenditure-switching effectis then E(u0 P2) E(u0 P1)

Finally the reduced competition effect aris-ing from higher prices for the remaining prod-

ucts is computed as the residual change in thecompensating variation once the product-varietyand expenditure-switching effects have been ac-counted for that is E(u0 P) E(u0 P2)where the price vector P is computed accordingto the formulas (14) and (15) to reflect theadjustment of prices to the new regime

To compute the standard errors associatedwith the counterfactual simulations (that is thestandard errors for the counterfactual prices andwelfare estimates) we again use bootstrappingWe first bootstrap the original sample (sales andprices of all the drugs) 300 times Next we esti-mate the AIDS model for each of these samplesand compute the counterfactual equilibrium pricesand welfare losses corresponding to each policyscenario for each of the 300 simulations In thefinal step we compute standard errors29

One limitation of our framework is that itdoes not allow for a heterogeneous response ofconsumers to the policy change Accordinglyour framework is not suited to addressing thequestion of how different groups in the popula-tion will be affected by patent law enforce-ment30 Along the same lines our frameworkdoes not accommodate the possibility of pricediscrimination which might lead to differentresults regarding welfare losses and profit gainsrelative to uniform pricing31 Still we believethat the results of the counterfactual simulationscan provide important insights into the likelyaggregate response to patent enforcement andthe factors that drive this response

III Results

A The Structure of Demand

Tables A2 and A3 in the Appendix display theresults from estimation of the lower- and upper-level AIDS systems respectively For ease of in-terpretation rather than discussing the coefficientestimates we focus our discussion on the implied

29 In cases some of the simulations do not converge wecompute the standard errors using only those simulationsthat converged

30 Addressing this question would require at a mini-mum micro data on consumer purchases as in Goldberg(1995) To our knowledge such data do not exist for theIndian pharmaceutical market

31 See Ernst R Berndt (1994) for an extensive discussionof uniform pricing

1496 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

unconditional price and expenditure elasticities re-ported in Table 6 Given that the region-specificproduct group effects ir reported in the last fourcolumns of Tables A2 and A3 imply region-spe-cific demand elasticities we report separate elas-ticities for each region in Tables 6A and 6B Asevident from the comparison of these tables theelasticities are very similar across regions so thatwe can focus the remaining discussion on theelasticities of one region only the northern regiondisplayed in Table 6A

The diagonal terms of Table 6A report theown-price elasticities which are in all but onecase negative and highly significant The oneexception is the foreign norfloxacin productgroupmdashwhose share of quinolone sales is 007percentmdashfor which we estimate a negative butinsignificant own-price elasticity For the remain-ing product groups demand appears to be highlyelastic with the estimated elasticities lower than2 in four out of the six cases The magnitude ofthe own-price elasticities matches the features ofthe Indian pharmaceutical market mentioned ear-lier which would suggest that Indian consumersare likely to be quite price-sensitive32 The elas-ticities appear especially large if one takes intoaccount that they refer to product groups (such asdomestic ciprofloxacin) and not individual drugsoffered by particular firms Their relative magni-tudes are also intuitive the drug with the largestmarket share and a relatively high price (domesticciprofloxacin) appears to be one of the least elas-tic In contrast foreign ciprofloxacin is highlyprice elastic this is plausible as ciprofloxacindrugs offered by subsidiaries of multinationalsface the stiffest competition in this market seg-ment from approximately 75 Indian firms offeringthe same molecule

The estimated expenditure elasticities appear inthe last column These are all positive indicatingthat the demand for all product groups is normalThe remaining cells display the estimated cross-price elasticities As one might perhaps expect forproducts within a therapeutic subsegment theseare mostly positive Out of a total of 49 priceelasticities we estimate there are six that do notconform to expectations these are the cross-price elasticities between different foreign prod-

uct groups which are estimated to be negativeand significant33 Fortunately these elasticitieshave negligible impact on the welfare analysisgiven that our counterfactuals focus on theeffect of withdrawing one or more domesticproducts from the market the most relevantelasticities are the ones that capture the responseof various product group shares to a change in theprice of one or more domestic groups these elas-ticities are the cross-price elasticities between var-ious domestic groups and the ones betweendomestic and foreign groups which are plausibleand precisely estimated34

Regarding these elasticities a striking aspectof our estimates is how large positive andsignificant the cross-price elasticities betweendifferent domestic product groups aremdashin factfor norfloxacin and ofloxacin we estimate thatdomestic product groups containing differentmolecules are closer substitutes for one anotherthan product groups that contain the same mol-ecule but are produced by foreign firms Incontrast for ciprofloxacin (the molecule withthe largest revenue share) we estimate a largepositive cross-price elasticity between the do-mestic and foreign versions

The fact that domestic products appear to beclose substitutes for other domestic productsthat contain different molecules truly representsan ldquoempiricalrdquo finding in the sense that we do

32 In developed economies elasticities of this magnitude havetypically been found only for generic drugs (and even then onlyrarely) or among consumers who lack health insurance

33 While these elasticities are clearly counterintuitive theyare not inconsistent with the underlying demand system whichimposes no restrictions on the sign of the cross-price elastici-ties We do not have a good explanation of why these elastic-ities are estimated to be negative A potential explanation isthat the shares of the foreign products are very small giventhis we observe very few consumers switching from one (verysmall) foreign group to another (very small) foreign groupwhen the price of the first foreign group goes up and hence theinference is not very reliable in this case

34 The cross-price elasticities between foreign drug groupscontaining different molecules will also have an effect on thewelfare estimates given that the withdrawal of domestic prod-ucts will generally lead to changes in the prices of all foreignproducts and market shares will be reallocated from someforeign products to other foreign products based on the newprices However this reallocation from ldquoforeign to foreignrdquo istruly second order in our case compared to the reallocationfrom ldquodomestic to domesticrdquo and ldquodomestic to foreignrdquo prod-ucts In scenarios in which we consider patent enforcementaccompanied by strict price regulation the cross-price elastic-ities between foreign product groups are in fact completelyirrelevant as the prices of foreign products are not allowed toincrease in these cases Still our welfare loss estimates remainsubstantial drivenmdashas beforemdashby the loss of product variety

1497VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

not impose it through any of our assumptionsregarding the demand function The questionthat naturally arises then is what might explainthis finding While we cannot formally addressthis question anecdotal accounts in various in-dustry studies suggest that the explanation maylie in the differences between domestic andforeign firms in the structure and coverage ofretail distribution networks

Distribution networks for pharmaceuticals inIndia are typically organized in a hierarchicalfashion Pharmaceutical companies deal mainlywith carrying and forwarding (CampF) agents inmany instances regionally based who each sup-ply a network of stockists (wholesalers) Thesestockists in turn deal with the retail pharma-cists through whom retail sales ultimately oc-cur35 The market share enjoyed by a particularpharmaceutical product therefore depends inpart on the number of retail pharmacists who

stock the product And it is here that thereappears to be a distinction between domesticfirms and multinational subsidiaries In particu-lar the retail reach of domestic firms as agroup tends to be much more comprehensivethan that of multinational subsidiaries (IndianCredit Rating Agency (ICRA) 1999)36

There appear to be two reasons for this Thefirst is that many of the larger Indian firmsbecause they have a much larger portfolio ofproducts over which to spread the associatedfixed costs typically have more extensive net-works of medical representatives The second issimply that there are many more domestic firms(and products) on the market At the retail levelthis would imply that local pharmacists mightbe more likely to stock domestic products con-taining two different molecules say ciprofloxa-cin and norfloxacin than they would domesticand foreign versions of the same molecule Tothe extent that patients (or their doctors) arewilling to substitute across molecules in order tosave on transport or search costs (eg going toanother pharmacy to check whether a particular

35 There are estimated to be some 300000 retail pharma-cists in India On average stockists deal with about 75 retailers(ICRA 1999) There are naturally variations in this structureand a host of specific exclusive dealing and other arrangementsexists in practice Pharmaceutical firms also maintain networksof medical representatives whose main function is to marketthe companyrsquos products to doctors who do the actual prescrib-ing of drugs In some instances firms do sell directly to thedoctors who then become the ldquoretailerrdquo as far as patients areconcerned but these are relatively rare

36 These differences were also highlighted in conversa-tions one of the authors had with CEOs and managingdirectors of several pharmaceutical firms as part of a sep-arate study

TABLE 6AmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN THE NORTHERN REGION

Product group

Elasticity with respect to

Prices of foreign productgroups Prices of domestic product groups Overall

quinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 557 013dagger 015 401 011dagger 011dagger 016 137(179) (007) (007) (184) (006) (006) (006) (029)

Foreign norfloxacin 427dagger 045 427dagger 350dagger 602 451 465 220(242) (112) (242) (210) (623) (184) (183) (105)

Foreign ofloxacin 011 010dagger 138 009 009dagger 023 011 116(005) (005) (031) (027) (005) (028) (004) (017)

Domestic ciprofloxacin 018 001 001 168 008 008 010 117(008) (000) (001) (023) (002) (002) (002) (003)

Domestic norfloxacin 004 003 004 058 223 042 040 073(001) (003) (001) (017) (011) (004) (003) (009)

Domestic ofloxacin 005 005 011 077 074 342 074 089(002) (002) (013) (028) (008) (025) (008) (021)

Domestic sparfloxacin 007 004 007 115 063 063 288 028(002) (001) (002) (015) (006) (006) (017) (012)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent significance level and dagger (dagger) denotes significance at the 10-percent level

1498 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE 6BmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN OTHER REGIONS

Demand patterns in the eastern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 594 014dagger 016 431 013 011 017 140(195) (008) (008) (201) (006) (007) (007) (031)

Foreign norfloxacin 329dagger 058 329dagger 264dagger 460 346 359 192(180) (083) (179) (153) (451) (138) (139) (082)

Foreign ofloxacin 012 010dagger 140 010 010 024 013 117(006) (006) (034) (031) (005) (030) (005) (019)

Domestic ciprofloxacin 018 001 001 172 009 008 011 117(008) (000) (002) (027) (002) (003) (002) (003)

Domestic norfloxacin 004 004 004 068 242 049 046 070(002) (004) (002) (020) (013) (005) (004) (010)

Domestic ofloxacin 004dagger 004 009 061 060 295 060 092(002) (001) (010) (028) (007) (022) (007) (017)

Domestic sparfloxacin 005 003 005 092 048 051 251 043(002) (001) (001) (017) (005) (006) (016) (010)

Demand patterns in the western region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 427 009dagger 011 286 008 006 011 126(129) (005) (005) (133) (004) (005) (004) (021)

Foreign norfloxacin 714dagger 008 710dagger 594dagger 995 742 770 299dagger

(369) (183) (368) (328) (1015) (297) (296) (178)Foreign ofloxacin 013 012dagger 145 009 012 026 013 119

(007) (006) (037) (029) (005) (033) (005) (021)Domestic ciprofloxacin 019 001 000 174 009 008 011 118

(008) (000) (001) (024) (002) (003) (002) (004)Domestic norfloxacin 004 004 004 067 243 050 047 069

(002) (004) (001) (018) (014) (005) (004) (010)Domestic ofloxacin 003 003 007 048 048 257 048 093

(002) (001) (008) (022) (005) (015) (004) (014)Domestic sparfloxacin 006 003 005 083 046 049 241 046

(002) (001) (001) (014) (004) (005) (012) (009)

Demand patterns in the southern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 560 013dagger 015 402 011 010 017 137(182) (007) (007) (187) (006) (006) (007) (029)

Foreign norfloxacin 533dagger 031 531dagger 431dagger 748 559 584 249dagger

(301) (134) (300) (245) (734) (219) (219) (128)Foreign ofloxacin 014 012dagger 148 011 012 029 016 120

(007) (007) (039) (033) (006) (035) (006) (022)Domestic ciprofloxacin 017 001 000 169 008 007 011 116

(008) (000) (001) (025) (002) (003) (002) (003)Domestic norfloxacin 004 003 003 059 225 043 040 073

(001) (003) (001) (019) (012) (004) (004) (009)Domestic ofloxacin 004dagger 004 009 062 060 296 060 091

(002) (001) (010) (026) (006) (019) (007) (017)Domestic sparfloxacin 008 005 008 138 073 075 323 016

(002) (001) (002) (018) (007) (008) (021) (014)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at 5-percent level anddagger (dagger) denotes significance at the 10-percent level

1499VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

foreign product is in stock) in aggregate datawe would expect to find precisely the substitu-tion patterns that we report in Table 6

Whether the particular explanation we provideabove is the correct one the high degree of sub-stitutability between domestic product groupsturns out to have important implications for thewelfare calculations We discuss these in moredetail below when we present the results of thecounterfactual welfare analysis Another elasticitywith important implications for the counterfactu-als is the price elasticity for the quinolone subseg-ment as a whole which indicates how likelyconsumers are to switch to other antibioticsgroups when faced with a price increase for quin-olones This elasticity is computed on the basis ofthe results in Table A3 and it is at 111 (stan-dard error 024) this is large in magnitudebutmdashas expectedmdashsmaller in absolute value thanthe own-price elasticities of the product groupswithin the quinolone subsegment

The results in Tables 6A and 6B are based onour preferred specification discussed in SectionII In Tables A4 to A6 in the Appendix weexperiment with some alternative specificationsTables A4(a)ndashA4(c) correspond to a specificationthat includes in addition to product-group-specificregional fixed effects product-group-specific (andfor the upper level antibiotics-segment-specific)seasonal effects We distinguish among three sea-sonsmdashthe summer monsoon and wintermdashandreport the unconditional demand elasticities for

the northern region for each of these seasons Asevident from the tables in the Appendix our elas-ticity estimates are robust to the inclusion of sea-sonal effects The demand elasticities in Table A5are based on estimation of the demand system byOLS Compared to the elasticities obtained by IVthe OLS elasticities are smaller in absolute valueimplying that welfare calculations based on theOLS estimates would produce larger welfare lossestimates Nevertheless some of the patterns re-garding the cross-price elasticities discussed ear-lier are also evident in the OLS results inparticular the cross-price elasticities between dif-ferent domestic product groups are all positivelarge and significant and in most instances largerthan the cross-price elasticities between drugs thatcontain the same molecule but are produced byfirms of different domesticforeign status Theclose substitutability of domestic products indi-cated by both the OLS and IV estimates seems tobe one of the most robust findings of the paper

B Cost and Markup Estimates

Table 7 displays the marginal costs markupsand profits implied by the price elasticity esti-mates of Tables 6A and 6B for each of the sevenproduct groups Given that our regional effectsimply different price elasticities for each regionour marginal cost and markup estimates alsodiffer by region Given however that based on

TABLE 7mdashUPPER AND LOWER BOUNDS FOR MARGINAL COST MARKUP AND ANNUAL PROFIT BY PRODUCT GROUPS WITHIN

THE QUINOLONE SUBSEGMENT

Product groupLower boundfor MC (Rs)

Upper boundfor markup

Upper boundfor profit(Rs mill)

Upper boundfor MC (Rs)

Lower boundfor markup

Lower boundfor profit

(Rs)

Foreign ciprofloxacin 83 19 269 103 0 00(123) (012) (1655)

Foreign norfloxacin NA NA NA 53 0 00Foreign ofloxacin 323 70 1061 1085 0 00

(2316) (021) (3185)Domestic ciprofloxacin 47 59 17019 112 0 00

(114) (010) (29858)Domestic norfloxacin 52 43 2807 90 0 00

(020) (002) (1532)Domestic ofloxacin 587 34 1612 901 0 00

(218) (002) (1280)Domestic sparfloxacin 495 37 1985 788 0 00

(157) (002) (1100)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level Estimated lower bound forforeign norfloxacinrsquos marginal cost is negative since the estimated price elasticity is less than one in absolute value

1500 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Tables 6A and 6B the price elasticities do notseem to differ substantially across regions wereport for ease of exposition only the nationalaverages for marginal costs and markups inTable 7

Table 7 has two parts In the left part (firstthree columns) we report the lower bound formarginal cost and the corresponding upperbound for markup and upper bound for totalannual profit for each product group Thesenumbers are based on the price elasticities weobtained from estimating the two-level AIDSdemand system Since we do not have a reliableestimate for the price elasticity of foreign nor-floxacin (the point estimate is negative but lessthan one in absolute value and insignificant)we cannot compute the lower bound for mar-ginal cost in this case It is important to note thatthese estimates do not reflect either the actualmarginal cost or the actual markup for thesedrugs both because the existence of price reg-ulation implies that the unconstrained first-orderconditions are unlikely to hold each period andbecause our aggregation across firms of thesame domesticforeign status supplying thesame molecule makes the interpretation of theseestimates problematic In particular the fact thatwe ignore competition among firms within eachproduct group implies that our estimates willtend to overstate market power These numberswill prove useful however in the counterfac-tual simulations as they can give us a sense ofhow large the maximum profit gains for multi-national firms and the maximum profit lossesfor domestic firms are likely to be under patentenforcement The right part of the table (lastthree columns) reports the upper bound for mar-ginal cost which we obtain by simply taking themarginal cost to be equal to the observed priceWe do not report standard errors in this casesince the numbers are based on actual data Themarkups corresponding to these marginal costupper bounds are of course zero We conductthe counterfactual simulations using both thelower and upper bounds for marginal costs

The (upper bounds for) markups on the leftside of the table are generally plausible Thedomestic ciprofloxacin product group that dom-inates the quinolone subsegment and for whichwe documented high prices a high marketshare and a relatively low elasticity of demandenjoys one of the highest markup upper bounds(60 percent) and accounts for nearly 70 percent

of all profits derived within the subsegmentForeign ciprofloxacin on the other hand whichfaces the stiffest competition from domesticfirms and for which we estimated a highly elas-tic demand has the lowest markup upper bound(19 percent)

C Counterfactual Estimates of the Impact onPrices and Welfare

With estimates of the key demand and costparameters in hand we turn to the counterfac-tuals We consider the five separate scenarioslisted in the previous section All of the scenar-ios involve the withdrawal of one or more of thedomestic product groups from the market Ta-ble 8 displays our estimates of the consumerwelfare losses that result under the differentscenarios The losses are expressed in billion ofrupees per year All numbers presented in Ta-ble 8 and subsequent tables are based on usingthe lower bounds for marginal cost and upperbounds for markup in the simulations As dis-cussed earlier these numbers are the more in-teresting to work with since they give us anupper bound for the changes in the profits ofdomestic and foreign firms that would resultfrom patent enforcement In the Appendix wealso present results based on using the upperbounds for marginal costs in the simulations inwhich case the pre-TRIPS profits of domesticand foreign firms are zero In all cases marginalcosts are assumed to be constant in outputWhile naturally the profit implications differdepending on whether one uses the upper orlower bounds for marginal costs (firm profits arezero if one assumes the upper bound of mar-ginal cost in which case price equals marginalcost) the estimated consumer welfare losses aresimilar in the two cases We discuss these re-sults in more detail at the end of this section

The first column presents our estimates of theconsumer welfare losses attributable to the pureloss of product variety effect where we fix theprices of all remaining products as well as theoverall expenditure on quinolones while with-drawing one or more of the domestic productgroups Note that in our initial specification ofthe demand system had we not allowed for thepossibility that consumers might differentiatebetween domestic and foreign products evenwhen they contain the same molecule this

1501VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

particular component of the loss of consumerwelfare would not have arisen

The estimates reported in the second columnincorporate the expenditure-switching effect ontop of the loss of product variety Here basedupon the price elasticity estimates from thehigher-level AIDS system we adjust (down-ward) the expenditures allocated to the quino-lone subsegment as the composite price ofquinolones effectively increases as a consequenceof the higher virtual prices of the domestic productgroups that are withdrawn from the market Be-cause the estimates in this column are generatedassuming that the prices of the products that re-main in the market are not adjusted upward theyprovide a sense of what consumer losses would beif the introduction of product patents were coupledwith strict price regulation aimed at maintainingprices at prepatent levels Alternatively they canbe thought of as the relevant welfare numbers ifintense competition among firms within the re-maining product groups kept the prices of theproducts that were still offered in the market closeto the firmsrsquo marginal costs

The last column of Table 8 displays the esti-mated consumer welfare losses when bothcross-segment expenditure-switching and with-in-segment upward price adjustments are takeninto account

If we compare the results across the firstsecond and third columns all the counterfac-tual scenarios produce qualitatively similar pat-terns patterns that are consistent with what we

would expect Starting from the initial loss ofwelfare attributable to the loss of product vari-ety the option of switching expenditures outof the quinolone subsegment to other subseg-ments mitigates some of the initial welfareloss But if we then incorporate the upwardprice adjustments that result in response to thereduced competition the welfare losses aremagnified

Of particular interest from a policy perspec-tive are the relative magnitudes of these threeeffects which are similar under all the counter-factual scenarios though absolute levels varyconsiderably First despite the fact that the de-mand for quinolones is quite sensitive to thecomposite price of quinolonesmdashthe upper levelprice elasticity is 111mdashthe cross-subsegmentexpenditure-switching effects are in all thecases small (in absolute value terms) relative tothe other two effects For instance under thescenario where all the domestic quinolone prod-uct groups are withdrawn from the market theoverall consumer welfare loss of Rs 1781 bil-lion per year can be decomposed into an initialloss of Rs 1176 billion (66 percent) attribut-able to the loss of product variety a slightreduction in this initial loss of Rs 041 billion(2 percent) from Rs 1176 billion to Rs1135 billion because of expenditure switchingand a subsequent additional loss of Rs 646billion (36 percent) from Rs 1135 billion toRs 1781 billion because of the reduced com-petition and consequent price increases

TABLE 8mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure loss ofvariety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 732(087) (089) (146)

Only ofloxacin 008 008 023(008) (008) (010)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1253(177) (180) (415)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1058(142) (145) (331)

All four domestic quinolones products 1176dagger 1135dagger 1781(643) (634) (1270)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent significance level and dagger (dagger)denotes significance at the 10-percent level

1502 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 12: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

From an empirical point of view this is notparticularly useful as it eliminates all compu-tational advantages of the two-stage approachTo address this problem we adopt an approxi-mate solution to model the higher-level expen-diture allocation along the lines suggested byDeaton and Muellbauer (1980b pp 131ndash32)This gives rise to a two-level AIDS specification

Consider the lower-level estimation firstwhich refers to the allocation of a particularsubsegmentrsquos expenditure to the product groupswithin the subsegment In our application therelevant subsegment is quinolones which weindex with Q Let the product groups within thissubsegment be indexed by i 1 N pi be theprice of product group i (where as noted abovei refers to a particular molecule and domesticforeign status combination) and XQ be the totalexpenditure on the quinolone segment The rev-enue share of each product group is given by

(2) i i j

ijln pj ilnXQ

PQ

where i the revenue share of product group iis defined as

(3) i pi qi

yenj

pj qj

xi

XQ with i j Q

XQ is the overall expenditure on the quinolonesubsegment and PQ is a price index given by

(4) ln PQ ap 0 i

iln pi

1

2 i

j

ijln piln pj

With a limited number of product groups and asufficiently large number of time-series observa-tions the flexibility implied by the AIDS modeldoes not impose too many demands on the data Inthe present application however where the num-ber of observations is limited the AIDS model isnot estimable in this general form To reduce thenumber of parameters that need to be estimatedwe impose two sets of restrictions

The first set of restrictions is implied by thetheory of utility maximization Specificallythese restrictions are

Adding-up yenk k 1 yenk k 0 yenk kj 0 j

Homogeneity yenk jk 0 j Symmetry ij 1frasl2 [ij ji] ji This last

restriction by itself reduces the number of parameters to [N(N 1)]2

The second set of restrictions we impose aimsat further reducing the number of parameters tobe estimated by exploiting our knowledge of thisparticular market Specifically for each productgroup i we allow one ij parameter for all productgroups j that have different molecules from prod-uct group i and are produced by foreign firms andone ij for all product groups j with differentmolecules produced by domestic firms We donrsquotimpose any restrictions on the ij parameter whenproduct group j has the same molecule as productgroup i (By construction product groups i and jcontain products produced by firms with differentdomesticforeign status)

To better illustrate the nature of the restric-tions we impose on the patterns of substitutionacross products some additional notation isneeded Let d(i j) be an indicator of the degreeof similarity (or difference) between productgroup i and product group j along the dimen-sions we are able to observe (molecule M anddomesticforeign status DF) For any two prod-uct groups i and j d(i j) can take on one of thefollowing three values17

(5)

di j (1 0) if Mi Mj DFi DFj

(0 1) if Mi Mj DFi DFj

(0 0) if Mi Mj DFi DFj

Let

(6) Diab j di j a b

17 The sequence (1 1) is not possible for two differentproducts in this case the parameter corresponds to theproductrsquos own-price effect that is ii

1488 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The equation at the lower level becomes

(7) i i iiln pi i10ln pj j Di10

j Di

01

i01ln pj

j Di

00

i00ln pj ilnXQ

PQ

Note that

The parameter ii captures a product grouprsquosown-price effect (there will be as many iiparameters as the number of product groups)

The parameter i10 captures the cross-priceeffect of the product group containing prod-ucts with the same molecule but produced byfirms of different nationality

The parameter i01 captures the cross-priceeffects of product groups containing productswith different molecules but produced byfirms with the same nationality

The parameter i00 captures the cross-priceeffects of product groups containing productswith different molecules produced by firms ofdifferent nationality

Before we take the demand equation to thedata we make two modifications The first oneis to let the product-specific effects i vary byregion r The resulting product-specific regionaleffects ir have two interpretations first theycontrol for the ldquoqualityrdquo of each drug withquality differences being allowed to vary acrossregions second they proxy for demographicsand other demand shifters which vary by re-gion and may affect the demand of each prod-uct group differently18 Note that by includingproduct-specific regional effects in the demandspecification we estimate the price parametersbased on the within-product-group variation ofprices in each region In an earlier version of thepaper we also estimated the demand systemwithout regional dummies and obtained similarresults

The second modification that allows us to gofrom a deterministic to a stochastic specificationof the demand equation is to include an additiveerror term in (7) The latter takes into accountthe fact that (7) is not expected to fit the dataexactly The error term irt accounts for mea-surement error (due to the fact that the productgroup prices pjrt we employ in the estimationare not exact price indices but approximationsthereof) and (potentially region-specific) de-mand shocks that may affect the demand for aproduct in a particular period Examples of suchshocks include an advertising campaign for aparticular product which temporarily increasesthe demand for this product or the outbreak ofa (potentially region-specific) epidemic whichcalls for the use of a particular drug We discussthe interpretation and properties of this errorterm in more detail in the next section

The final form of the equation we estimate atthe lower level becomes (with subscript t de-noting month and subscript r denoting region)

(8) irt i ir iiln pirt i10ln pjrt j Di10

j Di

01

i01ln pjrt

j Di

00

i00ln pjrt

ilnXQrt

PQrt irt

The analysis so far has conditioned on the ex-penditure allocated to the quinolone subseg-ment XQ The upper level of the estimationconsiders the problem of allocating total expen-diture across the different systemic antibioticssubsegments one of which is quinolones Theupper-level demand function is given by

(9) G G H

GHln PH GlnX

P

where all variables denoted by capital letters aredefined as before but now refer to subsegments(G H ) rather than product groups within asubsegment and the total expenditures on sys-temic antibiotics X are deflated by the Stone priceindex log P yenH Hlog PH When estimating thesystem above we impose all the restrictions

18 Given the short time span of our sample typicaldemand shifters such as age distribution income distribu-tion and education hardly change over our sample periodSuch shifters are therefore absorbed by the region-specificproduct fixed effects ir

1489VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

implied by utility maximization as we do withthe estimation of the lower-level AIDS We donot however impose any additional restrictionson the substitution patterns at this stage so thatthe cross-price effects across segments remainrelatively unconstrained

Estimation of the higher-level AIDS allowsus to obtain the unconditional own- and cross-price elasticities that are used in the formulationof the supply problem and welfare analysisThese are given by the formula

(10) ij ijXQXQ

ln qi

ln XQ

ln XQ

ln PQ

ln PQ

ln pj

with the conditional cross price elasticitiesgiven by

ijXQXQ ij ij jlnXQ

PQ

i

As in the lower stage we include subsegment-specific regional dummies in the specificationof the upper-stage demand system so that thefinal form of the estimating equation at theupper stage becomes

(11) Grt G Gr H

GHln PHrt

GlnXrt

Prt Grt

In sum the demand system we take to thedata is represented by equations (8) and (11)and the associated parameter restrictions im-plied by economic theory

B Modelling the Supply Side of the Market

Counterfactual simulations concerning theeffects of domestic product withdrawal requireknowledge of the marginal costs of pharmaceu-tical firms operating in the Indian market Theseare unobservable The usual approach in theNew Empirical Industrial Organization litera-ture has been to exploit the firm equilibriumconditions to infer marginal cost For exampleit is often assumed that the marginal cost ci isconstant and that the industry is an oligopoly

engaging in Bertrand competition with differen-tiated products Assuming that firms myopicallymaximize profits each period one can then de-rive the firmsrsquo first-order conditions that corre-spond to the assumptions above about costsmarket structure and firm behavior

We deviate from this procedure as the presenceof price regulation renders the assumption of un-constrained period-by-period maximization un-tenable Ideally one would like to explicitlyincorporate the price and other administrative con-trols into the firmrsquos optimization problem andderive the first-order conditions under the assump-tion of constrained maximization The complexityof the price regulation however makes this ap-proach infeasible Therefore we adopt an alterna-tive approach that does not rely on modelling theprice-setting process but is instead based on de-riving upper and lower bounds for marginal costsand markups

In particular an upper bound for marginalcosts and a lower bound for the markups (zero)can be derived under the assumption of perfectcompetition While this assumption is clearlyunrealistic in the pharmaceuticals market it isuseful in providing an upper bound for costs ci

Uwhich will be given by

(12) ciU pi

On the other hand a lower bound for marginalcosts ci

L (and upper bound for markups) can bederived by assuming that there is perfect collu-sion within each product group i (where i refershere to a moleculedomestic-foreign combina-tion)19 and ignoring price controls so thatprices are determined by the first-order condi-tion of the jointly profit-maximizing firmswithin product group i Solving this first-ordercondition for the (lower bound of) marginal costthen gives

(13) ciL pi 1

1

iipipj

19 The reason that this assumption will lead to an under-statement of marginal costs (and hence overstatement ofmarket power) is that it assumes away competition amongfirms within each drug group

1490 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

where ii(pipj) the own-price elasticity of de-

mand for product i will depend on the productrsquosown-price pi and all other productsrsquo prices pjwith i j

Once we have obtained the demand elastic-ities through estimation of the demand sys-tem we can calculate the upper and lowerbounds for marginal costs and correspondingmarkups according to (12) or (13) These willthen be employed in the counterfactual sim-ulations While our counterfactuals use boththe lower and upper bounds for costs most ofour discussion will be based on using thelower bounds for costs since this is the moreinteresting case in the policy analysis it givesus the largest possible profits for pharmaceuticalfirms and hence corresponds to the worst pos-sible scenario facing Indian firms and the bestpossible scenario facing multinationals underTRIPS

C Identification Assumptions and EstimationApproach

The discussion of the demand system has sofar abstracted from the issue of price endogene-ity The usual premise in the industrial organi-zation literature is that correlation of prices withthe error term in the demand equation arises byvirtue of the first-order conditions of profit-maximizing firms20 As we discuss in this sec-tion this source of simultaneity bias is unlikelyto be of major concern in the present contextbecause of the existence of price regulation Ourprimary concern is instead with the simultane-ity bias that is implied by the particular wayprices are constructed

To understand the sources of the simultaneitybias note first that the price of product group piwe employ in the demand estimation should bethought of as a proxy for an exact price indexthat we do not observe As such pi contains bydefinition measurement error and it will becorrelated with the error term of the demandequation Specifically the demand equation weare interested in estimating can be written insimplified form (suppressing the subscripts rand t and ignoring parameter restrictions forconvenience) as

i i j

ijln pjT ilnXQ

PQ i

where pjT denotes the true (exact) price index for

product group i and i captures unobservedvariables that may affect demand in a particularperiod (eg an advertising campaign or an ep-idemic) The price index pj

T depends on theSKU prices pk Note that in the presence ofSKU entry and exit into the sample the priceindex pj

T will vary over time even if the pricesof the individual SKUs remain stable as theproducts (SKUs) that comprise each productgroup change each period

Ideally we would like to compute the exactprice index pj

T for each group and employ theseindices in the demand estimation Under theassumption of predetermined SKU prices eachperiod the exact price indices would be uncor-related with the error term and the demandparameters could be estimated by simple OLSUnfortunately this strategy is not feasible in thecurrent context

To derive an exact price index for each prod-uct group it is necessary to model consumerrsquoschoice among SKUs conditional on the choiceof the group To this end it is necessary toexplicitly introduce a third stage in the demandspecification and estimate the parameters asso-ciated with the choice at that stage Howeverany specification based on three-stage budget-ing (that would require directly specifying ademand function associated with the SKUchoice at the third stage and imposing additiveseparability at the higher stages) would be in-feasible to estimate for three main reasonsFirst there is a large number of SKUs withinsome groups so that the demand parameters can-not be estimated even if one imposes parameterrestrictions second the SKU prices exhibit littlevariation over time lastly the choice sets at thislowest stage vary over time rendering estimationof the price parameters associated with SKUs thatare not available in all periods infeasible Alterna-tively one could abandon the multistage budget-ing altogether and adopt a discrete choiceapproach which is inherently better suited to deal-ing with varying choice sets and limited pricevariation A discrete choice approach would beparticularly appealing for modeling the SKUchoice conditional on product group choice giventhat within each group all SKUs contain the same

20 See Steven Berry et al (1995) or Aviv Nevo (2001)for a related discussion

1491VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

molecule and are hence comparable As notedearlier however a discrete choice approach at thehigher stages of the demand estimation requiresconverting revenue to physical shares which ischallenging in the case of systemic drugs such asantibiotics for which the appropriate dosage is notwell defined Finally a combination of a multi-stage budgeting approach for the higher stages anda discrete choice approach for the lowest stagewould be feasible to estimate but inconsistentwith a model of utility maximization For thesereasons we have no alternative but to adopt anapproximation for measuring the price index foreach product group

Let us denote this approximation by pjA and let

13j denote the proportional ldquoapproximationrdquo errorassociated with measuring the true price index forgroup j so that ln pj

A ln pjT ln 13j The

estimating demand equation can then be written as

i i j

ijln pjA ilnXQ

PQ

j

ijln 13j i

or

i i j

ijln pjA ilnXQ

PQ i

where the new error term i yenj ijln 13j iis comprised of two components the first onereflects ldquomeasurementrdquo error due to the fact thatthe product group prices we employ are approx-imations and not exact price indices and thesecond one captures conventional demandshocks The demand equation as written abovecorresponds to the equation we take to the data

The particular proxy we use for the exact priceindex for each product group j is the revenue-share-weighted average of the prices of multipleSKUs available within this group that is

pjA

k j

k pk

The notation above illustrates the two source ofprice variation variation in the SKU prices pkand variation in the weights k

We are concerned with potential correlation ofthe so-constructed product group prices with bothcomponents of the product group demand error

term The correlation between pjA and the first

component yenj ijln 13j is independent of theparticular way we construct pj

A and inherent in thefact that we measure the exact price index witherror The potential correlation between pj

A and thesecond component of the product group demanderror the shock i arises however because of thespecific way we construct the proxy pj

A and inparticular because of the presence of the reve-nue share weights k in pj

A since the SKUrevenue share weights k will generally dependon the product group expenditure they arelikely to be correlated with the product groupdemand shock i

To address the simultaneity bias we use in-strumental variables To this end we need vari-ables that are correlated with the proxies pj

A butuncorrelated with the error term i We use thenumber of SKUs within each product group j asan instrument as it does justice to the idea thatvariation in the product group price index stemsin part from variation in the set of SKUs that areavailable each period It is clearly correlatedwith the average group prices pj

A and plausiblyuncorrelated with the demand shock i Theassumption that the number of SKUs is uncor-related with the conventional demand error willhowever be violated if the introduction of anew SKU changes the perceived quality of adrug21 or if it is accompanied by promotionalactivities which will be reflected in i Our hopeis that this is not too often the case Assumingthat the SKU number for each group j is alsouncorrelated with the weighted sum of the mea-surement errors of all product groups yenj ijln 13j itcan be used as an instrument for the product groupprice pj

A22 If one accepts the premise that prices atthe SKU level are exogenous then SKU pricescan also be used as instruments The final list ofinstruments we use hence includes the number ofSKUs in each group the prices of the five largestSKUs for each group and all other exogenousvariables in the demand estimation such as re-gionalproduct dummy interactions and the upper-

21 Note that we already proxy for the quality of each prod-uct group through the product fixed effects i In the actualestimation we actually let these fixed effects be region-specific(ir) so that we allow quality to vary across regions

22 If the measurement error increases in the number ofincluded SKUs this assumption will be violated Unfortu-nately we have no way of checking the validity of thisassumption

1492 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

level total expenditure on antibiotics Regressionsof group prices on the instruments above yieldhigh R-squares with most regressors highly sig-nificant indicating that our instruments are highlycorrelated with prices23

Our sample includes four molecules cipro-floxacin norfloxacin ofloxacin and spar-floxacin Except for sparfloxacin all othermolecules are produced by both foreign anddomestic firms24 So we have seven products(domestic ciprofloxacin foreign ciprofloxa-cin etc) with 96 observations (two years ofmonthly data four geographical regions foreach period) for each product The parametersin the lower-level AIDS demand system asdefined in equation (8) are the product fixedeffects i the productndashregional dummy inter-actions ir the own revenue-share price elas-ticities ii the cross revenue-share priceelasticities i10 i01 i00 and the revenue-share expenditure elasticities i In estimatingthe parameters we first regress prices on allinstrumental variables and then plug the pre-dicted values for prices in the constrainedleast-square regression (For a detailed expla-nation of the constraints see the previoussection)

Given that we impose many cross-equationconstraints and employ instrumental variablesin the estimation it is difficult to derive stan-dard errors for the parameter estimates ana-lytically Our error term interpretation in (8)implies that the error terms for each productgroup are likely to be correlated across re-gions for example if a national advertisingcampaign increases the demand for a rela-tively expensive presentation in one regionsimultaneously increasing the aggregate de-mand and price index for the correspondingproduct group in that region it is likely that thesame effect will be observed in other regions Inprinciple we could exploit the cross-regionalcorrelation in the error terms to estimate thedemand system using seemingly unrelated re-

gression (SUR) This however would requireus to derive the variance-covariance matrix ofthe parameters analytically which as notedabove is cumbersome in the current contextInstead we use the bootstrap method choosingto remain agnostic about the structure of thevariance-covariance matrix The potential dis-advantage of this approach over SUR is a loss inefficiency but as will become apparent in theresults section the model parameters are fairlyprecisely estimated To maintain the marketstructure we randomly sample the periods (withreplacement) and use the same periods for allproducts Regarding the optimal number ofbootstrap repetitions ideally one would followthe three-step method proposed by Donald W KAndrews et al (2000) Empirical evidence sug-gests however that one rarely needs more than200 replications to estimate the standard er-rors25 To be safe we generate 300 bootstrapsamples (with replacement) based on the origi-nal data and estimate the standard errors usingthe standard errors of the bootstrap sampleestimates

The estimation of the top-level AIDS systemis similar The constraints imposed on this top-level demand system are adding-up homogene-ity and symmetry as defined on page 12Again bootstrapping is used to obtain the stan-dard errors of the parameter estimates

D The Counterfactual Scenarios

In assessing the effects of patent enforce-ment we start by focusing on the most extremecase in which compulsory licensing is not anoption and foreign firms are not subject to pricecontrols We use the results from the analysis ofthis case as a benchmark In reality the outcomeof the WTO negotiations is more likely to in-volve some constraints on the monopoly powerof foreign firms selling patented products indeveloping countries such as price caps orcompulsory licensing Our framework can eas-ily accommodate these cases as will becomeapparent in the next subsection

We now focus on the effects of potentialpatent enforcement in the quinolone segmentWe consider several scenarios which vary inthe number and size of domestic products that

23 Specifically the R-squares from the first-stage regres-sions of product group prices on instruments range from057 to 095 except for domestic ciprofloxacin for whichthe R-square is 017

24 Sparfloxacin is actually offered by one foreign subsid-iary in India (see Table 2) However its revenue share isminiscule We therefore treat Sparfloxacin as being pro-duced by domestic firms only 25 See Bradley Efron and Robert J Tibshirani (1993)

1493VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

will be removed from the market In particularwe consider the following five scenarios26

Withdrawal of one large domestic productgroup only domestic ciprofloxacin

Withdrawal of one relatively small domesticproduct group only domestic ofloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand norfloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand sparfloxacin

Withdrawal of all four domestic quinoloneproduct groups

As the list above suggests we proceed fromanalyzing the effects of single-product with-drawal to the analysis of eliminating the en-tire domestic segment This approach wasmotivated by early empirical results that in-dicated that the existence and extent of com-petition from domestic firms has a significantbearing on the predicted effects of patent en-forcement that is our predictions regardingprices and welfare vary substantially depend-ing on how many domestic products are af-fected by patent enforcement and on whethersome domestic competition will remain presentafter TRIPS In addition the size of the affecteddomestic groups is relevant for the welfare pre-dictions accordingly we examine both scenar-ios in which a large domestic product groupsuch as domestic ciprofloxacin is eliminatedfrom the market and scenarios in which the

eliminated domestic group is relatively small(eg domestic ofloxacin) In all scenarios wemaintain the assumption that the set of prod-ucts offered by the remaining firms in themarket does not change in response to patentenforcement

E Computation of Virtual Prices and NewEquilibrium Prices

The first step in the counterfactual analysisis to derive the new equilibrium prices underpatent enforcement27 In this context thereare two sets of prices that are relevant Thefirst set consists of the virtual prices of those(domestic) products that will not be availableonce TRIPS is put in effect To calculate thesevirtual prices we set the expenditures of therelevant products equal to zero The secondset of prices consists of the prices of thoseproducts that remain in the market In deriv-ing these prices we start by assuming profitmaximization without any form of price reg-ulation the firms remaining in the marketreoptimize in response to the policy changeand set new prices taking the prices of allother firms as given28 Of course at the equi-librium all prices change in response to thefact that some domestic products are nolonger present The new equilibrium pricesfor products that remain in the market are thuscomputed by utilizing the first-order condi-tions of profit-maximizing firms into whichthe virtual prices of the eliminated productsare substituted Hence to compute the newequilibrium prices we solve an equation sys-tem of the following form

26 As Table 2 indicates most of the patents for the drugsin the quinolone segment have expired by now so that noneof the scenarios described in this section is going to mate-rialize in practice The reason we consider these alternativescenarios in our counterfactual simulations is to get a senseof how the presence and extent of domestic competitionaffects welfare calculations Such calculations may becomerelevant in the future in other therapeutic classes in whichpatents have not expired yet and where domestic firmscompete with foreign patent holders While the particularexit scenarios will depend on the exact year of patent expiryof drugs in each therapeutic class and the specific way inwhich patents will be enforced (eg whether certain do-mestic drugs will be grandfathered in) the general conclu-sion that emerges from our calculations is that the consumerwelfare loss is substantially smaller in cases where somedomestic competition remains present in the market Hencethe particular way in which patent enforcement is imple-mented is essential for assessing its welfare impact

27 Of course until product patents are in fact introducedthese prices will not be observable Note also that we areassuming here that the range of available products will notchange with the introduction of patents

28 As mentioned above this first set of calculations ab-stracts from the existence of remaining price controls orother government regulations that would impose constraintson the firmsrsquo profit maximization problem Accordingly theresulting numbers should be interpreted as a benchmark ofwhat would happen if markets were completely unregulatedTo examine what the welfare effects of TRIPS would be inthe more realistic scenario of price regulation we subse-quently consider a scenario in which regulation keeps theprices of the products provided by patent-holders at theirpre-TRIPS level For a more detailed discussion see theresults section

1494 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

For products i that are withdrawn from themarket

(14) 0 i ir iiln pirV i10ln pjr j Di

10

j Di

01

i01ln pjr

j Di

00

i00ln pjr

ilnXQr

PQr

For products k that remain in the market

(15) pkr ckr 1 1

kk pkrpjr pirV 1

In the equations above pirV denotes the virtual

prices of the products that are removed fromthe market while pjr denotes the updatedprices of all other products Note that whensolving for the virtual prices we account forthe fact that both the price index for quino-lones PQr and the expenditure allocated tothis subsegment XQr need to be updated toreflect the fact that as a result of the pricechanges there may be substitution away fromthis subsegment To obtain the new quinoloneexpenditure XQr and the new price index PQrwe use the estimates and formulas for thehigher-level AIDS system In equation (15)ck refers to the marginal cost for product k thatwe have obtained from the previous estimationstage As mentioned in the previous subsectionwe conducted the simulations using both theupper and lower bounds for marginal cost ck

u

and ckL respectively The term kk(pkr

pjrpirV) re-

fers to the unconditional own-price elasticity forproduct k which is a function of the eliminatedproductsrsquo virtual prices and the remaining prod-uctsrsquo new equilibrium prices We conduct thecounterfactual simulations and welfare analysisat the regional level because the presence ofregion-specific product effects in the demandestimation implies region-specific demand elas-ticities and hence region-specific marginalcosts and markups The presentation and discus-sion of our results focus on national averages ofthe relevant variables that we construct by com-

puting weighted averages across regions usingthe population of each region as a weight

F Welfare Assessment

The simulation of the new equilibrium underpatent protection can provide important insightsinto how consumers and firms will respond tothe removal of domestic products in the market(for example toward which products consumerswill substitute or which prices will increase themost) To get a more precise idea of how peo-plersquos well-being will ultimately be affected byTRIPS we compute as a last step in our anal-ysis the welfare effects of the policy changeSocial welfare is defined as consumer welfareand the sum of domestic firm profits Thechange in domestic profits can easily be calcu-lated by comparing the domestic firm (variable)profits at the pre-TRIPS prices to the profitsthese firms will realize at the new simulatedprices Although foreign firm profits do notcount in domestic welfare calculations we alsocompute the effects of patent enforcement onforeign firm profits to get an idea of how largethe expected benefits of TRIPS are for thesefirms This provides in some sense an indirectway of assessing whether the claims that patentenforcement in countries like India will lead tomore research on developing-country-specificdiseases (such as malaria) have any validity iffor example we find that the effect of patentenforcement on the foreign firm profits realizedin India is small in magnitude it is unlikely thatforeign firms will engage in more developing-country-specific research in response to TRIPSIt is important to note that in all these calcula-tions we work with the lower-bound estimatesfor marginal costs since these give rise to thehighest possible markups Hence our estimateof profit loss for domestic producers most likelyoverstates this loss

On the consumer side we measure changes inconsumer welfare by the compensating variation(CV) defined as the additional expenditure thatconsumers need in order to achieve the sameutility level as before patent enforcement at thenew prices Specifically let P0 denote the pricevector before patent enforcement P the simulatedprice vector post-TRIPS (that we obtained usingthe methods described in the previous subsection)u0 the utility attained by consumers before TRIPS

1495VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

and E(u P) the higher-level expenditure functionThen the compensating variation is given by

(16) CV Eu0 P Eu0 P0

Note that the CV as computed in (16) representsthe combination of three effects

The pure product-variety effect that is theeffect that arises because one or more prod-ucts are not available to consumers anymoreholding the prices of all other remainingproducts and the total expenditure on thequinolone subsegment XQ constant

The expenditure-switching effect that is theeffect arising from substitution away fromquinolones and toward other subsegments ofthe antibiotics market again holding the pricesof all other remaining products constant

The reduced-competition effect that is theeffect that arises because the firms remainingin the market adjust (increase) their prices inresponse to the removal of domestic products

From both an analytical and a policy point ofview it is desirable to assess how large each ofthese effects is Accordingly we decompose thetotal effect on consumer welfare (the CV as givenby equation (16)) using the following procedure

To get the pure product variety effect wecompute virtual prices for the products that areremoved from the market holding the quino-lone expenditure XQ and the prices of all otherproducts fixed Let us call the resulting pricevector P1 Then the pure product variety effectis represented by E(u0 P1) E(u0 P0)

To compute the expenditure-switching effectwe compute another set of virtual prices againholding the prices of all remaining products fixedbut letting quinolone expenditure adjust in re-sponse to the new price index for the quinolonesegment (Given that the prices of the remainingproducts remain fixed the change in the priceindex arises only because of the removal of one ormore domestic products) Note that this scenariomost closely resembles the case in which patentlaws are enforced but strict price regulation keepsthe prices of the products offered in the market attheir pre-TRIP level Let us label the so-computedprice vector P2 The expenditure-switching effectis then E(u0 P2) E(u0 P1)

Finally the reduced competition effect aris-ing from higher prices for the remaining prod-

ucts is computed as the residual change in thecompensating variation once the product-varietyand expenditure-switching effects have been ac-counted for that is E(u0 P) E(u0 P2)where the price vector P is computed accordingto the formulas (14) and (15) to reflect theadjustment of prices to the new regime

To compute the standard errors associatedwith the counterfactual simulations (that is thestandard errors for the counterfactual prices andwelfare estimates) we again use bootstrappingWe first bootstrap the original sample (sales andprices of all the drugs) 300 times Next we esti-mate the AIDS model for each of these samplesand compute the counterfactual equilibrium pricesand welfare losses corresponding to each policyscenario for each of the 300 simulations In thefinal step we compute standard errors29

One limitation of our framework is that itdoes not allow for a heterogeneous response ofconsumers to the policy change Accordinglyour framework is not suited to addressing thequestion of how different groups in the popula-tion will be affected by patent law enforce-ment30 Along the same lines our frameworkdoes not accommodate the possibility of pricediscrimination which might lead to differentresults regarding welfare losses and profit gainsrelative to uniform pricing31 Still we believethat the results of the counterfactual simulationscan provide important insights into the likelyaggregate response to patent enforcement andthe factors that drive this response

III Results

A The Structure of Demand

Tables A2 and A3 in the Appendix display theresults from estimation of the lower- and upper-level AIDS systems respectively For ease of in-terpretation rather than discussing the coefficientestimates we focus our discussion on the implied

29 In cases some of the simulations do not converge wecompute the standard errors using only those simulationsthat converged

30 Addressing this question would require at a mini-mum micro data on consumer purchases as in Goldberg(1995) To our knowledge such data do not exist for theIndian pharmaceutical market

31 See Ernst R Berndt (1994) for an extensive discussionof uniform pricing

1496 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

unconditional price and expenditure elasticities re-ported in Table 6 Given that the region-specificproduct group effects ir reported in the last fourcolumns of Tables A2 and A3 imply region-spe-cific demand elasticities we report separate elas-ticities for each region in Tables 6A and 6B Asevident from the comparison of these tables theelasticities are very similar across regions so thatwe can focus the remaining discussion on theelasticities of one region only the northern regiondisplayed in Table 6A

The diagonal terms of Table 6A report theown-price elasticities which are in all but onecase negative and highly significant The oneexception is the foreign norfloxacin productgroupmdashwhose share of quinolone sales is 007percentmdashfor which we estimate a negative butinsignificant own-price elasticity For the remain-ing product groups demand appears to be highlyelastic with the estimated elasticities lower than2 in four out of the six cases The magnitude ofthe own-price elasticities matches the features ofthe Indian pharmaceutical market mentioned ear-lier which would suggest that Indian consumersare likely to be quite price-sensitive32 The elas-ticities appear especially large if one takes intoaccount that they refer to product groups (such asdomestic ciprofloxacin) and not individual drugsoffered by particular firms Their relative magni-tudes are also intuitive the drug with the largestmarket share and a relatively high price (domesticciprofloxacin) appears to be one of the least elas-tic In contrast foreign ciprofloxacin is highlyprice elastic this is plausible as ciprofloxacindrugs offered by subsidiaries of multinationalsface the stiffest competition in this market seg-ment from approximately 75 Indian firms offeringthe same molecule

The estimated expenditure elasticities appear inthe last column These are all positive indicatingthat the demand for all product groups is normalThe remaining cells display the estimated cross-price elasticities As one might perhaps expect forproducts within a therapeutic subsegment theseare mostly positive Out of a total of 49 priceelasticities we estimate there are six that do notconform to expectations these are the cross-price elasticities between different foreign prod-

uct groups which are estimated to be negativeand significant33 Fortunately these elasticitieshave negligible impact on the welfare analysisgiven that our counterfactuals focus on theeffect of withdrawing one or more domesticproducts from the market the most relevantelasticities are the ones that capture the responseof various product group shares to a change in theprice of one or more domestic groups these elas-ticities are the cross-price elasticities between var-ious domestic groups and the ones betweendomestic and foreign groups which are plausibleand precisely estimated34

Regarding these elasticities a striking aspectof our estimates is how large positive andsignificant the cross-price elasticities betweendifferent domestic product groups aremdashin factfor norfloxacin and ofloxacin we estimate thatdomestic product groups containing differentmolecules are closer substitutes for one anotherthan product groups that contain the same mol-ecule but are produced by foreign firms Incontrast for ciprofloxacin (the molecule withthe largest revenue share) we estimate a largepositive cross-price elasticity between the do-mestic and foreign versions

The fact that domestic products appear to beclose substitutes for other domestic productsthat contain different molecules truly representsan ldquoempiricalrdquo finding in the sense that we do

32 In developed economies elasticities of this magnitude havetypically been found only for generic drugs (and even then onlyrarely) or among consumers who lack health insurance

33 While these elasticities are clearly counterintuitive theyare not inconsistent with the underlying demand system whichimposes no restrictions on the sign of the cross-price elastici-ties We do not have a good explanation of why these elastic-ities are estimated to be negative A potential explanation isthat the shares of the foreign products are very small giventhis we observe very few consumers switching from one (verysmall) foreign group to another (very small) foreign groupwhen the price of the first foreign group goes up and hence theinference is not very reliable in this case

34 The cross-price elasticities between foreign drug groupscontaining different molecules will also have an effect on thewelfare estimates given that the withdrawal of domestic prod-ucts will generally lead to changes in the prices of all foreignproducts and market shares will be reallocated from someforeign products to other foreign products based on the newprices However this reallocation from ldquoforeign to foreignrdquo istruly second order in our case compared to the reallocationfrom ldquodomestic to domesticrdquo and ldquodomestic to foreignrdquo prod-ucts In scenarios in which we consider patent enforcementaccompanied by strict price regulation the cross-price elastic-ities between foreign product groups are in fact completelyirrelevant as the prices of foreign products are not allowed toincrease in these cases Still our welfare loss estimates remainsubstantial drivenmdashas beforemdashby the loss of product variety

1497VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

not impose it through any of our assumptionsregarding the demand function The questionthat naturally arises then is what might explainthis finding While we cannot formally addressthis question anecdotal accounts in various in-dustry studies suggest that the explanation maylie in the differences between domestic andforeign firms in the structure and coverage ofretail distribution networks

Distribution networks for pharmaceuticals inIndia are typically organized in a hierarchicalfashion Pharmaceutical companies deal mainlywith carrying and forwarding (CampF) agents inmany instances regionally based who each sup-ply a network of stockists (wholesalers) Thesestockists in turn deal with the retail pharma-cists through whom retail sales ultimately oc-cur35 The market share enjoyed by a particularpharmaceutical product therefore depends inpart on the number of retail pharmacists who

stock the product And it is here that thereappears to be a distinction between domesticfirms and multinational subsidiaries In particu-lar the retail reach of domestic firms as agroup tends to be much more comprehensivethan that of multinational subsidiaries (IndianCredit Rating Agency (ICRA) 1999)36

There appear to be two reasons for this Thefirst is that many of the larger Indian firmsbecause they have a much larger portfolio ofproducts over which to spread the associatedfixed costs typically have more extensive net-works of medical representatives The second issimply that there are many more domestic firms(and products) on the market At the retail levelthis would imply that local pharmacists mightbe more likely to stock domestic products con-taining two different molecules say ciprofloxa-cin and norfloxacin than they would domesticand foreign versions of the same molecule Tothe extent that patients (or their doctors) arewilling to substitute across molecules in order tosave on transport or search costs (eg going toanother pharmacy to check whether a particular

35 There are estimated to be some 300000 retail pharma-cists in India On average stockists deal with about 75 retailers(ICRA 1999) There are naturally variations in this structureand a host of specific exclusive dealing and other arrangementsexists in practice Pharmaceutical firms also maintain networksof medical representatives whose main function is to marketthe companyrsquos products to doctors who do the actual prescrib-ing of drugs In some instances firms do sell directly to thedoctors who then become the ldquoretailerrdquo as far as patients areconcerned but these are relatively rare

36 These differences were also highlighted in conversa-tions one of the authors had with CEOs and managingdirectors of several pharmaceutical firms as part of a sep-arate study

TABLE 6AmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN THE NORTHERN REGION

Product group

Elasticity with respect to

Prices of foreign productgroups Prices of domestic product groups Overall

quinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 557 013dagger 015 401 011dagger 011dagger 016 137(179) (007) (007) (184) (006) (006) (006) (029)

Foreign norfloxacin 427dagger 045 427dagger 350dagger 602 451 465 220(242) (112) (242) (210) (623) (184) (183) (105)

Foreign ofloxacin 011 010dagger 138 009 009dagger 023 011 116(005) (005) (031) (027) (005) (028) (004) (017)

Domestic ciprofloxacin 018 001 001 168 008 008 010 117(008) (000) (001) (023) (002) (002) (002) (003)

Domestic norfloxacin 004 003 004 058 223 042 040 073(001) (003) (001) (017) (011) (004) (003) (009)

Domestic ofloxacin 005 005 011 077 074 342 074 089(002) (002) (013) (028) (008) (025) (008) (021)

Domestic sparfloxacin 007 004 007 115 063 063 288 028(002) (001) (002) (015) (006) (006) (017) (012)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent significance level and dagger (dagger) denotes significance at the 10-percent level

1498 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE 6BmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN OTHER REGIONS

Demand patterns in the eastern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 594 014dagger 016 431 013 011 017 140(195) (008) (008) (201) (006) (007) (007) (031)

Foreign norfloxacin 329dagger 058 329dagger 264dagger 460 346 359 192(180) (083) (179) (153) (451) (138) (139) (082)

Foreign ofloxacin 012 010dagger 140 010 010 024 013 117(006) (006) (034) (031) (005) (030) (005) (019)

Domestic ciprofloxacin 018 001 001 172 009 008 011 117(008) (000) (002) (027) (002) (003) (002) (003)

Domestic norfloxacin 004 004 004 068 242 049 046 070(002) (004) (002) (020) (013) (005) (004) (010)

Domestic ofloxacin 004dagger 004 009 061 060 295 060 092(002) (001) (010) (028) (007) (022) (007) (017)

Domestic sparfloxacin 005 003 005 092 048 051 251 043(002) (001) (001) (017) (005) (006) (016) (010)

Demand patterns in the western region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 427 009dagger 011 286 008 006 011 126(129) (005) (005) (133) (004) (005) (004) (021)

Foreign norfloxacin 714dagger 008 710dagger 594dagger 995 742 770 299dagger

(369) (183) (368) (328) (1015) (297) (296) (178)Foreign ofloxacin 013 012dagger 145 009 012 026 013 119

(007) (006) (037) (029) (005) (033) (005) (021)Domestic ciprofloxacin 019 001 000 174 009 008 011 118

(008) (000) (001) (024) (002) (003) (002) (004)Domestic norfloxacin 004 004 004 067 243 050 047 069

(002) (004) (001) (018) (014) (005) (004) (010)Domestic ofloxacin 003 003 007 048 048 257 048 093

(002) (001) (008) (022) (005) (015) (004) (014)Domestic sparfloxacin 006 003 005 083 046 049 241 046

(002) (001) (001) (014) (004) (005) (012) (009)

Demand patterns in the southern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 560 013dagger 015 402 011 010 017 137(182) (007) (007) (187) (006) (006) (007) (029)

Foreign norfloxacin 533dagger 031 531dagger 431dagger 748 559 584 249dagger

(301) (134) (300) (245) (734) (219) (219) (128)Foreign ofloxacin 014 012dagger 148 011 012 029 016 120

(007) (007) (039) (033) (006) (035) (006) (022)Domestic ciprofloxacin 017 001 000 169 008 007 011 116

(008) (000) (001) (025) (002) (003) (002) (003)Domestic norfloxacin 004 003 003 059 225 043 040 073

(001) (003) (001) (019) (012) (004) (004) (009)Domestic ofloxacin 004dagger 004 009 062 060 296 060 091

(002) (001) (010) (026) (006) (019) (007) (017)Domestic sparfloxacin 008 005 008 138 073 075 323 016

(002) (001) (002) (018) (007) (008) (021) (014)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at 5-percent level anddagger (dagger) denotes significance at the 10-percent level

1499VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

foreign product is in stock) in aggregate datawe would expect to find precisely the substitu-tion patterns that we report in Table 6

Whether the particular explanation we provideabove is the correct one the high degree of sub-stitutability between domestic product groupsturns out to have important implications for thewelfare calculations We discuss these in moredetail below when we present the results of thecounterfactual welfare analysis Another elasticitywith important implications for the counterfactu-als is the price elasticity for the quinolone subseg-ment as a whole which indicates how likelyconsumers are to switch to other antibioticsgroups when faced with a price increase for quin-olones This elasticity is computed on the basis ofthe results in Table A3 and it is at 111 (stan-dard error 024) this is large in magnitudebutmdashas expectedmdashsmaller in absolute value thanthe own-price elasticities of the product groupswithin the quinolone subsegment

The results in Tables 6A and 6B are based onour preferred specification discussed in SectionII In Tables A4 to A6 in the Appendix weexperiment with some alternative specificationsTables A4(a)ndashA4(c) correspond to a specificationthat includes in addition to product-group-specificregional fixed effects product-group-specific (andfor the upper level antibiotics-segment-specific)seasonal effects We distinguish among three sea-sonsmdashthe summer monsoon and wintermdashandreport the unconditional demand elasticities for

the northern region for each of these seasons Asevident from the tables in the Appendix our elas-ticity estimates are robust to the inclusion of sea-sonal effects The demand elasticities in Table A5are based on estimation of the demand system byOLS Compared to the elasticities obtained by IVthe OLS elasticities are smaller in absolute valueimplying that welfare calculations based on theOLS estimates would produce larger welfare lossestimates Nevertheless some of the patterns re-garding the cross-price elasticities discussed ear-lier are also evident in the OLS results inparticular the cross-price elasticities between dif-ferent domestic product groups are all positivelarge and significant and in most instances largerthan the cross-price elasticities between drugs thatcontain the same molecule but are produced byfirms of different domesticforeign status Theclose substitutability of domestic products indi-cated by both the OLS and IV estimates seems tobe one of the most robust findings of the paper

B Cost and Markup Estimates

Table 7 displays the marginal costs markupsand profits implied by the price elasticity esti-mates of Tables 6A and 6B for each of the sevenproduct groups Given that our regional effectsimply different price elasticities for each regionour marginal cost and markup estimates alsodiffer by region Given however that based on

TABLE 7mdashUPPER AND LOWER BOUNDS FOR MARGINAL COST MARKUP AND ANNUAL PROFIT BY PRODUCT GROUPS WITHIN

THE QUINOLONE SUBSEGMENT

Product groupLower boundfor MC (Rs)

Upper boundfor markup

Upper boundfor profit(Rs mill)

Upper boundfor MC (Rs)

Lower boundfor markup

Lower boundfor profit

(Rs)

Foreign ciprofloxacin 83 19 269 103 0 00(123) (012) (1655)

Foreign norfloxacin NA NA NA 53 0 00Foreign ofloxacin 323 70 1061 1085 0 00

(2316) (021) (3185)Domestic ciprofloxacin 47 59 17019 112 0 00

(114) (010) (29858)Domestic norfloxacin 52 43 2807 90 0 00

(020) (002) (1532)Domestic ofloxacin 587 34 1612 901 0 00

(218) (002) (1280)Domestic sparfloxacin 495 37 1985 788 0 00

(157) (002) (1100)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level Estimated lower bound forforeign norfloxacinrsquos marginal cost is negative since the estimated price elasticity is less than one in absolute value

1500 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Tables 6A and 6B the price elasticities do notseem to differ substantially across regions wereport for ease of exposition only the nationalaverages for marginal costs and markups inTable 7

Table 7 has two parts In the left part (firstthree columns) we report the lower bound formarginal cost and the corresponding upperbound for markup and upper bound for totalannual profit for each product group Thesenumbers are based on the price elasticities weobtained from estimating the two-level AIDSdemand system Since we do not have a reliableestimate for the price elasticity of foreign nor-floxacin (the point estimate is negative but lessthan one in absolute value and insignificant)we cannot compute the lower bound for mar-ginal cost in this case It is important to note thatthese estimates do not reflect either the actualmarginal cost or the actual markup for thesedrugs both because the existence of price reg-ulation implies that the unconstrained first-orderconditions are unlikely to hold each period andbecause our aggregation across firms of thesame domesticforeign status supplying thesame molecule makes the interpretation of theseestimates problematic In particular the fact thatwe ignore competition among firms within eachproduct group implies that our estimates willtend to overstate market power These numberswill prove useful however in the counterfac-tual simulations as they can give us a sense ofhow large the maximum profit gains for multi-national firms and the maximum profit lossesfor domestic firms are likely to be under patentenforcement The right part of the table (lastthree columns) reports the upper bound for mar-ginal cost which we obtain by simply taking themarginal cost to be equal to the observed priceWe do not report standard errors in this casesince the numbers are based on actual data Themarkups corresponding to these marginal costupper bounds are of course zero We conductthe counterfactual simulations using both thelower and upper bounds for marginal costs

The (upper bounds for) markups on the leftside of the table are generally plausible Thedomestic ciprofloxacin product group that dom-inates the quinolone subsegment and for whichwe documented high prices a high marketshare and a relatively low elasticity of demandenjoys one of the highest markup upper bounds(60 percent) and accounts for nearly 70 percent

of all profits derived within the subsegmentForeign ciprofloxacin on the other hand whichfaces the stiffest competition from domesticfirms and for which we estimated a highly elas-tic demand has the lowest markup upper bound(19 percent)

C Counterfactual Estimates of the Impact onPrices and Welfare

With estimates of the key demand and costparameters in hand we turn to the counterfac-tuals We consider the five separate scenarioslisted in the previous section All of the scenar-ios involve the withdrawal of one or more of thedomestic product groups from the market Ta-ble 8 displays our estimates of the consumerwelfare losses that result under the differentscenarios The losses are expressed in billion ofrupees per year All numbers presented in Ta-ble 8 and subsequent tables are based on usingthe lower bounds for marginal cost and upperbounds for markup in the simulations As dis-cussed earlier these numbers are the more in-teresting to work with since they give us anupper bound for the changes in the profits ofdomestic and foreign firms that would resultfrom patent enforcement In the Appendix wealso present results based on using the upperbounds for marginal costs in the simulations inwhich case the pre-TRIPS profits of domesticand foreign firms are zero In all cases marginalcosts are assumed to be constant in outputWhile naturally the profit implications differdepending on whether one uses the upper orlower bounds for marginal costs (firm profits arezero if one assumes the upper bound of mar-ginal cost in which case price equals marginalcost) the estimated consumer welfare losses aresimilar in the two cases We discuss these re-sults in more detail at the end of this section

The first column presents our estimates of theconsumer welfare losses attributable to the pureloss of product variety effect where we fix theprices of all remaining products as well as theoverall expenditure on quinolones while with-drawing one or more of the domestic productgroups Note that in our initial specification ofthe demand system had we not allowed for thepossibility that consumers might differentiatebetween domestic and foreign products evenwhen they contain the same molecule this

1501VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

particular component of the loss of consumerwelfare would not have arisen

The estimates reported in the second columnincorporate the expenditure-switching effect ontop of the loss of product variety Here basedupon the price elasticity estimates from thehigher-level AIDS system we adjust (down-ward) the expenditures allocated to the quino-lone subsegment as the composite price ofquinolones effectively increases as a consequenceof the higher virtual prices of the domestic productgroups that are withdrawn from the market Be-cause the estimates in this column are generatedassuming that the prices of the products that re-main in the market are not adjusted upward theyprovide a sense of what consumer losses would beif the introduction of product patents were coupledwith strict price regulation aimed at maintainingprices at prepatent levels Alternatively they canbe thought of as the relevant welfare numbers ifintense competition among firms within the re-maining product groups kept the prices of theproducts that were still offered in the market closeto the firmsrsquo marginal costs

The last column of Table 8 displays the esti-mated consumer welfare losses when bothcross-segment expenditure-switching and with-in-segment upward price adjustments are takeninto account

If we compare the results across the firstsecond and third columns all the counterfac-tual scenarios produce qualitatively similar pat-terns patterns that are consistent with what we

would expect Starting from the initial loss ofwelfare attributable to the loss of product vari-ety the option of switching expenditures outof the quinolone subsegment to other subseg-ments mitigates some of the initial welfareloss But if we then incorporate the upwardprice adjustments that result in response to thereduced competition the welfare losses aremagnified

Of particular interest from a policy perspec-tive are the relative magnitudes of these threeeffects which are similar under all the counter-factual scenarios though absolute levels varyconsiderably First despite the fact that the de-mand for quinolones is quite sensitive to thecomposite price of quinolonesmdashthe upper levelprice elasticity is 111mdashthe cross-subsegmentexpenditure-switching effects are in all thecases small (in absolute value terms) relative tothe other two effects For instance under thescenario where all the domestic quinolone prod-uct groups are withdrawn from the market theoverall consumer welfare loss of Rs 1781 bil-lion per year can be decomposed into an initialloss of Rs 1176 billion (66 percent) attribut-able to the loss of product variety a slightreduction in this initial loss of Rs 041 billion(2 percent) from Rs 1176 billion to Rs1135 billion because of expenditure switchingand a subsequent additional loss of Rs 646billion (36 percent) from Rs 1135 billion toRs 1781 billion because of the reduced com-petition and consequent price increases

TABLE 8mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure loss ofvariety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 732(087) (089) (146)

Only ofloxacin 008 008 023(008) (008) (010)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1253(177) (180) (415)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1058(142) (145) (331)

All four domestic quinolones products 1176dagger 1135dagger 1781(643) (634) (1270)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent significance level and dagger (dagger)denotes significance at the 10-percent level

1502 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 13: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

The equation at the lower level becomes

(7) i i iiln pi i10ln pj j Di10

j Di

01

i01ln pj

j Di

00

i00ln pj ilnXQ

PQ

Note that

The parameter ii captures a product grouprsquosown-price effect (there will be as many iiparameters as the number of product groups)

The parameter i10 captures the cross-priceeffect of the product group containing prod-ucts with the same molecule but produced byfirms of different nationality

The parameter i01 captures the cross-priceeffects of product groups containing productswith different molecules but produced byfirms with the same nationality

The parameter i00 captures the cross-priceeffects of product groups containing productswith different molecules produced by firms ofdifferent nationality

Before we take the demand equation to thedata we make two modifications The first oneis to let the product-specific effects i vary byregion r The resulting product-specific regionaleffects ir have two interpretations first theycontrol for the ldquoqualityrdquo of each drug withquality differences being allowed to vary acrossregions second they proxy for demographicsand other demand shifters which vary by re-gion and may affect the demand of each prod-uct group differently18 Note that by includingproduct-specific regional effects in the demandspecification we estimate the price parametersbased on the within-product-group variation ofprices in each region In an earlier version of thepaper we also estimated the demand systemwithout regional dummies and obtained similarresults

The second modification that allows us to gofrom a deterministic to a stochastic specificationof the demand equation is to include an additiveerror term in (7) The latter takes into accountthe fact that (7) is not expected to fit the dataexactly The error term irt accounts for mea-surement error (due to the fact that the productgroup prices pjrt we employ in the estimationare not exact price indices but approximationsthereof) and (potentially region-specific) de-mand shocks that may affect the demand for aproduct in a particular period Examples of suchshocks include an advertising campaign for aparticular product which temporarily increasesthe demand for this product or the outbreak ofa (potentially region-specific) epidemic whichcalls for the use of a particular drug We discussthe interpretation and properties of this errorterm in more detail in the next section

The final form of the equation we estimate atthe lower level becomes (with subscript t de-noting month and subscript r denoting region)

(8) irt i ir iiln pirt i10ln pjrt j Di10

j Di

01

i01ln pjrt

j Di

00

i00ln pjrt

ilnXQrt

PQrt irt

The analysis so far has conditioned on the ex-penditure allocated to the quinolone subseg-ment XQ The upper level of the estimationconsiders the problem of allocating total expen-diture across the different systemic antibioticssubsegments one of which is quinolones Theupper-level demand function is given by

(9) G G H

GHln PH GlnX

P

where all variables denoted by capital letters aredefined as before but now refer to subsegments(G H ) rather than product groups within asubsegment and the total expenditures on sys-temic antibiotics X are deflated by the Stone priceindex log P yenH Hlog PH When estimating thesystem above we impose all the restrictions

18 Given the short time span of our sample typicaldemand shifters such as age distribution income distribu-tion and education hardly change over our sample periodSuch shifters are therefore absorbed by the region-specificproduct fixed effects ir

1489VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

implied by utility maximization as we do withthe estimation of the lower-level AIDS We donot however impose any additional restrictionson the substitution patterns at this stage so thatthe cross-price effects across segments remainrelatively unconstrained

Estimation of the higher-level AIDS allowsus to obtain the unconditional own- and cross-price elasticities that are used in the formulationof the supply problem and welfare analysisThese are given by the formula

(10) ij ijXQXQ

ln qi

ln XQ

ln XQ

ln PQ

ln PQ

ln pj

with the conditional cross price elasticitiesgiven by

ijXQXQ ij ij jlnXQ

PQ

i

As in the lower stage we include subsegment-specific regional dummies in the specificationof the upper-stage demand system so that thefinal form of the estimating equation at theupper stage becomes

(11) Grt G Gr H

GHln PHrt

GlnXrt

Prt Grt

In sum the demand system we take to thedata is represented by equations (8) and (11)and the associated parameter restrictions im-plied by economic theory

B Modelling the Supply Side of the Market

Counterfactual simulations concerning theeffects of domestic product withdrawal requireknowledge of the marginal costs of pharmaceu-tical firms operating in the Indian market Theseare unobservable The usual approach in theNew Empirical Industrial Organization litera-ture has been to exploit the firm equilibriumconditions to infer marginal cost For exampleit is often assumed that the marginal cost ci isconstant and that the industry is an oligopoly

engaging in Bertrand competition with differen-tiated products Assuming that firms myopicallymaximize profits each period one can then de-rive the firmsrsquo first-order conditions that corre-spond to the assumptions above about costsmarket structure and firm behavior

We deviate from this procedure as the presenceof price regulation renders the assumption of un-constrained period-by-period maximization un-tenable Ideally one would like to explicitlyincorporate the price and other administrative con-trols into the firmrsquos optimization problem andderive the first-order conditions under the assump-tion of constrained maximization The complexityof the price regulation however makes this ap-proach infeasible Therefore we adopt an alterna-tive approach that does not rely on modelling theprice-setting process but is instead based on de-riving upper and lower bounds for marginal costsand markups

In particular an upper bound for marginalcosts and a lower bound for the markups (zero)can be derived under the assumption of perfectcompetition While this assumption is clearlyunrealistic in the pharmaceuticals market it isuseful in providing an upper bound for costs ci

Uwhich will be given by

(12) ciU pi

On the other hand a lower bound for marginalcosts ci

L (and upper bound for markups) can bederived by assuming that there is perfect collu-sion within each product group i (where i refershere to a moleculedomestic-foreign combina-tion)19 and ignoring price controls so thatprices are determined by the first-order condi-tion of the jointly profit-maximizing firmswithin product group i Solving this first-ordercondition for the (lower bound of) marginal costthen gives

(13) ciL pi 1

1

iipipj

19 The reason that this assumption will lead to an under-statement of marginal costs (and hence overstatement ofmarket power) is that it assumes away competition amongfirms within each drug group

1490 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

where ii(pipj) the own-price elasticity of de-

mand for product i will depend on the productrsquosown-price pi and all other productsrsquo prices pjwith i j

Once we have obtained the demand elastic-ities through estimation of the demand sys-tem we can calculate the upper and lowerbounds for marginal costs and correspondingmarkups according to (12) or (13) These willthen be employed in the counterfactual sim-ulations While our counterfactuals use boththe lower and upper bounds for costs most ofour discussion will be based on using thelower bounds for costs since this is the moreinteresting case in the policy analysis it givesus the largest possible profits for pharmaceuticalfirms and hence corresponds to the worst pos-sible scenario facing Indian firms and the bestpossible scenario facing multinationals underTRIPS

C Identification Assumptions and EstimationApproach

The discussion of the demand system has sofar abstracted from the issue of price endogene-ity The usual premise in the industrial organi-zation literature is that correlation of prices withthe error term in the demand equation arises byvirtue of the first-order conditions of profit-maximizing firms20 As we discuss in this sec-tion this source of simultaneity bias is unlikelyto be of major concern in the present contextbecause of the existence of price regulation Ourprimary concern is instead with the simultane-ity bias that is implied by the particular wayprices are constructed

To understand the sources of the simultaneitybias note first that the price of product group piwe employ in the demand estimation should bethought of as a proxy for an exact price indexthat we do not observe As such pi contains bydefinition measurement error and it will becorrelated with the error term of the demandequation Specifically the demand equation weare interested in estimating can be written insimplified form (suppressing the subscripts rand t and ignoring parameter restrictions forconvenience) as

i i j

ijln pjT ilnXQ

PQ i

where pjT denotes the true (exact) price index for

product group i and i captures unobservedvariables that may affect demand in a particularperiod (eg an advertising campaign or an ep-idemic) The price index pj

T depends on theSKU prices pk Note that in the presence ofSKU entry and exit into the sample the priceindex pj

T will vary over time even if the pricesof the individual SKUs remain stable as theproducts (SKUs) that comprise each productgroup change each period

Ideally we would like to compute the exactprice index pj

T for each group and employ theseindices in the demand estimation Under theassumption of predetermined SKU prices eachperiod the exact price indices would be uncor-related with the error term and the demandparameters could be estimated by simple OLSUnfortunately this strategy is not feasible in thecurrent context

To derive an exact price index for each prod-uct group it is necessary to model consumerrsquoschoice among SKUs conditional on the choiceof the group To this end it is necessary toexplicitly introduce a third stage in the demandspecification and estimate the parameters asso-ciated with the choice at that stage Howeverany specification based on three-stage budget-ing (that would require directly specifying ademand function associated with the SKUchoice at the third stage and imposing additiveseparability at the higher stages) would be in-feasible to estimate for three main reasonsFirst there is a large number of SKUs withinsome groups so that the demand parameters can-not be estimated even if one imposes parameterrestrictions second the SKU prices exhibit littlevariation over time lastly the choice sets at thislowest stage vary over time rendering estimationof the price parameters associated with SKUs thatare not available in all periods infeasible Alterna-tively one could abandon the multistage budget-ing altogether and adopt a discrete choiceapproach which is inherently better suited to deal-ing with varying choice sets and limited pricevariation A discrete choice approach would beparticularly appealing for modeling the SKUchoice conditional on product group choice giventhat within each group all SKUs contain the same

20 See Steven Berry et al (1995) or Aviv Nevo (2001)for a related discussion

1491VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

molecule and are hence comparable As notedearlier however a discrete choice approach at thehigher stages of the demand estimation requiresconverting revenue to physical shares which ischallenging in the case of systemic drugs such asantibiotics for which the appropriate dosage is notwell defined Finally a combination of a multi-stage budgeting approach for the higher stages anda discrete choice approach for the lowest stagewould be feasible to estimate but inconsistentwith a model of utility maximization For thesereasons we have no alternative but to adopt anapproximation for measuring the price index foreach product group

Let us denote this approximation by pjA and let

13j denote the proportional ldquoapproximationrdquo errorassociated with measuring the true price index forgroup j so that ln pj

A ln pjT ln 13j The

estimating demand equation can then be written as

i i j

ijln pjA ilnXQ

PQ

j

ijln 13j i

or

i i j

ijln pjA ilnXQ

PQ i

where the new error term i yenj ijln 13j iis comprised of two components the first onereflects ldquomeasurementrdquo error due to the fact thatthe product group prices we employ are approx-imations and not exact price indices and thesecond one captures conventional demandshocks The demand equation as written abovecorresponds to the equation we take to the data

The particular proxy we use for the exact priceindex for each product group j is the revenue-share-weighted average of the prices of multipleSKUs available within this group that is

pjA

k j

k pk

The notation above illustrates the two source ofprice variation variation in the SKU prices pkand variation in the weights k

We are concerned with potential correlation ofthe so-constructed product group prices with bothcomponents of the product group demand error

term The correlation between pjA and the first

component yenj ijln 13j is independent of theparticular way we construct pj

A and inherent in thefact that we measure the exact price index witherror The potential correlation between pj

A and thesecond component of the product group demanderror the shock i arises however because of thespecific way we construct the proxy pj

A and inparticular because of the presence of the reve-nue share weights k in pj

A since the SKUrevenue share weights k will generally dependon the product group expenditure they arelikely to be correlated with the product groupdemand shock i

To address the simultaneity bias we use in-strumental variables To this end we need vari-ables that are correlated with the proxies pj

A butuncorrelated with the error term i We use thenumber of SKUs within each product group j asan instrument as it does justice to the idea thatvariation in the product group price index stemsin part from variation in the set of SKUs that areavailable each period It is clearly correlatedwith the average group prices pj

A and plausiblyuncorrelated with the demand shock i Theassumption that the number of SKUs is uncor-related with the conventional demand error willhowever be violated if the introduction of anew SKU changes the perceived quality of adrug21 or if it is accompanied by promotionalactivities which will be reflected in i Our hopeis that this is not too often the case Assumingthat the SKU number for each group j is alsouncorrelated with the weighted sum of the mea-surement errors of all product groups yenj ijln 13j itcan be used as an instrument for the product groupprice pj

A22 If one accepts the premise that prices atthe SKU level are exogenous then SKU pricescan also be used as instruments The final list ofinstruments we use hence includes the number ofSKUs in each group the prices of the five largestSKUs for each group and all other exogenousvariables in the demand estimation such as re-gionalproduct dummy interactions and the upper-

21 Note that we already proxy for the quality of each prod-uct group through the product fixed effects i In the actualestimation we actually let these fixed effects be region-specific(ir) so that we allow quality to vary across regions

22 If the measurement error increases in the number ofincluded SKUs this assumption will be violated Unfortu-nately we have no way of checking the validity of thisassumption

1492 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

level total expenditure on antibiotics Regressionsof group prices on the instruments above yieldhigh R-squares with most regressors highly sig-nificant indicating that our instruments are highlycorrelated with prices23

Our sample includes four molecules cipro-floxacin norfloxacin ofloxacin and spar-floxacin Except for sparfloxacin all othermolecules are produced by both foreign anddomestic firms24 So we have seven products(domestic ciprofloxacin foreign ciprofloxa-cin etc) with 96 observations (two years ofmonthly data four geographical regions foreach period) for each product The parametersin the lower-level AIDS demand system asdefined in equation (8) are the product fixedeffects i the productndashregional dummy inter-actions ir the own revenue-share price elas-ticities ii the cross revenue-share priceelasticities i10 i01 i00 and the revenue-share expenditure elasticities i In estimatingthe parameters we first regress prices on allinstrumental variables and then plug the pre-dicted values for prices in the constrainedleast-square regression (For a detailed expla-nation of the constraints see the previoussection)

Given that we impose many cross-equationconstraints and employ instrumental variablesin the estimation it is difficult to derive stan-dard errors for the parameter estimates ana-lytically Our error term interpretation in (8)implies that the error terms for each productgroup are likely to be correlated across re-gions for example if a national advertisingcampaign increases the demand for a rela-tively expensive presentation in one regionsimultaneously increasing the aggregate de-mand and price index for the correspondingproduct group in that region it is likely that thesame effect will be observed in other regions Inprinciple we could exploit the cross-regionalcorrelation in the error terms to estimate thedemand system using seemingly unrelated re-

gression (SUR) This however would requireus to derive the variance-covariance matrix ofthe parameters analytically which as notedabove is cumbersome in the current contextInstead we use the bootstrap method choosingto remain agnostic about the structure of thevariance-covariance matrix The potential dis-advantage of this approach over SUR is a loss inefficiency but as will become apparent in theresults section the model parameters are fairlyprecisely estimated To maintain the marketstructure we randomly sample the periods (withreplacement) and use the same periods for allproducts Regarding the optimal number ofbootstrap repetitions ideally one would followthe three-step method proposed by Donald W KAndrews et al (2000) Empirical evidence sug-gests however that one rarely needs more than200 replications to estimate the standard er-rors25 To be safe we generate 300 bootstrapsamples (with replacement) based on the origi-nal data and estimate the standard errors usingthe standard errors of the bootstrap sampleestimates

The estimation of the top-level AIDS systemis similar The constraints imposed on this top-level demand system are adding-up homogene-ity and symmetry as defined on page 12Again bootstrapping is used to obtain the stan-dard errors of the parameter estimates

D The Counterfactual Scenarios

In assessing the effects of patent enforce-ment we start by focusing on the most extremecase in which compulsory licensing is not anoption and foreign firms are not subject to pricecontrols We use the results from the analysis ofthis case as a benchmark In reality the outcomeof the WTO negotiations is more likely to in-volve some constraints on the monopoly powerof foreign firms selling patented products indeveloping countries such as price caps orcompulsory licensing Our framework can eas-ily accommodate these cases as will becomeapparent in the next subsection

We now focus on the effects of potentialpatent enforcement in the quinolone segmentWe consider several scenarios which vary inthe number and size of domestic products that

23 Specifically the R-squares from the first-stage regres-sions of product group prices on instruments range from057 to 095 except for domestic ciprofloxacin for whichthe R-square is 017

24 Sparfloxacin is actually offered by one foreign subsid-iary in India (see Table 2) However its revenue share isminiscule We therefore treat Sparfloxacin as being pro-duced by domestic firms only 25 See Bradley Efron and Robert J Tibshirani (1993)

1493VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

will be removed from the market In particularwe consider the following five scenarios26

Withdrawal of one large domestic productgroup only domestic ciprofloxacin

Withdrawal of one relatively small domesticproduct group only domestic ofloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand norfloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand sparfloxacin

Withdrawal of all four domestic quinoloneproduct groups

As the list above suggests we proceed fromanalyzing the effects of single-product with-drawal to the analysis of eliminating the en-tire domestic segment This approach wasmotivated by early empirical results that in-dicated that the existence and extent of com-petition from domestic firms has a significantbearing on the predicted effects of patent en-forcement that is our predictions regardingprices and welfare vary substantially depend-ing on how many domestic products are af-fected by patent enforcement and on whethersome domestic competition will remain presentafter TRIPS In addition the size of the affecteddomestic groups is relevant for the welfare pre-dictions accordingly we examine both scenar-ios in which a large domestic product groupsuch as domestic ciprofloxacin is eliminatedfrom the market and scenarios in which the

eliminated domestic group is relatively small(eg domestic ofloxacin) In all scenarios wemaintain the assumption that the set of prod-ucts offered by the remaining firms in themarket does not change in response to patentenforcement

E Computation of Virtual Prices and NewEquilibrium Prices

The first step in the counterfactual analysisis to derive the new equilibrium prices underpatent enforcement27 In this context thereare two sets of prices that are relevant Thefirst set consists of the virtual prices of those(domestic) products that will not be availableonce TRIPS is put in effect To calculate thesevirtual prices we set the expenditures of therelevant products equal to zero The secondset of prices consists of the prices of thoseproducts that remain in the market In deriv-ing these prices we start by assuming profitmaximization without any form of price reg-ulation the firms remaining in the marketreoptimize in response to the policy changeand set new prices taking the prices of allother firms as given28 Of course at the equi-librium all prices change in response to thefact that some domestic products are nolonger present The new equilibrium pricesfor products that remain in the market are thuscomputed by utilizing the first-order condi-tions of profit-maximizing firms into whichthe virtual prices of the eliminated productsare substituted Hence to compute the newequilibrium prices we solve an equation sys-tem of the following form

26 As Table 2 indicates most of the patents for the drugsin the quinolone segment have expired by now so that noneof the scenarios described in this section is going to mate-rialize in practice The reason we consider these alternativescenarios in our counterfactual simulations is to get a senseof how the presence and extent of domestic competitionaffects welfare calculations Such calculations may becomerelevant in the future in other therapeutic classes in whichpatents have not expired yet and where domestic firmscompete with foreign patent holders While the particularexit scenarios will depend on the exact year of patent expiryof drugs in each therapeutic class and the specific way inwhich patents will be enforced (eg whether certain do-mestic drugs will be grandfathered in) the general conclu-sion that emerges from our calculations is that the consumerwelfare loss is substantially smaller in cases where somedomestic competition remains present in the market Hencethe particular way in which patent enforcement is imple-mented is essential for assessing its welfare impact

27 Of course until product patents are in fact introducedthese prices will not be observable Note also that we areassuming here that the range of available products will notchange with the introduction of patents

28 As mentioned above this first set of calculations ab-stracts from the existence of remaining price controls orother government regulations that would impose constraintson the firmsrsquo profit maximization problem Accordingly theresulting numbers should be interpreted as a benchmark ofwhat would happen if markets were completely unregulatedTo examine what the welfare effects of TRIPS would be inthe more realistic scenario of price regulation we subse-quently consider a scenario in which regulation keeps theprices of the products provided by patent-holders at theirpre-TRIPS level For a more detailed discussion see theresults section

1494 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

For products i that are withdrawn from themarket

(14) 0 i ir iiln pirV i10ln pjr j Di

10

j Di

01

i01ln pjr

j Di

00

i00ln pjr

ilnXQr

PQr

For products k that remain in the market

(15) pkr ckr 1 1

kk pkrpjr pirV 1

In the equations above pirV denotes the virtual

prices of the products that are removed fromthe market while pjr denotes the updatedprices of all other products Note that whensolving for the virtual prices we account forthe fact that both the price index for quino-lones PQr and the expenditure allocated tothis subsegment XQr need to be updated toreflect the fact that as a result of the pricechanges there may be substitution away fromthis subsegment To obtain the new quinoloneexpenditure XQr and the new price index PQrwe use the estimates and formulas for thehigher-level AIDS system In equation (15)ck refers to the marginal cost for product k thatwe have obtained from the previous estimationstage As mentioned in the previous subsectionwe conducted the simulations using both theupper and lower bounds for marginal cost ck

u

and ckL respectively The term kk(pkr

pjrpirV) re-

fers to the unconditional own-price elasticity forproduct k which is a function of the eliminatedproductsrsquo virtual prices and the remaining prod-uctsrsquo new equilibrium prices We conduct thecounterfactual simulations and welfare analysisat the regional level because the presence ofregion-specific product effects in the demandestimation implies region-specific demand elas-ticities and hence region-specific marginalcosts and markups The presentation and discus-sion of our results focus on national averages ofthe relevant variables that we construct by com-

puting weighted averages across regions usingthe population of each region as a weight

F Welfare Assessment

The simulation of the new equilibrium underpatent protection can provide important insightsinto how consumers and firms will respond tothe removal of domestic products in the market(for example toward which products consumerswill substitute or which prices will increase themost) To get a more precise idea of how peo-plersquos well-being will ultimately be affected byTRIPS we compute as a last step in our anal-ysis the welfare effects of the policy changeSocial welfare is defined as consumer welfareand the sum of domestic firm profits Thechange in domestic profits can easily be calcu-lated by comparing the domestic firm (variable)profits at the pre-TRIPS prices to the profitsthese firms will realize at the new simulatedprices Although foreign firm profits do notcount in domestic welfare calculations we alsocompute the effects of patent enforcement onforeign firm profits to get an idea of how largethe expected benefits of TRIPS are for thesefirms This provides in some sense an indirectway of assessing whether the claims that patentenforcement in countries like India will lead tomore research on developing-country-specificdiseases (such as malaria) have any validity iffor example we find that the effect of patentenforcement on the foreign firm profits realizedin India is small in magnitude it is unlikely thatforeign firms will engage in more developing-country-specific research in response to TRIPSIt is important to note that in all these calcula-tions we work with the lower-bound estimatesfor marginal costs since these give rise to thehighest possible markups Hence our estimateof profit loss for domestic producers most likelyoverstates this loss

On the consumer side we measure changes inconsumer welfare by the compensating variation(CV) defined as the additional expenditure thatconsumers need in order to achieve the sameutility level as before patent enforcement at thenew prices Specifically let P0 denote the pricevector before patent enforcement P the simulatedprice vector post-TRIPS (that we obtained usingthe methods described in the previous subsection)u0 the utility attained by consumers before TRIPS

1495VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

and E(u P) the higher-level expenditure functionThen the compensating variation is given by

(16) CV Eu0 P Eu0 P0

Note that the CV as computed in (16) representsthe combination of three effects

The pure product-variety effect that is theeffect that arises because one or more prod-ucts are not available to consumers anymoreholding the prices of all other remainingproducts and the total expenditure on thequinolone subsegment XQ constant

The expenditure-switching effect that is theeffect arising from substitution away fromquinolones and toward other subsegments ofthe antibiotics market again holding the pricesof all other remaining products constant

The reduced-competition effect that is theeffect that arises because the firms remainingin the market adjust (increase) their prices inresponse to the removal of domestic products

From both an analytical and a policy point ofview it is desirable to assess how large each ofthese effects is Accordingly we decompose thetotal effect on consumer welfare (the CV as givenby equation (16)) using the following procedure

To get the pure product variety effect wecompute virtual prices for the products that areremoved from the market holding the quino-lone expenditure XQ and the prices of all otherproducts fixed Let us call the resulting pricevector P1 Then the pure product variety effectis represented by E(u0 P1) E(u0 P0)

To compute the expenditure-switching effectwe compute another set of virtual prices againholding the prices of all remaining products fixedbut letting quinolone expenditure adjust in re-sponse to the new price index for the quinolonesegment (Given that the prices of the remainingproducts remain fixed the change in the priceindex arises only because of the removal of one ormore domestic products) Note that this scenariomost closely resembles the case in which patentlaws are enforced but strict price regulation keepsthe prices of the products offered in the market attheir pre-TRIP level Let us label the so-computedprice vector P2 The expenditure-switching effectis then E(u0 P2) E(u0 P1)

Finally the reduced competition effect aris-ing from higher prices for the remaining prod-

ucts is computed as the residual change in thecompensating variation once the product-varietyand expenditure-switching effects have been ac-counted for that is E(u0 P) E(u0 P2)where the price vector P is computed accordingto the formulas (14) and (15) to reflect theadjustment of prices to the new regime

To compute the standard errors associatedwith the counterfactual simulations (that is thestandard errors for the counterfactual prices andwelfare estimates) we again use bootstrappingWe first bootstrap the original sample (sales andprices of all the drugs) 300 times Next we esti-mate the AIDS model for each of these samplesand compute the counterfactual equilibrium pricesand welfare losses corresponding to each policyscenario for each of the 300 simulations In thefinal step we compute standard errors29

One limitation of our framework is that itdoes not allow for a heterogeneous response ofconsumers to the policy change Accordinglyour framework is not suited to addressing thequestion of how different groups in the popula-tion will be affected by patent law enforce-ment30 Along the same lines our frameworkdoes not accommodate the possibility of pricediscrimination which might lead to differentresults regarding welfare losses and profit gainsrelative to uniform pricing31 Still we believethat the results of the counterfactual simulationscan provide important insights into the likelyaggregate response to patent enforcement andthe factors that drive this response

III Results

A The Structure of Demand

Tables A2 and A3 in the Appendix display theresults from estimation of the lower- and upper-level AIDS systems respectively For ease of in-terpretation rather than discussing the coefficientestimates we focus our discussion on the implied

29 In cases some of the simulations do not converge wecompute the standard errors using only those simulationsthat converged

30 Addressing this question would require at a mini-mum micro data on consumer purchases as in Goldberg(1995) To our knowledge such data do not exist for theIndian pharmaceutical market

31 See Ernst R Berndt (1994) for an extensive discussionof uniform pricing

1496 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

unconditional price and expenditure elasticities re-ported in Table 6 Given that the region-specificproduct group effects ir reported in the last fourcolumns of Tables A2 and A3 imply region-spe-cific demand elasticities we report separate elas-ticities for each region in Tables 6A and 6B Asevident from the comparison of these tables theelasticities are very similar across regions so thatwe can focus the remaining discussion on theelasticities of one region only the northern regiondisplayed in Table 6A

The diagonal terms of Table 6A report theown-price elasticities which are in all but onecase negative and highly significant The oneexception is the foreign norfloxacin productgroupmdashwhose share of quinolone sales is 007percentmdashfor which we estimate a negative butinsignificant own-price elasticity For the remain-ing product groups demand appears to be highlyelastic with the estimated elasticities lower than2 in four out of the six cases The magnitude ofthe own-price elasticities matches the features ofthe Indian pharmaceutical market mentioned ear-lier which would suggest that Indian consumersare likely to be quite price-sensitive32 The elas-ticities appear especially large if one takes intoaccount that they refer to product groups (such asdomestic ciprofloxacin) and not individual drugsoffered by particular firms Their relative magni-tudes are also intuitive the drug with the largestmarket share and a relatively high price (domesticciprofloxacin) appears to be one of the least elas-tic In contrast foreign ciprofloxacin is highlyprice elastic this is plausible as ciprofloxacindrugs offered by subsidiaries of multinationalsface the stiffest competition in this market seg-ment from approximately 75 Indian firms offeringthe same molecule

The estimated expenditure elasticities appear inthe last column These are all positive indicatingthat the demand for all product groups is normalThe remaining cells display the estimated cross-price elasticities As one might perhaps expect forproducts within a therapeutic subsegment theseare mostly positive Out of a total of 49 priceelasticities we estimate there are six that do notconform to expectations these are the cross-price elasticities between different foreign prod-

uct groups which are estimated to be negativeand significant33 Fortunately these elasticitieshave negligible impact on the welfare analysisgiven that our counterfactuals focus on theeffect of withdrawing one or more domesticproducts from the market the most relevantelasticities are the ones that capture the responseof various product group shares to a change in theprice of one or more domestic groups these elas-ticities are the cross-price elasticities between var-ious domestic groups and the ones betweendomestic and foreign groups which are plausibleand precisely estimated34

Regarding these elasticities a striking aspectof our estimates is how large positive andsignificant the cross-price elasticities betweendifferent domestic product groups aremdashin factfor norfloxacin and ofloxacin we estimate thatdomestic product groups containing differentmolecules are closer substitutes for one anotherthan product groups that contain the same mol-ecule but are produced by foreign firms Incontrast for ciprofloxacin (the molecule withthe largest revenue share) we estimate a largepositive cross-price elasticity between the do-mestic and foreign versions

The fact that domestic products appear to beclose substitutes for other domestic productsthat contain different molecules truly representsan ldquoempiricalrdquo finding in the sense that we do

32 In developed economies elasticities of this magnitude havetypically been found only for generic drugs (and even then onlyrarely) or among consumers who lack health insurance

33 While these elasticities are clearly counterintuitive theyare not inconsistent with the underlying demand system whichimposes no restrictions on the sign of the cross-price elastici-ties We do not have a good explanation of why these elastic-ities are estimated to be negative A potential explanation isthat the shares of the foreign products are very small giventhis we observe very few consumers switching from one (verysmall) foreign group to another (very small) foreign groupwhen the price of the first foreign group goes up and hence theinference is not very reliable in this case

34 The cross-price elasticities between foreign drug groupscontaining different molecules will also have an effect on thewelfare estimates given that the withdrawal of domestic prod-ucts will generally lead to changes in the prices of all foreignproducts and market shares will be reallocated from someforeign products to other foreign products based on the newprices However this reallocation from ldquoforeign to foreignrdquo istruly second order in our case compared to the reallocationfrom ldquodomestic to domesticrdquo and ldquodomestic to foreignrdquo prod-ucts In scenarios in which we consider patent enforcementaccompanied by strict price regulation the cross-price elastic-ities between foreign product groups are in fact completelyirrelevant as the prices of foreign products are not allowed toincrease in these cases Still our welfare loss estimates remainsubstantial drivenmdashas beforemdashby the loss of product variety

1497VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

not impose it through any of our assumptionsregarding the demand function The questionthat naturally arises then is what might explainthis finding While we cannot formally addressthis question anecdotal accounts in various in-dustry studies suggest that the explanation maylie in the differences between domestic andforeign firms in the structure and coverage ofretail distribution networks

Distribution networks for pharmaceuticals inIndia are typically organized in a hierarchicalfashion Pharmaceutical companies deal mainlywith carrying and forwarding (CampF) agents inmany instances regionally based who each sup-ply a network of stockists (wholesalers) Thesestockists in turn deal with the retail pharma-cists through whom retail sales ultimately oc-cur35 The market share enjoyed by a particularpharmaceutical product therefore depends inpart on the number of retail pharmacists who

stock the product And it is here that thereappears to be a distinction between domesticfirms and multinational subsidiaries In particu-lar the retail reach of domestic firms as agroup tends to be much more comprehensivethan that of multinational subsidiaries (IndianCredit Rating Agency (ICRA) 1999)36

There appear to be two reasons for this Thefirst is that many of the larger Indian firmsbecause they have a much larger portfolio ofproducts over which to spread the associatedfixed costs typically have more extensive net-works of medical representatives The second issimply that there are many more domestic firms(and products) on the market At the retail levelthis would imply that local pharmacists mightbe more likely to stock domestic products con-taining two different molecules say ciprofloxa-cin and norfloxacin than they would domesticand foreign versions of the same molecule Tothe extent that patients (or their doctors) arewilling to substitute across molecules in order tosave on transport or search costs (eg going toanother pharmacy to check whether a particular

35 There are estimated to be some 300000 retail pharma-cists in India On average stockists deal with about 75 retailers(ICRA 1999) There are naturally variations in this structureand a host of specific exclusive dealing and other arrangementsexists in practice Pharmaceutical firms also maintain networksof medical representatives whose main function is to marketthe companyrsquos products to doctors who do the actual prescrib-ing of drugs In some instances firms do sell directly to thedoctors who then become the ldquoretailerrdquo as far as patients areconcerned but these are relatively rare

36 These differences were also highlighted in conversa-tions one of the authors had with CEOs and managingdirectors of several pharmaceutical firms as part of a sep-arate study

TABLE 6AmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN THE NORTHERN REGION

Product group

Elasticity with respect to

Prices of foreign productgroups Prices of domestic product groups Overall

quinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 557 013dagger 015 401 011dagger 011dagger 016 137(179) (007) (007) (184) (006) (006) (006) (029)

Foreign norfloxacin 427dagger 045 427dagger 350dagger 602 451 465 220(242) (112) (242) (210) (623) (184) (183) (105)

Foreign ofloxacin 011 010dagger 138 009 009dagger 023 011 116(005) (005) (031) (027) (005) (028) (004) (017)

Domestic ciprofloxacin 018 001 001 168 008 008 010 117(008) (000) (001) (023) (002) (002) (002) (003)

Domestic norfloxacin 004 003 004 058 223 042 040 073(001) (003) (001) (017) (011) (004) (003) (009)

Domestic ofloxacin 005 005 011 077 074 342 074 089(002) (002) (013) (028) (008) (025) (008) (021)

Domestic sparfloxacin 007 004 007 115 063 063 288 028(002) (001) (002) (015) (006) (006) (017) (012)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent significance level and dagger (dagger) denotes significance at the 10-percent level

1498 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE 6BmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN OTHER REGIONS

Demand patterns in the eastern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 594 014dagger 016 431 013 011 017 140(195) (008) (008) (201) (006) (007) (007) (031)

Foreign norfloxacin 329dagger 058 329dagger 264dagger 460 346 359 192(180) (083) (179) (153) (451) (138) (139) (082)

Foreign ofloxacin 012 010dagger 140 010 010 024 013 117(006) (006) (034) (031) (005) (030) (005) (019)

Domestic ciprofloxacin 018 001 001 172 009 008 011 117(008) (000) (002) (027) (002) (003) (002) (003)

Domestic norfloxacin 004 004 004 068 242 049 046 070(002) (004) (002) (020) (013) (005) (004) (010)

Domestic ofloxacin 004dagger 004 009 061 060 295 060 092(002) (001) (010) (028) (007) (022) (007) (017)

Domestic sparfloxacin 005 003 005 092 048 051 251 043(002) (001) (001) (017) (005) (006) (016) (010)

Demand patterns in the western region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 427 009dagger 011 286 008 006 011 126(129) (005) (005) (133) (004) (005) (004) (021)

Foreign norfloxacin 714dagger 008 710dagger 594dagger 995 742 770 299dagger

(369) (183) (368) (328) (1015) (297) (296) (178)Foreign ofloxacin 013 012dagger 145 009 012 026 013 119

(007) (006) (037) (029) (005) (033) (005) (021)Domestic ciprofloxacin 019 001 000 174 009 008 011 118

(008) (000) (001) (024) (002) (003) (002) (004)Domestic norfloxacin 004 004 004 067 243 050 047 069

(002) (004) (001) (018) (014) (005) (004) (010)Domestic ofloxacin 003 003 007 048 048 257 048 093

(002) (001) (008) (022) (005) (015) (004) (014)Domestic sparfloxacin 006 003 005 083 046 049 241 046

(002) (001) (001) (014) (004) (005) (012) (009)

Demand patterns in the southern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 560 013dagger 015 402 011 010 017 137(182) (007) (007) (187) (006) (006) (007) (029)

Foreign norfloxacin 533dagger 031 531dagger 431dagger 748 559 584 249dagger

(301) (134) (300) (245) (734) (219) (219) (128)Foreign ofloxacin 014 012dagger 148 011 012 029 016 120

(007) (007) (039) (033) (006) (035) (006) (022)Domestic ciprofloxacin 017 001 000 169 008 007 011 116

(008) (000) (001) (025) (002) (003) (002) (003)Domestic norfloxacin 004 003 003 059 225 043 040 073

(001) (003) (001) (019) (012) (004) (004) (009)Domestic ofloxacin 004dagger 004 009 062 060 296 060 091

(002) (001) (010) (026) (006) (019) (007) (017)Domestic sparfloxacin 008 005 008 138 073 075 323 016

(002) (001) (002) (018) (007) (008) (021) (014)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at 5-percent level anddagger (dagger) denotes significance at the 10-percent level

1499VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

foreign product is in stock) in aggregate datawe would expect to find precisely the substitu-tion patterns that we report in Table 6

Whether the particular explanation we provideabove is the correct one the high degree of sub-stitutability between domestic product groupsturns out to have important implications for thewelfare calculations We discuss these in moredetail below when we present the results of thecounterfactual welfare analysis Another elasticitywith important implications for the counterfactu-als is the price elasticity for the quinolone subseg-ment as a whole which indicates how likelyconsumers are to switch to other antibioticsgroups when faced with a price increase for quin-olones This elasticity is computed on the basis ofthe results in Table A3 and it is at 111 (stan-dard error 024) this is large in magnitudebutmdashas expectedmdashsmaller in absolute value thanthe own-price elasticities of the product groupswithin the quinolone subsegment

The results in Tables 6A and 6B are based onour preferred specification discussed in SectionII In Tables A4 to A6 in the Appendix weexperiment with some alternative specificationsTables A4(a)ndashA4(c) correspond to a specificationthat includes in addition to product-group-specificregional fixed effects product-group-specific (andfor the upper level antibiotics-segment-specific)seasonal effects We distinguish among three sea-sonsmdashthe summer monsoon and wintermdashandreport the unconditional demand elasticities for

the northern region for each of these seasons Asevident from the tables in the Appendix our elas-ticity estimates are robust to the inclusion of sea-sonal effects The demand elasticities in Table A5are based on estimation of the demand system byOLS Compared to the elasticities obtained by IVthe OLS elasticities are smaller in absolute valueimplying that welfare calculations based on theOLS estimates would produce larger welfare lossestimates Nevertheless some of the patterns re-garding the cross-price elasticities discussed ear-lier are also evident in the OLS results inparticular the cross-price elasticities between dif-ferent domestic product groups are all positivelarge and significant and in most instances largerthan the cross-price elasticities between drugs thatcontain the same molecule but are produced byfirms of different domesticforeign status Theclose substitutability of domestic products indi-cated by both the OLS and IV estimates seems tobe one of the most robust findings of the paper

B Cost and Markup Estimates

Table 7 displays the marginal costs markupsand profits implied by the price elasticity esti-mates of Tables 6A and 6B for each of the sevenproduct groups Given that our regional effectsimply different price elasticities for each regionour marginal cost and markup estimates alsodiffer by region Given however that based on

TABLE 7mdashUPPER AND LOWER BOUNDS FOR MARGINAL COST MARKUP AND ANNUAL PROFIT BY PRODUCT GROUPS WITHIN

THE QUINOLONE SUBSEGMENT

Product groupLower boundfor MC (Rs)

Upper boundfor markup

Upper boundfor profit(Rs mill)

Upper boundfor MC (Rs)

Lower boundfor markup

Lower boundfor profit

(Rs)

Foreign ciprofloxacin 83 19 269 103 0 00(123) (012) (1655)

Foreign norfloxacin NA NA NA 53 0 00Foreign ofloxacin 323 70 1061 1085 0 00

(2316) (021) (3185)Domestic ciprofloxacin 47 59 17019 112 0 00

(114) (010) (29858)Domestic norfloxacin 52 43 2807 90 0 00

(020) (002) (1532)Domestic ofloxacin 587 34 1612 901 0 00

(218) (002) (1280)Domestic sparfloxacin 495 37 1985 788 0 00

(157) (002) (1100)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level Estimated lower bound forforeign norfloxacinrsquos marginal cost is negative since the estimated price elasticity is less than one in absolute value

1500 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Tables 6A and 6B the price elasticities do notseem to differ substantially across regions wereport for ease of exposition only the nationalaverages for marginal costs and markups inTable 7

Table 7 has two parts In the left part (firstthree columns) we report the lower bound formarginal cost and the corresponding upperbound for markup and upper bound for totalannual profit for each product group Thesenumbers are based on the price elasticities weobtained from estimating the two-level AIDSdemand system Since we do not have a reliableestimate for the price elasticity of foreign nor-floxacin (the point estimate is negative but lessthan one in absolute value and insignificant)we cannot compute the lower bound for mar-ginal cost in this case It is important to note thatthese estimates do not reflect either the actualmarginal cost or the actual markup for thesedrugs both because the existence of price reg-ulation implies that the unconstrained first-orderconditions are unlikely to hold each period andbecause our aggregation across firms of thesame domesticforeign status supplying thesame molecule makes the interpretation of theseestimates problematic In particular the fact thatwe ignore competition among firms within eachproduct group implies that our estimates willtend to overstate market power These numberswill prove useful however in the counterfac-tual simulations as they can give us a sense ofhow large the maximum profit gains for multi-national firms and the maximum profit lossesfor domestic firms are likely to be under patentenforcement The right part of the table (lastthree columns) reports the upper bound for mar-ginal cost which we obtain by simply taking themarginal cost to be equal to the observed priceWe do not report standard errors in this casesince the numbers are based on actual data Themarkups corresponding to these marginal costupper bounds are of course zero We conductthe counterfactual simulations using both thelower and upper bounds for marginal costs

The (upper bounds for) markups on the leftside of the table are generally plausible Thedomestic ciprofloxacin product group that dom-inates the quinolone subsegment and for whichwe documented high prices a high marketshare and a relatively low elasticity of demandenjoys one of the highest markup upper bounds(60 percent) and accounts for nearly 70 percent

of all profits derived within the subsegmentForeign ciprofloxacin on the other hand whichfaces the stiffest competition from domesticfirms and for which we estimated a highly elas-tic demand has the lowest markup upper bound(19 percent)

C Counterfactual Estimates of the Impact onPrices and Welfare

With estimates of the key demand and costparameters in hand we turn to the counterfac-tuals We consider the five separate scenarioslisted in the previous section All of the scenar-ios involve the withdrawal of one or more of thedomestic product groups from the market Ta-ble 8 displays our estimates of the consumerwelfare losses that result under the differentscenarios The losses are expressed in billion ofrupees per year All numbers presented in Ta-ble 8 and subsequent tables are based on usingthe lower bounds for marginal cost and upperbounds for markup in the simulations As dis-cussed earlier these numbers are the more in-teresting to work with since they give us anupper bound for the changes in the profits ofdomestic and foreign firms that would resultfrom patent enforcement In the Appendix wealso present results based on using the upperbounds for marginal costs in the simulations inwhich case the pre-TRIPS profits of domesticand foreign firms are zero In all cases marginalcosts are assumed to be constant in outputWhile naturally the profit implications differdepending on whether one uses the upper orlower bounds for marginal costs (firm profits arezero if one assumes the upper bound of mar-ginal cost in which case price equals marginalcost) the estimated consumer welfare losses aresimilar in the two cases We discuss these re-sults in more detail at the end of this section

The first column presents our estimates of theconsumer welfare losses attributable to the pureloss of product variety effect where we fix theprices of all remaining products as well as theoverall expenditure on quinolones while with-drawing one or more of the domestic productgroups Note that in our initial specification ofthe demand system had we not allowed for thepossibility that consumers might differentiatebetween domestic and foreign products evenwhen they contain the same molecule this

1501VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

particular component of the loss of consumerwelfare would not have arisen

The estimates reported in the second columnincorporate the expenditure-switching effect ontop of the loss of product variety Here basedupon the price elasticity estimates from thehigher-level AIDS system we adjust (down-ward) the expenditures allocated to the quino-lone subsegment as the composite price ofquinolones effectively increases as a consequenceof the higher virtual prices of the domestic productgroups that are withdrawn from the market Be-cause the estimates in this column are generatedassuming that the prices of the products that re-main in the market are not adjusted upward theyprovide a sense of what consumer losses would beif the introduction of product patents were coupledwith strict price regulation aimed at maintainingprices at prepatent levels Alternatively they canbe thought of as the relevant welfare numbers ifintense competition among firms within the re-maining product groups kept the prices of theproducts that were still offered in the market closeto the firmsrsquo marginal costs

The last column of Table 8 displays the esti-mated consumer welfare losses when bothcross-segment expenditure-switching and with-in-segment upward price adjustments are takeninto account

If we compare the results across the firstsecond and third columns all the counterfac-tual scenarios produce qualitatively similar pat-terns patterns that are consistent with what we

would expect Starting from the initial loss ofwelfare attributable to the loss of product vari-ety the option of switching expenditures outof the quinolone subsegment to other subseg-ments mitigates some of the initial welfareloss But if we then incorporate the upwardprice adjustments that result in response to thereduced competition the welfare losses aremagnified

Of particular interest from a policy perspec-tive are the relative magnitudes of these threeeffects which are similar under all the counter-factual scenarios though absolute levels varyconsiderably First despite the fact that the de-mand for quinolones is quite sensitive to thecomposite price of quinolonesmdashthe upper levelprice elasticity is 111mdashthe cross-subsegmentexpenditure-switching effects are in all thecases small (in absolute value terms) relative tothe other two effects For instance under thescenario where all the domestic quinolone prod-uct groups are withdrawn from the market theoverall consumer welfare loss of Rs 1781 bil-lion per year can be decomposed into an initialloss of Rs 1176 billion (66 percent) attribut-able to the loss of product variety a slightreduction in this initial loss of Rs 041 billion(2 percent) from Rs 1176 billion to Rs1135 billion because of expenditure switchingand a subsequent additional loss of Rs 646billion (36 percent) from Rs 1135 billion toRs 1781 billion because of the reduced com-petition and consequent price increases

TABLE 8mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure loss ofvariety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 732(087) (089) (146)

Only ofloxacin 008 008 023(008) (008) (010)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1253(177) (180) (415)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1058(142) (145) (331)

All four domestic quinolones products 1176dagger 1135dagger 1781(643) (634) (1270)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent significance level and dagger (dagger)denotes significance at the 10-percent level

1502 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 14: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

implied by utility maximization as we do withthe estimation of the lower-level AIDS We donot however impose any additional restrictionson the substitution patterns at this stage so thatthe cross-price effects across segments remainrelatively unconstrained

Estimation of the higher-level AIDS allowsus to obtain the unconditional own- and cross-price elasticities that are used in the formulationof the supply problem and welfare analysisThese are given by the formula

(10) ij ijXQXQ

ln qi

ln XQ

ln XQ

ln PQ

ln PQ

ln pj

with the conditional cross price elasticitiesgiven by

ijXQXQ ij ij jlnXQ

PQ

i

As in the lower stage we include subsegment-specific regional dummies in the specificationof the upper-stage demand system so that thefinal form of the estimating equation at theupper stage becomes

(11) Grt G Gr H

GHln PHrt

GlnXrt

Prt Grt

In sum the demand system we take to thedata is represented by equations (8) and (11)and the associated parameter restrictions im-plied by economic theory

B Modelling the Supply Side of the Market

Counterfactual simulations concerning theeffects of domestic product withdrawal requireknowledge of the marginal costs of pharmaceu-tical firms operating in the Indian market Theseare unobservable The usual approach in theNew Empirical Industrial Organization litera-ture has been to exploit the firm equilibriumconditions to infer marginal cost For exampleit is often assumed that the marginal cost ci isconstant and that the industry is an oligopoly

engaging in Bertrand competition with differen-tiated products Assuming that firms myopicallymaximize profits each period one can then de-rive the firmsrsquo first-order conditions that corre-spond to the assumptions above about costsmarket structure and firm behavior

We deviate from this procedure as the presenceof price regulation renders the assumption of un-constrained period-by-period maximization un-tenable Ideally one would like to explicitlyincorporate the price and other administrative con-trols into the firmrsquos optimization problem andderive the first-order conditions under the assump-tion of constrained maximization The complexityof the price regulation however makes this ap-proach infeasible Therefore we adopt an alterna-tive approach that does not rely on modelling theprice-setting process but is instead based on de-riving upper and lower bounds for marginal costsand markups

In particular an upper bound for marginalcosts and a lower bound for the markups (zero)can be derived under the assumption of perfectcompetition While this assumption is clearlyunrealistic in the pharmaceuticals market it isuseful in providing an upper bound for costs ci

Uwhich will be given by

(12) ciU pi

On the other hand a lower bound for marginalcosts ci

L (and upper bound for markups) can bederived by assuming that there is perfect collu-sion within each product group i (where i refershere to a moleculedomestic-foreign combina-tion)19 and ignoring price controls so thatprices are determined by the first-order condi-tion of the jointly profit-maximizing firmswithin product group i Solving this first-ordercondition for the (lower bound of) marginal costthen gives

(13) ciL pi 1

1

iipipj

19 The reason that this assumption will lead to an under-statement of marginal costs (and hence overstatement ofmarket power) is that it assumes away competition amongfirms within each drug group

1490 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

where ii(pipj) the own-price elasticity of de-

mand for product i will depend on the productrsquosown-price pi and all other productsrsquo prices pjwith i j

Once we have obtained the demand elastic-ities through estimation of the demand sys-tem we can calculate the upper and lowerbounds for marginal costs and correspondingmarkups according to (12) or (13) These willthen be employed in the counterfactual sim-ulations While our counterfactuals use boththe lower and upper bounds for costs most ofour discussion will be based on using thelower bounds for costs since this is the moreinteresting case in the policy analysis it givesus the largest possible profits for pharmaceuticalfirms and hence corresponds to the worst pos-sible scenario facing Indian firms and the bestpossible scenario facing multinationals underTRIPS

C Identification Assumptions and EstimationApproach

The discussion of the demand system has sofar abstracted from the issue of price endogene-ity The usual premise in the industrial organi-zation literature is that correlation of prices withthe error term in the demand equation arises byvirtue of the first-order conditions of profit-maximizing firms20 As we discuss in this sec-tion this source of simultaneity bias is unlikelyto be of major concern in the present contextbecause of the existence of price regulation Ourprimary concern is instead with the simultane-ity bias that is implied by the particular wayprices are constructed

To understand the sources of the simultaneitybias note first that the price of product group piwe employ in the demand estimation should bethought of as a proxy for an exact price indexthat we do not observe As such pi contains bydefinition measurement error and it will becorrelated with the error term of the demandequation Specifically the demand equation weare interested in estimating can be written insimplified form (suppressing the subscripts rand t and ignoring parameter restrictions forconvenience) as

i i j

ijln pjT ilnXQ

PQ i

where pjT denotes the true (exact) price index for

product group i and i captures unobservedvariables that may affect demand in a particularperiod (eg an advertising campaign or an ep-idemic) The price index pj

T depends on theSKU prices pk Note that in the presence ofSKU entry and exit into the sample the priceindex pj

T will vary over time even if the pricesof the individual SKUs remain stable as theproducts (SKUs) that comprise each productgroup change each period

Ideally we would like to compute the exactprice index pj

T for each group and employ theseindices in the demand estimation Under theassumption of predetermined SKU prices eachperiod the exact price indices would be uncor-related with the error term and the demandparameters could be estimated by simple OLSUnfortunately this strategy is not feasible in thecurrent context

To derive an exact price index for each prod-uct group it is necessary to model consumerrsquoschoice among SKUs conditional on the choiceof the group To this end it is necessary toexplicitly introduce a third stage in the demandspecification and estimate the parameters asso-ciated with the choice at that stage Howeverany specification based on three-stage budget-ing (that would require directly specifying ademand function associated with the SKUchoice at the third stage and imposing additiveseparability at the higher stages) would be in-feasible to estimate for three main reasonsFirst there is a large number of SKUs withinsome groups so that the demand parameters can-not be estimated even if one imposes parameterrestrictions second the SKU prices exhibit littlevariation over time lastly the choice sets at thislowest stage vary over time rendering estimationof the price parameters associated with SKUs thatare not available in all periods infeasible Alterna-tively one could abandon the multistage budget-ing altogether and adopt a discrete choiceapproach which is inherently better suited to deal-ing with varying choice sets and limited pricevariation A discrete choice approach would beparticularly appealing for modeling the SKUchoice conditional on product group choice giventhat within each group all SKUs contain the same

20 See Steven Berry et al (1995) or Aviv Nevo (2001)for a related discussion

1491VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

molecule and are hence comparable As notedearlier however a discrete choice approach at thehigher stages of the demand estimation requiresconverting revenue to physical shares which ischallenging in the case of systemic drugs such asantibiotics for which the appropriate dosage is notwell defined Finally a combination of a multi-stage budgeting approach for the higher stages anda discrete choice approach for the lowest stagewould be feasible to estimate but inconsistentwith a model of utility maximization For thesereasons we have no alternative but to adopt anapproximation for measuring the price index foreach product group

Let us denote this approximation by pjA and let

13j denote the proportional ldquoapproximationrdquo errorassociated with measuring the true price index forgroup j so that ln pj

A ln pjT ln 13j The

estimating demand equation can then be written as

i i j

ijln pjA ilnXQ

PQ

j

ijln 13j i

or

i i j

ijln pjA ilnXQ

PQ i

where the new error term i yenj ijln 13j iis comprised of two components the first onereflects ldquomeasurementrdquo error due to the fact thatthe product group prices we employ are approx-imations and not exact price indices and thesecond one captures conventional demandshocks The demand equation as written abovecorresponds to the equation we take to the data

The particular proxy we use for the exact priceindex for each product group j is the revenue-share-weighted average of the prices of multipleSKUs available within this group that is

pjA

k j

k pk

The notation above illustrates the two source ofprice variation variation in the SKU prices pkand variation in the weights k

We are concerned with potential correlation ofthe so-constructed product group prices with bothcomponents of the product group demand error

term The correlation between pjA and the first

component yenj ijln 13j is independent of theparticular way we construct pj

A and inherent in thefact that we measure the exact price index witherror The potential correlation between pj

A and thesecond component of the product group demanderror the shock i arises however because of thespecific way we construct the proxy pj

A and inparticular because of the presence of the reve-nue share weights k in pj

A since the SKUrevenue share weights k will generally dependon the product group expenditure they arelikely to be correlated with the product groupdemand shock i

To address the simultaneity bias we use in-strumental variables To this end we need vari-ables that are correlated with the proxies pj

A butuncorrelated with the error term i We use thenumber of SKUs within each product group j asan instrument as it does justice to the idea thatvariation in the product group price index stemsin part from variation in the set of SKUs that areavailable each period It is clearly correlatedwith the average group prices pj

A and plausiblyuncorrelated with the demand shock i Theassumption that the number of SKUs is uncor-related with the conventional demand error willhowever be violated if the introduction of anew SKU changes the perceived quality of adrug21 or if it is accompanied by promotionalactivities which will be reflected in i Our hopeis that this is not too often the case Assumingthat the SKU number for each group j is alsouncorrelated with the weighted sum of the mea-surement errors of all product groups yenj ijln 13j itcan be used as an instrument for the product groupprice pj

A22 If one accepts the premise that prices atthe SKU level are exogenous then SKU pricescan also be used as instruments The final list ofinstruments we use hence includes the number ofSKUs in each group the prices of the five largestSKUs for each group and all other exogenousvariables in the demand estimation such as re-gionalproduct dummy interactions and the upper-

21 Note that we already proxy for the quality of each prod-uct group through the product fixed effects i In the actualestimation we actually let these fixed effects be region-specific(ir) so that we allow quality to vary across regions

22 If the measurement error increases in the number ofincluded SKUs this assumption will be violated Unfortu-nately we have no way of checking the validity of thisassumption

1492 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

level total expenditure on antibiotics Regressionsof group prices on the instruments above yieldhigh R-squares with most regressors highly sig-nificant indicating that our instruments are highlycorrelated with prices23

Our sample includes four molecules cipro-floxacin norfloxacin ofloxacin and spar-floxacin Except for sparfloxacin all othermolecules are produced by both foreign anddomestic firms24 So we have seven products(domestic ciprofloxacin foreign ciprofloxa-cin etc) with 96 observations (two years ofmonthly data four geographical regions foreach period) for each product The parametersin the lower-level AIDS demand system asdefined in equation (8) are the product fixedeffects i the productndashregional dummy inter-actions ir the own revenue-share price elas-ticities ii the cross revenue-share priceelasticities i10 i01 i00 and the revenue-share expenditure elasticities i In estimatingthe parameters we first regress prices on allinstrumental variables and then plug the pre-dicted values for prices in the constrainedleast-square regression (For a detailed expla-nation of the constraints see the previoussection)

Given that we impose many cross-equationconstraints and employ instrumental variablesin the estimation it is difficult to derive stan-dard errors for the parameter estimates ana-lytically Our error term interpretation in (8)implies that the error terms for each productgroup are likely to be correlated across re-gions for example if a national advertisingcampaign increases the demand for a rela-tively expensive presentation in one regionsimultaneously increasing the aggregate de-mand and price index for the correspondingproduct group in that region it is likely that thesame effect will be observed in other regions Inprinciple we could exploit the cross-regionalcorrelation in the error terms to estimate thedemand system using seemingly unrelated re-

gression (SUR) This however would requireus to derive the variance-covariance matrix ofthe parameters analytically which as notedabove is cumbersome in the current contextInstead we use the bootstrap method choosingto remain agnostic about the structure of thevariance-covariance matrix The potential dis-advantage of this approach over SUR is a loss inefficiency but as will become apparent in theresults section the model parameters are fairlyprecisely estimated To maintain the marketstructure we randomly sample the periods (withreplacement) and use the same periods for allproducts Regarding the optimal number ofbootstrap repetitions ideally one would followthe three-step method proposed by Donald W KAndrews et al (2000) Empirical evidence sug-gests however that one rarely needs more than200 replications to estimate the standard er-rors25 To be safe we generate 300 bootstrapsamples (with replacement) based on the origi-nal data and estimate the standard errors usingthe standard errors of the bootstrap sampleestimates

The estimation of the top-level AIDS systemis similar The constraints imposed on this top-level demand system are adding-up homogene-ity and symmetry as defined on page 12Again bootstrapping is used to obtain the stan-dard errors of the parameter estimates

D The Counterfactual Scenarios

In assessing the effects of patent enforce-ment we start by focusing on the most extremecase in which compulsory licensing is not anoption and foreign firms are not subject to pricecontrols We use the results from the analysis ofthis case as a benchmark In reality the outcomeof the WTO negotiations is more likely to in-volve some constraints on the monopoly powerof foreign firms selling patented products indeveloping countries such as price caps orcompulsory licensing Our framework can eas-ily accommodate these cases as will becomeapparent in the next subsection

We now focus on the effects of potentialpatent enforcement in the quinolone segmentWe consider several scenarios which vary inthe number and size of domestic products that

23 Specifically the R-squares from the first-stage regres-sions of product group prices on instruments range from057 to 095 except for domestic ciprofloxacin for whichthe R-square is 017

24 Sparfloxacin is actually offered by one foreign subsid-iary in India (see Table 2) However its revenue share isminiscule We therefore treat Sparfloxacin as being pro-duced by domestic firms only 25 See Bradley Efron and Robert J Tibshirani (1993)

1493VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

will be removed from the market In particularwe consider the following five scenarios26

Withdrawal of one large domestic productgroup only domestic ciprofloxacin

Withdrawal of one relatively small domesticproduct group only domestic ofloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand norfloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand sparfloxacin

Withdrawal of all four domestic quinoloneproduct groups

As the list above suggests we proceed fromanalyzing the effects of single-product with-drawal to the analysis of eliminating the en-tire domestic segment This approach wasmotivated by early empirical results that in-dicated that the existence and extent of com-petition from domestic firms has a significantbearing on the predicted effects of patent en-forcement that is our predictions regardingprices and welfare vary substantially depend-ing on how many domestic products are af-fected by patent enforcement and on whethersome domestic competition will remain presentafter TRIPS In addition the size of the affecteddomestic groups is relevant for the welfare pre-dictions accordingly we examine both scenar-ios in which a large domestic product groupsuch as domestic ciprofloxacin is eliminatedfrom the market and scenarios in which the

eliminated domestic group is relatively small(eg domestic ofloxacin) In all scenarios wemaintain the assumption that the set of prod-ucts offered by the remaining firms in themarket does not change in response to patentenforcement

E Computation of Virtual Prices and NewEquilibrium Prices

The first step in the counterfactual analysisis to derive the new equilibrium prices underpatent enforcement27 In this context thereare two sets of prices that are relevant Thefirst set consists of the virtual prices of those(domestic) products that will not be availableonce TRIPS is put in effect To calculate thesevirtual prices we set the expenditures of therelevant products equal to zero The secondset of prices consists of the prices of thoseproducts that remain in the market In deriv-ing these prices we start by assuming profitmaximization without any form of price reg-ulation the firms remaining in the marketreoptimize in response to the policy changeand set new prices taking the prices of allother firms as given28 Of course at the equi-librium all prices change in response to thefact that some domestic products are nolonger present The new equilibrium pricesfor products that remain in the market are thuscomputed by utilizing the first-order condi-tions of profit-maximizing firms into whichthe virtual prices of the eliminated productsare substituted Hence to compute the newequilibrium prices we solve an equation sys-tem of the following form

26 As Table 2 indicates most of the patents for the drugsin the quinolone segment have expired by now so that noneof the scenarios described in this section is going to mate-rialize in practice The reason we consider these alternativescenarios in our counterfactual simulations is to get a senseof how the presence and extent of domestic competitionaffects welfare calculations Such calculations may becomerelevant in the future in other therapeutic classes in whichpatents have not expired yet and where domestic firmscompete with foreign patent holders While the particularexit scenarios will depend on the exact year of patent expiryof drugs in each therapeutic class and the specific way inwhich patents will be enforced (eg whether certain do-mestic drugs will be grandfathered in) the general conclu-sion that emerges from our calculations is that the consumerwelfare loss is substantially smaller in cases where somedomestic competition remains present in the market Hencethe particular way in which patent enforcement is imple-mented is essential for assessing its welfare impact

27 Of course until product patents are in fact introducedthese prices will not be observable Note also that we areassuming here that the range of available products will notchange with the introduction of patents

28 As mentioned above this first set of calculations ab-stracts from the existence of remaining price controls orother government regulations that would impose constraintson the firmsrsquo profit maximization problem Accordingly theresulting numbers should be interpreted as a benchmark ofwhat would happen if markets were completely unregulatedTo examine what the welfare effects of TRIPS would be inthe more realistic scenario of price regulation we subse-quently consider a scenario in which regulation keeps theprices of the products provided by patent-holders at theirpre-TRIPS level For a more detailed discussion see theresults section

1494 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

For products i that are withdrawn from themarket

(14) 0 i ir iiln pirV i10ln pjr j Di

10

j Di

01

i01ln pjr

j Di

00

i00ln pjr

ilnXQr

PQr

For products k that remain in the market

(15) pkr ckr 1 1

kk pkrpjr pirV 1

In the equations above pirV denotes the virtual

prices of the products that are removed fromthe market while pjr denotes the updatedprices of all other products Note that whensolving for the virtual prices we account forthe fact that both the price index for quino-lones PQr and the expenditure allocated tothis subsegment XQr need to be updated toreflect the fact that as a result of the pricechanges there may be substitution away fromthis subsegment To obtain the new quinoloneexpenditure XQr and the new price index PQrwe use the estimates and formulas for thehigher-level AIDS system In equation (15)ck refers to the marginal cost for product k thatwe have obtained from the previous estimationstage As mentioned in the previous subsectionwe conducted the simulations using both theupper and lower bounds for marginal cost ck

u

and ckL respectively The term kk(pkr

pjrpirV) re-

fers to the unconditional own-price elasticity forproduct k which is a function of the eliminatedproductsrsquo virtual prices and the remaining prod-uctsrsquo new equilibrium prices We conduct thecounterfactual simulations and welfare analysisat the regional level because the presence ofregion-specific product effects in the demandestimation implies region-specific demand elas-ticities and hence region-specific marginalcosts and markups The presentation and discus-sion of our results focus on national averages ofthe relevant variables that we construct by com-

puting weighted averages across regions usingthe population of each region as a weight

F Welfare Assessment

The simulation of the new equilibrium underpatent protection can provide important insightsinto how consumers and firms will respond tothe removal of domestic products in the market(for example toward which products consumerswill substitute or which prices will increase themost) To get a more precise idea of how peo-plersquos well-being will ultimately be affected byTRIPS we compute as a last step in our anal-ysis the welfare effects of the policy changeSocial welfare is defined as consumer welfareand the sum of domestic firm profits Thechange in domestic profits can easily be calcu-lated by comparing the domestic firm (variable)profits at the pre-TRIPS prices to the profitsthese firms will realize at the new simulatedprices Although foreign firm profits do notcount in domestic welfare calculations we alsocompute the effects of patent enforcement onforeign firm profits to get an idea of how largethe expected benefits of TRIPS are for thesefirms This provides in some sense an indirectway of assessing whether the claims that patentenforcement in countries like India will lead tomore research on developing-country-specificdiseases (such as malaria) have any validity iffor example we find that the effect of patentenforcement on the foreign firm profits realizedin India is small in magnitude it is unlikely thatforeign firms will engage in more developing-country-specific research in response to TRIPSIt is important to note that in all these calcula-tions we work with the lower-bound estimatesfor marginal costs since these give rise to thehighest possible markups Hence our estimateof profit loss for domestic producers most likelyoverstates this loss

On the consumer side we measure changes inconsumer welfare by the compensating variation(CV) defined as the additional expenditure thatconsumers need in order to achieve the sameutility level as before patent enforcement at thenew prices Specifically let P0 denote the pricevector before patent enforcement P the simulatedprice vector post-TRIPS (that we obtained usingthe methods described in the previous subsection)u0 the utility attained by consumers before TRIPS

1495VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

and E(u P) the higher-level expenditure functionThen the compensating variation is given by

(16) CV Eu0 P Eu0 P0

Note that the CV as computed in (16) representsthe combination of three effects

The pure product-variety effect that is theeffect that arises because one or more prod-ucts are not available to consumers anymoreholding the prices of all other remainingproducts and the total expenditure on thequinolone subsegment XQ constant

The expenditure-switching effect that is theeffect arising from substitution away fromquinolones and toward other subsegments ofthe antibiotics market again holding the pricesof all other remaining products constant

The reduced-competition effect that is theeffect that arises because the firms remainingin the market adjust (increase) their prices inresponse to the removal of domestic products

From both an analytical and a policy point ofview it is desirable to assess how large each ofthese effects is Accordingly we decompose thetotal effect on consumer welfare (the CV as givenby equation (16)) using the following procedure

To get the pure product variety effect wecompute virtual prices for the products that areremoved from the market holding the quino-lone expenditure XQ and the prices of all otherproducts fixed Let us call the resulting pricevector P1 Then the pure product variety effectis represented by E(u0 P1) E(u0 P0)

To compute the expenditure-switching effectwe compute another set of virtual prices againholding the prices of all remaining products fixedbut letting quinolone expenditure adjust in re-sponse to the new price index for the quinolonesegment (Given that the prices of the remainingproducts remain fixed the change in the priceindex arises only because of the removal of one ormore domestic products) Note that this scenariomost closely resembles the case in which patentlaws are enforced but strict price regulation keepsthe prices of the products offered in the market attheir pre-TRIP level Let us label the so-computedprice vector P2 The expenditure-switching effectis then E(u0 P2) E(u0 P1)

Finally the reduced competition effect aris-ing from higher prices for the remaining prod-

ucts is computed as the residual change in thecompensating variation once the product-varietyand expenditure-switching effects have been ac-counted for that is E(u0 P) E(u0 P2)where the price vector P is computed accordingto the formulas (14) and (15) to reflect theadjustment of prices to the new regime

To compute the standard errors associatedwith the counterfactual simulations (that is thestandard errors for the counterfactual prices andwelfare estimates) we again use bootstrappingWe first bootstrap the original sample (sales andprices of all the drugs) 300 times Next we esti-mate the AIDS model for each of these samplesand compute the counterfactual equilibrium pricesand welfare losses corresponding to each policyscenario for each of the 300 simulations In thefinal step we compute standard errors29

One limitation of our framework is that itdoes not allow for a heterogeneous response ofconsumers to the policy change Accordinglyour framework is not suited to addressing thequestion of how different groups in the popula-tion will be affected by patent law enforce-ment30 Along the same lines our frameworkdoes not accommodate the possibility of pricediscrimination which might lead to differentresults regarding welfare losses and profit gainsrelative to uniform pricing31 Still we believethat the results of the counterfactual simulationscan provide important insights into the likelyaggregate response to patent enforcement andthe factors that drive this response

III Results

A The Structure of Demand

Tables A2 and A3 in the Appendix display theresults from estimation of the lower- and upper-level AIDS systems respectively For ease of in-terpretation rather than discussing the coefficientestimates we focus our discussion on the implied

29 In cases some of the simulations do not converge wecompute the standard errors using only those simulationsthat converged

30 Addressing this question would require at a mini-mum micro data on consumer purchases as in Goldberg(1995) To our knowledge such data do not exist for theIndian pharmaceutical market

31 See Ernst R Berndt (1994) for an extensive discussionof uniform pricing

1496 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

unconditional price and expenditure elasticities re-ported in Table 6 Given that the region-specificproduct group effects ir reported in the last fourcolumns of Tables A2 and A3 imply region-spe-cific demand elasticities we report separate elas-ticities for each region in Tables 6A and 6B Asevident from the comparison of these tables theelasticities are very similar across regions so thatwe can focus the remaining discussion on theelasticities of one region only the northern regiondisplayed in Table 6A

The diagonal terms of Table 6A report theown-price elasticities which are in all but onecase negative and highly significant The oneexception is the foreign norfloxacin productgroupmdashwhose share of quinolone sales is 007percentmdashfor which we estimate a negative butinsignificant own-price elasticity For the remain-ing product groups demand appears to be highlyelastic with the estimated elasticities lower than2 in four out of the six cases The magnitude ofthe own-price elasticities matches the features ofthe Indian pharmaceutical market mentioned ear-lier which would suggest that Indian consumersare likely to be quite price-sensitive32 The elas-ticities appear especially large if one takes intoaccount that they refer to product groups (such asdomestic ciprofloxacin) and not individual drugsoffered by particular firms Their relative magni-tudes are also intuitive the drug with the largestmarket share and a relatively high price (domesticciprofloxacin) appears to be one of the least elas-tic In contrast foreign ciprofloxacin is highlyprice elastic this is plausible as ciprofloxacindrugs offered by subsidiaries of multinationalsface the stiffest competition in this market seg-ment from approximately 75 Indian firms offeringthe same molecule

The estimated expenditure elasticities appear inthe last column These are all positive indicatingthat the demand for all product groups is normalThe remaining cells display the estimated cross-price elasticities As one might perhaps expect forproducts within a therapeutic subsegment theseare mostly positive Out of a total of 49 priceelasticities we estimate there are six that do notconform to expectations these are the cross-price elasticities between different foreign prod-

uct groups which are estimated to be negativeand significant33 Fortunately these elasticitieshave negligible impact on the welfare analysisgiven that our counterfactuals focus on theeffect of withdrawing one or more domesticproducts from the market the most relevantelasticities are the ones that capture the responseof various product group shares to a change in theprice of one or more domestic groups these elas-ticities are the cross-price elasticities between var-ious domestic groups and the ones betweendomestic and foreign groups which are plausibleand precisely estimated34

Regarding these elasticities a striking aspectof our estimates is how large positive andsignificant the cross-price elasticities betweendifferent domestic product groups aremdashin factfor norfloxacin and ofloxacin we estimate thatdomestic product groups containing differentmolecules are closer substitutes for one anotherthan product groups that contain the same mol-ecule but are produced by foreign firms Incontrast for ciprofloxacin (the molecule withthe largest revenue share) we estimate a largepositive cross-price elasticity between the do-mestic and foreign versions

The fact that domestic products appear to beclose substitutes for other domestic productsthat contain different molecules truly representsan ldquoempiricalrdquo finding in the sense that we do

32 In developed economies elasticities of this magnitude havetypically been found only for generic drugs (and even then onlyrarely) or among consumers who lack health insurance

33 While these elasticities are clearly counterintuitive theyare not inconsistent with the underlying demand system whichimposes no restrictions on the sign of the cross-price elastici-ties We do not have a good explanation of why these elastic-ities are estimated to be negative A potential explanation isthat the shares of the foreign products are very small giventhis we observe very few consumers switching from one (verysmall) foreign group to another (very small) foreign groupwhen the price of the first foreign group goes up and hence theinference is not very reliable in this case

34 The cross-price elasticities between foreign drug groupscontaining different molecules will also have an effect on thewelfare estimates given that the withdrawal of domestic prod-ucts will generally lead to changes in the prices of all foreignproducts and market shares will be reallocated from someforeign products to other foreign products based on the newprices However this reallocation from ldquoforeign to foreignrdquo istruly second order in our case compared to the reallocationfrom ldquodomestic to domesticrdquo and ldquodomestic to foreignrdquo prod-ucts In scenarios in which we consider patent enforcementaccompanied by strict price regulation the cross-price elastic-ities between foreign product groups are in fact completelyirrelevant as the prices of foreign products are not allowed toincrease in these cases Still our welfare loss estimates remainsubstantial drivenmdashas beforemdashby the loss of product variety

1497VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

not impose it through any of our assumptionsregarding the demand function The questionthat naturally arises then is what might explainthis finding While we cannot formally addressthis question anecdotal accounts in various in-dustry studies suggest that the explanation maylie in the differences between domestic andforeign firms in the structure and coverage ofretail distribution networks

Distribution networks for pharmaceuticals inIndia are typically organized in a hierarchicalfashion Pharmaceutical companies deal mainlywith carrying and forwarding (CampF) agents inmany instances regionally based who each sup-ply a network of stockists (wholesalers) Thesestockists in turn deal with the retail pharma-cists through whom retail sales ultimately oc-cur35 The market share enjoyed by a particularpharmaceutical product therefore depends inpart on the number of retail pharmacists who

stock the product And it is here that thereappears to be a distinction between domesticfirms and multinational subsidiaries In particu-lar the retail reach of domestic firms as agroup tends to be much more comprehensivethan that of multinational subsidiaries (IndianCredit Rating Agency (ICRA) 1999)36

There appear to be two reasons for this Thefirst is that many of the larger Indian firmsbecause they have a much larger portfolio ofproducts over which to spread the associatedfixed costs typically have more extensive net-works of medical representatives The second issimply that there are many more domestic firms(and products) on the market At the retail levelthis would imply that local pharmacists mightbe more likely to stock domestic products con-taining two different molecules say ciprofloxa-cin and norfloxacin than they would domesticand foreign versions of the same molecule Tothe extent that patients (or their doctors) arewilling to substitute across molecules in order tosave on transport or search costs (eg going toanother pharmacy to check whether a particular

35 There are estimated to be some 300000 retail pharma-cists in India On average stockists deal with about 75 retailers(ICRA 1999) There are naturally variations in this structureand a host of specific exclusive dealing and other arrangementsexists in practice Pharmaceutical firms also maintain networksof medical representatives whose main function is to marketthe companyrsquos products to doctors who do the actual prescrib-ing of drugs In some instances firms do sell directly to thedoctors who then become the ldquoretailerrdquo as far as patients areconcerned but these are relatively rare

36 These differences were also highlighted in conversa-tions one of the authors had with CEOs and managingdirectors of several pharmaceutical firms as part of a sep-arate study

TABLE 6AmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN THE NORTHERN REGION

Product group

Elasticity with respect to

Prices of foreign productgroups Prices of domestic product groups Overall

quinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 557 013dagger 015 401 011dagger 011dagger 016 137(179) (007) (007) (184) (006) (006) (006) (029)

Foreign norfloxacin 427dagger 045 427dagger 350dagger 602 451 465 220(242) (112) (242) (210) (623) (184) (183) (105)

Foreign ofloxacin 011 010dagger 138 009 009dagger 023 011 116(005) (005) (031) (027) (005) (028) (004) (017)

Domestic ciprofloxacin 018 001 001 168 008 008 010 117(008) (000) (001) (023) (002) (002) (002) (003)

Domestic norfloxacin 004 003 004 058 223 042 040 073(001) (003) (001) (017) (011) (004) (003) (009)

Domestic ofloxacin 005 005 011 077 074 342 074 089(002) (002) (013) (028) (008) (025) (008) (021)

Domestic sparfloxacin 007 004 007 115 063 063 288 028(002) (001) (002) (015) (006) (006) (017) (012)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent significance level and dagger (dagger) denotes significance at the 10-percent level

1498 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE 6BmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN OTHER REGIONS

Demand patterns in the eastern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 594 014dagger 016 431 013 011 017 140(195) (008) (008) (201) (006) (007) (007) (031)

Foreign norfloxacin 329dagger 058 329dagger 264dagger 460 346 359 192(180) (083) (179) (153) (451) (138) (139) (082)

Foreign ofloxacin 012 010dagger 140 010 010 024 013 117(006) (006) (034) (031) (005) (030) (005) (019)

Domestic ciprofloxacin 018 001 001 172 009 008 011 117(008) (000) (002) (027) (002) (003) (002) (003)

Domestic norfloxacin 004 004 004 068 242 049 046 070(002) (004) (002) (020) (013) (005) (004) (010)

Domestic ofloxacin 004dagger 004 009 061 060 295 060 092(002) (001) (010) (028) (007) (022) (007) (017)

Domestic sparfloxacin 005 003 005 092 048 051 251 043(002) (001) (001) (017) (005) (006) (016) (010)

Demand patterns in the western region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 427 009dagger 011 286 008 006 011 126(129) (005) (005) (133) (004) (005) (004) (021)

Foreign norfloxacin 714dagger 008 710dagger 594dagger 995 742 770 299dagger

(369) (183) (368) (328) (1015) (297) (296) (178)Foreign ofloxacin 013 012dagger 145 009 012 026 013 119

(007) (006) (037) (029) (005) (033) (005) (021)Domestic ciprofloxacin 019 001 000 174 009 008 011 118

(008) (000) (001) (024) (002) (003) (002) (004)Domestic norfloxacin 004 004 004 067 243 050 047 069

(002) (004) (001) (018) (014) (005) (004) (010)Domestic ofloxacin 003 003 007 048 048 257 048 093

(002) (001) (008) (022) (005) (015) (004) (014)Domestic sparfloxacin 006 003 005 083 046 049 241 046

(002) (001) (001) (014) (004) (005) (012) (009)

Demand patterns in the southern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 560 013dagger 015 402 011 010 017 137(182) (007) (007) (187) (006) (006) (007) (029)

Foreign norfloxacin 533dagger 031 531dagger 431dagger 748 559 584 249dagger

(301) (134) (300) (245) (734) (219) (219) (128)Foreign ofloxacin 014 012dagger 148 011 012 029 016 120

(007) (007) (039) (033) (006) (035) (006) (022)Domestic ciprofloxacin 017 001 000 169 008 007 011 116

(008) (000) (001) (025) (002) (003) (002) (003)Domestic norfloxacin 004 003 003 059 225 043 040 073

(001) (003) (001) (019) (012) (004) (004) (009)Domestic ofloxacin 004dagger 004 009 062 060 296 060 091

(002) (001) (010) (026) (006) (019) (007) (017)Domestic sparfloxacin 008 005 008 138 073 075 323 016

(002) (001) (002) (018) (007) (008) (021) (014)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at 5-percent level anddagger (dagger) denotes significance at the 10-percent level

1499VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

foreign product is in stock) in aggregate datawe would expect to find precisely the substitu-tion patterns that we report in Table 6

Whether the particular explanation we provideabove is the correct one the high degree of sub-stitutability between domestic product groupsturns out to have important implications for thewelfare calculations We discuss these in moredetail below when we present the results of thecounterfactual welfare analysis Another elasticitywith important implications for the counterfactu-als is the price elasticity for the quinolone subseg-ment as a whole which indicates how likelyconsumers are to switch to other antibioticsgroups when faced with a price increase for quin-olones This elasticity is computed on the basis ofthe results in Table A3 and it is at 111 (stan-dard error 024) this is large in magnitudebutmdashas expectedmdashsmaller in absolute value thanthe own-price elasticities of the product groupswithin the quinolone subsegment

The results in Tables 6A and 6B are based onour preferred specification discussed in SectionII In Tables A4 to A6 in the Appendix weexperiment with some alternative specificationsTables A4(a)ndashA4(c) correspond to a specificationthat includes in addition to product-group-specificregional fixed effects product-group-specific (andfor the upper level antibiotics-segment-specific)seasonal effects We distinguish among three sea-sonsmdashthe summer monsoon and wintermdashandreport the unconditional demand elasticities for

the northern region for each of these seasons Asevident from the tables in the Appendix our elas-ticity estimates are robust to the inclusion of sea-sonal effects The demand elasticities in Table A5are based on estimation of the demand system byOLS Compared to the elasticities obtained by IVthe OLS elasticities are smaller in absolute valueimplying that welfare calculations based on theOLS estimates would produce larger welfare lossestimates Nevertheless some of the patterns re-garding the cross-price elasticities discussed ear-lier are also evident in the OLS results inparticular the cross-price elasticities between dif-ferent domestic product groups are all positivelarge and significant and in most instances largerthan the cross-price elasticities between drugs thatcontain the same molecule but are produced byfirms of different domesticforeign status Theclose substitutability of domestic products indi-cated by both the OLS and IV estimates seems tobe one of the most robust findings of the paper

B Cost and Markup Estimates

Table 7 displays the marginal costs markupsand profits implied by the price elasticity esti-mates of Tables 6A and 6B for each of the sevenproduct groups Given that our regional effectsimply different price elasticities for each regionour marginal cost and markup estimates alsodiffer by region Given however that based on

TABLE 7mdashUPPER AND LOWER BOUNDS FOR MARGINAL COST MARKUP AND ANNUAL PROFIT BY PRODUCT GROUPS WITHIN

THE QUINOLONE SUBSEGMENT

Product groupLower boundfor MC (Rs)

Upper boundfor markup

Upper boundfor profit(Rs mill)

Upper boundfor MC (Rs)

Lower boundfor markup

Lower boundfor profit

(Rs)

Foreign ciprofloxacin 83 19 269 103 0 00(123) (012) (1655)

Foreign norfloxacin NA NA NA 53 0 00Foreign ofloxacin 323 70 1061 1085 0 00

(2316) (021) (3185)Domestic ciprofloxacin 47 59 17019 112 0 00

(114) (010) (29858)Domestic norfloxacin 52 43 2807 90 0 00

(020) (002) (1532)Domestic ofloxacin 587 34 1612 901 0 00

(218) (002) (1280)Domestic sparfloxacin 495 37 1985 788 0 00

(157) (002) (1100)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level Estimated lower bound forforeign norfloxacinrsquos marginal cost is negative since the estimated price elasticity is less than one in absolute value

1500 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Tables 6A and 6B the price elasticities do notseem to differ substantially across regions wereport for ease of exposition only the nationalaverages for marginal costs and markups inTable 7

Table 7 has two parts In the left part (firstthree columns) we report the lower bound formarginal cost and the corresponding upperbound for markup and upper bound for totalannual profit for each product group Thesenumbers are based on the price elasticities weobtained from estimating the two-level AIDSdemand system Since we do not have a reliableestimate for the price elasticity of foreign nor-floxacin (the point estimate is negative but lessthan one in absolute value and insignificant)we cannot compute the lower bound for mar-ginal cost in this case It is important to note thatthese estimates do not reflect either the actualmarginal cost or the actual markup for thesedrugs both because the existence of price reg-ulation implies that the unconstrained first-orderconditions are unlikely to hold each period andbecause our aggregation across firms of thesame domesticforeign status supplying thesame molecule makes the interpretation of theseestimates problematic In particular the fact thatwe ignore competition among firms within eachproduct group implies that our estimates willtend to overstate market power These numberswill prove useful however in the counterfac-tual simulations as they can give us a sense ofhow large the maximum profit gains for multi-national firms and the maximum profit lossesfor domestic firms are likely to be under patentenforcement The right part of the table (lastthree columns) reports the upper bound for mar-ginal cost which we obtain by simply taking themarginal cost to be equal to the observed priceWe do not report standard errors in this casesince the numbers are based on actual data Themarkups corresponding to these marginal costupper bounds are of course zero We conductthe counterfactual simulations using both thelower and upper bounds for marginal costs

The (upper bounds for) markups on the leftside of the table are generally plausible Thedomestic ciprofloxacin product group that dom-inates the quinolone subsegment and for whichwe documented high prices a high marketshare and a relatively low elasticity of demandenjoys one of the highest markup upper bounds(60 percent) and accounts for nearly 70 percent

of all profits derived within the subsegmentForeign ciprofloxacin on the other hand whichfaces the stiffest competition from domesticfirms and for which we estimated a highly elas-tic demand has the lowest markup upper bound(19 percent)

C Counterfactual Estimates of the Impact onPrices and Welfare

With estimates of the key demand and costparameters in hand we turn to the counterfac-tuals We consider the five separate scenarioslisted in the previous section All of the scenar-ios involve the withdrawal of one or more of thedomestic product groups from the market Ta-ble 8 displays our estimates of the consumerwelfare losses that result under the differentscenarios The losses are expressed in billion ofrupees per year All numbers presented in Ta-ble 8 and subsequent tables are based on usingthe lower bounds for marginal cost and upperbounds for markup in the simulations As dis-cussed earlier these numbers are the more in-teresting to work with since they give us anupper bound for the changes in the profits ofdomestic and foreign firms that would resultfrom patent enforcement In the Appendix wealso present results based on using the upperbounds for marginal costs in the simulations inwhich case the pre-TRIPS profits of domesticand foreign firms are zero In all cases marginalcosts are assumed to be constant in outputWhile naturally the profit implications differdepending on whether one uses the upper orlower bounds for marginal costs (firm profits arezero if one assumes the upper bound of mar-ginal cost in which case price equals marginalcost) the estimated consumer welfare losses aresimilar in the two cases We discuss these re-sults in more detail at the end of this section

The first column presents our estimates of theconsumer welfare losses attributable to the pureloss of product variety effect where we fix theprices of all remaining products as well as theoverall expenditure on quinolones while with-drawing one or more of the domestic productgroups Note that in our initial specification ofthe demand system had we not allowed for thepossibility that consumers might differentiatebetween domestic and foreign products evenwhen they contain the same molecule this

1501VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

particular component of the loss of consumerwelfare would not have arisen

The estimates reported in the second columnincorporate the expenditure-switching effect ontop of the loss of product variety Here basedupon the price elasticity estimates from thehigher-level AIDS system we adjust (down-ward) the expenditures allocated to the quino-lone subsegment as the composite price ofquinolones effectively increases as a consequenceof the higher virtual prices of the domestic productgroups that are withdrawn from the market Be-cause the estimates in this column are generatedassuming that the prices of the products that re-main in the market are not adjusted upward theyprovide a sense of what consumer losses would beif the introduction of product patents were coupledwith strict price regulation aimed at maintainingprices at prepatent levels Alternatively they canbe thought of as the relevant welfare numbers ifintense competition among firms within the re-maining product groups kept the prices of theproducts that were still offered in the market closeto the firmsrsquo marginal costs

The last column of Table 8 displays the esti-mated consumer welfare losses when bothcross-segment expenditure-switching and with-in-segment upward price adjustments are takeninto account

If we compare the results across the firstsecond and third columns all the counterfac-tual scenarios produce qualitatively similar pat-terns patterns that are consistent with what we

would expect Starting from the initial loss ofwelfare attributable to the loss of product vari-ety the option of switching expenditures outof the quinolone subsegment to other subseg-ments mitigates some of the initial welfareloss But if we then incorporate the upwardprice adjustments that result in response to thereduced competition the welfare losses aremagnified

Of particular interest from a policy perspec-tive are the relative magnitudes of these threeeffects which are similar under all the counter-factual scenarios though absolute levels varyconsiderably First despite the fact that the de-mand for quinolones is quite sensitive to thecomposite price of quinolonesmdashthe upper levelprice elasticity is 111mdashthe cross-subsegmentexpenditure-switching effects are in all thecases small (in absolute value terms) relative tothe other two effects For instance under thescenario where all the domestic quinolone prod-uct groups are withdrawn from the market theoverall consumer welfare loss of Rs 1781 bil-lion per year can be decomposed into an initialloss of Rs 1176 billion (66 percent) attribut-able to the loss of product variety a slightreduction in this initial loss of Rs 041 billion(2 percent) from Rs 1176 billion to Rs1135 billion because of expenditure switchingand a subsequent additional loss of Rs 646billion (36 percent) from Rs 1135 billion toRs 1781 billion because of the reduced com-petition and consequent price increases

TABLE 8mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure loss ofvariety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 732(087) (089) (146)

Only ofloxacin 008 008 023(008) (008) (010)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1253(177) (180) (415)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1058(142) (145) (331)

All four domestic quinolones products 1176dagger 1135dagger 1781(643) (634) (1270)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent significance level and dagger (dagger)denotes significance at the 10-percent level

1502 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 15: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

where ii(pipj) the own-price elasticity of de-

mand for product i will depend on the productrsquosown-price pi and all other productsrsquo prices pjwith i j

Once we have obtained the demand elastic-ities through estimation of the demand sys-tem we can calculate the upper and lowerbounds for marginal costs and correspondingmarkups according to (12) or (13) These willthen be employed in the counterfactual sim-ulations While our counterfactuals use boththe lower and upper bounds for costs most ofour discussion will be based on using thelower bounds for costs since this is the moreinteresting case in the policy analysis it givesus the largest possible profits for pharmaceuticalfirms and hence corresponds to the worst pos-sible scenario facing Indian firms and the bestpossible scenario facing multinationals underTRIPS

C Identification Assumptions and EstimationApproach

The discussion of the demand system has sofar abstracted from the issue of price endogene-ity The usual premise in the industrial organi-zation literature is that correlation of prices withthe error term in the demand equation arises byvirtue of the first-order conditions of profit-maximizing firms20 As we discuss in this sec-tion this source of simultaneity bias is unlikelyto be of major concern in the present contextbecause of the existence of price regulation Ourprimary concern is instead with the simultane-ity bias that is implied by the particular wayprices are constructed

To understand the sources of the simultaneitybias note first that the price of product group piwe employ in the demand estimation should bethought of as a proxy for an exact price indexthat we do not observe As such pi contains bydefinition measurement error and it will becorrelated with the error term of the demandequation Specifically the demand equation weare interested in estimating can be written insimplified form (suppressing the subscripts rand t and ignoring parameter restrictions forconvenience) as

i i j

ijln pjT ilnXQ

PQ i

where pjT denotes the true (exact) price index for

product group i and i captures unobservedvariables that may affect demand in a particularperiod (eg an advertising campaign or an ep-idemic) The price index pj

T depends on theSKU prices pk Note that in the presence ofSKU entry and exit into the sample the priceindex pj

T will vary over time even if the pricesof the individual SKUs remain stable as theproducts (SKUs) that comprise each productgroup change each period

Ideally we would like to compute the exactprice index pj

T for each group and employ theseindices in the demand estimation Under theassumption of predetermined SKU prices eachperiod the exact price indices would be uncor-related with the error term and the demandparameters could be estimated by simple OLSUnfortunately this strategy is not feasible in thecurrent context

To derive an exact price index for each prod-uct group it is necessary to model consumerrsquoschoice among SKUs conditional on the choiceof the group To this end it is necessary toexplicitly introduce a third stage in the demandspecification and estimate the parameters asso-ciated with the choice at that stage Howeverany specification based on three-stage budget-ing (that would require directly specifying ademand function associated with the SKUchoice at the third stage and imposing additiveseparability at the higher stages) would be in-feasible to estimate for three main reasonsFirst there is a large number of SKUs withinsome groups so that the demand parameters can-not be estimated even if one imposes parameterrestrictions second the SKU prices exhibit littlevariation over time lastly the choice sets at thislowest stage vary over time rendering estimationof the price parameters associated with SKUs thatare not available in all periods infeasible Alterna-tively one could abandon the multistage budget-ing altogether and adopt a discrete choiceapproach which is inherently better suited to deal-ing with varying choice sets and limited pricevariation A discrete choice approach would beparticularly appealing for modeling the SKUchoice conditional on product group choice giventhat within each group all SKUs contain the same

20 See Steven Berry et al (1995) or Aviv Nevo (2001)for a related discussion

1491VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

molecule and are hence comparable As notedearlier however a discrete choice approach at thehigher stages of the demand estimation requiresconverting revenue to physical shares which ischallenging in the case of systemic drugs such asantibiotics for which the appropriate dosage is notwell defined Finally a combination of a multi-stage budgeting approach for the higher stages anda discrete choice approach for the lowest stagewould be feasible to estimate but inconsistentwith a model of utility maximization For thesereasons we have no alternative but to adopt anapproximation for measuring the price index foreach product group

Let us denote this approximation by pjA and let

13j denote the proportional ldquoapproximationrdquo errorassociated with measuring the true price index forgroup j so that ln pj

A ln pjT ln 13j The

estimating demand equation can then be written as

i i j

ijln pjA ilnXQ

PQ

j

ijln 13j i

or

i i j

ijln pjA ilnXQ

PQ i

where the new error term i yenj ijln 13j iis comprised of two components the first onereflects ldquomeasurementrdquo error due to the fact thatthe product group prices we employ are approx-imations and not exact price indices and thesecond one captures conventional demandshocks The demand equation as written abovecorresponds to the equation we take to the data

The particular proxy we use for the exact priceindex for each product group j is the revenue-share-weighted average of the prices of multipleSKUs available within this group that is

pjA

k j

k pk

The notation above illustrates the two source ofprice variation variation in the SKU prices pkand variation in the weights k

We are concerned with potential correlation ofthe so-constructed product group prices with bothcomponents of the product group demand error

term The correlation between pjA and the first

component yenj ijln 13j is independent of theparticular way we construct pj

A and inherent in thefact that we measure the exact price index witherror The potential correlation between pj

A and thesecond component of the product group demanderror the shock i arises however because of thespecific way we construct the proxy pj

A and inparticular because of the presence of the reve-nue share weights k in pj

A since the SKUrevenue share weights k will generally dependon the product group expenditure they arelikely to be correlated with the product groupdemand shock i

To address the simultaneity bias we use in-strumental variables To this end we need vari-ables that are correlated with the proxies pj

A butuncorrelated with the error term i We use thenumber of SKUs within each product group j asan instrument as it does justice to the idea thatvariation in the product group price index stemsin part from variation in the set of SKUs that areavailable each period It is clearly correlatedwith the average group prices pj

A and plausiblyuncorrelated with the demand shock i Theassumption that the number of SKUs is uncor-related with the conventional demand error willhowever be violated if the introduction of anew SKU changes the perceived quality of adrug21 or if it is accompanied by promotionalactivities which will be reflected in i Our hopeis that this is not too often the case Assumingthat the SKU number for each group j is alsouncorrelated with the weighted sum of the mea-surement errors of all product groups yenj ijln 13j itcan be used as an instrument for the product groupprice pj

A22 If one accepts the premise that prices atthe SKU level are exogenous then SKU pricescan also be used as instruments The final list ofinstruments we use hence includes the number ofSKUs in each group the prices of the five largestSKUs for each group and all other exogenousvariables in the demand estimation such as re-gionalproduct dummy interactions and the upper-

21 Note that we already proxy for the quality of each prod-uct group through the product fixed effects i In the actualestimation we actually let these fixed effects be region-specific(ir) so that we allow quality to vary across regions

22 If the measurement error increases in the number ofincluded SKUs this assumption will be violated Unfortu-nately we have no way of checking the validity of thisassumption

1492 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

level total expenditure on antibiotics Regressionsof group prices on the instruments above yieldhigh R-squares with most regressors highly sig-nificant indicating that our instruments are highlycorrelated with prices23

Our sample includes four molecules cipro-floxacin norfloxacin ofloxacin and spar-floxacin Except for sparfloxacin all othermolecules are produced by both foreign anddomestic firms24 So we have seven products(domestic ciprofloxacin foreign ciprofloxa-cin etc) with 96 observations (two years ofmonthly data four geographical regions foreach period) for each product The parametersin the lower-level AIDS demand system asdefined in equation (8) are the product fixedeffects i the productndashregional dummy inter-actions ir the own revenue-share price elas-ticities ii the cross revenue-share priceelasticities i10 i01 i00 and the revenue-share expenditure elasticities i In estimatingthe parameters we first regress prices on allinstrumental variables and then plug the pre-dicted values for prices in the constrainedleast-square regression (For a detailed expla-nation of the constraints see the previoussection)

Given that we impose many cross-equationconstraints and employ instrumental variablesin the estimation it is difficult to derive stan-dard errors for the parameter estimates ana-lytically Our error term interpretation in (8)implies that the error terms for each productgroup are likely to be correlated across re-gions for example if a national advertisingcampaign increases the demand for a rela-tively expensive presentation in one regionsimultaneously increasing the aggregate de-mand and price index for the correspondingproduct group in that region it is likely that thesame effect will be observed in other regions Inprinciple we could exploit the cross-regionalcorrelation in the error terms to estimate thedemand system using seemingly unrelated re-

gression (SUR) This however would requireus to derive the variance-covariance matrix ofthe parameters analytically which as notedabove is cumbersome in the current contextInstead we use the bootstrap method choosingto remain agnostic about the structure of thevariance-covariance matrix The potential dis-advantage of this approach over SUR is a loss inefficiency but as will become apparent in theresults section the model parameters are fairlyprecisely estimated To maintain the marketstructure we randomly sample the periods (withreplacement) and use the same periods for allproducts Regarding the optimal number ofbootstrap repetitions ideally one would followthe three-step method proposed by Donald W KAndrews et al (2000) Empirical evidence sug-gests however that one rarely needs more than200 replications to estimate the standard er-rors25 To be safe we generate 300 bootstrapsamples (with replacement) based on the origi-nal data and estimate the standard errors usingthe standard errors of the bootstrap sampleestimates

The estimation of the top-level AIDS systemis similar The constraints imposed on this top-level demand system are adding-up homogene-ity and symmetry as defined on page 12Again bootstrapping is used to obtain the stan-dard errors of the parameter estimates

D The Counterfactual Scenarios

In assessing the effects of patent enforce-ment we start by focusing on the most extremecase in which compulsory licensing is not anoption and foreign firms are not subject to pricecontrols We use the results from the analysis ofthis case as a benchmark In reality the outcomeof the WTO negotiations is more likely to in-volve some constraints on the monopoly powerof foreign firms selling patented products indeveloping countries such as price caps orcompulsory licensing Our framework can eas-ily accommodate these cases as will becomeapparent in the next subsection

We now focus on the effects of potentialpatent enforcement in the quinolone segmentWe consider several scenarios which vary inthe number and size of domestic products that

23 Specifically the R-squares from the first-stage regres-sions of product group prices on instruments range from057 to 095 except for domestic ciprofloxacin for whichthe R-square is 017

24 Sparfloxacin is actually offered by one foreign subsid-iary in India (see Table 2) However its revenue share isminiscule We therefore treat Sparfloxacin as being pro-duced by domestic firms only 25 See Bradley Efron and Robert J Tibshirani (1993)

1493VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

will be removed from the market In particularwe consider the following five scenarios26

Withdrawal of one large domestic productgroup only domestic ciprofloxacin

Withdrawal of one relatively small domesticproduct group only domestic ofloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand norfloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand sparfloxacin

Withdrawal of all four domestic quinoloneproduct groups

As the list above suggests we proceed fromanalyzing the effects of single-product with-drawal to the analysis of eliminating the en-tire domestic segment This approach wasmotivated by early empirical results that in-dicated that the existence and extent of com-petition from domestic firms has a significantbearing on the predicted effects of patent en-forcement that is our predictions regardingprices and welfare vary substantially depend-ing on how many domestic products are af-fected by patent enforcement and on whethersome domestic competition will remain presentafter TRIPS In addition the size of the affecteddomestic groups is relevant for the welfare pre-dictions accordingly we examine both scenar-ios in which a large domestic product groupsuch as domestic ciprofloxacin is eliminatedfrom the market and scenarios in which the

eliminated domestic group is relatively small(eg domestic ofloxacin) In all scenarios wemaintain the assumption that the set of prod-ucts offered by the remaining firms in themarket does not change in response to patentenforcement

E Computation of Virtual Prices and NewEquilibrium Prices

The first step in the counterfactual analysisis to derive the new equilibrium prices underpatent enforcement27 In this context thereare two sets of prices that are relevant Thefirst set consists of the virtual prices of those(domestic) products that will not be availableonce TRIPS is put in effect To calculate thesevirtual prices we set the expenditures of therelevant products equal to zero The secondset of prices consists of the prices of thoseproducts that remain in the market In deriv-ing these prices we start by assuming profitmaximization without any form of price reg-ulation the firms remaining in the marketreoptimize in response to the policy changeand set new prices taking the prices of allother firms as given28 Of course at the equi-librium all prices change in response to thefact that some domestic products are nolonger present The new equilibrium pricesfor products that remain in the market are thuscomputed by utilizing the first-order condi-tions of profit-maximizing firms into whichthe virtual prices of the eliminated productsare substituted Hence to compute the newequilibrium prices we solve an equation sys-tem of the following form

26 As Table 2 indicates most of the patents for the drugsin the quinolone segment have expired by now so that noneof the scenarios described in this section is going to mate-rialize in practice The reason we consider these alternativescenarios in our counterfactual simulations is to get a senseof how the presence and extent of domestic competitionaffects welfare calculations Such calculations may becomerelevant in the future in other therapeutic classes in whichpatents have not expired yet and where domestic firmscompete with foreign patent holders While the particularexit scenarios will depend on the exact year of patent expiryof drugs in each therapeutic class and the specific way inwhich patents will be enforced (eg whether certain do-mestic drugs will be grandfathered in) the general conclu-sion that emerges from our calculations is that the consumerwelfare loss is substantially smaller in cases where somedomestic competition remains present in the market Hencethe particular way in which patent enforcement is imple-mented is essential for assessing its welfare impact

27 Of course until product patents are in fact introducedthese prices will not be observable Note also that we areassuming here that the range of available products will notchange with the introduction of patents

28 As mentioned above this first set of calculations ab-stracts from the existence of remaining price controls orother government regulations that would impose constraintson the firmsrsquo profit maximization problem Accordingly theresulting numbers should be interpreted as a benchmark ofwhat would happen if markets were completely unregulatedTo examine what the welfare effects of TRIPS would be inthe more realistic scenario of price regulation we subse-quently consider a scenario in which regulation keeps theprices of the products provided by patent-holders at theirpre-TRIPS level For a more detailed discussion see theresults section

1494 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

For products i that are withdrawn from themarket

(14) 0 i ir iiln pirV i10ln pjr j Di

10

j Di

01

i01ln pjr

j Di

00

i00ln pjr

ilnXQr

PQr

For products k that remain in the market

(15) pkr ckr 1 1

kk pkrpjr pirV 1

In the equations above pirV denotes the virtual

prices of the products that are removed fromthe market while pjr denotes the updatedprices of all other products Note that whensolving for the virtual prices we account forthe fact that both the price index for quino-lones PQr and the expenditure allocated tothis subsegment XQr need to be updated toreflect the fact that as a result of the pricechanges there may be substitution away fromthis subsegment To obtain the new quinoloneexpenditure XQr and the new price index PQrwe use the estimates and formulas for thehigher-level AIDS system In equation (15)ck refers to the marginal cost for product k thatwe have obtained from the previous estimationstage As mentioned in the previous subsectionwe conducted the simulations using both theupper and lower bounds for marginal cost ck

u

and ckL respectively The term kk(pkr

pjrpirV) re-

fers to the unconditional own-price elasticity forproduct k which is a function of the eliminatedproductsrsquo virtual prices and the remaining prod-uctsrsquo new equilibrium prices We conduct thecounterfactual simulations and welfare analysisat the regional level because the presence ofregion-specific product effects in the demandestimation implies region-specific demand elas-ticities and hence region-specific marginalcosts and markups The presentation and discus-sion of our results focus on national averages ofthe relevant variables that we construct by com-

puting weighted averages across regions usingthe population of each region as a weight

F Welfare Assessment

The simulation of the new equilibrium underpatent protection can provide important insightsinto how consumers and firms will respond tothe removal of domestic products in the market(for example toward which products consumerswill substitute or which prices will increase themost) To get a more precise idea of how peo-plersquos well-being will ultimately be affected byTRIPS we compute as a last step in our anal-ysis the welfare effects of the policy changeSocial welfare is defined as consumer welfareand the sum of domestic firm profits Thechange in domestic profits can easily be calcu-lated by comparing the domestic firm (variable)profits at the pre-TRIPS prices to the profitsthese firms will realize at the new simulatedprices Although foreign firm profits do notcount in domestic welfare calculations we alsocompute the effects of patent enforcement onforeign firm profits to get an idea of how largethe expected benefits of TRIPS are for thesefirms This provides in some sense an indirectway of assessing whether the claims that patentenforcement in countries like India will lead tomore research on developing-country-specificdiseases (such as malaria) have any validity iffor example we find that the effect of patentenforcement on the foreign firm profits realizedin India is small in magnitude it is unlikely thatforeign firms will engage in more developing-country-specific research in response to TRIPSIt is important to note that in all these calcula-tions we work with the lower-bound estimatesfor marginal costs since these give rise to thehighest possible markups Hence our estimateof profit loss for domestic producers most likelyoverstates this loss

On the consumer side we measure changes inconsumer welfare by the compensating variation(CV) defined as the additional expenditure thatconsumers need in order to achieve the sameutility level as before patent enforcement at thenew prices Specifically let P0 denote the pricevector before patent enforcement P the simulatedprice vector post-TRIPS (that we obtained usingthe methods described in the previous subsection)u0 the utility attained by consumers before TRIPS

1495VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

and E(u P) the higher-level expenditure functionThen the compensating variation is given by

(16) CV Eu0 P Eu0 P0

Note that the CV as computed in (16) representsthe combination of three effects

The pure product-variety effect that is theeffect that arises because one or more prod-ucts are not available to consumers anymoreholding the prices of all other remainingproducts and the total expenditure on thequinolone subsegment XQ constant

The expenditure-switching effect that is theeffect arising from substitution away fromquinolones and toward other subsegments ofthe antibiotics market again holding the pricesof all other remaining products constant

The reduced-competition effect that is theeffect that arises because the firms remainingin the market adjust (increase) their prices inresponse to the removal of domestic products

From both an analytical and a policy point ofview it is desirable to assess how large each ofthese effects is Accordingly we decompose thetotal effect on consumer welfare (the CV as givenby equation (16)) using the following procedure

To get the pure product variety effect wecompute virtual prices for the products that areremoved from the market holding the quino-lone expenditure XQ and the prices of all otherproducts fixed Let us call the resulting pricevector P1 Then the pure product variety effectis represented by E(u0 P1) E(u0 P0)

To compute the expenditure-switching effectwe compute another set of virtual prices againholding the prices of all remaining products fixedbut letting quinolone expenditure adjust in re-sponse to the new price index for the quinolonesegment (Given that the prices of the remainingproducts remain fixed the change in the priceindex arises only because of the removal of one ormore domestic products) Note that this scenariomost closely resembles the case in which patentlaws are enforced but strict price regulation keepsthe prices of the products offered in the market attheir pre-TRIP level Let us label the so-computedprice vector P2 The expenditure-switching effectis then E(u0 P2) E(u0 P1)

Finally the reduced competition effect aris-ing from higher prices for the remaining prod-

ucts is computed as the residual change in thecompensating variation once the product-varietyand expenditure-switching effects have been ac-counted for that is E(u0 P) E(u0 P2)where the price vector P is computed accordingto the formulas (14) and (15) to reflect theadjustment of prices to the new regime

To compute the standard errors associatedwith the counterfactual simulations (that is thestandard errors for the counterfactual prices andwelfare estimates) we again use bootstrappingWe first bootstrap the original sample (sales andprices of all the drugs) 300 times Next we esti-mate the AIDS model for each of these samplesand compute the counterfactual equilibrium pricesand welfare losses corresponding to each policyscenario for each of the 300 simulations In thefinal step we compute standard errors29

One limitation of our framework is that itdoes not allow for a heterogeneous response ofconsumers to the policy change Accordinglyour framework is not suited to addressing thequestion of how different groups in the popula-tion will be affected by patent law enforce-ment30 Along the same lines our frameworkdoes not accommodate the possibility of pricediscrimination which might lead to differentresults regarding welfare losses and profit gainsrelative to uniform pricing31 Still we believethat the results of the counterfactual simulationscan provide important insights into the likelyaggregate response to patent enforcement andthe factors that drive this response

III Results

A The Structure of Demand

Tables A2 and A3 in the Appendix display theresults from estimation of the lower- and upper-level AIDS systems respectively For ease of in-terpretation rather than discussing the coefficientestimates we focus our discussion on the implied

29 In cases some of the simulations do not converge wecompute the standard errors using only those simulationsthat converged

30 Addressing this question would require at a mini-mum micro data on consumer purchases as in Goldberg(1995) To our knowledge such data do not exist for theIndian pharmaceutical market

31 See Ernst R Berndt (1994) for an extensive discussionof uniform pricing

1496 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

unconditional price and expenditure elasticities re-ported in Table 6 Given that the region-specificproduct group effects ir reported in the last fourcolumns of Tables A2 and A3 imply region-spe-cific demand elasticities we report separate elas-ticities for each region in Tables 6A and 6B Asevident from the comparison of these tables theelasticities are very similar across regions so thatwe can focus the remaining discussion on theelasticities of one region only the northern regiondisplayed in Table 6A

The diagonal terms of Table 6A report theown-price elasticities which are in all but onecase negative and highly significant The oneexception is the foreign norfloxacin productgroupmdashwhose share of quinolone sales is 007percentmdashfor which we estimate a negative butinsignificant own-price elasticity For the remain-ing product groups demand appears to be highlyelastic with the estimated elasticities lower than2 in four out of the six cases The magnitude ofthe own-price elasticities matches the features ofthe Indian pharmaceutical market mentioned ear-lier which would suggest that Indian consumersare likely to be quite price-sensitive32 The elas-ticities appear especially large if one takes intoaccount that they refer to product groups (such asdomestic ciprofloxacin) and not individual drugsoffered by particular firms Their relative magni-tudes are also intuitive the drug with the largestmarket share and a relatively high price (domesticciprofloxacin) appears to be one of the least elas-tic In contrast foreign ciprofloxacin is highlyprice elastic this is plausible as ciprofloxacindrugs offered by subsidiaries of multinationalsface the stiffest competition in this market seg-ment from approximately 75 Indian firms offeringthe same molecule

The estimated expenditure elasticities appear inthe last column These are all positive indicatingthat the demand for all product groups is normalThe remaining cells display the estimated cross-price elasticities As one might perhaps expect forproducts within a therapeutic subsegment theseare mostly positive Out of a total of 49 priceelasticities we estimate there are six that do notconform to expectations these are the cross-price elasticities between different foreign prod-

uct groups which are estimated to be negativeand significant33 Fortunately these elasticitieshave negligible impact on the welfare analysisgiven that our counterfactuals focus on theeffect of withdrawing one or more domesticproducts from the market the most relevantelasticities are the ones that capture the responseof various product group shares to a change in theprice of one or more domestic groups these elas-ticities are the cross-price elasticities between var-ious domestic groups and the ones betweendomestic and foreign groups which are plausibleand precisely estimated34

Regarding these elasticities a striking aspectof our estimates is how large positive andsignificant the cross-price elasticities betweendifferent domestic product groups aremdashin factfor norfloxacin and ofloxacin we estimate thatdomestic product groups containing differentmolecules are closer substitutes for one anotherthan product groups that contain the same mol-ecule but are produced by foreign firms Incontrast for ciprofloxacin (the molecule withthe largest revenue share) we estimate a largepositive cross-price elasticity between the do-mestic and foreign versions

The fact that domestic products appear to beclose substitutes for other domestic productsthat contain different molecules truly representsan ldquoempiricalrdquo finding in the sense that we do

32 In developed economies elasticities of this magnitude havetypically been found only for generic drugs (and even then onlyrarely) or among consumers who lack health insurance

33 While these elasticities are clearly counterintuitive theyare not inconsistent with the underlying demand system whichimposes no restrictions on the sign of the cross-price elastici-ties We do not have a good explanation of why these elastic-ities are estimated to be negative A potential explanation isthat the shares of the foreign products are very small giventhis we observe very few consumers switching from one (verysmall) foreign group to another (very small) foreign groupwhen the price of the first foreign group goes up and hence theinference is not very reliable in this case

34 The cross-price elasticities between foreign drug groupscontaining different molecules will also have an effect on thewelfare estimates given that the withdrawal of domestic prod-ucts will generally lead to changes in the prices of all foreignproducts and market shares will be reallocated from someforeign products to other foreign products based on the newprices However this reallocation from ldquoforeign to foreignrdquo istruly second order in our case compared to the reallocationfrom ldquodomestic to domesticrdquo and ldquodomestic to foreignrdquo prod-ucts In scenarios in which we consider patent enforcementaccompanied by strict price regulation the cross-price elastic-ities between foreign product groups are in fact completelyirrelevant as the prices of foreign products are not allowed toincrease in these cases Still our welfare loss estimates remainsubstantial drivenmdashas beforemdashby the loss of product variety

1497VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

not impose it through any of our assumptionsregarding the demand function The questionthat naturally arises then is what might explainthis finding While we cannot formally addressthis question anecdotal accounts in various in-dustry studies suggest that the explanation maylie in the differences between domestic andforeign firms in the structure and coverage ofretail distribution networks

Distribution networks for pharmaceuticals inIndia are typically organized in a hierarchicalfashion Pharmaceutical companies deal mainlywith carrying and forwarding (CampF) agents inmany instances regionally based who each sup-ply a network of stockists (wholesalers) Thesestockists in turn deal with the retail pharma-cists through whom retail sales ultimately oc-cur35 The market share enjoyed by a particularpharmaceutical product therefore depends inpart on the number of retail pharmacists who

stock the product And it is here that thereappears to be a distinction between domesticfirms and multinational subsidiaries In particu-lar the retail reach of domestic firms as agroup tends to be much more comprehensivethan that of multinational subsidiaries (IndianCredit Rating Agency (ICRA) 1999)36

There appear to be two reasons for this Thefirst is that many of the larger Indian firmsbecause they have a much larger portfolio ofproducts over which to spread the associatedfixed costs typically have more extensive net-works of medical representatives The second issimply that there are many more domestic firms(and products) on the market At the retail levelthis would imply that local pharmacists mightbe more likely to stock domestic products con-taining two different molecules say ciprofloxa-cin and norfloxacin than they would domesticand foreign versions of the same molecule Tothe extent that patients (or their doctors) arewilling to substitute across molecules in order tosave on transport or search costs (eg going toanother pharmacy to check whether a particular

35 There are estimated to be some 300000 retail pharma-cists in India On average stockists deal with about 75 retailers(ICRA 1999) There are naturally variations in this structureand a host of specific exclusive dealing and other arrangementsexists in practice Pharmaceutical firms also maintain networksof medical representatives whose main function is to marketthe companyrsquos products to doctors who do the actual prescrib-ing of drugs In some instances firms do sell directly to thedoctors who then become the ldquoretailerrdquo as far as patients areconcerned but these are relatively rare

36 These differences were also highlighted in conversa-tions one of the authors had with CEOs and managingdirectors of several pharmaceutical firms as part of a sep-arate study

TABLE 6AmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN THE NORTHERN REGION

Product group

Elasticity with respect to

Prices of foreign productgroups Prices of domestic product groups Overall

quinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 557 013dagger 015 401 011dagger 011dagger 016 137(179) (007) (007) (184) (006) (006) (006) (029)

Foreign norfloxacin 427dagger 045 427dagger 350dagger 602 451 465 220(242) (112) (242) (210) (623) (184) (183) (105)

Foreign ofloxacin 011 010dagger 138 009 009dagger 023 011 116(005) (005) (031) (027) (005) (028) (004) (017)

Domestic ciprofloxacin 018 001 001 168 008 008 010 117(008) (000) (001) (023) (002) (002) (002) (003)

Domestic norfloxacin 004 003 004 058 223 042 040 073(001) (003) (001) (017) (011) (004) (003) (009)

Domestic ofloxacin 005 005 011 077 074 342 074 089(002) (002) (013) (028) (008) (025) (008) (021)

Domestic sparfloxacin 007 004 007 115 063 063 288 028(002) (001) (002) (015) (006) (006) (017) (012)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent significance level and dagger (dagger) denotes significance at the 10-percent level

1498 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE 6BmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN OTHER REGIONS

Demand patterns in the eastern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 594 014dagger 016 431 013 011 017 140(195) (008) (008) (201) (006) (007) (007) (031)

Foreign norfloxacin 329dagger 058 329dagger 264dagger 460 346 359 192(180) (083) (179) (153) (451) (138) (139) (082)

Foreign ofloxacin 012 010dagger 140 010 010 024 013 117(006) (006) (034) (031) (005) (030) (005) (019)

Domestic ciprofloxacin 018 001 001 172 009 008 011 117(008) (000) (002) (027) (002) (003) (002) (003)

Domestic norfloxacin 004 004 004 068 242 049 046 070(002) (004) (002) (020) (013) (005) (004) (010)

Domestic ofloxacin 004dagger 004 009 061 060 295 060 092(002) (001) (010) (028) (007) (022) (007) (017)

Domestic sparfloxacin 005 003 005 092 048 051 251 043(002) (001) (001) (017) (005) (006) (016) (010)

Demand patterns in the western region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 427 009dagger 011 286 008 006 011 126(129) (005) (005) (133) (004) (005) (004) (021)

Foreign norfloxacin 714dagger 008 710dagger 594dagger 995 742 770 299dagger

(369) (183) (368) (328) (1015) (297) (296) (178)Foreign ofloxacin 013 012dagger 145 009 012 026 013 119

(007) (006) (037) (029) (005) (033) (005) (021)Domestic ciprofloxacin 019 001 000 174 009 008 011 118

(008) (000) (001) (024) (002) (003) (002) (004)Domestic norfloxacin 004 004 004 067 243 050 047 069

(002) (004) (001) (018) (014) (005) (004) (010)Domestic ofloxacin 003 003 007 048 048 257 048 093

(002) (001) (008) (022) (005) (015) (004) (014)Domestic sparfloxacin 006 003 005 083 046 049 241 046

(002) (001) (001) (014) (004) (005) (012) (009)

Demand patterns in the southern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 560 013dagger 015 402 011 010 017 137(182) (007) (007) (187) (006) (006) (007) (029)

Foreign norfloxacin 533dagger 031 531dagger 431dagger 748 559 584 249dagger

(301) (134) (300) (245) (734) (219) (219) (128)Foreign ofloxacin 014 012dagger 148 011 012 029 016 120

(007) (007) (039) (033) (006) (035) (006) (022)Domestic ciprofloxacin 017 001 000 169 008 007 011 116

(008) (000) (001) (025) (002) (003) (002) (003)Domestic norfloxacin 004 003 003 059 225 043 040 073

(001) (003) (001) (019) (012) (004) (004) (009)Domestic ofloxacin 004dagger 004 009 062 060 296 060 091

(002) (001) (010) (026) (006) (019) (007) (017)Domestic sparfloxacin 008 005 008 138 073 075 323 016

(002) (001) (002) (018) (007) (008) (021) (014)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at 5-percent level anddagger (dagger) denotes significance at the 10-percent level

1499VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

foreign product is in stock) in aggregate datawe would expect to find precisely the substitu-tion patterns that we report in Table 6

Whether the particular explanation we provideabove is the correct one the high degree of sub-stitutability between domestic product groupsturns out to have important implications for thewelfare calculations We discuss these in moredetail below when we present the results of thecounterfactual welfare analysis Another elasticitywith important implications for the counterfactu-als is the price elasticity for the quinolone subseg-ment as a whole which indicates how likelyconsumers are to switch to other antibioticsgroups when faced with a price increase for quin-olones This elasticity is computed on the basis ofthe results in Table A3 and it is at 111 (stan-dard error 024) this is large in magnitudebutmdashas expectedmdashsmaller in absolute value thanthe own-price elasticities of the product groupswithin the quinolone subsegment

The results in Tables 6A and 6B are based onour preferred specification discussed in SectionII In Tables A4 to A6 in the Appendix weexperiment with some alternative specificationsTables A4(a)ndashA4(c) correspond to a specificationthat includes in addition to product-group-specificregional fixed effects product-group-specific (andfor the upper level antibiotics-segment-specific)seasonal effects We distinguish among three sea-sonsmdashthe summer monsoon and wintermdashandreport the unconditional demand elasticities for

the northern region for each of these seasons Asevident from the tables in the Appendix our elas-ticity estimates are robust to the inclusion of sea-sonal effects The demand elasticities in Table A5are based on estimation of the demand system byOLS Compared to the elasticities obtained by IVthe OLS elasticities are smaller in absolute valueimplying that welfare calculations based on theOLS estimates would produce larger welfare lossestimates Nevertheless some of the patterns re-garding the cross-price elasticities discussed ear-lier are also evident in the OLS results inparticular the cross-price elasticities between dif-ferent domestic product groups are all positivelarge and significant and in most instances largerthan the cross-price elasticities between drugs thatcontain the same molecule but are produced byfirms of different domesticforeign status Theclose substitutability of domestic products indi-cated by both the OLS and IV estimates seems tobe one of the most robust findings of the paper

B Cost and Markup Estimates

Table 7 displays the marginal costs markupsand profits implied by the price elasticity esti-mates of Tables 6A and 6B for each of the sevenproduct groups Given that our regional effectsimply different price elasticities for each regionour marginal cost and markup estimates alsodiffer by region Given however that based on

TABLE 7mdashUPPER AND LOWER BOUNDS FOR MARGINAL COST MARKUP AND ANNUAL PROFIT BY PRODUCT GROUPS WITHIN

THE QUINOLONE SUBSEGMENT

Product groupLower boundfor MC (Rs)

Upper boundfor markup

Upper boundfor profit(Rs mill)

Upper boundfor MC (Rs)

Lower boundfor markup

Lower boundfor profit

(Rs)

Foreign ciprofloxacin 83 19 269 103 0 00(123) (012) (1655)

Foreign norfloxacin NA NA NA 53 0 00Foreign ofloxacin 323 70 1061 1085 0 00

(2316) (021) (3185)Domestic ciprofloxacin 47 59 17019 112 0 00

(114) (010) (29858)Domestic norfloxacin 52 43 2807 90 0 00

(020) (002) (1532)Domestic ofloxacin 587 34 1612 901 0 00

(218) (002) (1280)Domestic sparfloxacin 495 37 1985 788 0 00

(157) (002) (1100)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level Estimated lower bound forforeign norfloxacinrsquos marginal cost is negative since the estimated price elasticity is less than one in absolute value

1500 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Tables 6A and 6B the price elasticities do notseem to differ substantially across regions wereport for ease of exposition only the nationalaverages for marginal costs and markups inTable 7

Table 7 has two parts In the left part (firstthree columns) we report the lower bound formarginal cost and the corresponding upperbound for markup and upper bound for totalannual profit for each product group Thesenumbers are based on the price elasticities weobtained from estimating the two-level AIDSdemand system Since we do not have a reliableestimate for the price elasticity of foreign nor-floxacin (the point estimate is negative but lessthan one in absolute value and insignificant)we cannot compute the lower bound for mar-ginal cost in this case It is important to note thatthese estimates do not reflect either the actualmarginal cost or the actual markup for thesedrugs both because the existence of price reg-ulation implies that the unconstrained first-orderconditions are unlikely to hold each period andbecause our aggregation across firms of thesame domesticforeign status supplying thesame molecule makes the interpretation of theseestimates problematic In particular the fact thatwe ignore competition among firms within eachproduct group implies that our estimates willtend to overstate market power These numberswill prove useful however in the counterfac-tual simulations as they can give us a sense ofhow large the maximum profit gains for multi-national firms and the maximum profit lossesfor domestic firms are likely to be under patentenforcement The right part of the table (lastthree columns) reports the upper bound for mar-ginal cost which we obtain by simply taking themarginal cost to be equal to the observed priceWe do not report standard errors in this casesince the numbers are based on actual data Themarkups corresponding to these marginal costupper bounds are of course zero We conductthe counterfactual simulations using both thelower and upper bounds for marginal costs

The (upper bounds for) markups on the leftside of the table are generally plausible Thedomestic ciprofloxacin product group that dom-inates the quinolone subsegment and for whichwe documented high prices a high marketshare and a relatively low elasticity of demandenjoys one of the highest markup upper bounds(60 percent) and accounts for nearly 70 percent

of all profits derived within the subsegmentForeign ciprofloxacin on the other hand whichfaces the stiffest competition from domesticfirms and for which we estimated a highly elas-tic demand has the lowest markup upper bound(19 percent)

C Counterfactual Estimates of the Impact onPrices and Welfare

With estimates of the key demand and costparameters in hand we turn to the counterfac-tuals We consider the five separate scenarioslisted in the previous section All of the scenar-ios involve the withdrawal of one or more of thedomestic product groups from the market Ta-ble 8 displays our estimates of the consumerwelfare losses that result under the differentscenarios The losses are expressed in billion ofrupees per year All numbers presented in Ta-ble 8 and subsequent tables are based on usingthe lower bounds for marginal cost and upperbounds for markup in the simulations As dis-cussed earlier these numbers are the more in-teresting to work with since they give us anupper bound for the changes in the profits ofdomestic and foreign firms that would resultfrom patent enforcement In the Appendix wealso present results based on using the upperbounds for marginal costs in the simulations inwhich case the pre-TRIPS profits of domesticand foreign firms are zero In all cases marginalcosts are assumed to be constant in outputWhile naturally the profit implications differdepending on whether one uses the upper orlower bounds for marginal costs (firm profits arezero if one assumes the upper bound of mar-ginal cost in which case price equals marginalcost) the estimated consumer welfare losses aresimilar in the two cases We discuss these re-sults in more detail at the end of this section

The first column presents our estimates of theconsumer welfare losses attributable to the pureloss of product variety effect where we fix theprices of all remaining products as well as theoverall expenditure on quinolones while with-drawing one or more of the domestic productgroups Note that in our initial specification ofthe demand system had we not allowed for thepossibility that consumers might differentiatebetween domestic and foreign products evenwhen they contain the same molecule this

1501VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

particular component of the loss of consumerwelfare would not have arisen

The estimates reported in the second columnincorporate the expenditure-switching effect ontop of the loss of product variety Here basedupon the price elasticity estimates from thehigher-level AIDS system we adjust (down-ward) the expenditures allocated to the quino-lone subsegment as the composite price ofquinolones effectively increases as a consequenceof the higher virtual prices of the domestic productgroups that are withdrawn from the market Be-cause the estimates in this column are generatedassuming that the prices of the products that re-main in the market are not adjusted upward theyprovide a sense of what consumer losses would beif the introduction of product patents were coupledwith strict price regulation aimed at maintainingprices at prepatent levels Alternatively they canbe thought of as the relevant welfare numbers ifintense competition among firms within the re-maining product groups kept the prices of theproducts that were still offered in the market closeto the firmsrsquo marginal costs

The last column of Table 8 displays the esti-mated consumer welfare losses when bothcross-segment expenditure-switching and with-in-segment upward price adjustments are takeninto account

If we compare the results across the firstsecond and third columns all the counterfac-tual scenarios produce qualitatively similar pat-terns patterns that are consistent with what we

would expect Starting from the initial loss ofwelfare attributable to the loss of product vari-ety the option of switching expenditures outof the quinolone subsegment to other subseg-ments mitigates some of the initial welfareloss But if we then incorporate the upwardprice adjustments that result in response to thereduced competition the welfare losses aremagnified

Of particular interest from a policy perspec-tive are the relative magnitudes of these threeeffects which are similar under all the counter-factual scenarios though absolute levels varyconsiderably First despite the fact that the de-mand for quinolones is quite sensitive to thecomposite price of quinolonesmdashthe upper levelprice elasticity is 111mdashthe cross-subsegmentexpenditure-switching effects are in all thecases small (in absolute value terms) relative tothe other two effects For instance under thescenario where all the domestic quinolone prod-uct groups are withdrawn from the market theoverall consumer welfare loss of Rs 1781 bil-lion per year can be decomposed into an initialloss of Rs 1176 billion (66 percent) attribut-able to the loss of product variety a slightreduction in this initial loss of Rs 041 billion(2 percent) from Rs 1176 billion to Rs1135 billion because of expenditure switchingand a subsequent additional loss of Rs 646billion (36 percent) from Rs 1135 billion toRs 1781 billion because of the reduced com-petition and consequent price increases

TABLE 8mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure loss ofvariety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 732(087) (089) (146)

Only ofloxacin 008 008 023(008) (008) (010)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1253(177) (180) (415)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1058(142) (145) (331)

All four domestic quinolones products 1176dagger 1135dagger 1781(643) (634) (1270)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent significance level and dagger (dagger)denotes significance at the 10-percent level

1502 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 16: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

molecule and are hence comparable As notedearlier however a discrete choice approach at thehigher stages of the demand estimation requiresconverting revenue to physical shares which ischallenging in the case of systemic drugs such asantibiotics for which the appropriate dosage is notwell defined Finally a combination of a multi-stage budgeting approach for the higher stages anda discrete choice approach for the lowest stagewould be feasible to estimate but inconsistentwith a model of utility maximization For thesereasons we have no alternative but to adopt anapproximation for measuring the price index foreach product group

Let us denote this approximation by pjA and let

13j denote the proportional ldquoapproximationrdquo errorassociated with measuring the true price index forgroup j so that ln pj

A ln pjT ln 13j The

estimating demand equation can then be written as

i i j

ijln pjA ilnXQ

PQ

j

ijln 13j i

or

i i j

ijln pjA ilnXQ

PQ i

where the new error term i yenj ijln 13j iis comprised of two components the first onereflects ldquomeasurementrdquo error due to the fact thatthe product group prices we employ are approx-imations and not exact price indices and thesecond one captures conventional demandshocks The demand equation as written abovecorresponds to the equation we take to the data

The particular proxy we use for the exact priceindex for each product group j is the revenue-share-weighted average of the prices of multipleSKUs available within this group that is

pjA

k j

k pk

The notation above illustrates the two source ofprice variation variation in the SKU prices pkand variation in the weights k

We are concerned with potential correlation ofthe so-constructed product group prices with bothcomponents of the product group demand error

term The correlation between pjA and the first

component yenj ijln 13j is independent of theparticular way we construct pj

A and inherent in thefact that we measure the exact price index witherror The potential correlation between pj

A and thesecond component of the product group demanderror the shock i arises however because of thespecific way we construct the proxy pj

A and inparticular because of the presence of the reve-nue share weights k in pj

A since the SKUrevenue share weights k will generally dependon the product group expenditure they arelikely to be correlated with the product groupdemand shock i

To address the simultaneity bias we use in-strumental variables To this end we need vari-ables that are correlated with the proxies pj

A butuncorrelated with the error term i We use thenumber of SKUs within each product group j asan instrument as it does justice to the idea thatvariation in the product group price index stemsin part from variation in the set of SKUs that areavailable each period It is clearly correlatedwith the average group prices pj

A and plausiblyuncorrelated with the demand shock i Theassumption that the number of SKUs is uncor-related with the conventional demand error willhowever be violated if the introduction of anew SKU changes the perceived quality of adrug21 or if it is accompanied by promotionalactivities which will be reflected in i Our hopeis that this is not too often the case Assumingthat the SKU number for each group j is alsouncorrelated with the weighted sum of the mea-surement errors of all product groups yenj ijln 13j itcan be used as an instrument for the product groupprice pj

A22 If one accepts the premise that prices atthe SKU level are exogenous then SKU pricescan also be used as instruments The final list ofinstruments we use hence includes the number ofSKUs in each group the prices of the five largestSKUs for each group and all other exogenousvariables in the demand estimation such as re-gionalproduct dummy interactions and the upper-

21 Note that we already proxy for the quality of each prod-uct group through the product fixed effects i In the actualestimation we actually let these fixed effects be region-specific(ir) so that we allow quality to vary across regions

22 If the measurement error increases in the number ofincluded SKUs this assumption will be violated Unfortu-nately we have no way of checking the validity of thisassumption

1492 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

level total expenditure on antibiotics Regressionsof group prices on the instruments above yieldhigh R-squares with most regressors highly sig-nificant indicating that our instruments are highlycorrelated with prices23

Our sample includes four molecules cipro-floxacin norfloxacin ofloxacin and spar-floxacin Except for sparfloxacin all othermolecules are produced by both foreign anddomestic firms24 So we have seven products(domestic ciprofloxacin foreign ciprofloxa-cin etc) with 96 observations (two years ofmonthly data four geographical regions foreach period) for each product The parametersin the lower-level AIDS demand system asdefined in equation (8) are the product fixedeffects i the productndashregional dummy inter-actions ir the own revenue-share price elas-ticities ii the cross revenue-share priceelasticities i10 i01 i00 and the revenue-share expenditure elasticities i In estimatingthe parameters we first regress prices on allinstrumental variables and then plug the pre-dicted values for prices in the constrainedleast-square regression (For a detailed expla-nation of the constraints see the previoussection)

Given that we impose many cross-equationconstraints and employ instrumental variablesin the estimation it is difficult to derive stan-dard errors for the parameter estimates ana-lytically Our error term interpretation in (8)implies that the error terms for each productgroup are likely to be correlated across re-gions for example if a national advertisingcampaign increases the demand for a rela-tively expensive presentation in one regionsimultaneously increasing the aggregate de-mand and price index for the correspondingproduct group in that region it is likely that thesame effect will be observed in other regions Inprinciple we could exploit the cross-regionalcorrelation in the error terms to estimate thedemand system using seemingly unrelated re-

gression (SUR) This however would requireus to derive the variance-covariance matrix ofthe parameters analytically which as notedabove is cumbersome in the current contextInstead we use the bootstrap method choosingto remain agnostic about the structure of thevariance-covariance matrix The potential dis-advantage of this approach over SUR is a loss inefficiency but as will become apparent in theresults section the model parameters are fairlyprecisely estimated To maintain the marketstructure we randomly sample the periods (withreplacement) and use the same periods for allproducts Regarding the optimal number ofbootstrap repetitions ideally one would followthe three-step method proposed by Donald W KAndrews et al (2000) Empirical evidence sug-gests however that one rarely needs more than200 replications to estimate the standard er-rors25 To be safe we generate 300 bootstrapsamples (with replacement) based on the origi-nal data and estimate the standard errors usingthe standard errors of the bootstrap sampleestimates

The estimation of the top-level AIDS systemis similar The constraints imposed on this top-level demand system are adding-up homogene-ity and symmetry as defined on page 12Again bootstrapping is used to obtain the stan-dard errors of the parameter estimates

D The Counterfactual Scenarios

In assessing the effects of patent enforce-ment we start by focusing on the most extremecase in which compulsory licensing is not anoption and foreign firms are not subject to pricecontrols We use the results from the analysis ofthis case as a benchmark In reality the outcomeof the WTO negotiations is more likely to in-volve some constraints on the monopoly powerof foreign firms selling patented products indeveloping countries such as price caps orcompulsory licensing Our framework can eas-ily accommodate these cases as will becomeapparent in the next subsection

We now focus on the effects of potentialpatent enforcement in the quinolone segmentWe consider several scenarios which vary inthe number and size of domestic products that

23 Specifically the R-squares from the first-stage regres-sions of product group prices on instruments range from057 to 095 except for domestic ciprofloxacin for whichthe R-square is 017

24 Sparfloxacin is actually offered by one foreign subsid-iary in India (see Table 2) However its revenue share isminiscule We therefore treat Sparfloxacin as being pro-duced by domestic firms only 25 See Bradley Efron and Robert J Tibshirani (1993)

1493VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

will be removed from the market In particularwe consider the following five scenarios26

Withdrawal of one large domestic productgroup only domestic ciprofloxacin

Withdrawal of one relatively small domesticproduct group only domestic ofloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand norfloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand sparfloxacin

Withdrawal of all four domestic quinoloneproduct groups

As the list above suggests we proceed fromanalyzing the effects of single-product with-drawal to the analysis of eliminating the en-tire domestic segment This approach wasmotivated by early empirical results that in-dicated that the existence and extent of com-petition from domestic firms has a significantbearing on the predicted effects of patent en-forcement that is our predictions regardingprices and welfare vary substantially depend-ing on how many domestic products are af-fected by patent enforcement and on whethersome domestic competition will remain presentafter TRIPS In addition the size of the affecteddomestic groups is relevant for the welfare pre-dictions accordingly we examine both scenar-ios in which a large domestic product groupsuch as domestic ciprofloxacin is eliminatedfrom the market and scenarios in which the

eliminated domestic group is relatively small(eg domestic ofloxacin) In all scenarios wemaintain the assumption that the set of prod-ucts offered by the remaining firms in themarket does not change in response to patentenforcement

E Computation of Virtual Prices and NewEquilibrium Prices

The first step in the counterfactual analysisis to derive the new equilibrium prices underpatent enforcement27 In this context thereare two sets of prices that are relevant Thefirst set consists of the virtual prices of those(domestic) products that will not be availableonce TRIPS is put in effect To calculate thesevirtual prices we set the expenditures of therelevant products equal to zero The secondset of prices consists of the prices of thoseproducts that remain in the market In deriv-ing these prices we start by assuming profitmaximization without any form of price reg-ulation the firms remaining in the marketreoptimize in response to the policy changeand set new prices taking the prices of allother firms as given28 Of course at the equi-librium all prices change in response to thefact that some domestic products are nolonger present The new equilibrium pricesfor products that remain in the market are thuscomputed by utilizing the first-order condi-tions of profit-maximizing firms into whichthe virtual prices of the eliminated productsare substituted Hence to compute the newequilibrium prices we solve an equation sys-tem of the following form

26 As Table 2 indicates most of the patents for the drugsin the quinolone segment have expired by now so that noneof the scenarios described in this section is going to mate-rialize in practice The reason we consider these alternativescenarios in our counterfactual simulations is to get a senseof how the presence and extent of domestic competitionaffects welfare calculations Such calculations may becomerelevant in the future in other therapeutic classes in whichpatents have not expired yet and where domestic firmscompete with foreign patent holders While the particularexit scenarios will depend on the exact year of patent expiryof drugs in each therapeutic class and the specific way inwhich patents will be enforced (eg whether certain do-mestic drugs will be grandfathered in) the general conclu-sion that emerges from our calculations is that the consumerwelfare loss is substantially smaller in cases where somedomestic competition remains present in the market Hencethe particular way in which patent enforcement is imple-mented is essential for assessing its welfare impact

27 Of course until product patents are in fact introducedthese prices will not be observable Note also that we areassuming here that the range of available products will notchange with the introduction of patents

28 As mentioned above this first set of calculations ab-stracts from the existence of remaining price controls orother government regulations that would impose constraintson the firmsrsquo profit maximization problem Accordingly theresulting numbers should be interpreted as a benchmark ofwhat would happen if markets were completely unregulatedTo examine what the welfare effects of TRIPS would be inthe more realistic scenario of price regulation we subse-quently consider a scenario in which regulation keeps theprices of the products provided by patent-holders at theirpre-TRIPS level For a more detailed discussion see theresults section

1494 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

For products i that are withdrawn from themarket

(14) 0 i ir iiln pirV i10ln pjr j Di

10

j Di

01

i01ln pjr

j Di

00

i00ln pjr

ilnXQr

PQr

For products k that remain in the market

(15) pkr ckr 1 1

kk pkrpjr pirV 1

In the equations above pirV denotes the virtual

prices of the products that are removed fromthe market while pjr denotes the updatedprices of all other products Note that whensolving for the virtual prices we account forthe fact that both the price index for quino-lones PQr and the expenditure allocated tothis subsegment XQr need to be updated toreflect the fact that as a result of the pricechanges there may be substitution away fromthis subsegment To obtain the new quinoloneexpenditure XQr and the new price index PQrwe use the estimates and formulas for thehigher-level AIDS system In equation (15)ck refers to the marginal cost for product k thatwe have obtained from the previous estimationstage As mentioned in the previous subsectionwe conducted the simulations using both theupper and lower bounds for marginal cost ck

u

and ckL respectively The term kk(pkr

pjrpirV) re-

fers to the unconditional own-price elasticity forproduct k which is a function of the eliminatedproductsrsquo virtual prices and the remaining prod-uctsrsquo new equilibrium prices We conduct thecounterfactual simulations and welfare analysisat the regional level because the presence ofregion-specific product effects in the demandestimation implies region-specific demand elas-ticities and hence region-specific marginalcosts and markups The presentation and discus-sion of our results focus on national averages ofthe relevant variables that we construct by com-

puting weighted averages across regions usingthe population of each region as a weight

F Welfare Assessment

The simulation of the new equilibrium underpatent protection can provide important insightsinto how consumers and firms will respond tothe removal of domestic products in the market(for example toward which products consumerswill substitute or which prices will increase themost) To get a more precise idea of how peo-plersquos well-being will ultimately be affected byTRIPS we compute as a last step in our anal-ysis the welfare effects of the policy changeSocial welfare is defined as consumer welfareand the sum of domestic firm profits Thechange in domestic profits can easily be calcu-lated by comparing the domestic firm (variable)profits at the pre-TRIPS prices to the profitsthese firms will realize at the new simulatedprices Although foreign firm profits do notcount in domestic welfare calculations we alsocompute the effects of patent enforcement onforeign firm profits to get an idea of how largethe expected benefits of TRIPS are for thesefirms This provides in some sense an indirectway of assessing whether the claims that patentenforcement in countries like India will lead tomore research on developing-country-specificdiseases (such as malaria) have any validity iffor example we find that the effect of patentenforcement on the foreign firm profits realizedin India is small in magnitude it is unlikely thatforeign firms will engage in more developing-country-specific research in response to TRIPSIt is important to note that in all these calcula-tions we work with the lower-bound estimatesfor marginal costs since these give rise to thehighest possible markups Hence our estimateof profit loss for domestic producers most likelyoverstates this loss

On the consumer side we measure changes inconsumer welfare by the compensating variation(CV) defined as the additional expenditure thatconsumers need in order to achieve the sameutility level as before patent enforcement at thenew prices Specifically let P0 denote the pricevector before patent enforcement P the simulatedprice vector post-TRIPS (that we obtained usingthe methods described in the previous subsection)u0 the utility attained by consumers before TRIPS

1495VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

and E(u P) the higher-level expenditure functionThen the compensating variation is given by

(16) CV Eu0 P Eu0 P0

Note that the CV as computed in (16) representsthe combination of three effects

The pure product-variety effect that is theeffect that arises because one or more prod-ucts are not available to consumers anymoreholding the prices of all other remainingproducts and the total expenditure on thequinolone subsegment XQ constant

The expenditure-switching effect that is theeffect arising from substitution away fromquinolones and toward other subsegments ofthe antibiotics market again holding the pricesof all other remaining products constant

The reduced-competition effect that is theeffect that arises because the firms remainingin the market adjust (increase) their prices inresponse to the removal of domestic products

From both an analytical and a policy point ofview it is desirable to assess how large each ofthese effects is Accordingly we decompose thetotal effect on consumer welfare (the CV as givenby equation (16)) using the following procedure

To get the pure product variety effect wecompute virtual prices for the products that areremoved from the market holding the quino-lone expenditure XQ and the prices of all otherproducts fixed Let us call the resulting pricevector P1 Then the pure product variety effectis represented by E(u0 P1) E(u0 P0)

To compute the expenditure-switching effectwe compute another set of virtual prices againholding the prices of all remaining products fixedbut letting quinolone expenditure adjust in re-sponse to the new price index for the quinolonesegment (Given that the prices of the remainingproducts remain fixed the change in the priceindex arises only because of the removal of one ormore domestic products) Note that this scenariomost closely resembles the case in which patentlaws are enforced but strict price regulation keepsthe prices of the products offered in the market attheir pre-TRIP level Let us label the so-computedprice vector P2 The expenditure-switching effectis then E(u0 P2) E(u0 P1)

Finally the reduced competition effect aris-ing from higher prices for the remaining prod-

ucts is computed as the residual change in thecompensating variation once the product-varietyand expenditure-switching effects have been ac-counted for that is E(u0 P) E(u0 P2)where the price vector P is computed accordingto the formulas (14) and (15) to reflect theadjustment of prices to the new regime

To compute the standard errors associatedwith the counterfactual simulations (that is thestandard errors for the counterfactual prices andwelfare estimates) we again use bootstrappingWe first bootstrap the original sample (sales andprices of all the drugs) 300 times Next we esti-mate the AIDS model for each of these samplesand compute the counterfactual equilibrium pricesand welfare losses corresponding to each policyscenario for each of the 300 simulations In thefinal step we compute standard errors29

One limitation of our framework is that itdoes not allow for a heterogeneous response ofconsumers to the policy change Accordinglyour framework is not suited to addressing thequestion of how different groups in the popula-tion will be affected by patent law enforce-ment30 Along the same lines our frameworkdoes not accommodate the possibility of pricediscrimination which might lead to differentresults regarding welfare losses and profit gainsrelative to uniform pricing31 Still we believethat the results of the counterfactual simulationscan provide important insights into the likelyaggregate response to patent enforcement andthe factors that drive this response

III Results

A The Structure of Demand

Tables A2 and A3 in the Appendix display theresults from estimation of the lower- and upper-level AIDS systems respectively For ease of in-terpretation rather than discussing the coefficientestimates we focus our discussion on the implied

29 In cases some of the simulations do not converge wecompute the standard errors using only those simulationsthat converged

30 Addressing this question would require at a mini-mum micro data on consumer purchases as in Goldberg(1995) To our knowledge such data do not exist for theIndian pharmaceutical market

31 See Ernst R Berndt (1994) for an extensive discussionof uniform pricing

1496 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

unconditional price and expenditure elasticities re-ported in Table 6 Given that the region-specificproduct group effects ir reported in the last fourcolumns of Tables A2 and A3 imply region-spe-cific demand elasticities we report separate elas-ticities for each region in Tables 6A and 6B Asevident from the comparison of these tables theelasticities are very similar across regions so thatwe can focus the remaining discussion on theelasticities of one region only the northern regiondisplayed in Table 6A

The diagonal terms of Table 6A report theown-price elasticities which are in all but onecase negative and highly significant The oneexception is the foreign norfloxacin productgroupmdashwhose share of quinolone sales is 007percentmdashfor which we estimate a negative butinsignificant own-price elasticity For the remain-ing product groups demand appears to be highlyelastic with the estimated elasticities lower than2 in four out of the six cases The magnitude ofthe own-price elasticities matches the features ofthe Indian pharmaceutical market mentioned ear-lier which would suggest that Indian consumersare likely to be quite price-sensitive32 The elas-ticities appear especially large if one takes intoaccount that they refer to product groups (such asdomestic ciprofloxacin) and not individual drugsoffered by particular firms Their relative magni-tudes are also intuitive the drug with the largestmarket share and a relatively high price (domesticciprofloxacin) appears to be one of the least elas-tic In contrast foreign ciprofloxacin is highlyprice elastic this is plausible as ciprofloxacindrugs offered by subsidiaries of multinationalsface the stiffest competition in this market seg-ment from approximately 75 Indian firms offeringthe same molecule

The estimated expenditure elasticities appear inthe last column These are all positive indicatingthat the demand for all product groups is normalThe remaining cells display the estimated cross-price elasticities As one might perhaps expect forproducts within a therapeutic subsegment theseare mostly positive Out of a total of 49 priceelasticities we estimate there are six that do notconform to expectations these are the cross-price elasticities between different foreign prod-

uct groups which are estimated to be negativeand significant33 Fortunately these elasticitieshave negligible impact on the welfare analysisgiven that our counterfactuals focus on theeffect of withdrawing one or more domesticproducts from the market the most relevantelasticities are the ones that capture the responseof various product group shares to a change in theprice of one or more domestic groups these elas-ticities are the cross-price elasticities between var-ious domestic groups and the ones betweendomestic and foreign groups which are plausibleand precisely estimated34

Regarding these elasticities a striking aspectof our estimates is how large positive andsignificant the cross-price elasticities betweendifferent domestic product groups aremdashin factfor norfloxacin and ofloxacin we estimate thatdomestic product groups containing differentmolecules are closer substitutes for one anotherthan product groups that contain the same mol-ecule but are produced by foreign firms Incontrast for ciprofloxacin (the molecule withthe largest revenue share) we estimate a largepositive cross-price elasticity between the do-mestic and foreign versions

The fact that domestic products appear to beclose substitutes for other domestic productsthat contain different molecules truly representsan ldquoempiricalrdquo finding in the sense that we do

32 In developed economies elasticities of this magnitude havetypically been found only for generic drugs (and even then onlyrarely) or among consumers who lack health insurance

33 While these elasticities are clearly counterintuitive theyare not inconsistent with the underlying demand system whichimposes no restrictions on the sign of the cross-price elastici-ties We do not have a good explanation of why these elastic-ities are estimated to be negative A potential explanation isthat the shares of the foreign products are very small giventhis we observe very few consumers switching from one (verysmall) foreign group to another (very small) foreign groupwhen the price of the first foreign group goes up and hence theinference is not very reliable in this case

34 The cross-price elasticities between foreign drug groupscontaining different molecules will also have an effect on thewelfare estimates given that the withdrawal of domestic prod-ucts will generally lead to changes in the prices of all foreignproducts and market shares will be reallocated from someforeign products to other foreign products based on the newprices However this reallocation from ldquoforeign to foreignrdquo istruly second order in our case compared to the reallocationfrom ldquodomestic to domesticrdquo and ldquodomestic to foreignrdquo prod-ucts In scenarios in which we consider patent enforcementaccompanied by strict price regulation the cross-price elastic-ities between foreign product groups are in fact completelyirrelevant as the prices of foreign products are not allowed toincrease in these cases Still our welfare loss estimates remainsubstantial drivenmdashas beforemdashby the loss of product variety

1497VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

not impose it through any of our assumptionsregarding the demand function The questionthat naturally arises then is what might explainthis finding While we cannot formally addressthis question anecdotal accounts in various in-dustry studies suggest that the explanation maylie in the differences between domestic andforeign firms in the structure and coverage ofretail distribution networks

Distribution networks for pharmaceuticals inIndia are typically organized in a hierarchicalfashion Pharmaceutical companies deal mainlywith carrying and forwarding (CampF) agents inmany instances regionally based who each sup-ply a network of stockists (wholesalers) Thesestockists in turn deal with the retail pharma-cists through whom retail sales ultimately oc-cur35 The market share enjoyed by a particularpharmaceutical product therefore depends inpart on the number of retail pharmacists who

stock the product And it is here that thereappears to be a distinction between domesticfirms and multinational subsidiaries In particu-lar the retail reach of domestic firms as agroup tends to be much more comprehensivethan that of multinational subsidiaries (IndianCredit Rating Agency (ICRA) 1999)36

There appear to be two reasons for this Thefirst is that many of the larger Indian firmsbecause they have a much larger portfolio ofproducts over which to spread the associatedfixed costs typically have more extensive net-works of medical representatives The second issimply that there are many more domestic firms(and products) on the market At the retail levelthis would imply that local pharmacists mightbe more likely to stock domestic products con-taining two different molecules say ciprofloxa-cin and norfloxacin than they would domesticand foreign versions of the same molecule Tothe extent that patients (or their doctors) arewilling to substitute across molecules in order tosave on transport or search costs (eg going toanother pharmacy to check whether a particular

35 There are estimated to be some 300000 retail pharma-cists in India On average stockists deal with about 75 retailers(ICRA 1999) There are naturally variations in this structureand a host of specific exclusive dealing and other arrangementsexists in practice Pharmaceutical firms also maintain networksof medical representatives whose main function is to marketthe companyrsquos products to doctors who do the actual prescrib-ing of drugs In some instances firms do sell directly to thedoctors who then become the ldquoretailerrdquo as far as patients areconcerned but these are relatively rare

36 These differences were also highlighted in conversa-tions one of the authors had with CEOs and managingdirectors of several pharmaceutical firms as part of a sep-arate study

TABLE 6AmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN THE NORTHERN REGION

Product group

Elasticity with respect to

Prices of foreign productgroups Prices of domestic product groups Overall

quinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 557 013dagger 015 401 011dagger 011dagger 016 137(179) (007) (007) (184) (006) (006) (006) (029)

Foreign norfloxacin 427dagger 045 427dagger 350dagger 602 451 465 220(242) (112) (242) (210) (623) (184) (183) (105)

Foreign ofloxacin 011 010dagger 138 009 009dagger 023 011 116(005) (005) (031) (027) (005) (028) (004) (017)

Domestic ciprofloxacin 018 001 001 168 008 008 010 117(008) (000) (001) (023) (002) (002) (002) (003)

Domestic norfloxacin 004 003 004 058 223 042 040 073(001) (003) (001) (017) (011) (004) (003) (009)

Domestic ofloxacin 005 005 011 077 074 342 074 089(002) (002) (013) (028) (008) (025) (008) (021)

Domestic sparfloxacin 007 004 007 115 063 063 288 028(002) (001) (002) (015) (006) (006) (017) (012)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent significance level and dagger (dagger) denotes significance at the 10-percent level

1498 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE 6BmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN OTHER REGIONS

Demand patterns in the eastern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 594 014dagger 016 431 013 011 017 140(195) (008) (008) (201) (006) (007) (007) (031)

Foreign norfloxacin 329dagger 058 329dagger 264dagger 460 346 359 192(180) (083) (179) (153) (451) (138) (139) (082)

Foreign ofloxacin 012 010dagger 140 010 010 024 013 117(006) (006) (034) (031) (005) (030) (005) (019)

Domestic ciprofloxacin 018 001 001 172 009 008 011 117(008) (000) (002) (027) (002) (003) (002) (003)

Domestic norfloxacin 004 004 004 068 242 049 046 070(002) (004) (002) (020) (013) (005) (004) (010)

Domestic ofloxacin 004dagger 004 009 061 060 295 060 092(002) (001) (010) (028) (007) (022) (007) (017)

Domestic sparfloxacin 005 003 005 092 048 051 251 043(002) (001) (001) (017) (005) (006) (016) (010)

Demand patterns in the western region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 427 009dagger 011 286 008 006 011 126(129) (005) (005) (133) (004) (005) (004) (021)

Foreign norfloxacin 714dagger 008 710dagger 594dagger 995 742 770 299dagger

(369) (183) (368) (328) (1015) (297) (296) (178)Foreign ofloxacin 013 012dagger 145 009 012 026 013 119

(007) (006) (037) (029) (005) (033) (005) (021)Domestic ciprofloxacin 019 001 000 174 009 008 011 118

(008) (000) (001) (024) (002) (003) (002) (004)Domestic norfloxacin 004 004 004 067 243 050 047 069

(002) (004) (001) (018) (014) (005) (004) (010)Domestic ofloxacin 003 003 007 048 048 257 048 093

(002) (001) (008) (022) (005) (015) (004) (014)Domestic sparfloxacin 006 003 005 083 046 049 241 046

(002) (001) (001) (014) (004) (005) (012) (009)

Demand patterns in the southern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 560 013dagger 015 402 011 010 017 137(182) (007) (007) (187) (006) (006) (007) (029)

Foreign norfloxacin 533dagger 031 531dagger 431dagger 748 559 584 249dagger

(301) (134) (300) (245) (734) (219) (219) (128)Foreign ofloxacin 014 012dagger 148 011 012 029 016 120

(007) (007) (039) (033) (006) (035) (006) (022)Domestic ciprofloxacin 017 001 000 169 008 007 011 116

(008) (000) (001) (025) (002) (003) (002) (003)Domestic norfloxacin 004 003 003 059 225 043 040 073

(001) (003) (001) (019) (012) (004) (004) (009)Domestic ofloxacin 004dagger 004 009 062 060 296 060 091

(002) (001) (010) (026) (006) (019) (007) (017)Domestic sparfloxacin 008 005 008 138 073 075 323 016

(002) (001) (002) (018) (007) (008) (021) (014)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at 5-percent level anddagger (dagger) denotes significance at the 10-percent level

1499VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

foreign product is in stock) in aggregate datawe would expect to find precisely the substitu-tion patterns that we report in Table 6

Whether the particular explanation we provideabove is the correct one the high degree of sub-stitutability between domestic product groupsturns out to have important implications for thewelfare calculations We discuss these in moredetail below when we present the results of thecounterfactual welfare analysis Another elasticitywith important implications for the counterfactu-als is the price elasticity for the quinolone subseg-ment as a whole which indicates how likelyconsumers are to switch to other antibioticsgroups when faced with a price increase for quin-olones This elasticity is computed on the basis ofthe results in Table A3 and it is at 111 (stan-dard error 024) this is large in magnitudebutmdashas expectedmdashsmaller in absolute value thanthe own-price elasticities of the product groupswithin the quinolone subsegment

The results in Tables 6A and 6B are based onour preferred specification discussed in SectionII In Tables A4 to A6 in the Appendix weexperiment with some alternative specificationsTables A4(a)ndashA4(c) correspond to a specificationthat includes in addition to product-group-specificregional fixed effects product-group-specific (andfor the upper level antibiotics-segment-specific)seasonal effects We distinguish among three sea-sonsmdashthe summer monsoon and wintermdashandreport the unconditional demand elasticities for

the northern region for each of these seasons Asevident from the tables in the Appendix our elas-ticity estimates are robust to the inclusion of sea-sonal effects The demand elasticities in Table A5are based on estimation of the demand system byOLS Compared to the elasticities obtained by IVthe OLS elasticities are smaller in absolute valueimplying that welfare calculations based on theOLS estimates would produce larger welfare lossestimates Nevertheless some of the patterns re-garding the cross-price elasticities discussed ear-lier are also evident in the OLS results inparticular the cross-price elasticities between dif-ferent domestic product groups are all positivelarge and significant and in most instances largerthan the cross-price elasticities between drugs thatcontain the same molecule but are produced byfirms of different domesticforeign status Theclose substitutability of domestic products indi-cated by both the OLS and IV estimates seems tobe one of the most robust findings of the paper

B Cost and Markup Estimates

Table 7 displays the marginal costs markupsand profits implied by the price elasticity esti-mates of Tables 6A and 6B for each of the sevenproduct groups Given that our regional effectsimply different price elasticities for each regionour marginal cost and markup estimates alsodiffer by region Given however that based on

TABLE 7mdashUPPER AND LOWER BOUNDS FOR MARGINAL COST MARKUP AND ANNUAL PROFIT BY PRODUCT GROUPS WITHIN

THE QUINOLONE SUBSEGMENT

Product groupLower boundfor MC (Rs)

Upper boundfor markup

Upper boundfor profit(Rs mill)

Upper boundfor MC (Rs)

Lower boundfor markup

Lower boundfor profit

(Rs)

Foreign ciprofloxacin 83 19 269 103 0 00(123) (012) (1655)

Foreign norfloxacin NA NA NA 53 0 00Foreign ofloxacin 323 70 1061 1085 0 00

(2316) (021) (3185)Domestic ciprofloxacin 47 59 17019 112 0 00

(114) (010) (29858)Domestic norfloxacin 52 43 2807 90 0 00

(020) (002) (1532)Domestic ofloxacin 587 34 1612 901 0 00

(218) (002) (1280)Domestic sparfloxacin 495 37 1985 788 0 00

(157) (002) (1100)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level Estimated lower bound forforeign norfloxacinrsquos marginal cost is negative since the estimated price elasticity is less than one in absolute value

1500 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Tables 6A and 6B the price elasticities do notseem to differ substantially across regions wereport for ease of exposition only the nationalaverages for marginal costs and markups inTable 7

Table 7 has two parts In the left part (firstthree columns) we report the lower bound formarginal cost and the corresponding upperbound for markup and upper bound for totalannual profit for each product group Thesenumbers are based on the price elasticities weobtained from estimating the two-level AIDSdemand system Since we do not have a reliableestimate for the price elasticity of foreign nor-floxacin (the point estimate is negative but lessthan one in absolute value and insignificant)we cannot compute the lower bound for mar-ginal cost in this case It is important to note thatthese estimates do not reflect either the actualmarginal cost or the actual markup for thesedrugs both because the existence of price reg-ulation implies that the unconstrained first-orderconditions are unlikely to hold each period andbecause our aggregation across firms of thesame domesticforeign status supplying thesame molecule makes the interpretation of theseestimates problematic In particular the fact thatwe ignore competition among firms within eachproduct group implies that our estimates willtend to overstate market power These numberswill prove useful however in the counterfac-tual simulations as they can give us a sense ofhow large the maximum profit gains for multi-national firms and the maximum profit lossesfor domestic firms are likely to be under patentenforcement The right part of the table (lastthree columns) reports the upper bound for mar-ginal cost which we obtain by simply taking themarginal cost to be equal to the observed priceWe do not report standard errors in this casesince the numbers are based on actual data Themarkups corresponding to these marginal costupper bounds are of course zero We conductthe counterfactual simulations using both thelower and upper bounds for marginal costs

The (upper bounds for) markups on the leftside of the table are generally plausible Thedomestic ciprofloxacin product group that dom-inates the quinolone subsegment and for whichwe documented high prices a high marketshare and a relatively low elasticity of demandenjoys one of the highest markup upper bounds(60 percent) and accounts for nearly 70 percent

of all profits derived within the subsegmentForeign ciprofloxacin on the other hand whichfaces the stiffest competition from domesticfirms and for which we estimated a highly elas-tic demand has the lowest markup upper bound(19 percent)

C Counterfactual Estimates of the Impact onPrices and Welfare

With estimates of the key demand and costparameters in hand we turn to the counterfac-tuals We consider the five separate scenarioslisted in the previous section All of the scenar-ios involve the withdrawal of one or more of thedomestic product groups from the market Ta-ble 8 displays our estimates of the consumerwelfare losses that result under the differentscenarios The losses are expressed in billion ofrupees per year All numbers presented in Ta-ble 8 and subsequent tables are based on usingthe lower bounds for marginal cost and upperbounds for markup in the simulations As dis-cussed earlier these numbers are the more in-teresting to work with since they give us anupper bound for the changes in the profits ofdomestic and foreign firms that would resultfrom patent enforcement In the Appendix wealso present results based on using the upperbounds for marginal costs in the simulations inwhich case the pre-TRIPS profits of domesticand foreign firms are zero In all cases marginalcosts are assumed to be constant in outputWhile naturally the profit implications differdepending on whether one uses the upper orlower bounds for marginal costs (firm profits arezero if one assumes the upper bound of mar-ginal cost in which case price equals marginalcost) the estimated consumer welfare losses aresimilar in the two cases We discuss these re-sults in more detail at the end of this section

The first column presents our estimates of theconsumer welfare losses attributable to the pureloss of product variety effect where we fix theprices of all remaining products as well as theoverall expenditure on quinolones while with-drawing one or more of the domestic productgroups Note that in our initial specification ofthe demand system had we not allowed for thepossibility that consumers might differentiatebetween domestic and foreign products evenwhen they contain the same molecule this

1501VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

particular component of the loss of consumerwelfare would not have arisen

The estimates reported in the second columnincorporate the expenditure-switching effect ontop of the loss of product variety Here basedupon the price elasticity estimates from thehigher-level AIDS system we adjust (down-ward) the expenditures allocated to the quino-lone subsegment as the composite price ofquinolones effectively increases as a consequenceof the higher virtual prices of the domestic productgroups that are withdrawn from the market Be-cause the estimates in this column are generatedassuming that the prices of the products that re-main in the market are not adjusted upward theyprovide a sense of what consumer losses would beif the introduction of product patents were coupledwith strict price regulation aimed at maintainingprices at prepatent levels Alternatively they canbe thought of as the relevant welfare numbers ifintense competition among firms within the re-maining product groups kept the prices of theproducts that were still offered in the market closeto the firmsrsquo marginal costs

The last column of Table 8 displays the esti-mated consumer welfare losses when bothcross-segment expenditure-switching and with-in-segment upward price adjustments are takeninto account

If we compare the results across the firstsecond and third columns all the counterfac-tual scenarios produce qualitatively similar pat-terns patterns that are consistent with what we

would expect Starting from the initial loss ofwelfare attributable to the loss of product vari-ety the option of switching expenditures outof the quinolone subsegment to other subseg-ments mitigates some of the initial welfareloss But if we then incorporate the upwardprice adjustments that result in response to thereduced competition the welfare losses aremagnified

Of particular interest from a policy perspec-tive are the relative magnitudes of these threeeffects which are similar under all the counter-factual scenarios though absolute levels varyconsiderably First despite the fact that the de-mand for quinolones is quite sensitive to thecomposite price of quinolonesmdashthe upper levelprice elasticity is 111mdashthe cross-subsegmentexpenditure-switching effects are in all thecases small (in absolute value terms) relative tothe other two effects For instance under thescenario where all the domestic quinolone prod-uct groups are withdrawn from the market theoverall consumer welfare loss of Rs 1781 bil-lion per year can be decomposed into an initialloss of Rs 1176 billion (66 percent) attribut-able to the loss of product variety a slightreduction in this initial loss of Rs 041 billion(2 percent) from Rs 1176 billion to Rs1135 billion because of expenditure switchingand a subsequent additional loss of Rs 646billion (36 percent) from Rs 1135 billion toRs 1781 billion because of the reduced com-petition and consequent price increases

TABLE 8mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure loss ofvariety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 732(087) (089) (146)

Only ofloxacin 008 008 023(008) (008) (010)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1253(177) (180) (415)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1058(142) (145) (331)

All four domestic quinolones products 1176dagger 1135dagger 1781(643) (634) (1270)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent significance level and dagger (dagger)denotes significance at the 10-percent level

1502 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 17: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

level total expenditure on antibiotics Regressionsof group prices on the instruments above yieldhigh R-squares with most regressors highly sig-nificant indicating that our instruments are highlycorrelated with prices23

Our sample includes four molecules cipro-floxacin norfloxacin ofloxacin and spar-floxacin Except for sparfloxacin all othermolecules are produced by both foreign anddomestic firms24 So we have seven products(domestic ciprofloxacin foreign ciprofloxa-cin etc) with 96 observations (two years ofmonthly data four geographical regions foreach period) for each product The parametersin the lower-level AIDS demand system asdefined in equation (8) are the product fixedeffects i the productndashregional dummy inter-actions ir the own revenue-share price elas-ticities ii the cross revenue-share priceelasticities i10 i01 i00 and the revenue-share expenditure elasticities i In estimatingthe parameters we first regress prices on allinstrumental variables and then plug the pre-dicted values for prices in the constrainedleast-square regression (For a detailed expla-nation of the constraints see the previoussection)

Given that we impose many cross-equationconstraints and employ instrumental variablesin the estimation it is difficult to derive stan-dard errors for the parameter estimates ana-lytically Our error term interpretation in (8)implies that the error terms for each productgroup are likely to be correlated across re-gions for example if a national advertisingcampaign increases the demand for a rela-tively expensive presentation in one regionsimultaneously increasing the aggregate de-mand and price index for the correspondingproduct group in that region it is likely that thesame effect will be observed in other regions Inprinciple we could exploit the cross-regionalcorrelation in the error terms to estimate thedemand system using seemingly unrelated re-

gression (SUR) This however would requireus to derive the variance-covariance matrix ofthe parameters analytically which as notedabove is cumbersome in the current contextInstead we use the bootstrap method choosingto remain agnostic about the structure of thevariance-covariance matrix The potential dis-advantage of this approach over SUR is a loss inefficiency but as will become apparent in theresults section the model parameters are fairlyprecisely estimated To maintain the marketstructure we randomly sample the periods (withreplacement) and use the same periods for allproducts Regarding the optimal number ofbootstrap repetitions ideally one would followthe three-step method proposed by Donald W KAndrews et al (2000) Empirical evidence sug-gests however that one rarely needs more than200 replications to estimate the standard er-rors25 To be safe we generate 300 bootstrapsamples (with replacement) based on the origi-nal data and estimate the standard errors usingthe standard errors of the bootstrap sampleestimates

The estimation of the top-level AIDS systemis similar The constraints imposed on this top-level demand system are adding-up homogene-ity and symmetry as defined on page 12Again bootstrapping is used to obtain the stan-dard errors of the parameter estimates

D The Counterfactual Scenarios

In assessing the effects of patent enforce-ment we start by focusing on the most extremecase in which compulsory licensing is not anoption and foreign firms are not subject to pricecontrols We use the results from the analysis ofthis case as a benchmark In reality the outcomeof the WTO negotiations is more likely to in-volve some constraints on the monopoly powerof foreign firms selling patented products indeveloping countries such as price caps orcompulsory licensing Our framework can eas-ily accommodate these cases as will becomeapparent in the next subsection

We now focus on the effects of potentialpatent enforcement in the quinolone segmentWe consider several scenarios which vary inthe number and size of domestic products that

23 Specifically the R-squares from the first-stage regres-sions of product group prices on instruments range from057 to 095 except for domestic ciprofloxacin for whichthe R-square is 017

24 Sparfloxacin is actually offered by one foreign subsid-iary in India (see Table 2) However its revenue share isminiscule We therefore treat Sparfloxacin as being pro-duced by domestic firms only 25 See Bradley Efron and Robert J Tibshirani (1993)

1493VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

will be removed from the market In particularwe consider the following five scenarios26

Withdrawal of one large domestic productgroup only domestic ciprofloxacin

Withdrawal of one relatively small domesticproduct group only domestic ofloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand norfloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand sparfloxacin

Withdrawal of all four domestic quinoloneproduct groups

As the list above suggests we proceed fromanalyzing the effects of single-product with-drawal to the analysis of eliminating the en-tire domestic segment This approach wasmotivated by early empirical results that in-dicated that the existence and extent of com-petition from domestic firms has a significantbearing on the predicted effects of patent en-forcement that is our predictions regardingprices and welfare vary substantially depend-ing on how many domestic products are af-fected by patent enforcement and on whethersome domestic competition will remain presentafter TRIPS In addition the size of the affecteddomestic groups is relevant for the welfare pre-dictions accordingly we examine both scenar-ios in which a large domestic product groupsuch as domestic ciprofloxacin is eliminatedfrom the market and scenarios in which the

eliminated domestic group is relatively small(eg domestic ofloxacin) In all scenarios wemaintain the assumption that the set of prod-ucts offered by the remaining firms in themarket does not change in response to patentenforcement

E Computation of Virtual Prices and NewEquilibrium Prices

The first step in the counterfactual analysisis to derive the new equilibrium prices underpatent enforcement27 In this context thereare two sets of prices that are relevant Thefirst set consists of the virtual prices of those(domestic) products that will not be availableonce TRIPS is put in effect To calculate thesevirtual prices we set the expenditures of therelevant products equal to zero The secondset of prices consists of the prices of thoseproducts that remain in the market In deriv-ing these prices we start by assuming profitmaximization without any form of price reg-ulation the firms remaining in the marketreoptimize in response to the policy changeand set new prices taking the prices of allother firms as given28 Of course at the equi-librium all prices change in response to thefact that some domestic products are nolonger present The new equilibrium pricesfor products that remain in the market are thuscomputed by utilizing the first-order condi-tions of profit-maximizing firms into whichthe virtual prices of the eliminated productsare substituted Hence to compute the newequilibrium prices we solve an equation sys-tem of the following form

26 As Table 2 indicates most of the patents for the drugsin the quinolone segment have expired by now so that noneof the scenarios described in this section is going to mate-rialize in practice The reason we consider these alternativescenarios in our counterfactual simulations is to get a senseof how the presence and extent of domestic competitionaffects welfare calculations Such calculations may becomerelevant in the future in other therapeutic classes in whichpatents have not expired yet and where domestic firmscompete with foreign patent holders While the particularexit scenarios will depend on the exact year of patent expiryof drugs in each therapeutic class and the specific way inwhich patents will be enforced (eg whether certain do-mestic drugs will be grandfathered in) the general conclu-sion that emerges from our calculations is that the consumerwelfare loss is substantially smaller in cases where somedomestic competition remains present in the market Hencethe particular way in which patent enforcement is imple-mented is essential for assessing its welfare impact

27 Of course until product patents are in fact introducedthese prices will not be observable Note also that we areassuming here that the range of available products will notchange with the introduction of patents

28 As mentioned above this first set of calculations ab-stracts from the existence of remaining price controls orother government regulations that would impose constraintson the firmsrsquo profit maximization problem Accordingly theresulting numbers should be interpreted as a benchmark ofwhat would happen if markets were completely unregulatedTo examine what the welfare effects of TRIPS would be inthe more realistic scenario of price regulation we subse-quently consider a scenario in which regulation keeps theprices of the products provided by patent-holders at theirpre-TRIPS level For a more detailed discussion see theresults section

1494 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

For products i that are withdrawn from themarket

(14) 0 i ir iiln pirV i10ln pjr j Di

10

j Di

01

i01ln pjr

j Di

00

i00ln pjr

ilnXQr

PQr

For products k that remain in the market

(15) pkr ckr 1 1

kk pkrpjr pirV 1

In the equations above pirV denotes the virtual

prices of the products that are removed fromthe market while pjr denotes the updatedprices of all other products Note that whensolving for the virtual prices we account forthe fact that both the price index for quino-lones PQr and the expenditure allocated tothis subsegment XQr need to be updated toreflect the fact that as a result of the pricechanges there may be substitution away fromthis subsegment To obtain the new quinoloneexpenditure XQr and the new price index PQrwe use the estimates and formulas for thehigher-level AIDS system In equation (15)ck refers to the marginal cost for product k thatwe have obtained from the previous estimationstage As mentioned in the previous subsectionwe conducted the simulations using both theupper and lower bounds for marginal cost ck

u

and ckL respectively The term kk(pkr

pjrpirV) re-

fers to the unconditional own-price elasticity forproduct k which is a function of the eliminatedproductsrsquo virtual prices and the remaining prod-uctsrsquo new equilibrium prices We conduct thecounterfactual simulations and welfare analysisat the regional level because the presence ofregion-specific product effects in the demandestimation implies region-specific demand elas-ticities and hence region-specific marginalcosts and markups The presentation and discus-sion of our results focus on national averages ofthe relevant variables that we construct by com-

puting weighted averages across regions usingthe population of each region as a weight

F Welfare Assessment

The simulation of the new equilibrium underpatent protection can provide important insightsinto how consumers and firms will respond tothe removal of domestic products in the market(for example toward which products consumerswill substitute or which prices will increase themost) To get a more precise idea of how peo-plersquos well-being will ultimately be affected byTRIPS we compute as a last step in our anal-ysis the welfare effects of the policy changeSocial welfare is defined as consumer welfareand the sum of domestic firm profits Thechange in domestic profits can easily be calcu-lated by comparing the domestic firm (variable)profits at the pre-TRIPS prices to the profitsthese firms will realize at the new simulatedprices Although foreign firm profits do notcount in domestic welfare calculations we alsocompute the effects of patent enforcement onforeign firm profits to get an idea of how largethe expected benefits of TRIPS are for thesefirms This provides in some sense an indirectway of assessing whether the claims that patentenforcement in countries like India will lead tomore research on developing-country-specificdiseases (such as malaria) have any validity iffor example we find that the effect of patentenforcement on the foreign firm profits realizedin India is small in magnitude it is unlikely thatforeign firms will engage in more developing-country-specific research in response to TRIPSIt is important to note that in all these calcula-tions we work with the lower-bound estimatesfor marginal costs since these give rise to thehighest possible markups Hence our estimateof profit loss for domestic producers most likelyoverstates this loss

On the consumer side we measure changes inconsumer welfare by the compensating variation(CV) defined as the additional expenditure thatconsumers need in order to achieve the sameutility level as before patent enforcement at thenew prices Specifically let P0 denote the pricevector before patent enforcement P the simulatedprice vector post-TRIPS (that we obtained usingthe methods described in the previous subsection)u0 the utility attained by consumers before TRIPS

1495VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

and E(u P) the higher-level expenditure functionThen the compensating variation is given by

(16) CV Eu0 P Eu0 P0

Note that the CV as computed in (16) representsthe combination of three effects

The pure product-variety effect that is theeffect that arises because one or more prod-ucts are not available to consumers anymoreholding the prices of all other remainingproducts and the total expenditure on thequinolone subsegment XQ constant

The expenditure-switching effect that is theeffect arising from substitution away fromquinolones and toward other subsegments ofthe antibiotics market again holding the pricesof all other remaining products constant

The reduced-competition effect that is theeffect that arises because the firms remainingin the market adjust (increase) their prices inresponse to the removal of domestic products

From both an analytical and a policy point ofview it is desirable to assess how large each ofthese effects is Accordingly we decompose thetotal effect on consumer welfare (the CV as givenby equation (16)) using the following procedure

To get the pure product variety effect wecompute virtual prices for the products that areremoved from the market holding the quino-lone expenditure XQ and the prices of all otherproducts fixed Let us call the resulting pricevector P1 Then the pure product variety effectis represented by E(u0 P1) E(u0 P0)

To compute the expenditure-switching effectwe compute another set of virtual prices againholding the prices of all remaining products fixedbut letting quinolone expenditure adjust in re-sponse to the new price index for the quinolonesegment (Given that the prices of the remainingproducts remain fixed the change in the priceindex arises only because of the removal of one ormore domestic products) Note that this scenariomost closely resembles the case in which patentlaws are enforced but strict price regulation keepsthe prices of the products offered in the market attheir pre-TRIP level Let us label the so-computedprice vector P2 The expenditure-switching effectis then E(u0 P2) E(u0 P1)

Finally the reduced competition effect aris-ing from higher prices for the remaining prod-

ucts is computed as the residual change in thecompensating variation once the product-varietyand expenditure-switching effects have been ac-counted for that is E(u0 P) E(u0 P2)where the price vector P is computed accordingto the formulas (14) and (15) to reflect theadjustment of prices to the new regime

To compute the standard errors associatedwith the counterfactual simulations (that is thestandard errors for the counterfactual prices andwelfare estimates) we again use bootstrappingWe first bootstrap the original sample (sales andprices of all the drugs) 300 times Next we esti-mate the AIDS model for each of these samplesand compute the counterfactual equilibrium pricesand welfare losses corresponding to each policyscenario for each of the 300 simulations In thefinal step we compute standard errors29

One limitation of our framework is that itdoes not allow for a heterogeneous response ofconsumers to the policy change Accordinglyour framework is not suited to addressing thequestion of how different groups in the popula-tion will be affected by patent law enforce-ment30 Along the same lines our frameworkdoes not accommodate the possibility of pricediscrimination which might lead to differentresults regarding welfare losses and profit gainsrelative to uniform pricing31 Still we believethat the results of the counterfactual simulationscan provide important insights into the likelyaggregate response to patent enforcement andthe factors that drive this response

III Results

A The Structure of Demand

Tables A2 and A3 in the Appendix display theresults from estimation of the lower- and upper-level AIDS systems respectively For ease of in-terpretation rather than discussing the coefficientestimates we focus our discussion on the implied

29 In cases some of the simulations do not converge wecompute the standard errors using only those simulationsthat converged

30 Addressing this question would require at a mini-mum micro data on consumer purchases as in Goldberg(1995) To our knowledge such data do not exist for theIndian pharmaceutical market

31 See Ernst R Berndt (1994) for an extensive discussionof uniform pricing

1496 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

unconditional price and expenditure elasticities re-ported in Table 6 Given that the region-specificproduct group effects ir reported in the last fourcolumns of Tables A2 and A3 imply region-spe-cific demand elasticities we report separate elas-ticities for each region in Tables 6A and 6B Asevident from the comparison of these tables theelasticities are very similar across regions so thatwe can focus the remaining discussion on theelasticities of one region only the northern regiondisplayed in Table 6A

The diagonal terms of Table 6A report theown-price elasticities which are in all but onecase negative and highly significant The oneexception is the foreign norfloxacin productgroupmdashwhose share of quinolone sales is 007percentmdashfor which we estimate a negative butinsignificant own-price elasticity For the remain-ing product groups demand appears to be highlyelastic with the estimated elasticities lower than2 in four out of the six cases The magnitude ofthe own-price elasticities matches the features ofthe Indian pharmaceutical market mentioned ear-lier which would suggest that Indian consumersare likely to be quite price-sensitive32 The elas-ticities appear especially large if one takes intoaccount that they refer to product groups (such asdomestic ciprofloxacin) and not individual drugsoffered by particular firms Their relative magni-tudes are also intuitive the drug with the largestmarket share and a relatively high price (domesticciprofloxacin) appears to be one of the least elas-tic In contrast foreign ciprofloxacin is highlyprice elastic this is plausible as ciprofloxacindrugs offered by subsidiaries of multinationalsface the stiffest competition in this market seg-ment from approximately 75 Indian firms offeringthe same molecule

The estimated expenditure elasticities appear inthe last column These are all positive indicatingthat the demand for all product groups is normalThe remaining cells display the estimated cross-price elasticities As one might perhaps expect forproducts within a therapeutic subsegment theseare mostly positive Out of a total of 49 priceelasticities we estimate there are six that do notconform to expectations these are the cross-price elasticities between different foreign prod-

uct groups which are estimated to be negativeand significant33 Fortunately these elasticitieshave negligible impact on the welfare analysisgiven that our counterfactuals focus on theeffect of withdrawing one or more domesticproducts from the market the most relevantelasticities are the ones that capture the responseof various product group shares to a change in theprice of one or more domestic groups these elas-ticities are the cross-price elasticities between var-ious domestic groups and the ones betweendomestic and foreign groups which are plausibleand precisely estimated34

Regarding these elasticities a striking aspectof our estimates is how large positive andsignificant the cross-price elasticities betweendifferent domestic product groups aremdashin factfor norfloxacin and ofloxacin we estimate thatdomestic product groups containing differentmolecules are closer substitutes for one anotherthan product groups that contain the same mol-ecule but are produced by foreign firms Incontrast for ciprofloxacin (the molecule withthe largest revenue share) we estimate a largepositive cross-price elasticity between the do-mestic and foreign versions

The fact that domestic products appear to beclose substitutes for other domestic productsthat contain different molecules truly representsan ldquoempiricalrdquo finding in the sense that we do

32 In developed economies elasticities of this magnitude havetypically been found only for generic drugs (and even then onlyrarely) or among consumers who lack health insurance

33 While these elasticities are clearly counterintuitive theyare not inconsistent with the underlying demand system whichimposes no restrictions on the sign of the cross-price elastici-ties We do not have a good explanation of why these elastic-ities are estimated to be negative A potential explanation isthat the shares of the foreign products are very small giventhis we observe very few consumers switching from one (verysmall) foreign group to another (very small) foreign groupwhen the price of the first foreign group goes up and hence theinference is not very reliable in this case

34 The cross-price elasticities between foreign drug groupscontaining different molecules will also have an effect on thewelfare estimates given that the withdrawal of domestic prod-ucts will generally lead to changes in the prices of all foreignproducts and market shares will be reallocated from someforeign products to other foreign products based on the newprices However this reallocation from ldquoforeign to foreignrdquo istruly second order in our case compared to the reallocationfrom ldquodomestic to domesticrdquo and ldquodomestic to foreignrdquo prod-ucts In scenarios in which we consider patent enforcementaccompanied by strict price regulation the cross-price elastic-ities between foreign product groups are in fact completelyirrelevant as the prices of foreign products are not allowed toincrease in these cases Still our welfare loss estimates remainsubstantial drivenmdashas beforemdashby the loss of product variety

1497VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

not impose it through any of our assumptionsregarding the demand function The questionthat naturally arises then is what might explainthis finding While we cannot formally addressthis question anecdotal accounts in various in-dustry studies suggest that the explanation maylie in the differences between domestic andforeign firms in the structure and coverage ofretail distribution networks

Distribution networks for pharmaceuticals inIndia are typically organized in a hierarchicalfashion Pharmaceutical companies deal mainlywith carrying and forwarding (CampF) agents inmany instances regionally based who each sup-ply a network of stockists (wholesalers) Thesestockists in turn deal with the retail pharma-cists through whom retail sales ultimately oc-cur35 The market share enjoyed by a particularpharmaceutical product therefore depends inpart on the number of retail pharmacists who

stock the product And it is here that thereappears to be a distinction between domesticfirms and multinational subsidiaries In particu-lar the retail reach of domestic firms as agroup tends to be much more comprehensivethan that of multinational subsidiaries (IndianCredit Rating Agency (ICRA) 1999)36

There appear to be two reasons for this Thefirst is that many of the larger Indian firmsbecause they have a much larger portfolio ofproducts over which to spread the associatedfixed costs typically have more extensive net-works of medical representatives The second issimply that there are many more domestic firms(and products) on the market At the retail levelthis would imply that local pharmacists mightbe more likely to stock domestic products con-taining two different molecules say ciprofloxa-cin and norfloxacin than they would domesticand foreign versions of the same molecule Tothe extent that patients (or their doctors) arewilling to substitute across molecules in order tosave on transport or search costs (eg going toanother pharmacy to check whether a particular

35 There are estimated to be some 300000 retail pharma-cists in India On average stockists deal with about 75 retailers(ICRA 1999) There are naturally variations in this structureand a host of specific exclusive dealing and other arrangementsexists in practice Pharmaceutical firms also maintain networksof medical representatives whose main function is to marketthe companyrsquos products to doctors who do the actual prescrib-ing of drugs In some instances firms do sell directly to thedoctors who then become the ldquoretailerrdquo as far as patients areconcerned but these are relatively rare

36 These differences were also highlighted in conversa-tions one of the authors had with CEOs and managingdirectors of several pharmaceutical firms as part of a sep-arate study

TABLE 6AmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN THE NORTHERN REGION

Product group

Elasticity with respect to

Prices of foreign productgroups Prices of domestic product groups Overall

quinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 557 013dagger 015 401 011dagger 011dagger 016 137(179) (007) (007) (184) (006) (006) (006) (029)

Foreign norfloxacin 427dagger 045 427dagger 350dagger 602 451 465 220(242) (112) (242) (210) (623) (184) (183) (105)

Foreign ofloxacin 011 010dagger 138 009 009dagger 023 011 116(005) (005) (031) (027) (005) (028) (004) (017)

Domestic ciprofloxacin 018 001 001 168 008 008 010 117(008) (000) (001) (023) (002) (002) (002) (003)

Domestic norfloxacin 004 003 004 058 223 042 040 073(001) (003) (001) (017) (011) (004) (003) (009)

Domestic ofloxacin 005 005 011 077 074 342 074 089(002) (002) (013) (028) (008) (025) (008) (021)

Domestic sparfloxacin 007 004 007 115 063 063 288 028(002) (001) (002) (015) (006) (006) (017) (012)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent significance level and dagger (dagger) denotes significance at the 10-percent level

1498 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE 6BmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN OTHER REGIONS

Demand patterns in the eastern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 594 014dagger 016 431 013 011 017 140(195) (008) (008) (201) (006) (007) (007) (031)

Foreign norfloxacin 329dagger 058 329dagger 264dagger 460 346 359 192(180) (083) (179) (153) (451) (138) (139) (082)

Foreign ofloxacin 012 010dagger 140 010 010 024 013 117(006) (006) (034) (031) (005) (030) (005) (019)

Domestic ciprofloxacin 018 001 001 172 009 008 011 117(008) (000) (002) (027) (002) (003) (002) (003)

Domestic norfloxacin 004 004 004 068 242 049 046 070(002) (004) (002) (020) (013) (005) (004) (010)

Domestic ofloxacin 004dagger 004 009 061 060 295 060 092(002) (001) (010) (028) (007) (022) (007) (017)

Domestic sparfloxacin 005 003 005 092 048 051 251 043(002) (001) (001) (017) (005) (006) (016) (010)

Demand patterns in the western region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 427 009dagger 011 286 008 006 011 126(129) (005) (005) (133) (004) (005) (004) (021)

Foreign norfloxacin 714dagger 008 710dagger 594dagger 995 742 770 299dagger

(369) (183) (368) (328) (1015) (297) (296) (178)Foreign ofloxacin 013 012dagger 145 009 012 026 013 119

(007) (006) (037) (029) (005) (033) (005) (021)Domestic ciprofloxacin 019 001 000 174 009 008 011 118

(008) (000) (001) (024) (002) (003) (002) (004)Domestic norfloxacin 004 004 004 067 243 050 047 069

(002) (004) (001) (018) (014) (005) (004) (010)Domestic ofloxacin 003 003 007 048 048 257 048 093

(002) (001) (008) (022) (005) (015) (004) (014)Domestic sparfloxacin 006 003 005 083 046 049 241 046

(002) (001) (001) (014) (004) (005) (012) (009)

Demand patterns in the southern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 560 013dagger 015 402 011 010 017 137(182) (007) (007) (187) (006) (006) (007) (029)

Foreign norfloxacin 533dagger 031 531dagger 431dagger 748 559 584 249dagger

(301) (134) (300) (245) (734) (219) (219) (128)Foreign ofloxacin 014 012dagger 148 011 012 029 016 120

(007) (007) (039) (033) (006) (035) (006) (022)Domestic ciprofloxacin 017 001 000 169 008 007 011 116

(008) (000) (001) (025) (002) (003) (002) (003)Domestic norfloxacin 004 003 003 059 225 043 040 073

(001) (003) (001) (019) (012) (004) (004) (009)Domestic ofloxacin 004dagger 004 009 062 060 296 060 091

(002) (001) (010) (026) (006) (019) (007) (017)Domestic sparfloxacin 008 005 008 138 073 075 323 016

(002) (001) (002) (018) (007) (008) (021) (014)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at 5-percent level anddagger (dagger) denotes significance at the 10-percent level

1499VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

foreign product is in stock) in aggregate datawe would expect to find precisely the substitu-tion patterns that we report in Table 6

Whether the particular explanation we provideabove is the correct one the high degree of sub-stitutability between domestic product groupsturns out to have important implications for thewelfare calculations We discuss these in moredetail below when we present the results of thecounterfactual welfare analysis Another elasticitywith important implications for the counterfactu-als is the price elasticity for the quinolone subseg-ment as a whole which indicates how likelyconsumers are to switch to other antibioticsgroups when faced with a price increase for quin-olones This elasticity is computed on the basis ofthe results in Table A3 and it is at 111 (stan-dard error 024) this is large in magnitudebutmdashas expectedmdashsmaller in absolute value thanthe own-price elasticities of the product groupswithin the quinolone subsegment

The results in Tables 6A and 6B are based onour preferred specification discussed in SectionII In Tables A4 to A6 in the Appendix weexperiment with some alternative specificationsTables A4(a)ndashA4(c) correspond to a specificationthat includes in addition to product-group-specificregional fixed effects product-group-specific (andfor the upper level antibiotics-segment-specific)seasonal effects We distinguish among three sea-sonsmdashthe summer monsoon and wintermdashandreport the unconditional demand elasticities for

the northern region for each of these seasons Asevident from the tables in the Appendix our elas-ticity estimates are robust to the inclusion of sea-sonal effects The demand elasticities in Table A5are based on estimation of the demand system byOLS Compared to the elasticities obtained by IVthe OLS elasticities are smaller in absolute valueimplying that welfare calculations based on theOLS estimates would produce larger welfare lossestimates Nevertheless some of the patterns re-garding the cross-price elasticities discussed ear-lier are also evident in the OLS results inparticular the cross-price elasticities between dif-ferent domestic product groups are all positivelarge and significant and in most instances largerthan the cross-price elasticities between drugs thatcontain the same molecule but are produced byfirms of different domesticforeign status Theclose substitutability of domestic products indi-cated by both the OLS and IV estimates seems tobe one of the most robust findings of the paper

B Cost and Markup Estimates

Table 7 displays the marginal costs markupsand profits implied by the price elasticity esti-mates of Tables 6A and 6B for each of the sevenproduct groups Given that our regional effectsimply different price elasticities for each regionour marginal cost and markup estimates alsodiffer by region Given however that based on

TABLE 7mdashUPPER AND LOWER BOUNDS FOR MARGINAL COST MARKUP AND ANNUAL PROFIT BY PRODUCT GROUPS WITHIN

THE QUINOLONE SUBSEGMENT

Product groupLower boundfor MC (Rs)

Upper boundfor markup

Upper boundfor profit(Rs mill)

Upper boundfor MC (Rs)

Lower boundfor markup

Lower boundfor profit

(Rs)

Foreign ciprofloxacin 83 19 269 103 0 00(123) (012) (1655)

Foreign norfloxacin NA NA NA 53 0 00Foreign ofloxacin 323 70 1061 1085 0 00

(2316) (021) (3185)Domestic ciprofloxacin 47 59 17019 112 0 00

(114) (010) (29858)Domestic norfloxacin 52 43 2807 90 0 00

(020) (002) (1532)Domestic ofloxacin 587 34 1612 901 0 00

(218) (002) (1280)Domestic sparfloxacin 495 37 1985 788 0 00

(157) (002) (1100)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level Estimated lower bound forforeign norfloxacinrsquos marginal cost is negative since the estimated price elasticity is less than one in absolute value

1500 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Tables 6A and 6B the price elasticities do notseem to differ substantially across regions wereport for ease of exposition only the nationalaverages for marginal costs and markups inTable 7

Table 7 has two parts In the left part (firstthree columns) we report the lower bound formarginal cost and the corresponding upperbound for markup and upper bound for totalannual profit for each product group Thesenumbers are based on the price elasticities weobtained from estimating the two-level AIDSdemand system Since we do not have a reliableestimate for the price elasticity of foreign nor-floxacin (the point estimate is negative but lessthan one in absolute value and insignificant)we cannot compute the lower bound for mar-ginal cost in this case It is important to note thatthese estimates do not reflect either the actualmarginal cost or the actual markup for thesedrugs both because the existence of price reg-ulation implies that the unconstrained first-orderconditions are unlikely to hold each period andbecause our aggregation across firms of thesame domesticforeign status supplying thesame molecule makes the interpretation of theseestimates problematic In particular the fact thatwe ignore competition among firms within eachproduct group implies that our estimates willtend to overstate market power These numberswill prove useful however in the counterfac-tual simulations as they can give us a sense ofhow large the maximum profit gains for multi-national firms and the maximum profit lossesfor domestic firms are likely to be under patentenforcement The right part of the table (lastthree columns) reports the upper bound for mar-ginal cost which we obtain by simply taking themarginal cost to be equal to the observed priceWe do not report standard errors in this casesince the numbers are based on actual data Themarkups corresponding to these marginal costupper bounds are of course zero We conductthe counterfactual simulations using both thelower and upper bounds for marginal costs

The (upper bounds for) markups on the leftside of the table are generally plausible Thedomestic ciprofloxacin product group that dom-inates the quinolone subsegment and for whichwe documented high prices a high marketshare and a relatively low elasticity of demandenjoys one of the highest markup upper bounds(60 percent) and accounts for nearly 70 percent

of all profits derived within the subsegmentForeign ciprofloxacin on the other hand whichfaces the stiffest competition from domesticfirms and for which we estimated a highly elas-tic demand has the lowest markup upper bound(19 percent)

C Counterfactual Estimates of the Impact onPrices and Welfare

With estimates of the key demand and costparameters in hand we turn to the counterfac-tuals We consider the five separate scenarioslisted in the previous section All of the scenar-ios involve the withdrawal of one or more of thedomestic product groups from the market Ta-ble 8 displays our estimates of the consumerwelfare losses that result under the differentscenarios The losses are expressed in billion ofrupees per year All numbers presented in Ta-ble 8 and subsequent tables are based on usingthe lower bounds for marginal cost and upperbounds for markup in the simulations As dis-cussed earlier these numbers are the more in-teresting to work with since they give us anupper bound for the changes in the profits ofdomestic and foreign firms that would resultfrom patent enforcement In the Appendix wealso present results based on using the upperbounds for marginal costs in the simulations inwhich case the pre-TRIPS profits of domesticand foreign firms are zero In all cases marginalcosts are assumed to be constant in outputWhile naturally the profit implications differdepending on whether one uses the upper orlower bounds for marginal costs (firm profits arezero if one assumes the upper bound of mar-ginal cost in which case price equals marginalcost) the estimated consumer welfare losses aresimilar in the two cases We discuss these re-sults in more detail at the end of this section

The first column presents our estimates of theconsumer welfare losses attributable to the pureloss of product variety effect where we fix theprices of all remaining products as well as theoverall expenditure on quinolones while with-drawing one or more of the domestic productgroups Note that in our initial specification ofthe demand system had we not allowed for thepossibility that consumers might differentiatebetween domestic and foreign products evenwhen they contain the same molecule this

1501VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

particular component of the loss of consumerwelfare would not have arisen

The estimates reported in the second columnincorporate the expenditure-switching effect ontop of the loss of product variety Here basedupon the price elasticity estimates from thehigher-level AIDS system we adjust (down-ward) the expenditures allocated to the quino-lone subsegment as the composite price ofquinolones effectively increases as a consequenceof the higher virtual prices of the domestic productgroups that are withdrawn from the market Be-cause the estimates in this column are generatedassuming that the prices of the products that re-main in the market are not adjusted upward theyprovide a sense of what consumer losses would beif the introduction of product patents were coupledwith strict price regulation aimed at maintainingprices at prepatent levels Alternatively they canbe thought of as the relevant welfare numbers ifintense competition among firms within the re-maining product groups kept the prices of theproducts that were still offered in the market closeto the firmsrsquo marginal costs

The last column of Table 8 displays the esti-mated consumer welfare losses when bothcross-segment expenditure-switching and with-in-segment upward price adjustments are takeninto account

If we compare the results across the firstsecond and third columns all the counterfac-tual scenarios produce qualitatively similar pat-terns patterns that are consistent with what we

would expect Starting from the initial loss ofwelfare attributable to the loss of product vari-ety the option of switching expenditures outof the quinolone subsegment to other subseg-ments mitigates some of the initial welfareloss But if we then incorporate the upwardprice adjustments that result in response to thereduced competition the welfare losses aremagnified

Of particular interest from a policy perspec-tive are the relative magnitudes of these threeeffects which are similar under all the counter-factual scenarios though absolute levels varyconsiderably First despite the fact that the de-mand for quinolones is quite sensitive to thecomposite price of quinolonesmdashthe upper levelprice elasticity is 111mdashthe cross-subsegmentexpenditure-switching effects are in all thecases small (in absolute value terms) relative tothe other two effects For instance under thescenario where all the domestic quinolone prod-uct groups are withdrawn from the market theoverall consumer welfare loss of Rs 1781 bil-lion per year can be decomposed into an initialloss of Rs 1176 billion (66 percent) attribut-able to the loss of product variety a slightreduction in this initial loss of Rs 041 billion(2 percent) from Rs 1176 billion to Rs1135 billion because of expenditure switchingand a subsequent additional loss of Rs 646billion (36 percent) from Rs 1135 billion toRs 1781 billion because of the reduced com-petition and consequent price increases

TABLE 8mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure loss ofvariety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 732(087) (089) (146)

Only ofloxacin 008 008 023(008) (008) (010)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1253(177) (180) (415)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1058(142) (145) (331)

All four domestic quinolones products 1176dagger 1135dagger 1781(643) (634) (1270)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent significance level and dagger (dagger)denotes significance at the 10-percent level

1502 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 18: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

will be removed from the market In particularwe consider the following five scenarios26

Withdrawal of one large domestic productgroup only domestic ciprofloxacin

Withdrawal of one relatively small domesticproduct group only domestic ofloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand norfloxacin

Withdrawal of three domestic productgroups domestic ciprofloxacin ofloxacinand sparfloxacin

Withdrawal of all four domestic quinoloneproduct groups

As the list above suggests we proceed fromanalyzing the effects of single-product with-drawal to the analysis of eliminating the en-tire domestic segment This approach wasmotivated by early empirical results that in-dicated that the existence and extent of com-petition from domestic firms has a significantbearing on the predicted effects of patent en-forcement that is our predictions regardingprices and welfare vary substantially depend-ing on how many domestic products are af-fected by patent enforcement and on whethersome domestic competition will remain presentafter TRIPS In addition the size of the affecteddomestic groups is relevant for the welfare pre-dictions accordingly we examine both scenar-ios in which a large domestic product groupsuch as domestic ciprofloxacin is eliminatedfrom the market and scenarios in which the

eliminated domestic group is relatively small(eg domestic ofloxacin) In all scenarios wemaintain the assumption that the set of prod-ucts offered by the remaining firms in themarket does not change in response to patentenforcement

E Computation of Virtual Prices and NewEquilibrium Prices

The first step in the counterfactual analysisis to derive the new equilibrium prices underpatent enforcement27 In this context thereare two sets of prices that are relevant Thefirst set consists of the virtual prices of those(domestic) products that will not be availableonce TRIPS is put in effect To calculate thesevirtual prices we set the expenditures of therelevant products equal to zero The secondset of prices consists of the prices of thoseproducts that remain in the market In deriv-ing these prices we start by assuming profitmaximization without any form of price reg-ulation the firms remaining in the marketreoptimize in response to the policy changeand set new prices taking the prices of allother firms as given28 Of course at the equi-librium all prices change in response to thefact that some domestic products are nolonger present The new equilibrium pricesfor products that remain in the market are thuscomputed by utilizing the first-order condi-tions of profit-maximizing firms into whichthe virtual prices of the eliminated productsare substituted Hence to compute the newequilibrium prices we solve an equation sys-tem of the following form

26 As Table 2 indicates most of the patents for the drugsin the quinolone segment have expired by now so that noneof the scenarios described in this section is going to mate-rialize in practice The reason we consider these alternativescenarios in our counterfactual simulations is to get a senseof how the presence and extent of domestic competitionaffects welfare calculations Such calculations may becomerelevant in the future in other therapeutic classes in whichpatents have not expired yet and where domestic firmscompete with foreign patent holders While the particularexit scenarios will depend on the exact year of patent expiryof drugs in each therapeutic class and the specific way inwhich patents will be enforced (eg whether certain do-mestic drugs will be grandfathered in) the general conclu-sion that emerges from our calculations is that the consumerwelfare loss is substantially smaller in cases where somedomestic competition remains present in the market Hencethe particular way in which patent enforcement is imple-mented is essential for assessing its welfare impact

27 Of course until product patents are in fact introducedthese prices will not be observable Note also that we areassuming here that the range of available products will notchange with the introduction of patents

28 As mentioned above this first set of calculations ab-stracts from the existence of remaining price controls orother government regulations that would impose constraintson the firmsrsquo profit maximization problem Accordingly theresulting numbers should be interpreted as a benchmark ofwhat would happen if markets were completely unregulatedTo examine what the welfare effects of TRIPS would be inthe more realistic scenario of price regulation we subse-quently consider a scenario in which regulation keeps theprices of the products provided by patent-holders at theirpre-TRIPS level For a more detailed discussion see theresults section

1494 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

For products i that are withdrawn from themarket

(14) 0 i ir iiln pirV i10ln pjr j Di

10

j Di

01

i01ln pjr

j Di

00

i00ln pjr

ilnXQr

PQr

For products k that remain in the market

(15) pkr ckr 1 1

kk pkrpjr pirV 1

In the equations above pirV denotes the virtual

prices of the products that are removed fromthe market while pjr denotes the updatedprices of all other products Note that whensolving for the virtual prices we account forthe fact that both the price index for quino-lones PQr and the expenditure allocated tothis subsegment XQr need to be updated toreflect the fact that as a result of the pricechanges there may be substitution away fromthis subsegment To obtain the new quinoloneexpenditure XQr and the new price index PQrwe use the estimates and formulas for thehigher-level AIDS system In equation (15)ck refers to the marginal cost for product k thatwe have obtained from the previous estimationstage As mentioned in the previous subsectionwe conducted the simulations using both theupper and lower bounds for marginal cost ck

u

and ckL respectively The term kk(pkr

pjrpirV) re-

fers to the unconditional own-price elasticity forproduct k which is a function of the eliminatedproductsrsquo virtual prices and the remaining prod-uctsrsquo new equilibrium prices We conduct thecounterfactual simulations and welfare analysisat the regional level because the presence ofregion-specific product effects in the demandestimation implies region-specific demand elas-ticities and hence region-specific marginalcosts and markups The presentation and discus-sion of our results focus on national averages ofthe relevant variables that we construct by com-

puting weighted averages across regions usingthe population of each region as a weight

F Welfare Assessment

The simulation of the new equilibrium underpatent protection can provide important insightsinto how consumers and firms will respond tothe removal of domestic products in the market(for example toward which products consumerswill substitute or which prices will increase themost) To get a more precise idea of how peo-plersquos well-being will ultimately be affected byTRIPS we compute as a last step in our anal-ysis the welfare effects of the policy changeSocial welfare is defined as consumer welfareand the sum of domestic firm profits Thechange in domestic profits can easily be calcu-lated by comparing the domestic firm (variable)profits at the pre-TRIPS prices to the profitsthese firms will realize at the new simulatedprices Although foreign firm profits do notcount in domestic welfare calculations we alsocompute the effects of patent enforcement onforeign firm profits to get an idea of how largethe expected benefits of TRIPS are for thesefirms This provides in some sense an indirectway of assessing whether the claims that patentenforcement in countries like India will lead tomore research on developing-country-specificdiseases (such as malaria) have any validity iffor example we find that the effect of patentenforcement on the foreign firm profits realizedin India is small in magnitude it is unlikely thatforeign firms will engage in more developing-country-specific research in response to TRIPSIt is important to note that in all these calcula-tions we work with the lower-bound estimatesfor marginal costs since these give rise to thehighest possible markups Hence our estimateof profit loss for domestic producers most likelyoverstates this loss

On the consumer side we measure changes inconsumer welfare by the compensating variation(CV) defined as the additional expenditure thatconsumers need in order to achieve the sameutility level as before patent enforcement at thenew prices Specifically let P0 denote the pricevector before patent enforcement P the simulatedprice vector post-TRIPS (that we obtained usingthe methods described in the previous subsection)u0 the utility attained by consumers before TRIPS

1495VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

and E(u P) the higher-level expenditure functionThen the compensating variation is given by

(16) CV Eu0 P Eu0 P0

Note that the CV as computed in (16) representsthe combination of three effects

The pure product-variety effect that is theeffect that arises because one or more prod-ucts are not available to consumers anymoreholding the prices of all other remainingproducts and the total expenditure on thequinolone subsegment XQ constant

The expenditure-switching effect that is theeffect arising from substitution away fromquinolones and toward other subsegments ofthe antibiotics market again holding the pricesof all other remaining products constant

The reduced-competition effect that is theeffect that arises because the firms remainingin the market adjust (increase) their prices inresponse to the removal of domestic products

From both an analytical and a policy point ofview it is desirable to assess how large each ofthese effects is Accordingly we decompose thetotal effect on consumer welfare (the CV as givenby equation (16)) using the following procedure

To get the pure product variety effect wecompute virtual prices for the products that areremoved from the market holding the quino-lone expenditure XQ and the prices of all otherproducts fixed Let us call the resulting pricevector P1 Then the pure product variety effectis represented by E(u0 P1) E(u0 P0)

To compute the expenditure-switching effectwe compute another set of virtual prices againholding the prices of all remaining products fixedbut letting quinolone expenditure adjust in re-sponse to the new price index for the quinolonesegment (Given that the prices of the remainingproducts remain fixed the change in the priceindex arises only because of the removal of one ormore domestic products) Note that this scenariomost closely resembles the case in which patentlaws are enforced but strict price regulation keepsthe prices of the products offered in the market attheir pre-TRIP level Let us label the so-computedprice vector P2 The expenditure-switching effectis then E(u0 P2) E(u0 P1)

Finally the reduced competition effect aris-ing from higher prices for the remaining prod-

ucts is computed as the residual change in thecompensating variation once the product-varietyand expenditure-switching effects have been ac-counted for that is E(u0 P) E(u0 P2)where the price vector P is computed accordingto the formulas (14) and (15) to reflect theadjustment of prices to the new regime

To compute the standard errors associatedwith the counterfactual simulations (that is thestandard errors for the counterfactual prices andwelfare estimates) we again use bootstrappingWe first bootstrap the original sample (sales andprices of all the drugs) 300 times Next we esti-mate the AIDS model for each of these samplesand compute the counterfactual equilibrium pricesand welfare losses corresponding to each policyscenario for each of the 300 simulations In thefinal step we compute standard errors29

One limitation of our framework is that itdoes not allow for a heterogeneous response ofconsumers to the policy change Accordinglyour framework is not suited to addressing thequestion of how different groups in the popula-tion will be affected by patent law enforce-ment30 Along the same lines our frameworkdoes not accommodate the possibility of pricediscrimination which might lead to differentresults regarding welfare losses and profit gainsrelative to uniform pricing31 Still we believethat the results of the counterfactual simulationscan provide important insights into the likelyaggregate response to patent enforcement andthe factors that drive this response

III Results

A The Structure of Demand

Tables A2 and A3 in the Appendix display theresults from estimation of the lower- and upper-level AIDS systems respectively For ease of in-terpretation rather than discussing the coefficientestimates we focus our discussion on the implied

29 In cases some of the simulations do not converge wecompute the standard errors using only those simulationsthat converged

30 Addressing this question would require at a mini-mum micro data on consumer purchases as in Goldberg(1995) To our knowledge such data do not exist for theIndian pharmaceutical market

31 See Ernst R Berndt (1994) for an extensive discussionof uniform pricing

1496 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

unconditional price and expenditure elasticities re-ported in Table 6 Given that the region-specificproduct group effects ir reported in the last fourcolumns of Tables A2 and A3 imply region-spe-cific demand elasticities we report separate elas-ticities for each region in Tables 6A and 6B Asevident from the comparison of these tables theelasticities are very similar across regions so thatwe can focus the remaining discussion on theelasticities of one region only the northern regiondisplayed in Table 6A

The diagonal terms of Table 6A report theown-price elasticities which are in all but onecase negative and highly significant The oneexception is the foreign norfloxacin productgroupmdashwhose share of quinolone sales is 007percentmdashfor which we estimate a negative butinsignificant own-price elasticity For the remain-ing product groups demand appears to be highlyelastic with the estimated elasticities lower than2 in four out of the six cases The magnitude ofthe own-price elasticities matches the features ofthe Indian pharmaceutical market mentioned ear-lier which would suggest that Indian consumersare likely to be quite price-sensitive32 The elas-ticities appear especially large if one takes intoaccount that they refer to product groups (such asdomestic ciprofloxacin) and not individual drugsoffered by particular firms Their relative magni-tudes are also intuitive the drug with the largestmarket share and a relatively high price (domesticciprofloxacin) appears to be one of the least elas-tic In contrast foreign ciprofloxacin is highlyprice elastic this is plausible as ciprofloxacindrugs offered by subsidiaries of multinationalsface the stiffest competition in this market seg-ment from approximately 75 Indian firms offeringthe same molecule

The estimated expenditure elasticities appear inthe last column These are all positive indicatingthat the demand for all product groups is normalThe remaining cells display the estimated cross-price elasticities As one might perhaps expect forproducts within a therapeutic subsegment theseare mostly positive Out of a total of 49 priceelasticities we estimate there are six that do notconform to expectations these are the cross-price elasticities between different foreign prod-

uct groups which are estimated to be negativeand significant33 Fortunately these elasticitieshave negligible impact on the welfare analysisgiven that our counterfactuals focus on theeffect of withdrawing one or more domesticproducts from the market the most relevantelasticities are the ones that capture the responseof various product group shares to a change in theprice of one or more domestic groups these elas-ticities are the cross-price elasticities between var-ious domestic groups and the ones betweendomestic and foreign groups which are plausibleand precisely estimated34

Regarding these elasticities a striking aspectof our estimates is how large positive andsignificant the cross-price elasticities betweendifferent domestic product groups aremdashin factfor norfloxacin and ofloxacin we estimate thatdomestic product groups containing differentmolecules are closer substitutes for one anotherthan product groups that contain the same mol-ecule but are produced by foreign firms Incontrast for ciprofloxacin (the molecule withthe largest revenue share) we estimate a largepositive cross-price elasticity between the do-mestic and foreign versions

The fact that domestic products appear to beclose substitutes for other domestic productsthat contain different molecules truly representsan ldquoempiricalrdquo finding in the sense that we do

32 In developed economies elasticities of this magnitude havetypically been found only for generic drugs (and even then onlyrarely) or among consumers who lack health insurance

33 While these elasticities are clearly counterintuitive theyare not inconsistent with the underlying demand system whichimposes no restrictions on the sign of the cross-price elastici-ties We do not have a good explanation of why these elastic-ities are estimated to be negative A potential explanation isthat the shares of the foreign products are very small giventhis we observe very few consumers switching from one (verysmall) foreign group to another (very small) foreign groupwhen the price of the first foreign group goes up and hence theinference is not very reliable in this case

34 The cross-price elasticities between foreign drug groupscontaining different molecules will also have an effect on thewelfare estimates given that the withdrawal of domestic prod-ucts will generally lead to changes in the prices of all foreignproducts and market shares will be reallocated from someforeign products to other foreign products based on the newprices However this reallocation from ldquoforeign to foreignrdquo istruly second order in our case compared to the reallocationfrom ldquodomestic to domesticrdquo and ldquodomestic to foreignrdquo prod-ucts In scenarios in which we consider patent enforcementaccompanied by strict price regulation the cross-price elastic-ities between foreign product groups are in fact completelyirrelevant as the prices of foreign products are not allowed toincrease in these cases Still our welfare loss estimates remainsubstantial drivenmdashas beforemdashby the loss of product variety

1497VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

not impose it through any of our assumptionsregarding the demand function The questionthat naturally arises then is what might explainthis finding While we cannot formally addressthis question anecdotal accounts in various in-dustry studies suggest that the explanation maylie in the differences between domestic andforeign firms in the structure and coverage ofretail distribution networks

Distribution networks for pharmaceuticals inIndia are typically organized in a hierarchicalfashion Pharmaceutical companies deal mainlywith carrying and forwarding (CampF) agents inmany instances regionally based who each sup-ply a network of stockists (wholesalers) Thesestockists in turn deal with the retail pharma-cists through whom retail sales ultimately oc-cur35 The market share enjoyed by a particularpharmaceutical product therefore depends inpart on the number of retail pharmacists who

stock the product And it is here that thereappears to be a distinction between domesticfirms and multinational subsidiaries In particu-lar the retail reach of domestic firms as agroup tends to be much more comprehensivethan that of multinational subsidiaries (IndianCredit Rating Agency (ICRA) 1999)36

There appear to be two reasons for this Thefirst is that many of the larger Indian firmsbecause they have a much larger portfolio ofproducts over which to spread the associatedfixed costs typically have more extensive net-works of medical representatives The second issimply that there are many more domestic firms(and products) on the market At the retail levelthis would imply that local pharmacists mightbe more likely to stock domestic products con-taining two different molecules say ciprofloxa-cin and norfloxacin than they would domesticand foreign versions of the same molecule Tothe extent that patients (or their doctors) arewilling to substitute across molecules in order tosave on transport or search costs (eg going toanother pharmacy to check whether a particular

35 There are estimated to be some 300000 retail pharma-cists in India On average stockists deal with about 75 retailers(ICRA 1999) There are naturally variations in this structureand a host of specific exclusive dealing and other arrangementsexists in practice Pharmaceutical firms also maintain networksof medical representatives whose main function is to marketthe companyrsquos products to doctors who do the actual prescrib-ing of drugs In some instances firms do sell directly to thedoctors who then become the ldquoretailerrdquo as far as patients areconcerned but these are relatively rare

36 These differences were also highlighted in conversa-tions one of the authors had with CEOs and managingdirectors of several pharmaceutical firms as part of a sep-arate study

TABLE 6AmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN THE NORTHERN REGION

Product group

Elasticity with respect to

Prices of foreign productgroups Prices of domestic product groups Overall

quinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 557 013dagger 015 401 011dagger 011dagger 016 137(179) (007) (007) (184) (006) (006) (006) (029)

Foreign norfloxacin 427dagger 045 427dagger 350dagger 602 451 465 220(242) (112) (242) (210) (623) (184) (183) (105)

Foreign ofloxacin 011 010dagger 138 009 009dagger 023 011 116(005) (005) (031) (027) (005) (028) (004) (017)

Domestic ciprofloxacin 018 001 001 168 008 008 010 117(008) (000) (001) (023) (002) (002) (002) (003)

Domestic norfloxacin 004 003 004 058 223 042 040 073(001) (003) (001) (017) (011) (004) (003) (009)

Domestic ofloxacin 005 005 011 077 074 342 074 089(002) (002) (013) (028) (008) (025) (008) (021)

Domestic sparfloxacin 007 004 007 115 063 063 288 028(002) (001) (002) (015) (006) (006) (017) (012)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent significance level and dagger (dagger) denotes significance at the 10-percent level

1498 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE 6BmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN OTHER REGIONS

Demand patterns in the eastern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 594 014dagger 016 431 013 011 017 140(195) (008) (008) (201) (006) (007) (007) (031)

Foreign norfloxacin 329dagger 058 329dagger 264dagger 460 346 359 192(180) (083) (179) (153) (451) (138) (139) (082)

Foreign ofloxacin 012 010dagger 140 010 010 024 013 117(006) (006) (034) (031) (005) (030) (005) (019)

Domestic ciprofloxacin 018 001 001 172 009 008 011 117(008) (000) (002) (027) (002) (003) (002) (003)

Domestic norfloxacin 004 004 004 068 242 049 046 070(002) (004) (002) (020) (013) (005) (004) (010)

Domestic ofloxacin 004dagger 004 009 061 060 295 060 092(002) (001) (010) (028) (007) (022) (007) (017)

Domestic sparfloxacin 005 003 005 092 048 051 251 043(002) (001) (001) (017) (005) (006) (016) (010)

Demand patterns in the western region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 427 009dagger 011 286 008 006 011 126(129) (005) (005) (133) (004) (005) (004) (021)

Foreign norfloxacin 714dagger 008 710dagger 594dagger 995 742 770 299dagger

(369) (183) (368) (328) (1015) (297) (296) (178)Foreign ofloxacin 013 012dagger 145 009 012 026 013 119

(007) (006) (037) (029) (005) (033) (005) (021)Domestic ciprofloxacin 019 001 000 174 009 008 011 118

(008) (000) (001) (024) (002) (003) (002) (004)Domestic norfloxacin 004 004 004 067 243 050 047 069

(002) (004) (001) (018) (014) (005) (004) (010)Domestic ofloxacin 003 003 007 048 048 257 048 093

(002) (001) (008) (022) (005) (015) (004) (014)Domestic sparfloxacin 006 003 005 083 046 049 241 046

(002) (001) (001) (014) (004) (005) (012) (009)

Demand patterns in the southern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 560 013dagger 015 402 011 010 017 137(182) (007) (007) (187) (006) (006) (007) (029)

Foreign norfloxacin 533dagger 031 531dagger 431dagger 748 559 584 249dagger

(301) (134) (300) (245) (734) (219) (219) (128)Foreign ofloxacin 014 012dagger 148 011 012 029 016 120

(007) (007) (039) (033) (006) (035) (006) (022)Domestic ciprofloxacin 017 001 000 169 008 007 011 116

(008) (000) (001) (025) (002) (003) (002) (003)Domestic norfloxacin 004 003 003 059 225 043 040 073

(001) (003) (001) (019) (012) (004) (004) (009)Domestic ofloxacin 004dagger 004 009 062 060 296 060 091

(002) (001) (010) (026) (006) (019) (007) (017)Domestic sparfloxacin 008 005 008 138 073 075 323 016

(002) (001) (002) (018) (007) (008) (021) (014)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at 5-percent level anddagger (dagger) denotes significance at the 10-percent level

1499VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

foreign product is in stock) in aggregate datawe would expect to find precisely the substitu-tion patterns that we report in Table 6

Whether the particular explanation we provideabove is the correct one the high degree of sub-stitutability between domestic product groupsturns out to have important implications for thewelfare calculations We discuss these in moredetail below when we present the results of thecounterfactual welfare analysis Another elasticitywith important implications for the counterfactu-als is the price elasticity for the quinolone subseg-ment as a whole which indicates how likelyconsumers are to switch to other antibioticsgroups when faced with a price increase for quin-olones This elasticity is computed on the basis ofthe results in Table A3 and it is at 111 (stan-dard error 024) this is large in magnitudebutmdashas expectedmdashsmaller in absolute value thanthe own-price elasticities of the product groupswithin the quinolone subsegment

The results in Tables 6A and 6B are based onour preferred specification discussed in SectionII In Tables A4 to A6 in the Appendix weexperiment with some alternative specificationsTables A4(a)ndashA4(c) correspond to a specificationthat includes in addition to product-group-specificregional fixed effects product-group-specific (andfor the upper level antibiotics-segment-specific)seasonal effects We distinguish among three sea-sonsmdashthe summer monsoon and wintermdashandreport the unconditional demand elasticities for

the northern region for each of these seasons Asevident from the tables in the Appendix our elas-ticity estimates are robust to the inclusion of sea-sonal effects The demand elasticities in Table A5are based on estimation of the demand system byOLS Compared to the elasticities obtained by IVthe OLS elasticities are smaller in absolute valueimplying that welfare calculations based on theOLS estimates would produce larger welfare lossestimates Nevertheless some of the patterns re-garding the cross-price elasticities discussed ear-lier are also evident in the OLS results inparticular the cross-price elasticities between dif-ferent domestic product groups are all positivelarge and significant and in most instances largerthan the cross-price elasticities between drugs thatcontain the same molecule but are produced byfirms of different domesticforeign status Theclose substitutability of domestic products indi-cated by both the OLS and IV estimates seems tobe one of the most robust findings of the paper

B Cost and Markup Estimates

Table 7 displays the marginal costs markupsand profits implied by the price elasticity esti-mates of Tables 6A and 6B for each of the sevenproduct groups Given that our regional effectsimply different price elasticities for each regionour marginal cost and markup estimates alsodiffer by region Given however that based on

TABLE 7mdashUPPER AND LOWER BOUNDS FOR MARGINAL COST MARKUP AND ANNUAL PROFIT BY PRODUCT GROUPS WITHIN

THE QUINOLONE SUBSEGMENT

Product groupLower boundfor MC (Rs)

Upper boundfor markup

Upper boundfor profit(Rs mill)

Upper boundfor MC (Rs)

Lower boundfor markup

Lower boundfor profit

(Rs)

Foreign ciprofloxacin 83 19 269 103 0 00(123) (012) (1655)

Foreign norfloxacin NA NA NA 53 0 00Foreign ofloxacin 323 70 1061 1085 0 00

(2316) (021) (3185)Domestic ciprofloxacin 47 59 17019 112 0 00

(114) (010) (29858)Domestic norfloxacin 52 43 2807 90 0 00

(020) (002) (1532)Domestic ofloxacin 587 34 1612 901 0 00

(218) (002) (1280)Domestic sparfloxacin 495 37 1985 788 0 00

(157) (002) (1100)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level Estimated lower bound forforeign norfloxacinrsquos marginal cost is negative since the estimated price elasticity is less than one in absolute value

1500 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Tables 6A and 6B the price elasticities do notseem to differ substantially across regions wereport for ease of exposition only the nationalaverages for marginal costs and markups inTable 7

Table 7 has two parts In the left part (firstthree columns) we report the lower bound formarginal cost and the corresponding upperbound for markup and upper bound for totalannual profit for each product group Thesenumbers are based on the price elasticities weobtained from estimating the two-level AIDSdemand system Since we do not have a reliableestimate for the price elasticity of foreign nor-floxacin (the point estimate is negative but lessthan one in absolute value and insignificant)we cannot compute the lower bound for mar-ginal cost in this case It is important to note thatthese estimates do not reflect either the actualmarginal cost or the actual markup for thesedrugs both because the existence of price reg-ulation implies that the unconstrained first-orderconditions are unlikely to hold each period andbecause our aggregation across firms of thesame domesticforeign status supplying thesame molecule makes the interpretation of theseestimates problematic In particular the fact thatwe ignore competition among firms within eachproduct group implies that our estimates willtend to overstate market power These numberswill prove useful however in the counterfac-tual simulations as they can give us a sense ofhow large the maximum profit gains for multi-national firms and the maximum profit lossesfor domestic firms are likely to be under patentenforcement The right part of the table (lastthree columns) reports the upper bound for mar-ginal cost which we obtain by simply taking themarginal cost to be equal to the observed priceWe do not report standard errors in this casesince the numbers are based on actual data Themarkups corresponding to these marginal costupper bounds are of course zero We conductthe counterfactual simulations using both thelower and upper bounds for marginal costs

The (upper bounds for) markups on the leftside of the table are generally plausible Thedomestic ciprofloxacin product group that dom-inates the quinolone subsegment and for whichwe documented high prices a high marketshare and a relatively low elasticity of demandenjoys one of the highest markup upper bounds(60 percent) and accounts for nearly 70 percent

of all profits derived within the subsegmentForeign ciprofloxacin on the other hand whichfaces the stiffest competition from domesticfirms and for which we estimated a highly elas-tic demand has the lowest markup upper bound(19 percent)

C Counterfactual Estimates of the Impact onPrices and Welfare

With estimates of the key demand and costparameters in hand we turn to the counterfac-tuals We consider the five separate scenarioslisted in the previous section All of the scenar-ios involve the withdrawal of one or more of thedomestic product groups from the market Ta-ble 8 displays our estimates of the consumerwelfare losses that result under the differentscenarios The losses are expressed in billion ofrupees per year All numbers presented in Ta-ble 8 and subsequent tables are based on usingthe lower bounds for marginal cost and upperbounds for markup in the simulations As dis-cussed earlier these numbers are the more in-teresting to work with since they give us anupper bound for the changes in the profits ofdomestic and foreign firms that would resultfrom patent enforcement In the Appendix wealso present results based on using the upperbounds for marginal costs in the simulations inwhich case the pre-TRIPS profits of domesticand foreign firms are zero In all cases marginalcosts are assumed to be constant in outputWhile naturally the profit implications differdepending on whether one uses the upper orlower bounds for marginal costs (firm profits arezero if one assumes the upper bound of mar-ginal cost in which case price equals marginalcost) the estimated consumer welfare losses aresimilar in the two cases We discuss these re-sults in more detail at the end of this section

The first column presents our estimates of theconsumer welfare losses attributable to the pureloss of product variety effect where we fix theprices of all remaining products as well as theoverall expenditure on quinolones while with-drawing one or more of the domestic productgroups Note that in our initial specification ofthe demand system had we not allowed for thepossibility that consumers might differentiatebetween domestic and foreign products evenwhen they contain the same molecule this

1501VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

particular component of the loss of consumerwelfare would not have arisen

The estimates reported in the second columnincorporate the expenditure-switching effect ontop of the loss of product variety Here basedupon the price elasticity estimates from thehigher-level AIDS system we adjust (down-ward) the expenditures allocated to the quino-lone subsegment as the composite price ofquinolones effectively increases as a consequenceof the higher virtual prices of the domestic productgroups that are withdrawn from the market Be-cause the estimates in this column are generatedassuming that the prices of the products that re-main in the market are not adjusted upward theyprovide a sense of what consumer losses would beif the introduction of product patents were coupledwith strict price regulation aimed at maintainingprices at prepatent levels Alternatively they canbe thought of as the relevant welfare numbers ifintense competition among firms within the re-maining product groups kept the prices of theproducts that were still offered in the market closeto the firmsrsquo marginal costs

The last column of Table 8 displays the esti-mated consumer welfare losses when bothcross-segment expenditure-switching and with-in-segment upward price adjustments are takeninto account

If we compare the results across the firstsecond and third columns all the counterfac-tual scenarios produce qualitatively similar pat-terns patterns that are consistent with what we

would expect Starting from the initial loss ofwelfare attributable to the loss of product vari-ety the option of switching expenditures outof the quinolone subsegment to other subseg-ments mitigates some of the initial welfareloss But if we then incorporate the upwardprice adjustments that result in response to thereduced competition the welfare losses aremagnified

Of particular interest from a policy perspec-tive are the relative magnitudes of these threeeffects which are similar under all the counter-factual scenarios though absolute levels varyconsiderably First despite the fact that the de-mand for quinolones is quite sensitive to thecomposite price of quinolonesmdashthe upper levelprice elasticity is 111mdashthe cross-subsegmentexpenditure-switching effects are in all thecases small (in absolute value terms) relative tothe other two effects For instance under thescenario where all the domestic quinolone prod-uct groups are withdrawn from the market theoverall consumer welfare loss of Rs 1781 bil-lion per year can be decomposed into an initialloss of Rs 1176 billion (66 percent) attribut-able to the loss of product variety a slightreduction in this initial loss of Rs 041 billion(2 percent) from Rs 1176 billion to Rs1135 billion because of expenditure switchingand a subsequent additional loss of Rs 646billion (36 percent) from Rs 1135 billion toRs 1781 billion because of the reduced com-petition and consequent price increases

TABLE 8mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure loss ofvariety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 732(087) (089) (146)

Only ofloxacin 008 008 023(008) (008) (010)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1253(177) (180) (415)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1058(142) (145) (331)

All four domestic quinolones products 1176dagger 1135dagger 1781(643) (634) (1270)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent significance level and dagger (dagger)denotes significance at the 10-percent level

1502 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 19: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

For products i that are withdrawn from themarket

(14) 0 i ir iiln pirV i10ln pjr j Di

10

j Di

01

i01ln pjr

j Di

00

i00ln pjr

ilnXQr

PQr

For products k that remain in the market

(15) pkr ckr 1 1

kk pkrpjr pirV 1

In the equations above pirV denotes the virtual

prices of the products that are removed fromthe market while pjr denotes the updatedprices of all other products Note that whensolving for the virtual prices we account forthe fact that both the price index for quino-lones PQr and the expenditure allocated tothis subsegment XQr need to be updated toreflect the fact that as a result of the pricechanges there may be substitution away fromthis subsegment To obtain the new quinoloneexpenditure XQr and the new price index PQrwe use the estimates and formulas for thehigher-level AIDS system In equation (15)ck refers to the marginal cost for product k thatwe have obtained from the previous estimationstage As mentioned in the previous subsectionwe conducted the simulations using both theupper and lower bounds for marginal cost ck

u

and ckL respectively The term kk(pkr

pjrpirV) re-

fers to the unconditional own-price elasticity forproduct k which is a function of the eliminatedproductsrsquo virtual prices and the remaining prod-uctsrsquo new equilibrium prices We conduct thecounterfactual simulations and welfare analysisat the regional level because the presence ofregion-specific product effects in the demandestimation implies region-specific demand elas-ticities and hence region-specific marginalcosts and markups The presentation and discus-sion of our results focus on national averages ofthe relevant variables that we construct by com-

puting weighted averages across regions usingthe population of each region as a weight

F Welfare Assessment

The simulation of the new equilibrium underpatent protection can provide important insightsinto how consumers and firms will respond tothe removal of domestic products in the market(for example toward which products consumerswill substitute or which prices will increase themost) To get a more precise idea of how peo-plersquos well-being will ultimately be affected byTRIPS we compute as a last step in our anal-ysis the welfare effects of the policy changeSocial welfare is defined as consumer welfareand the sum of domestic firm profits Thechange in domestic profits can easily be calcu-lated by comparing the domestic firm (variable)profits at the pre-TRIPS prices to the profitsthese firms will realize at the new simulatedprices Although foreign firm profits do notcount in domestic welfare calculations we alsocompute the effects of patent enforcement onforeign firm profits to get an idea of how largethe expected benefits of TRIPS are for thesefirms This provides in some sense an indirectway of assessing whether the claims that patentenforcement in countries like India will lead tomore research on developing-country-specificdiseases (such as malaria) have any validity iffor example we find that the effect of patentenforcement on the foreign firm profits realizedin India is small in magnitude it is unlikely thatforeign firms will engage in more developing-country-specific research in response to TRIPSIt is important to note that in all these calcula-tions we work with the lower-bound estimatesfor marginal costs since these give rise to thehighest possible markups Hence our estimateof profit loss for domestic producers most likelyoverstates this loss

On the consumer side we measure changes inconsumer welfare by the compensating variation(CV) defined as the additional expenditure thatconsumers need in order to achieve the sameutility level as before patent enforcement at thenew prices Specifically let P0 denote the pricevector before patent enforcement P the simulatedprice vector post-TRIPS (that we obtained usingthe methods described in the previous subsection)u0 the utility attained by consumers before TRIPS

1495VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

and E(u P) the higher-level expenditure functionThen the compensating variation is given by

(16) CV Eu0 P Eu0 P0

Note that the CV as computed in (16) representsthe combination of three effects

The pure product-variety effect that is theeffect that arises because one or more prod-ucts are not available to consumers anymoreholding the prices of all other remainingproducts and the total expenditure on thequinolone subsegment XQ constant

The expenditure-switching effect that is theeffect arising from substitution away fromquinolones and toward other subsegments ofthe antibiotics market again holding the pricesof all other remaining products constant

The reduced-competition effect that is theeffect that arises because the firms remainingin the market adjust (increase) their prices inresponse to the removal of domestic products

From both an analytical and a policy point ofview it is desirable to assess how large each ofthese effects is Accordingly we decompose thetotal effect on consumer welfare (the CV as givenby equation (16)) using the following procedure

To get the pure product variety effect wecompute virtual prices for the products that areremoved from the market holding the quino-lone expenditure XQ and the prices of all otherproducts fixed Let us call the resulting pricevector P1 Then the pure product variety effectis represented by E(u0 P1) E(u0 P0)

To compute the expenditure-switching effectwe compute another set of virtual prices againholding the prices of all remaining products fixedbut letting quinolone expenditure adjust in re-sponse to the new price index for the quinolonesegment (Given that the prices of the remainingproducts remain fixed the change in the priceindex arises only because of the removal of one ormore domestic products) Note that this scenariomost closely resembles the case in which patentlaws are enforced but strict price regulation keepsthe prices of the products offered in the market attheir pre-TRIP level Let us label the so-computedprice vector P2 The expenditure-switching effectis then E(u0 P2) E(u0 P1)

Finally the reduced competition effect aris-ing from higher prices for the remaining prod-

ucts is computed as the residual change in thecompensating variation once the product-varietyand expenditure-switching effects have been ac-counted for that is E(u0 P) E(u0 P2)where the price vector P is computed accordingto the formulas (14) and (15) to reflect theadjustment of prices to the new regime

To compute the standard errors associatedwith the counterfactual simulations (that is thestandard errors for the counterfactual prices andwelfare estimates) we again use bootstrappingWe first bootstrap the original sample (sales andprices of all the drugs) 300 times Next we esti-mate the AIDS model for each of these samplesand compute the counterfactual equilibrium pricesand welfare losses corresponding to each policyscenario for each of the 300 simulations In thefinal step we compute standard errors29

One limitation of our framework is that itdoes not allow for a heterogeneous response ofconsumers to the policy change Accordinglyour framework is not suited to addressing thequestion of how different groups in the popula-tion will be affected by patent law enforce-ment30 Along the same lines our frameworkdoes not accommodate the possibility of pricediscrimination which might lead to differentresults regarding welfare losses and profit gainsrelative to uniform pricing31 Still we believethat the results of the counterfactual simulationscan provide important insights into the likelyaggregate response to patent enforcement andthe factors that drive this response

III Results

A The Structure of Demand

Tables A2 and A3 in the Appendix display theresults from estimation of the lower- and upper-level AIDS systems respectively For ease of in-terpretation rather than discussing the coefficientestimates we focus our discussion on the implied

29 In cases some of the simulations do not converge wecompute the standard errors using only those simulationsthat converged

30 Addressing this question would require at a mini-mum micro data on consumer purchases as in Goldberg(1995) To our knowledge such data do not exist for theIndian pharmaceutical market

31 See Ernst R Berndt (1994) for an extensive discussionof uniform pricing

1496 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

unconditional price and expenditure elasticities re-ported in Table 6 Given that the region-specificproduct group effects ir reported in the last fourcolumns of Tables A2 and A3 imply region-spe-cific demand elasticities we report separate elas-ticities for each region in Tables 6A and 6B Asevident from the comparison of these tables theelasticities are very similar across regions so thatwe can focus the remaining discussion on theelasticities of one region only the northern regiondisplayed in Table 6A

The diagonal terms of Table 6A report theown-price elasticities which are in all but onecase negative and highly significant The oneexception is the foreign norfloxacin productgroupmdashwhose share of quinolone sales is 007percentmdashfor which we estimate a negative butinsignificant own-price elasticity For the remain-ing product groups demand appears to be highlyelastic with the estimated elasticities lower than2 in four out of the six cases The magnitude ofthe own-price elasticities matches the features ofthe Indian pharmaceutical market mentioned ear-lier which would suggest that Indian consumersare likely to be quite price-sensitive32 The elas-ticities appear especially large if one takes intoaccount that they refer to product groups (such asdomestic ciprofloxacin) and not individual drugsoffered by particular firms Their relative magni-tudes are also intuitive the drug with the largestmarket share and a relatively high price (domesticciprofloxacin) appears to be one of the least elas-tic In contrast foreign ciprofloxacin is highlyprice elastic this is plausible as ciprofloxacindrugs offered by subsidiaries of multinationalsface the stiffest competition in this market seg-ment from approximately 75 Indian firms offeringthe same molecule

The estimated expenditure elasticities appear inthe last column These are all positive indicatingthat the demand for all product groups is normalThe remaining cells display the estimated cross-price elasticities As one might perhaps expect forproducts within a therapeutic subsegment theseare mostly positive Out of a total of 49 priceelasticities we estimate there are six that do notconform to expectations these are the cross-price elasticities between different foreign prod-

uct groups which are estimated to be negativeand significant33 Fortunately these elasticitieshave negligible impact on the welfare analysisgiven that our counterfactuals focus on theeffect of withdrawing one or more domesticproducts from the market the most relevantelasticities are the ones that capture the responseof various product group shares to a change in theprice of one or more domestic groups these elas-ticities are the cross-price elasticities between var-ious domestic groups and the ones betweendomestic and foreign groups which are plausibleand precisely estimated34

Regarding these elasticities a striking aspectof our estimates is how large positive andsignificant the cross-price elasticities betweendifferent domestic product groups aremdashin factfor norfloxacin and ofloxacin we estimate thatdomestic product groups containing differentmolecules are closer substitutes for one anotherthan product groups that contain the same mol-ecule but are produced by foreign firms Incontrast for ciprofloxacin (the molecule withthe largest revenue share) we estimate a largepositive cross-price elasticity between the do-mestic and foreign versions

The fact that domestic products appear to beclose substitutes for other domestic productsthat contain different molecules truly representsan ldquoempiricalrdquo finding in the sense that we do

32 In developed economies elasticities of this magnitude havetypically been found only for generic drugs (and even then onlyrarely) or among consumers who lack health insurance

33 While these elasticities are clearly counterintuitive theyare not inconsistent with the underlying demand system whichimposes no restrictions on the sign of the cross-price elastici-ties We do not have a good explanation of why these elastic-ities are estimated to be negative A potential explanation isthat the shares of the foreign products are very small giventhis we observe very few consumers switching from one (verysmall) foreign group to another (very small) foreign groupwhen the price of the first foreign group goes up and hence theinference is not very reliable in this case

34 The cross-price elasticities between foreign drug groupscontaining different molecules will also have an effect on thewelfare estimates given that the withdrawal of domestic prod-ucts will generally lead to changes in the prices of all foreignproducts and market shares will be reallocated from someforeign products to other foreign products based on the newprices However this reallocation from ldquoforeign to foreignrdquo istruly second order in our case compared to the reallocationfrom ldquodomestic to domesticrdquo and ldquodomestic to foreignrdquo prod-ucts In scenarios in which we consider patent enforcementaccompanied by strict price regulation the cross-price elastic-ities between foreign product groups are in fact completelyirrelevant as the prices of foreign products are not allowed toincrease in these cases Still our welfare loss estimates remainsubstantial drivenmdashas beforemdashby the loss of product variety

1497VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

not impose it through any of our assumptionsregarding the demand function The questionthat naturally arises then is what might explainthis finding While we cannot formally addressthis question anecdotal accounts in various in-dustry studies suggest that the explanation maylie in the differences between domestic andforeign firms in the structure and coverage ofretail distribution networks

Distribution networks for pharmaceuticals inIndia are typically organized in a hierarchicalfashion Pharmaceutical companies deal mainlywith carrying and forwarding (CampF) agents inmany instances regionally based who each sup-ply a network of stockists (wholesalers) Thesestockists in turn deal with the retail pharma-cists through whom retail sales ultimately oc-cur35 The market share enjoyed by a particularpharmaceutical product therefore depends inpart on the number of retail pharmacists who

stock the product And it is here that thereappears to be a distinction between domesticfirms and multinational subsidiaries In particu-lar the retail reach of domestic firms as agroup tends to be much more comprehensivethan that of multinational subsidiaries (IndianCredit Rating Agency (ICRA) 1999)36

There appear to be two reasons for this Thefirst is that many of the larger Indian firmsbecause they have a much larger portfolio ofproducts over which to spread the associatedfixed costs typically have more extensive net-works of medical representatives The second issimply that there are many more domestic firms(and products) on the market At the retail levelthis would imply that local pharmacists mightbe more likely to stock domestic products con-taining two different molecules say ciprofloxa-cin and norfloxacin than they would domesticand foreign versions of the same molecule Tothe extent that patients (or their doctors) arewilling to substitute across molecules in order tosave on transport or search costs (eg going toanother pharmacy to check whether a particular

35 There are estimated to be some 300000 retail pharma-cists in India On average stockists deal with about 75 retailers(ICRA 1999) There are naturally variations in this structureand a host of specific exclusive dealing and other arrangementsexists in practice Pharmaceutical firms also maintain networksof medical representatives whose main function is to marketthe companyrsquos products to doctors who do the actual prescrib-ing of drugs In some instances firms do sell directly to thedoctors who then become the ldquoretailerrdquo as far as patients areconcerned but these are relatively rare

36 These differences were also highlighted in conversa-tions one of the authors had with CEOs and managingdirectors of several pharmaceutical firms as part of a sep-arate study

TABLE 6AmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN THE NORTHERN REGION

Product group

Elasticity with respect to

Prices of foreign productgroups Prices of domestic product groups Overall

quinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 557 013dagger 015 401 011dagger 011dagger 016 137(179) (007) (007) (184) (006) (006) (006) (029)

Foreign norfloxacin 427dagger 045 427dagger 350dagger 602 451 465 220(242) (112) (242) (210) (623) (184) (183) (105)

Foreign ofloxacin 011 010dagger 138 009 009dagger 023 011 116(005) (005) (031) (027) (005) (028) (004) (017)

Domestic ciprofloxacin 018 001 001 168 008 008 010 117(008) (000) (001) (023) (002) (002) (002) (003)

Domestic norfloxacin 004 003 004 058 223 042 040 073(001) (003) (001) (017) (011) (004) (003) (009)

Domestic ofloxacin 005 005 011 077 074 342 074 089(002) (002) (013) (028) (008) (025) (008) (021)

Domestic sparfloxacin 007 004 007 115 063 063 288 028(002) (001) (002) (015) (006) (006) (017) (012)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent significance level and dagger (dagger) denotes significance at the 10-percent level

1498 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE 6BmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN OTHER REGIONS

Demand patterns in the eastern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 594 014dagger 016 431 013 011 017 140(195) (008) (008) (201) (006) (007) (007) (031)

Foreign norfloxacin 329dagger 058 329dagger 264dagger 460 346 359 192(180) (083) (179) (153) (451) (138) (139) (082)

Foreign ofloxacin 012 010dagger 140 010 010 024 013 117(006) (006) (034) (031) (005) (030) (005) (019)

Domestic ciprofloxacin 018 001 001 172 009 008 011 117(008) (000) (002) (027) (002) (003) (002) (003)

Domestic norfloxacin 004 004 004 068 242 049 046 070(002) (004) (002) (020) (013) (005) (004) (010)

Domestic ofloxacin 004dagger 004 009 061 060 295 060 092(002) (001) (010) (028) (007) (022) (007) (017)

Domestic sparfloxacin 005 003 005 092 048 051 251 043(002) (001) (001) (017) (005) (006) (016) (010)

Demand patterns in the western region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 427 009dagger 011 286 008 006 011 126(129) (005) (005) (133) (004) (005) (004) (021)

Foreign norfloxacin 714dagger 008 710dagger 594dagger 995 742 770 299dagger

(369) (183) (368) (328) (1015) (297) (296) (178)Foreign ofloxacin 013 012dagger 145 009 012 026 013 119

(007) (006) (037) (029) (005) (033) (005) (021)Domestic ciprofloxacin 019 001 000 174 009 008 011 118

(008) (000) (001) (024) (002) (003) (002) (004)Domestic norfloxacin 004 004 004 067 243 050 047 069

(002) (004) (001) (018) (014) (005) (004) (010)Domestic ofloxacin 003 003 007 048 048 257 048 093

(002) (001) (008) (022) (005) (015) (004) (014)Domestic sparfloxacin 006 003 005 083 046 049 241 046

(002) (001) (001) (014) (004) (005) (012) (009)

Demand patterns in the southern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 560 013dagger 015 402 011 010 017 137(182) (007) (007) (187) (006) (006) (007) (029)

Foreign norfloxacin 533dagger 031 531dagger 431dagger 748 559 584 249dagger

(301) (134) (300) (245) (734) (219) (219) (128)Foreign ofloxacin 014 012dagger 148 011 012 029 016 120

(007) (007) (039) (033) (006) (035) (006) (022)Domestic ciprofloxacin 017 001 000 169 008 007 011 116

(008) (000) (001) (025) (002) (003) (002) (003)Domestic norfloxacin 004 003 003 059 225 043 040 073

(001) (003) (001) (019) (012) (004) (004) (009)Domestic ofloxacin 004dagger 004 009 062 060 296 060 091

(002) (001) (010) (026) (006) (019) (007) (017)Domestic sparfloxacin 008 005 008 138 073 075 323 016

(002) (001) (002) (018) (007) (008) (021) (014)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at 5-percent level anddagger (dagger) denotes significance at the 10-percent level

1499VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

foreign product is in stock) in aggregate datawe would expect to find precisely the substitu-tion patterns that we report in Table 6

Whether the particular explanation we provideabove is the correct one the high degree of sub-stitutability between domestic product groupsturns out to have important implications for thewelfare calculations We discuss these in moredetail below when we present the results of thecounterfactual welfare analysis Another elasticitywith important implications for the counterfactu-als is the price elasticity for the quinolone subseg-ment as a whole which indicates how likelyconsumers are to switch to other antibioticsgroups when faced with a price increase for quin-olones This elasticity is computed on the basis ofthe results in Table A3 and it is at 111 (stan-dard error 024) this is large in magnitudebutmdashas expectedmdashsmaller in absolute value thanthe own-price elasticities of the product groupswithin the quinolone subsegment

The results in Tables 6A and 6B are based onour preferred specification discussed in SectionII In Tables A4 to A6 in the Appendix weexperiment with some alternative specificationsTables A4(a)ndashA4(c) correspond to a specificationthat includes in addition to product-group-specificregional fixed effects product-group-specific (andfor the upper level antibiotics-segment-specific)seasonal effects We distinguish among three sea-sonsmdashthe summer monsoon and wintermdashandreport the unconditional demand elasticities for

the northern region for each of these seasons Asevident from the tables in the Appendix our elas-ticity estimates are robust to the inclusion of sea-sonal effects The demand elasticities in Table A5are based on estimation of the demand system byOLS Compared to the elasticities obtained by IVthe OLS elasticities are smaller in absolute valueimplying that welfare calculations based on theOLS estimates would produce larger welfare lossestimates Nevertheless some of the patterns re-garding the cross-price elasticities discussed ear-lier are also evident in the OLS results inparticular the cross-price elasticities between dif-ferent domestic product groups are all positivelarge and significant and in most instances largerthan the cross-price elasticities between drugs thatcontain the same molecule but are produced byfirms of different domesticforeign status Theclose substitutability of domestic products indi-cated by both the OLS and IV estimates seems tobe one of the most robust findings of the paper

B Cost and Markup Estimates

Table 7 displays the marginal costs markupsand profits implied by the price elasticity esti-mates of Tables 6A and 6B for each of the sevenproduct groups Given that our regional effectsimply different price elasticities for each regionour marginal cost and markup estimates alsodiffer by region Given however that based on

TABLE 7mdashUPPER AND LOWER BOUNDS FOR MARGINAL COST MARKUP AND ANNUAL PROFIT BY PRODUCT GROUPS WITHIN

THE QUINOLONE SUBSEGMENT

Product groupLower boundfor MC (Rs)

Upper boundfor markup

Upper boundfor profit(Rs mill)

Upper boundfor MC (Rs)

Lower boundfor markup

Lower boundfor profit

(Rs)

Foreign ciprofloxacin 83 19 269 103 0 00(123) (012) (1655)

Foreign norfloxacin NA NA NA 53 0 00Foreign ofloxacin 323 70 1061 1085 0 00

(2316) (021) (3185)Domestic ciprofloxacin 47 59 17019 112 0 00

(114) (010) (29858)Domestic norfloxacin 52 43 2807 90 0 00

(020) (002) (1532)Domestic ofloxacin 587 34 1612 901 0 00

(218) (002) (1280)Domestic sparfloxacin 495 37 1985 788 0 00

(157) (002) (1100)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level Estimated lower bound forforeign norfloxacinrsquos marginal cost is negative since the estimated price elasticity is less than one in absolute value

1500 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Tables 6A and 6B the price elasticities do notseem to differ substantially across regions wereport for ease of exposition only the nationalaverages for marginal costs and markups inTable 7

Table 7 has two parts In the left part (firstthree columns) we report the lower bound formarginal cost and the corresponding upperbound for markup and upper bound for totalannual profit for each product group Thesenumbers are based on the price elasticities weobtained from estimating the two-level AIDSdemand system Since we do not have a reliableestimate for the price elasticity of foreign nor-floxacin (the point estimate is negative but lessthan one in absolute value and insignificant)we cannot compute the lower bound for mar-ginal cost in this case It is important to note thatthese estimates do not reflect either the actualmarginal cost or the actual markup for thesedrugs both because the existence of price reg-ulation implies that the unconstrained first-orderconditions are unlikely to hold each period andbecause our aggregation across firms of thesame domesticforeign status supplying thesame molecule makes the interpretation of theseestimates problematic In particular the fact thatwe ignore competition among firms within eachproduct group implies that our estimates willtend to overstate market power These numberswill prove useful however in the counterfac-tual simulations as they can give us a sense ofhow large the maximum profit gains for multi-national firms and the maximum profit lossesfor domestic firms are likely to be under patentenforcement The right part of the table (lastthree columns) reports the upper bound for mar-ginal cost which we obtain by simply taking themarginal cost to be equal to the observed priceWe do not report standard errors in this casesince the numbers are based on actual data Themarkups corresponding to these marginal costupper bounds are of course zero We conductthe counterfactual simulations using both thelower and upper bounds for marginal costs

The (upper bounds for) markups on the leftside of the table are generally plausible Thedomestic ciprofloxacin product group that dom-inates the quinolone subsegment and for whichwe documented high prices a high marketshare and a relatively low elasticity of demandenjoys one of the highest markup upper bounds(60 percent) and accounts for nearly 70 percent

of all profits derived within the subsegmentForeign ciprofloxacin on the other hand whichfaces the stiffest competition from domesticfirms and for which we estimated a highly elas-tic demand has the lowest markup upper bound(19 percent)

C Counterfactual Estimates of the Impact onPrices and Welfare

With estimates of the key demand and costparameters in hand we turn to the counterfac-tuals We consider the five separate scenarioslisted in the previous section All of the scenar-ios involve the withdrawal of one or more of thedomestic product groups from the market Ta-ble 8 displays our estimates of the consumerwelfare losses that result under the differentscenarios The losses are expressed in billion ofrupees per year All numbers presented in Ta-ble 8 and subsequent tables are based on usingthe lower bounds for marginal cost and upperbounds for markup in the simulations As dis-cussed earlier these numbers are the more in-teresting to work with since they give us anupper bound for the changes in the profits ofdomestic and foreign firms that would resultfrom patent enforcement In the Appendix wealso present results based on using the upperbounds for marginal costs in the simulations inwhich case the pre-TRIPS profits of domesticand foreign firms are zero In all cases marginalcosts are assumed to be constant in outputWhile naturally the profit implications differdepending on whether one uses the upper orlower bounds for marginal costs (firm profits arezero if one assumes the upper bound of mar-ginal cost in which case price equals marginalcost) the estimated consumer welfare losses aresimilar in the two cases We discuss these re-sults in more detail at the end of this section

The first column presents our estimates of theconsumer welfare losses attributable to the pureloss of product variety effect where we fix theprices of all remaining products as well as theoverall expenditure on quinolones while with-drawing one or more of the domestic productgroups Note that in our initial specification ofthe demand system had we not allowed for thepossibility that consumers might differentiatebetween domestic and foreign products evenwhen they contain the same molecule this

1501VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

particular component of the loss of consumerwelfare would not have arisen

The estimates reported in the second columnincorporate the expenditure-switching effect ontop of the loss of product variety Here basedupon the price elasticity estimates from thehigher-level AIDS system we adjust (down-ward) the expenditures allocated to the quino-lone subsegment as the composite price ofquinolones effectively increases as a consequenceof the higher virtual prices of the domestic productgroups that are withdrawn from the market Be-cause the estimates in this column are generatedassuming that the prices of the products that re-main in the market are not adjusted upward theyprovide a sense of what consumer losses would beif the introduction of product patents were coupledwith strict price regulation aimed at maintainingprices at prepatent levels Alternatively they canbe thought of as the relevant welfare numbers ifintense competition among firms within the re-maining product groups kept the prices of theproducts that were still offered in the market closeto the firmsrsquo marginal costs

The last column of Table 8 displays the esti-mated consumer welfare losses when bothcross-segment expenditure-switching and with-in-segment upward price adjustments are takeninto account

If we compare the results across the firstsecond and third columns all the counterfac-tual scenarios produce qualitatively similar pat-terns patterns that are consistent with what we

would expect Starting from the initial loss ofwelfare attributable to the loss of product vari-ety the option of switching expenditures outof the quinolone subsegment to other subseg-ments mitigates some of the initial welfareloss But if we then incorporate the upwardprice adjustments that result in response to thereduced competition the welfare losses aremagnified

Of particular interest from a policy perspec-tive are the relative magnitudes of these threeeffects which are similar under all the counter-factual scenarios though absolute levels varyconsiderably First despite the fact that the de-mand for quinolones is quite sensitive to thecomposite price of quinolonesmdashthe upper levelprice elasticity is 111mdashthe cross-subsegmentexpenditure-switching effects are in all thecases small (in absolute value terms) relative tothe other two effects For instance under thescenario where all the domestic quinolone prod-uct groups are withdrawn from the market theoverall consumer welfare loss of Rs 1781 bil-lion per year can be decomposed into an initialloss of Rs 1176 billion (66 percent) attribut-able to the loss of product variety a slightreduction in this initial loss of Rs 041 billion(2 percent) from Rs 1176 billion to Rs1135 billion because of expenditure switchingand a subsequent additional loss of Rs 646billion (36 percent) from Rs 1135 billion toRs 1781 billion because of the reduced com-petition and consequent price increases

TABLE 8mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure loss ofvariety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 732(087) (089) (146)

Only ofloxacin 008 008 023(008) (008) (010)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1253(177) (180) (415)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1058(142) (145) (331)

All four domestic quinolones products 1176dagger 1135dagger 1781(643) (634) (1270)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent significance level and dagger (dagger)denotes significance at the 10-percent level

1502 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 20: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

and E(u P) the higher-level expenditure functionThen the compensating variation is given by

(16) CV Eu0 P Eu0 P0

Note that the CV as computed in (16) representsthe combination of three effects

The pure product-variety effect that is theeffect that arises because one or more prod-ucts are not available to consumers anymoreholding the prices of all other remainingproducts and the total expenditure on thequinolone subsegment XQ constant

The expenditure-switching effect that is theeffect arising from substitution away fromquinolones and toward other subsegments ofthe antibiotics market again holding the pricesof all other remaining products constant

The reduced-competition effect that is theeffect that arises because the firms remainingin the market adjust (increase) their prices inresponse to the removal of domestic products

From both an analytical and a policy point ofview it is desirable to assess how large each ofthese effects is Accordingly we decompose thetotal effect on consumer welfare (the CV as givenby equation (16)) using the following procedure

To get the pure product variety effect wecompute virtual prices for the products that areremoved from the market holding the quino-lone expenditure XQ and the prices of all otherproducts fixed Let us call the resulting pricevector P1 Then the pure product variety effectis represented by E(u0 P1) E(u0 P0)

To compute the expenditure-switching effectwe compute another set of virtual prices againholding the prices of all remaining products fixedbut letting quinolone expenditure adjust in re-sponse to the new price index for the quinolonesegment (Given that the prices of the remainingproducts remain fixed the change in the priceindex arises only because of the removal of one ormore domestic products) Note that this scenariomost closely resembles the case in which patentlaws are enforced but strict price regulation keepsthe prices of the products offered in the market attheir pre-TRIP level Let us label the so-computedprice vector P2 The expenditure-switching effectis then E(u0 P2) E(u0 P1)

Finally the reduced competition effect aris-ing from higher prices for the remaining prod-

ucts is computed as the residual change in thecompensating variation once the product-varietyand expenditure-switching effects have been ac-counted for that is E(u0 P) E(u0 P2)where the price vector P is computed accordingto the formulas (14) and (15) to reflect theadjustment of prices to the new regime

To compute the standard errors associatedwith the counterfactual simulations (that is thestandard errors for the counterfactual prices andwelfare estimates) we again use bootstrappingWe first bootstrap the original sample (sales andprices of all the drugs) 300 times Next we esti-mate the AIDS model for each of these samplesand compute the counterfactual equilibrium pricesand welfare losses corresponding to each policyscenario for each of the 300 simulations In thefinal step we compute standard errors29

One limitation of our framework is that itdoes not allow for a heterogeneous response ofconsumers to the policy change Accordinglyour framework is not suited to addressing thequestion of how different groups in the popula-tion will be affected by patent law enforce-ment30 Along the same lines our frameworkdoes not accommodate the possibility of pricediscrimination which might lead to differentresults regarding welfare losses and profit gainsrelative to uniform pricing31 Still we believethat the results of the counterfactual simulationscan provide important insights into the likelyaggregate response to patent enforcement andthe factors that drive this response

III Results

A The Structure of Demand

Tables A2 and A3 in the Appendix display theresults from estimation of the lower- and upper-level AIDS systems respectively For ease of in-terpretation rather than discussing the coefficientestimates we focus our discussion on the implied

29 In cases some of the simulations do not converge wecompute the standard errors using only those simulationsthat converged

30 Addressing this question would require at a mini-mum micro data on consumer purchases as in Goldberg(1995) To our knowledge such data do not exist for theIndian pharmaceutical market

31 See Ernst R Berndt (1994) for an extensive discussionof uniform pricing

1496 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

unconditional price and expenditure elasticities re-ported in Table 6 Given that the region-specificproduct group effects ir reported in the last fourcolumns of Tables A2 and A3 imply region-spe-cific demand elasticities we report separate elas-ticities for each region in Tables 6A and 6B Asevident from the comparison of these tables theelasticities are very similar across regions so thatwe can focus the remaining discussion on theelasticities of one region only the northern regiondisplayed in Table 6A

The diagonal terms of Table 6A report theown-price elasticities which are in all but onecase negative and highly significant The oneexception is the foreign norfloxacin productgroupmdashwhose share of quinolone sales is 007percentmdashfor which we estimate a negative butinsignificant own-price elasticity For the remain-ing product groups demand appears to be highlyelastic with the estimated elasticities lower than2 in four out of the six cases The magnitude ofthe own-price elasticities matches the features ofthe Indian pharmaceutical market mentioned ear-lier which would suggest that Indian consumersare likely to be quite price-sensitive32 The elas-ticities appear especially large if one takes intoaccount that they refer to product groups (such asdomestic ciprofloxacin) and not individual drugsoffered by particular firms Their relative magni-tudes are also intuitive the drug with the largestmarket share and a relatively high price (domesticciprofloxacin) appears to be one of the least elas-tic In contrast foreign ciprofloxacin is highlyprice elastic this is plausible as ciprofloxacindrugs offered by subsidiaries of multinationalsface the stiffest competition in this market seg-ment from approximately 75 Indian firms offeringthe same molecule

The estimated expenditure elasticities appear inthe last column These are all positive indicatingthat the demand for all product groups is normalThe remaining cells display the estimated cross-price elasticities As one might perhaps expect forproducts within a therapeutic subsegment theseare mostly positive Out of a total of 49 priceelasticities we estimate there are six that do notconform to expectations these are the cross-price elasticities between different foreign prod-

uct groups which are estimated to be negativeand significant33 Fortunately these elasticitieshave negligible impact on the welfare analysisgiven that our counterfactuals focus on theeffect of withdrawing one or more domesticproducts from the market the most relevantelasticities are the ones that capture the responseof various product group shares to a change in theprice of one or more domestic groups these elas-ticities are the cross-price elasticities between var-ious domestic groups and the ones betweendomestic and foreign groups which are plausibleand precisely estimated34

Regarding these elasticities a striking aspectof our estimates is how large positive andsignificant the cross-price elasticities betweendifferent domestic product groups aremdashin factfor norfloxacin and ofloxacin we estimate thatdomestic product groups containing differentmolecules are closer substitutes for one anotherthan product groups that contain the same mol-ecule but are produced by foreign firms Incontrast for ciprofloxacin (the molecule withthe largest revenue share) we estimate a largepositive cross-price elasticity between the do-mestic and foreign versions

The fact that domestic products appear to beclose substitutes for other domestic productsthat contain different molecules truly representsan ldquoempiricalrdquo finding in the sense that we do

32 In developed economies elasticities of this magnitude havetypically been found only for generic drugs (and even then onlyrarely) or among consumers who lack health insurance

33 While these elasticities are clearly counterintuitive theyare not inconsistent with the underlying demand system whichimposes no restrictions on the sign of the cross-price elastici-ties We do not have a good explanation of why these elastic-ities are estimated to be negative A potential explanation isthat the shares of the foreign products are very small giventhis we observe very few consumers switching from one (verysmall) foreign group to another (very small) foreign groupwhen the price of the first foreign group goes up and hence theinference is not very reliable in this case

34 The cross-price elasticities between foreign drug groupscontaining different molecules will also have an effect on thewelfare estimates given that the withdrawal of domestic prod-ucts will generally lead to changes in the prices of all foreignproducts and market shares will be reallocated from someforeign products to other foreign products based on the newprices However this reallocation from ldquoforeign to foreignrdquo istruly second order in our case compared to the reallocationfrom ldquodomestic to domesticrdquo and ldquodomestic to foreignrdquo prod-ucts In scenarios in which we consider patent enforcementaccompanied by strict price regulation the cross-price elastic-ities between foreign product groups are in fact completelyirrelevant as the prices of foreign products are not allowed toincrease in these cases Still our welfare loss estimates remainsubstantial drivenmdashas beforemdashby the loss of product variety

1497VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

not impose it through any of our assumptionsregarding the demand function The questionthat naturally arises then is what might explainthis finding While we cannot formally addressthis question anecdotal accounts in various in-dustry studies suggest that the explanation maylie in the differences between domestic andforeign firms in the structure and coverage ofretail distribution networks

Distribution networks for pharmaceuticals inIndia are typically organized in a hierarchicalfashion Pharmaceutical companies deal mainlywith carrying and forwarding (CampF) agents inmany instances regionally based who each sup-ply a network of stockists (wholesalers) Thesestockists in turn deal with the retail pharma-cists through whom retail sales ultimately oc-cur35 The market share enjoyed by a particularpharmaceutical product therefore depends inpart on the number of retail pharmacists who

stock the product And it is here that thereappears to be a distinction between domesticfirms and multinational subsidiaries In particu-lar the retail reach of domestic firms as agroup tends to be much more comprehensivethan that of multinational subsidiaries (IndianCredit Rating Agency (ICRA) 1999)36

There appear to be two reasons for this Thefirst is that many of the larger Indian firmsbecause they have a much larger portfolio ofproducts over which to spread the associatedfixed costs typically have more extensive net-works of medical representatives The second issimply that there are many more domestic firms(and products) on the market At the retail levelthis would imply that local pharmacists mightbe more likely to stock domestic products con-taining two different molecules say ciprofloxa-cin and norfloxacin than they would domesticand foreign versions of the same molecule Tothe extent that patients (or their doctors) arewilling to substitute across molecules in order tosave on transport or search costs (eg going toanother pharmacy to check whether a particular

35 There are estimated to be some 300000 retail pharma-cists in India On average stockists deal with about 75 retailers(ICRA 1999) There are naturally variations in this structureand a host of specific exclusive dealing and other arrangementsexists in practice Pharmaceutical firms also maintain networksof medical representatives whose main function is to marketthe companyrsquos products to doctors who do the actual prescrib-ing of drugs In some instances firms do sell directly to thedoctors who then become the ldquoretailerrdquo as far as patients areconcerned but these are relatively rare

36 These differences were also highlighted in conversa-tions one of the authors had with CEOs and managingdirectors of several pharmaceutical firms as part of a sep-arate study

TABLE 6AmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN THE NORTHERN REGION

Product group

Elasticity with respect to

Prices of foreign productgroups Prices of domestic product groups Overall

quinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 557 013dagger 015 401 011dagger 011dagger 016 137(179) (007) (007) (184) (006) (006) (006) (029)

Foreign norfloxacin 427dagger 045 427dagger 350dagger 602 451 465 220(242) (112) (242) (210) (623) (184) (183) (105)

Foreign ofloxacin 011 010dagger 138 009 009dagger 023 011 116(005) (005) (031) (027) (005) (028) (004) (017)

Domestic ciprofloxacin 018 001 001 168 008 008 010 117(008) (000) (001) (023) (002) (002) (002) (003)

Domestic norfloxacin 004 003 004 058 223 042 040 073(001) (003) (001) (017) (011) (004) (003) (009)

Domestic ofloxacin 005 005 011 077 074 342 074 089(002) (002) (013) (028) (008) (025) (008) (021)

Domestic sparfloxacin 007 004 007 115 063 063 288 028(002) (001) (002) (015) (006) (006) (017) (012)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent significance level and dagger (dagger) denotes significance at the 10-percent level

1498 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE 6BmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN OTHER REGIONS

Demand patterns in the eastern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 594 014dagger 016 431 013 011 017 140(195) (008) (008) (201) (006) (007) (007) (031)

Foreign norfloxacin 329dagger 058 329dagger 264dagger 460 346 359 192(180) (083) (179) (153) (451) (138) (139) (082)

Foreign ofloxacin 012 010dagger 140 010 010 024 013 117(006) (006) (034) (031) (005) (030) (005) (019)

Domestic ciprofloxacin 018 001 001 172 009 008 011 117(008) (000) (002) (027) (002) (003) (002) (003)

Domestic norfloxacin 004 004 004 068 242 049 046 070(002) (004) (002) (020) (013) (005) (004) (010)

Domestic ofloxacin 004dagger 004 009 061 060 295 060 092(002) (001) (010) (028) (007) (022) (007) (017)

Domestic sparfloxacin 005 003 005 092 048 051 251 043(002) (001) (001) (017) (005) (006) (016) (010)

Demand patterns in the western region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 427 009dagger 011 286 008 006 011 126(129) (005) (005) (133) (004) (005) (004) (021)

Foreign norfloxacin 714dagger 008 710dagger 594dagger 995 742 770 299dagger

(369) (183) (368) (328) (1015) (297) (296) (178)Foreign ofloxacin 013 012dagger 145 009 012 026 013 119

(007) (006) (037) (029) (005) (033) (005) (021)Domestic ciprofloxacin 019 001 000 174 009 008 011 118

(008) (000) (001) (024) (002) (003) (002) (004)Domestic norfloxacin 004 004 004 067 243 050 047 069

(002) (004) (001) (018) (014) (005) (004) (010)Domestic ofloxacin 003 003 007 048 048 257 048 093

(002) (001) (008) (022) (005) (015) (004) (014)Domestic sparfloxacin 006 003 005 083 046 049 241 046

(002) (001) (001) (014) (004) (005) (012) (009)

Demand patterns in the southern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 560 013dagger 015 402 011 010 017 137(182) (007) (007) (187) (006) (006) (007) (029)

Foreign norfloxacin 533dagger 031 531dagger 431dagger 748 559 584 249dagger

(301) (134) (300) (245) (734) (219) (219) (128)Foreign ofloxacin 014 012dagger 148 011 012 029 016 120

(007) (007) (039) (033) (006) (035) (006) (022)Domestic ciprofloxacin 017 001 000 169 008 007 011 116

(008) (000) (001) (025) (002) (003) (002) (003)Domestic norfloxacin 004 003 003 059 225 043 040 073

(001) (003) (001) (019) (012) (004) (004) (009)Domestic ofloxacin 004dagger 004 009 062 060 296 060 091

(002) (001) (010) (026) (006) (019) (007) (017)Domestic sparfloxacin 008 005 008 138 073 075 323 016

(002) (001) (002) (018) (007) (008) (021) (014)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at 5-percent level anddagger (dagger) denotes significance at the 10-percent level

1499VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

foreign product is in stock) in aggregate datawe would expect to find precisely the substitu-tion patterns that we report in Table 6

Whether the particular explanation we provideabove is the correct one the high degree of sub-stitutability between domestic product groupsturns out to have important implications for thewelfare calculations We discuss these in moredetail below when we present the results of thecounterfactual welfare analysis Another elasticitywith important implications for the counterfactu-als is the price elasticity for the quinolone subseg-ment as a whole which indicates how likelyconsumers are to switch to other antibioticsgroups when faced with a price increase for quin-olones This elasticity is computed on the basis ofthe results in Table A3 and it is at 111 (stan-dard error 024) this is large in magnitudebutmdashas expectedmdashsmaller in absolute value thanthe own-price elasticities of the product groupswithin the quinolone subsegment

The results in Tables 6A and 6B are based onour preferred specification discussed in SectionII In Tables A4 to A6 in the Appendix weexperiment with some alternative specificationsTables A4(a)ndashA4(c) correspond to a specificationthat includes in addition to product-group-specificregional fixed effects product-group-specific (andfor the upper level antibiotics-segment-specific)seasonal effects We distinguish among three sea-sonsmdashthe summer monsoon and wintermdashandreport the unconditional demand elasticities for

the northern region for each of these seasons Asevident from the tables in the Appendix our elas-ticity estimates are robust to the inclusion of sea-sonal effects The demand elasticities in Table A5are based on estimation of the demand system byOLS Compared to the elasticities obtained by IVthe OLS elasticities are smaller in absolute valueimplying that welfare calculations based on theOLS estimates would produce larger welfare lossestimates Nevertheless some of the patterns re-garding the cross-price elasticities discussed ear-lier are also evident in the OLS results inparticular the cross-price elasticities between dif-ferent domestic product groups are all positivelarge and significant and in most instances largerthan the cross-price elasticities between drugs thatcontain the same molecule but are produced byfirms of different domesticforeign status Theclose substitutability of domestic products indi-cated by both the OLS and IV estimates seems tobe one of the most robust findings of the paper

B Cost and Markup Estimates

Table 7 displays the marginal costs markupsand profits implied by the price elasticity esti-mates of Tables 6A and 6B for each of the sevenproduct groups Given that our regional effectsimply different price elasticities for each regionour marginal cost and markup estimates alsodiffer by region Given however that based on

TABLE 7mdashUPPER AND LOWER BOUNDS FOR MARGINAL COST MARKUP AND ANNUAL PROFIT BY PRODUCT GROUPS WITHIN

THE QUINOLONE SUBSEGMENT

Product groupLower boundfor MC (Rs)

Upper boundfor markup

Upper boundfor profit(Rs mill)

Upper boundfor MC (Rs)

Lower boundfor markup

Lower boundfor profit

(Rs)

Foreign ciprofloxacin 83 19 269 103 0 00(123) (012) (1655)

Foreign norfloxacin NA NA NA 53 0 00Foreign ofloxacin 323 70 1061 1085 0 00

(2316) (021) (3185)Domestic ciprofloxacin 47 59 17019 112 0 00

(114) (010) (29858)Domestic norfloxacin 52 43 2807 90 0 00

(020) (002) (1532)Domestic ofloxacin 587 34 1612 901 0 00

(218) (002) (1280)Domestic sparfloxacin 495 37 1985 788 0 00

(157) (002) (1100)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level Estimated lower bound forforeign norfloxacinrsquos marginal cost is negative since the estimated price elasticity is less than one in absolute value

1500 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Tables 6A and 6B the price elasticities do notseem to differ substantially across regions wereport for ease of exposition only the nationalaverages for marginal costs and markups inTable 7

Table 7 has two parts In the left part (firstthree columns) we report the lower bound formarginal cost and the corresponding upperbound for markup and upper bound for totalannual profit for each product group Thesenumbers are based on the price elasticities weobtained from estimating the two-level AIDSdemand system Since we do not have a reliableestimate for the price elasticity of foreign nor-floxacin (the point estimate is negative but lessthan one in absolute value and insignificant)we cannot compute the lower bound for mar-ginal cost in this case It is important to note thatthese estimates do not reflect either the actualmarginal cost or the actual markup for thesedrugs both because the existence of price reg-ulation implies that the unconstrained first-orderconditions are unlikely to hold each period andbecause our aggregation across firms of thesame domesticforeign status supplying thesame molecule makes the interpretation of theseestimates problematic In particular the fact thatwe ignore competition among firms within eachproduct group implies that our estimates willtend to overstate market power These numberswill prove useful however in the counterfac-tual simulations as they can give us a sense ofhow large the maximum profit gains for multi-national firms and the maximum profit lossesfor domestic firms are likely to be under patentenforcement The right part of the table (lastthree columns) reports the upper bound for mar-ginal cost which we obtain by simply taking themarginal cost to be equal to the observed priceWe do not report standard errors in this casesince the numbers are based on actual data Themarkups corresponding to these marginal costupper bounds are of course zero We conductthe counterfactual simulations using both thelower and upper bounds for marginal costs

The (upper bounds for) markups on the leftside of the table are generally plausible Thedomestic ciprofloxacin product group that dom-inates the quinolone subsegment and for whichwe documented high prices a high marketshare and a relatively low elasticity of demandenjoys one of the highest markup upper bounds(60 percent) and accounts for nearly 70 percent

of all profits derived within the subsegmentForeign ciprofloxacin on the other hand whichfaces the stiffest competition from domesticfirms and for which we estimated a highly elas-tic demand has the lowest markup upper bound(19 percent)

C Counterfactual Estimates of the Impact onPrices and Welfare

With estimates of the key demand and costparameters in hand we turn to the counterfac-tuals We consider the five separate scenarioslisted in the previous section All of the scenar-ios involve the withdrawal of one or more of thedomestic product groups from the market Ta-ble 8 displays our estimates of the consumerwelfare losses that result under the differentscenarios The losses are expressed in billion ofrupees per year All numbers presented in Ta-ble 8 and subsequent tables are based on usingthe lower bounds for marginal cost and upperbounds for markup in the simulations As dis-cussed earlier these numbers are the more in-teresting to work with since they give us anupper bound for the changes in the profits ofdomestic and foreign firms that would resultfrom patent enforcement In the Appendix wealso present results based on using the upperbounds for marginal costs in the simulations inwhich case the pre-TRIPS profits of domesticand foreign firms are zero In all cases marginalcosts are assumed to be constant in outputWhile naturally the profit implications differdepending on whether one uses the upper orlower bounds for marginal costs (firm profits arezero if one assumes the upper bound of mar-ginal cost in which case price equals marginalcost) the estimated consumer welfare losses aresimilar in the two cases We discuss these re-sults in more detail at the end of this section

The first column presents our estimates of theconsumer welfare losses attributable to the pureloss of product variety effect where we fix theprices of all remaining products as well as theoverall expenditure on quinolones while with-drawing one or more of the domestic productgroups Note that in our initial specification ofthe demand system had we not allowed for thepossibility that consumers might differentiatebetween domestic and foreign products evenwhen they contain the same molecule this

1501VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

particular component of the loss of consumerwelfare would not have arisen

The estimates reported in the second columnincorporate the expenditure-switching effect ontop of the loss of product variety Here basedupon the price elasticity estimates from thehigher-level AIDS system we adjust (down-ward) the expenditures allocated to the quino-lone subsegment as the composite price ofquinolones effectively increases as a consequenceof the higher virtual prices of the domestic productgroups that are withdrawn from the market Be-cause the estimates in this column are generatedassuming that the prices of the products that re-main in the market are not adjusted upward theyprovide a sense of what consumer losses would beif the introduction of product patents were coupledwith strict price regulation aimed at maintainingprices at prepatent levels Alternatively they canbe thought of as the relevant welfare numbers ifintense competition among firms within the re-maining product groups kept the prices of theproducts that were still offered in the market closeto the firmsrsquo marginal costs

The last column of Table 8 displays the esti-mated consumer welfare losses when bothcross-segment expenditure-switching and with-in-segment upward price adjustments are takeninto account

If we compare the results across the firstsecond and third columns all the counterfac-tual scenarios produce qualitatively similar pat-terns patterns that are consistent with what we

would expect Starting from the initial loss ofwelfare attributable to the loss of product vari-ety the option of switching expenditures outof the quinolone subsegment to other subseg-ments mitigates some of the initial welfareloss But if we then incorporate the upwardprice adjustments that result in response to thereduced competition the welfare losses aremagnified

Of particular interest from a policy perspec-tive are the relative magnitudes of these threeeffects which are similar under all the counter-factual scenarios though absolute levels varyconsiderably First despite the fact that the de-mand for quinolones is quite sensitive to thecomposite price of quinolonesmdashthe upper levelprice elasticity is 111mdashthe cross-subsegmentexpenditure-switching effects are in all thecases small (in absolute value terms) relative tothe other two effects For instance under thescenario where all the domestic quinolone prod-uct groups are withdrawn from the market theoverall consumer welfare loss of Rs 1781 bil-lion per year can be decomposed into an initialloss of Rs 1176 billion (66 percent) attribut-able to the loss of product variety a slightreduction in this initial loss of Rs 041 billion(2 percent) from Rs 1176 billion to Rs1135 billion because of expenditure switchingand a subsequent additional loss of Rs 646billion (36 percent) from Rs 1135 billion toRs 1781 billion because of the reduced com-petition and consequent price increases

TABLE 8mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure loss ofvariety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 732(087) (089) (146)

Only ofloxacin 008 008 023(008) (008) (010)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1253(177) (180) (415)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1058(142) (145) (331)

All four domestic quinolones products 1176dagger 1135dagger 1781(643) (634) (1270)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent significance level and dagger (dagger)denotes significance at the 10-percent level

1502 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 21: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

unconditional price and expenditure elasticities re-ported in Table 6 Given that the region-specificproduct group effects ir reported in the last fourcolumns of Tables A2 and A3 imply region-spe-cific demand elasticities we report separate elas-ticities for each region in Tables 6A and 6B Asevident from the comparison of these tables theelasticities are very similar across regions so thatwe can focus the remaining discussion on theelasticities of one region only the northern regiondisplayed in Table 6A

The diagonal terms of Table 6A report theown-price elasticities which are in all but onecase negative and highly significant The oneexception is the foreign norfloxacin productgroupmdashwhose share of quinolone sales is 007percentmdashfor which we estimate a negative butinsignificant own-price elasticity For the remain-ing product groups demand appears to be highlyelastic with the estimated elasticities lower than2 in four out of the six cases The magnitude ofthe own-price elasticities matches the features ofthe Indian pharmaceutical market mentioned ear-lier which would suggest that Indian consumersare likely to be quite price-sensitive32 The elas-ticities appear especially large if one takes intoaccount that they refer to product groups (such asdomestic ciprofloxacin) and not individual drugsoffered by particular firms Their relative magni-tudes are also intuitive the drug with the largestmarket share and a relatively high price (domesticciprofloxacin) appears to be one of the least elas-tic In contrast foreign ciprofloxacin is highlyprice elastic this is plausible as ciprofloxacindrugs offered by subsidiaries of multinationalsface the stiffest competition in this market seg-ment from approximately 75 Indian firms offeringthe same molecule

The estimated expenditure elasticities appear inthe last column These are all positive indicatingthat the demand for all product groups is normalThe remaining cells display the estimated cross-price elasticities As one might perhaps expect forproducts within a therapeutic subsegment theseare mostly positive Out of a total of 49 priceelasticities we estimate there are six that do notconform to expectations these are the cross-price elasticities between different foreign prod-

uct groups which are estimated to be negativeand significant33 Fortunately these elasticitieshave negligible impact on the welfare analysisgiven that our counterfactuals focus on theeffect of withdrawing one or more domesticproducts from the market the most relevantelasticities are the ones that capture the responseof various product group shares to a change in theprice of one or more domestic groups these elas-ticities are the cross-price elasticities between var-ious domestic groups and the ones betweendomestic and foreign groups which are plausibleand precisely estimated34

Regarding these elasticities a striking aspectof our estimates is how large positive andsignificant the cross-price elasticities betweendifferent domestic product groups aremdashin factfor norfloxacin and ofloxacin we estimate thatdomestic product groups containing differentmolecules are closer substitutes for one anotherthan product groups that contain the same mol-ecule but are produced by foreign firms Incontrast for ciprofloxacin (the molecule withthe largest revenue share) we estimate a largepositive cross-price elasticity between the do-mestic and foreign versions

The fact that domestic products appear to beclose substitutes for other domestic productsthat contain different molecules truly representsan ldquoempiricalrdquo finding in the sense that we do

32 In developed economies elasticities of this magnitude havetypically been found only for generic drugs (and even then onlyrarely) or among consumers who lack health insurance

33 While these elasticities are clearly counterintuitive theyare not inconsistent with the underlying demand system whichimposes no restrictions on the sign of the cross-price elastici-ties We do not have a good explanation of why these elastic-ities are estimated to be negative A potential explanation isthat the shares of the foreign products are very small giventhis we observe very few consumers switching from one (verysmall) foreign group to another (very small) foreign groupwhen the price of the first foreign group goes up and hence theinference is not very reliable in this case

34 The cross-price elasticities between foreign drug groupscontaining different molecules will also have an effect on thewelfare estimates given that the withdrawal of domestic prod-ucts will generally lead to changes in the prices of all foreignproducts and market shares will be reallocated from someforeign products to other foreign products based on the newprices However this reallocation from ldquoforeign to foreignrdquo istruly second order in our case compared to the reallocationfrom ldquodomestic to domesticrdquo and ldquodomestic to foreignrdquo prod-ucts In scenarios in which we consider patent enforcementaccompanied by strict price regulation the cross-price elastic-ities between foreign product groups are in fact completelyirrelevant as the prices of foreign products are not allowed toincrease in these cases Still our welfare loss estimates remainsubstantial drivenmdashas beforemdashby the loss of product variety

1497VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

not impose it through any of our assumptionsregarding the demand function The questionthat naturally arises then is what might explainthis finding While we cannot formally addressthis question anecdotal accounts in various in-dustry studies suggest that the explanation maylie in the differences between domestic andforeign firms in the structure and coverage ofretail distribution networks

Distribution networks for pharmaceuticals inIndia are typically organized in a hierarchicalfashion Pharmaceutical companies deal mainlywith carrying and forwarding (CampF) agents inmany instances regionally based who each sup-ply a network of stockists (wholesalers) Thesestockists in turn deal with the retail pharma-cists through whom retail sales ultimately oc-cur35 The market share enjoyed by a particularpharmaceutical product therefore depends inpart on the number of retail pharmacists who

stock the product And it is here that thereappears to be a distinction between domesticfirms and multinational subsidiaries In particu-lar the retail reach of domestic firms as agroup tends to be much more comprehensivethan that of multinational subsidiaries (IndianCredit Rating Agency (ICRA) 1999)36

There appear to be two reasons for this Thefirst is that many of the larger Indian firmsbecause they have a much larger portfolio ofproducts over which to spread the associatedfixed costs typically have more extensive net-works of medical representatives The second issimply that there are many more domestic firms(and products) on the market At the retail levelthis would imply that local pharmacists mightbe more likely to stock domestic products con-taining two different molecules say ciprofloxa-cin and norfloxacin than they would domesticand foreign versions of the same molecule Tothe extent that patients (or their doctors) arewilling to substitute across molecules in order tosave on transport or search costs (eg going toanother pharmacy to check whether a particular

35 There are estimated to be some 300000 retail pharma-cists in India On average stockists deal with about 75 retailers(ICRA 1999) There are naturally variations in this structureand a host of specific exclusive dealing and other arrangementsexists in practice Pharmaceutical firms also maintain networksof medical representatives whose main function is to marketthe companyrsquos products to doctors who do the actual prescrib-ing of drugs In some instances firms do sell directly to thedoctors who then become the ldquoretailerrdquo as far as patients areconcerned but these are relatively rare

36 These differences were also highlighted in conversa-tions one of the authors had with CEOs and managingdirectors of several pharmaceutical firms as part of a sep-arate study

TABLE 6AmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN THE NORTHERN REGION

Product group

Elasticity with respect to

Prices of foreign productgroups Prices of domestic product groups Overall

quinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 557 013dagger 015 401 011dagger 011dagger 016 137(179) (007) (007) (184) (006) (006) (006) (029)

Foreign norfloxacin 427dagger 045 427dagger 350dagger 602 451 465 220(242) (112) (242) (210) (623) (184) (183) (105)

Foreign ofloxacin 011 010dagger 138 009 009dagger 023 011 116(005) (005) (031) (027) (005) (028) (004) (017)

Domestic ciprofloxacin 018 001 001 168 008 008 010 117(008) (000) (001) (023) (002) (002) (002) (003)

Domestic norfloxacin 004 003 004 058 223 042 040 073(001) (003) (001) (017) (011) (004) (003) (009)

Domestic ofloxacin 005 005 011 077 074 342 074 089(002) (002) (013) (028) (008) (025) (008) (021)

Domestic sparfloxacin 007 004 007 115 063 063 288 028(002) (001) (002) (015) (006) (006) (017) (012)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent significance level and dagger (dagger) denotes significance at the 10-percent level

1498 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE 6BmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN OTHER REGIONS

Demand patterns in the eastern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 594 014dagger 016 431 013 011 017 140(195) (008) (008) (201) (006) (007) (007) (031)

Foreign norfloxacin 329dagger 058 329dagger 264dagger 460 346 359 192(180) (083) (179) (153) (451) (138) (139) (082)

Foreign ofloxacin 012 010dagger 140 010 010 024 013 117(006) (006) (034) (031) (005) (030) (005) (019)

Domestic ciprofloxacin 018 001 001 172 009 008 011 117(008) (000) (002) (027) (002) (003) (002) (003)

Domestic norfloxacin 004 004 004 068 242 049 046 070(002) (004) (002) (020) (013) (005) (004) (010)

Domestic ofloxacin 004dagger 004 009 061 060 295 060 092(002) (001) (010) (028) (007) (022) (007) (017)

Domestic sparfloxacin 005 003 005 092 048 051 251 043(002) (001) (001) (017) (005) (006) (016) (010)

Demand patterns in the western region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 427 009dagger 011 286 008 006 011 126(129) (005) (005) (133) (004) (005) (004) (021)

Foreign norfloxacin 714dagger 008 710dagger 594dagger 995 742 770 299dagger

(369) (183) (368) (328) (1015) (297) (296) (178)Foreign ofloxacin 013 012dagger 145 009 012 026 013 119

(007) (006) (037) (029) (005) (033) (005) (021)Domestic ciprofloxacin 019 001 000 174 009 008 011 118

(008) (000) (001) (024) (002) (003) (002) (004)Domestic norfloxacin 004 004 004 067 243 050 047 069

(002) (004) (001) (018) (014) (005) (004) (010)Domestic ofloxacin 003 003 007 048 048 257 048 093

(002) (001) (008) (022) (005) (015) (004) (014)Domestic sparfloxacin 006 003 005 083 046 049 241 046

(002) (001) (001) (014) (004) (005) (012) (009)

Demand patterns in the southern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 560 013dagger 015 402 011 010 017 137(182) (007) (007) (187) (006) (006) (007) (029)

Foreign norfloxacin 533dagger 031 531dagger 431dagger 748 559 584 249dagger

(301) (134) (300) (245) (734) (219) (219) (128)Foreign ofloxacin 014 012dagger 148 011 012 029 016 120

(007) (007) (039) (033) (006) (035) (006) (022)Domestic ciprofloxacin 017 001 000 169 008 007 011 116

(008) (000) (001) (025) (002) (003) (002) (003)Domestic norfloxacin 004 003 003 059 225 043 040 073

(001) (003) (001) (019) (012) (004) (004) (009)Domestic ofloxacin 004dagger 004 009 062 060 296 060 091

(002) (001) (010) (026) (006) (019) (007) (017)Domestic sparfloxacin 008 005 008 138 073 075 323 016

(002) (001) (002) (018) (007) (008) (021) (014)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at 5-percent level anddagger (dagger) denotes significance at the 10-percent level

1499VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

foreign product is in stock) in aggregate datawe would expect to find precisely the substitu-tion patterns that we report in Table 6

Whether the particular explanation we provideabove is the correct one the high degree of sub-stitutability between domestic product groupsturns out to have important implications for thewelfare calculations We discuss these in moredetail below when we present the results of thecounterfactual welfare analysis Another elasticitywith important implications for the counterfactu-als is the price elasticity for the quinolone subseg-ment as a whole which indicates how likelyconsumers are to switch to other antibioticsgroups when faced with a price increase for quin-olones This elasticity is computed on the basis ofthe results in Table A3 and it is at 111 (stan-dard error 024) this is large in magnitudebutmdashas expectedmdashsmaller in absolute value thanthe own-price elasticities of the product groupswithin the quinolone subsegment

The results in Tables 6A and 6B are based onour preferred specification discussed in SectionII In Tables A4 to A6 in the Appendix weexperiment with some alternative specificationsTables A4(a)ndashA4(c) correspond to a specificationthat includes in addition to product-group-specificregional fixed effects product-group-specific (andfor the upper level antibiotics-segment-specific)seasonal effects We distinguish among three sea-sonsmdashthe summer monsoon and wintermdashandreport the unconditional demand elasticities for

the northern region for each of these seasons Asevident from the tables in the Appendix our elas-ticity estimates are robust to the inclusion of sea-sonal effects The demand elasticities in Table A5are based on estimation of the demand system byOLS Compared to the elasticities obtained by IVthe OLS elasticities are smaller in absolute valueimplying that welfare calculations based on theOLS estimates would produce larger welfare lossestimates Nevertheless some of the patterns re-garding the cross-price elasticities discussed ear-lier are also evident in the OLS results inparticular the cross-price elasticities between dif-ferent domestic product groups are all positivelarge and significant and in most instances largerthan the cross-price elasticities between drugs thatcontain the same molecule but are produced byfirms of different domesticforeign status Theclose substitutability of domestic products indi-cated by both the OLS and IV estimates seems tobe one of the most robust findings of the paper

B Cost and Markup Estimates

Table 7 displays the marginal costs markupsand profits implied by the price elasticity esti-mates of Tables 6A and 6B for each of the sevenproduct groups Given that our regional effectsimply different price elasticities for each regionour marginal cost and markup estimates alsodiffer by region Given however that based on

TABLE 7mdashUPPER AND LOWER BOUNDS FOR MARGINAL COST MARKUP AND ANNUAL PROFIT BY PRODUCT GROUPS WITHIN

THE QUINOLONE SUBSEGMENT

Product groupLower boundfor MC (Rs)

Upper boundfor markup

Upper boundfor profit(Rs mill)

Upper boundfor MC (Rs)

Lower boundfor markup

Lower boundfor profit

(Rs)

Foreign ciprofloxacin 83 19 269 103 0 00(123) (012) (1655)

Foreign norfloxacin NA NA NA 53 0 00Foreign ofloxacin 323 70 1061 1085 0 00

(2316) (021) (3185)Domestic ciprofloxacin 47 59 17019 112 0 00

(114) (010) (29858)Domestic norfloxacin 52 43 2807 90 0 00

(020) (002) (1532)Domestic ofloxacin 587 34 1612 901 0 00

(218) (002) (1280)Domestic sparfloxacin 495 37 1985 788 0 00

(157) (002) (1100)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level Estimated lower bound forforeign norfloxacinrsquos marginal cost is negative since the estimated price elasticity is less than one in absolute value

1500 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Tables 6A and 6B the price elasticities do notseem to differ substantially across regions wereport for ease of exposition only the nationalaverages for marginal costs and markups inTable 7

Table 7 has two parts In the left part (firstthree columns) we report the lower bound formarginal cost and the corresponding upperbound for markup and upper bound for totalannual profit for each product group Thesenumbers are based on the price elasticities weobtained from estimating the two-level AIDSdemand system Since we do not have a reliableestimate for the price elasticity of foreign nor-floxacin (the point estimate is negative but lessthan one in absolute value and insignificant)we cannot compute the lower bound for mar-ginal cost in this case It is important to note thatthese estimates do not reflect either the actualmarginal cost or the actual markup for thesedrugs both because the existence of price reg-ulation implies that the unconstrained first-orderconditions are unlikely to hold each period andbecause our aggregation across firms of thesame domesticforeign status supplying thesame molecule makes the interpretation of theseestimates problematic In particular the fact thatwe ignore competition among firms within eachproduct group implies that our estimates willtend to overstate market power These numberswill prove useful however in the counterfac-tual simulations as they can give us a sense ofhow large the maximum profit gains for multi-national firms and the maximum profit lossesfor domestic firms are likely to be under patentenforcement The right part of the table (lastthree columns) reports the upper bound for mar-ginal cost which we obtain by simply taking themarginal cost to be equal to the observed priceWe do not report standard errors in this casesince the numbers are based on actual data Themarkups corresponding to these marginal costupper bounds are of course zero We conductthe counterfactual simulations using both thelower and upper bounds for marginal costs

The (upper bounds for) markups on the leftside of the table are generally plausible Thedomestic ciprofloxacin product group that dom-inates the quinolone subsegment and for whichwe documented high prices a high marketshare and a relatively low elasticity of demandenjoys one of the highest markup upper bounds(60 percent) and accounts for nearly 70 percent

of all profits derived within the subsegmentForeign ciprofloxacin on the other hand whichfaces the stiffest competition from domesticfirms and for which we estimated a highly elas-tic demand has the lowest markup upper bound(19 percent)

C Counterfactual Estimates of the Impact onPrices and Welfare

With estimates of the key demand and costparameters in hand we turn to the counterfac-tuals We consider the five separate scenarioslisted in the previous section All of the scenar-ios involve the withdrawal of one or more of thedomestic product groups from the market Ta-ble 8 displays our estimates of the consumerwelfare losses that result under the differentscenarios The losses are expressed in billion ofrupees per year All numbers presented in Ta-ble 8 and subsequent tables are based on usingthe lower bounds for marginal cost and upperbounds for markup in the simulations As dis-cussed earlier these numbers are the more in-teresting to work with since they give us anupper bound for the changes in the profits ofdomestic and foreign firms that would resultfrom patent enforcement In the Appendix wealso present results based on using the upperbounds for marginal costs in the simulations inwhich case the pre-TRIPS profits of domesticand foreign firms are zero In all cases marginalcosts are assumed to be constant in outputWhile naturally the profit implications differdepending on whether one uses the upper orlower bounds for marginal costs (firm profits arezero if one assumes the upper bound of mar-ginal cost in which case price equals marginalcost) the estimated consumer welfare losses aresimilar in the two cases We discuss these re-sults in more detail at the end of this section

The first column presents our estimates of theconsumer welfare losses attributable to the pureloss of product variety effect where we fix theprices of all remaining products as well as theoverall expenditure on quinolones while with-drawing one or more of the domestic productgroups Note that in our initial specification ofthe demand system had we not allowed for thepossibility that consumers might differentiatebetween domestic and foreign products evenwhen they contain the same molecule this

1501VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

particular component of the loss of consumerwelfare would not have arisen

The estimates reported in the second columnincorporate the expenditure-switching effect ontop of the loss of product variety Here basedupon the price elasticity estimates from thehigher-level AIDS system we adjust (down-ward) the expenditures allocated to the quino-lone subsegment as the composite price ofquinolones effectively increases as a consequenceof the higher virtual prices of the domestic productgroups that are withdrawn from the market Be-cause the estimates in this column are generatedassuming that the prices of the products that re-main in the market are not adjusted upward theyprovide a sense of what consumer losses would beif the introduction of product patents were coupledwith strict price regulation aimed at maintainingprices at prepatent levels Alternatively they canbe thought of as the relevant welfare numbers ifintense competition among firms within the re-maining product groups kept the prices of theproducts that were still offered in the market closeto the firmsrsquo marginal costs

The last column of Table 8 displays the esti-mated consumer welfare losses when bothcross-segment expenditure-switching and with-in-segment upward price adjustments are takeninto account

If we compare the results across the firstsecond and third columns all the counterfac-tual scenarios produce qualitatively similar pat-terns patterns that are consistent with what we

would expect Starting from the initial loss ofwelfare attributable to the loss of product vari-ety the option of switching expenditures outof the quinolone subsegment to other subseg-ments mitigates some of the initial welfareloss But if we then incorporate the upwardprice adjustments that result in response to thereduced competition the welfare losses aremagnified

Of particular interest from a policy perspec-tive are the relative magnitudes of these threeeffects which are similar under all the counter-factual scenarios though absolute levels varyconsiderably First despite the fact that the de-mand for quinolones is quite sensitive to thecomposite price of quinolonesmdashthe upper levelprice elasticity is 111mdashthe cross-subsegmentexpenditure-switching effects are in all thecases small (in absolute value terms) relative tothe other two effects For instance under thescenario where all the domestic quinolone prod-uct groups are withdrawn from the market theoverall consumer welfare loss of Rs 1781 bil-lion per year can be decomposed into an initialloss of Rs 1176 billion (66 percent) attribut-able to the loss of product variety a slightreduction in this initial loss of Rs 041 billion(2 percent) from Rs 1176 billion to Rs1135 billion because of expenditure switchingand a subsequent additional loss of Rs 646billion (36 percent) from Rs 1135 billion toRs 1781 billion because of the reduced com-petition and consequent price increases

TABLE 8mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure loss ofvariety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 732(087) (089) (146)

Only ofloxacin 008 008 023(008) (008) (010)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1253(177) (180) (415)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1058(142) (145) (331)

All four domestic quinolones products 1176dagger 1135dagger 1781(643) (634) (1270)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent significance level and dagger (dagger)denotes significance at the 10-percent level

1502 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 22: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

not impose it through any of our assumptionsregarding the demand function The questionthat naturally arises then is what might explainthis finding While we cannot formally addressthis question anecdotal accounts in various in-dustry studies suggest that the explanation maylie in the differences between domestic andforeign firms in the structure and coverage ofretail distribution networks

Distribution networks for pharmaceuticals inIndia are typically organized in a hierarchicalfashion Pharmaceutical companies deal mainlywith carrying and forwarding (CampF) agents inmany instances regionally based who each sup-ply a network of stockists (wholesalers) Thesestockists in turn deal with the retail pharma-cists through whom retail sales ultimately oc-cur35 The market share enjoyed by a particularpharmaceutical product therefore depends inpart on the number of retail pharmacists who

stock the product And it is here that thereappears to be a distinction between domesticfirms and multinational subsidiaries In particu-lar the retail reach of domestic firms as agroup tends to be much more comprehensivethan that of multinational subsidiaries (IndianCredit Rating Agency (ICRA) 1999)36

There appear to be two reasons for this Thefirst is that many of the larger Indian firmsbecause they have a much larger portfolio ofproducts over which to spread the associatedfixed costs typically have more extensive net-works of medical representatives The second issimply that there are many more domestic firms(and products) on the market At the retail levelthis would imply that local pharmacists mightbe more likely to stock domestic products con-taining two different molecules say ciprofloxa-cin and norfloxacin than they would domesticand foreign versions of the same molecule Tothe extent that patients (or their doctors) arewilling to substitute across molecules in order tosave on transport or search costs (eg going toanother pharmacy to check whether a particular

35 There are estimated to be some 300000 retail pharma-cists in India On average stockists deal with about 75 retailers(ICRA 1999) There are naturally variations in this structureand a host of specific exclusive dealing and other arrangementsexists in practice Pharmaceutical firms also maintain networksof medical representatives whose main function is to marketthe companyrsquos products to doctors who do the actual prescrib-ing of drugs In some instances firms do sell directly to thedoctors who then become the ldquoretailerrdquo as far as patients areconcerned but these are relatively rare

36 These differences were also highlighted in conversa-tions one of the authors had with CEOs and managingdirectors of several pharmaceutical firms as part of a sep-arate study

TABLE 6AmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN THE NORTHERN REGION

Product group

Elasticity with respect to

Prices of foreign productgroups Prices of domestic product groups Overall

quinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 557 013dagger 015 401 011dagger 011dagger 016 137(179) (007) (007) (184) (006) (006) (006) (029)

Foreign norfloxacin 427dagger 045 427dagger 350dagger 602 451 465 220(242) (112) (242) (210) (623) (184) (183) (105)

Foreign ofloxacin 011 010dagger 138 009 009dagger 023 011 116(005) (005) (031) (027) (005) (028) (004) (017)

Domestic ciprofloxacin 018 001 001 168 008 008 010 117(008) (000) (001) (023) (002) (002) (002) (003)

Domestic norfloxacin 004 003 004 058 223 042 040 073(001) (003) (001) (017) (011) (004) (003) (009)

Domestic ofloxacin 005 005 011 077 074 342 074 089(002) (002) (013) (028) (008) (025) (008) (021)

Domestic sparfloxacin 007 004 007 115 063 063 288 028(002) (001) (002) (015) (006) (006) (017) (012)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent significance level and dagger (dagger) denotes significance at the 10-percent level

1498 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE 6BmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN OTHER REGIONS

Demand patterns in the eastern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 594 014dagger 016 431 013 011 017 140(195) (008) (008) (201) (006) (007) (007) (031)

Foreign norfloxacin 329dagger 058 329dagger 264dagger 460 346 359 192(180) (083) (179) (153) (451) (138) (139) (082)

Foreign ofloxacin 012 010dagger 140 010 010 024 013 117(006) (006) (034) (031) (005) (030) (005) (019)

Domestic ciprofloxacin 018 001 001 172 009 008 011 117(008) (000) (002) (027) (002) (003) (002) (003)

Domestic norfloxacin 004 004 004 068 242 049 046 070(002) (004) (002) (020) (013) (005) (004) (010)

Domestic ofloxacin 004dagger 004 009 061 060 295 060 092(002) (001) (010) (028) (007) (022) (007) (017)

Domestic sparfloxacin 005 003 005 092 048 051 251 043(002) (001) (001) (017) (005) (006) (016) (010)

Demand patterns in the western region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 427 009dagger 011 286 008 006 011 126(129) (005) (005) (133) (004) (005) (004) (021)

Foreign norfloxacin 714dagger 008 710dagger 594dagger 995 742 770 299dagger

(369) (183) (368) (328) (1015) (297) (296) (178)Foreign ofloxacin 013 012dagger 145 009 012 026 013 119

(007) (006) (037) (029) (005) (033) (005) (021)Domestic ciprofloxacin 019 001 000 174 009 008 011 118

(008) (000) (001) (024) (002) (003) (002) (004)Domestic norfloxacin 004 004 004 067 243 050 047 069

(002) (004) (001) (018) (014) (005) (004) (010)Domestic ofloxacin 003 003 007 048 048 257 048 093

(002) (001) (008) (022) (005) (015) (004) (014)Domestic sparfloxacin 006 003 005 083 046 049 241 046

(002) (001) (001) (014) (004) (005) (012) (009)

Demand patterns in the southern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 560 013dagger 015 402 011 010 017 137(182) (007) (007) (187) (006) (006) (007) (029)

Foreign norfloxacin 533dagger 031 531dagger 431dagger 748 559 584 249dagger

(301) (134) (300) (245) (734) (219) (219) (128)Foreign ofloxacin 014 012dagger 148 011 012 029 016 120

(007) (007) (039) (033) (006) (035) (006) (022)Domestic ciprofloxacin 017 001 000 169 008 007 011 116

(008) (000) (001) (025) (002) (003) (002) (003)Domestic norfloxacin 004 003 003 059 225 043 040 073

(001) (003) (001) (019) (012) (004) (004) (009)Domestic ofloxacin 004dagger 004 009 062 060 296 060 091

(002) (001) (010) (026) (006) (019) (007) (017)Domestic sparfloxacin 008 005 008 138 073 075 323 016

(002) (001) (002) (018) (007) (008) (021) (014)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at 5-percent level anddagger (dagger) denotes significance at the 10-percent level

1499VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

foreign product is in stock) in aggregate datawe would expect to find precisely the substitu-tion patterns that we report in Table 6

Whether the particular explanation we provideabove is the correct one the high degree of sub-stitutability between domestic product groupsturns out to have important implications for thewelfare calculations We discuss these in moredetail below when we present the results of thecounterfactual welfare analysis Another elasticitywith important implications for the counterfactu-als is the price elasticity for the quinolone subseg-ment as a whole which indicates how likelyconsumers are to switch to other antibioticsgroups when faced with a price increase for quin-olones This elasticity is computed on the basis ofthe results in Table A3 and it is at 111 (stan-dard error 024) this is large in magnitudebutmdashas expectedmdashsmaller in absolute value thanthe own-price elasticities of the product groupswithin the quinolone subsegment

The results in Tables 6A and 6B are based onour preferred specification discussed in SectionII In Tables A4 to A6 in the Appendix weexperiment with some alternative specificationsTables A4(a)ndashA4(c) correspond to a specificationthat includes in addition to product-group-specificregional fixed effects product-group-specific (andfor the upper level antibiotics-segment-specific)seasonal effects We distinguish among three sea-sonsmdashthe summer monsoon and wintermdashandreport the unconditional demand elasticities for

the northern region for each of these seasons Asevident from the tables in the Appendix our elas-ticity estimates are robust to the inclusion of sea-sonal effects The demand elasticities in Table A5are based on estimation of the demand system byOLS Compared to the elasticities obtained by IVthe OLS elasticities are smaller in absolute valueimplying that welfare calculations based on theOLS estimates would produce larger welfare lossestimates Nevertheless some of the patterns re-garding the cross-price elasticities discussed ear-lier are also evident in the OLS results inparticular the cross-price elasticities between dif-ferent domestic product groups are all positivelarge and significant and in most instances largerthan the cross-price elasticities between drugs thatcontain the same molecule but are produced byfirms of different domesticforeign status Theclose substitutability of domestic products indi-cated by both the OLS and IV estimates seems tobe one of the most robust findings of the paper

B Cost and Markup Estimates

Table 7 displays the marginal costs markupsand profits implied by the price elasticity esti-mates of Tables 6A and 6B for each of the sevenproduct groups Given that our regional effectsimply different price elasticities for each regionour marginal cost and markup estimates alsodiffer by region Given however that based on

TABLE 7mdashUPPER AND LOWER BOUNDS FOR MARGINAL COST MARKUP AND ANNUAL PROFIT BY PRODUCT GROUPS WITHIN

THE QUINOLONE SUBSEGMENT

Product groupLower boundfor MC (Rs)

Upper boundfor markup

Upper boundfor profit(Rs mill)

Upper boundfor MC (Rs)

Lower boundfor markup

Lower boundfor profit

(Rs)

Foreign ciprofloxacin 83 19 269 103 0 00(123) (012) (1655)

Foreign norfloxacin NA NA NA 53 0 00Foreign ofloxacin 323 70 1061 1085 0 00

(2316) (021) (3185)Domestic ciprofloxacin 47 59 17019 112 0 00

(114) (010) (29858)Domestic norfloxacin 52 43 2807 90 0 00

(020) (002) (1532)Domestic ofloxacin 587 34 1612 901 0 00

(218) (002) (1280)Domestic sparfloxacin 495 37 1985 788 0 00

(157) (002) (1100)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level Estimated lower bound forforeign norfloxacinrsquos marginal cost is negative since the estimated price elasticity is less than one in absolute value

1500 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Tables 6A and 6B the price elasticities do notseem to differ substantially across regions wereport for ease of exposition only the nationalaverages for marginal costs and markups inTable 7

Table 7 has two parts In the left part (firstthree columns) we report the lower bound formarginal cost and the corresponding upperbound for markup and upper bound for totalannual profit for each product group Thesenumbers are based on the price elasticities weobtained from estimating the two-level AIDSdemand system Since we do not have a reliableestimate for the price elasticity of foreign nor-floxacin (the point estimate is negative but lessthan one in absolute value and insignificant)we cannot compute the lower bound for mar-ginal cost in this case It is important to note thatthese estimates do not reflect either the actualmarginal cost or the actual markup for thesedrugs both because the existence of price reg-ulation implies that the unconstrained first-orderconditions are unlikely to hold each period andbecause our aggregation across firms of thesame domesticforeign status supplying thesame molecule makes the interpretation of theseestimates problematic In particular the fact thatwe ignore competition among firms within eachproduct group implies that our estimates willtend to overstate market power These numberswill prove useful however in the counterfac-tual simulations as they can give us a sense ofhow large the maximum profit gains for multi-national firms and the maximum profit lossesfor domestic firms are likely to be under patentenforcement The right part of the table (lastthree columns) reports the upper bound for mar-ginal cost which we obtain by simply taking themarginal cost to be equal to the observed priceWe do not report standard errors in this casesince the numbers are based on actual data Themarkups corresponding to these marginal costupper bounds are of course zero We conductthe counterfactual simulations using both thelower and upper bounds for marginal costs

The (upper bounds for) markups on the leftside of the table are generally plausible Thedomestic ciprofloxacin product group that dom-inates the quinolone subsegment and for whichwe documented high prices a high marketshare and a relatively low elasticity of demandenjoys one of the highest markup upper bounds(60 percent) and accounts for nearly 70 percent

of all profits derived within the subsegmentForeign ciprofloxacin on the other hand whichfaces the stiffest competition from domesticfirms and for which we estimated a highly elas-tic demand has the lowest markup upper bound(19 percent)

C Counterfactual Estimates of the Impact onPrices and Welfare

With estimates of the key demand and costparameters in hand we turn to the counterfac-tuals We consider the five separate scenarioslisted in the previous section All of the scenar-ios involve the withdrawal of one or more of thedomestic product groups from the market Ta-ble 8 displays our estimates of the consumerwelfare losses that result under the differentscenarios The losses are expressed in billion ofrupees per year All numbers presented in Ta-ble 8 and subsequent tables are based on usingthe lower bounds for marginal cost and upperbounds for markup in the simulations As dis-cussed earlier these numbers are the more in-teresting to work with since they give us anupper bound for the changes in the profits ofdomestic and foreign firms that would resultfrom patent enforcement In the Appendix wealso present results based on using the upperbounds for marginal costs in the simulations inwhich case the pre-TRIPS profits of domesticand foreign firms are zero In all cases marginalcosts are assumed to be constant in outputWhile naturally the profit implications differdepending on whether one uses the upper orlower bounds for marginal costs (firm profits arezero if one assumes the upper bound of mar-ginal cost in which case price equals marginalcost) the estimated consumer welfare losses aresimilar in the two cases We discuss these re-sults in more detail at the end of this section

The first column presents our estimates of theconsumer welfare losses attributable to the pureloss of product variety effect where we fix theprices of all remaining products as well as theoverall expenditure on quinolones while with-drawing one or more of the domestic productgroups Note that in our initial specification ofthe demand system had we not allowed for thepossibility that consumers might differentiatebetween domestic and foreign products evenwhen they contain the same molecule this

1501VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

particular component of the loss of consumerwelfare would not have arisen

The estimates reported in the second columnincorporate the expenditure-switching effect ontop of the loss of product variety Here basedupon the price elasticity estimates from thehigher-level AIDS system we adjust (down-ward) the expenditures allocated to the quino-lone subsegment as the composite price ofquinolones effectively increases as a consequenceof the higher virtual prices of the domestic productgroups that are withdrawn from the market Be-cause the estimates in this column are generatedassuming that the prices of the products that re-main in the market are not adjusted upward theyprovide a sense of what consumer losses would beif the introduction of product patents were coupledwith strict price regulation aimed at maintainingprices at prepatent levels Alternatively they canbe thought of as the relevant welfare numbers ifintense competition among firms within the re-maining product groups kept the prices of theproducts that were still offered in the market closeto the firmsrsquo marginal costs

The last column of Table 8 displays the esti-mated consumer welfare losses when bothcross-segment expenditure-switching and with-in-segment upward price adjustments are takeninto account

If we compare the results across the firstsecond and third columns all the counterfac-tual scenarios produce qualitatively similar pat-terns patterns that are consistent with what we

would expect Starting from the initial loss ofwelfare attributable to the loss of product vari-ety the option of switching expenditures outof the quinolone subsegment to other subseg-ments mitigates some of the initial welfareloss But if we then incorporate the upwardprice adjustments that result in response to thereduced competition the welfare losses aremagnified

Of particular interest from a policy perspec-tive are the relative magnitudes of these threeeffects which are similar under all the counter-factual scenarios though absolute levels varyconsiderably First despite the fact that the de-mand for quinolones is quite sensitive to thecomposite price of quinolonesmdashthe upper levelprice elasticity is 111mdashthe cross-subsegmentexpenditure-switching effects are in all thecases small (in absolute value terms) relative tothe other two effects For instance under thescenario where all the domestic quinolone prod-uct groups are withdrawn from the market theoverall consumer welfare loss of Rs 1781 bil-lion per year can be decomposed into an initialloss of Rs 1176 billion (66 percent) attribut-able to the loss of product variety a slightreduction in this initial loss of Rs 041 billion(2 percent) from Rs 1176 billion to Rs1135 billion because of expenditure switchingand a subsequent additional loss of Rs 646billion (36 percent) from Rs 1135 billion toRs 1781 billion because of the reduced com-petition and consequent price increases

TABLE 8mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure loss ofvariety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 732(087) (089) (146)

Only ofloxacin 008 008 023(008) (008) (010)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1253(177) (180) (415)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1058(142) (145) (331)

All four domestic quinolones products 1176dagger 1135dagger 1781(643) (634) (1270)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent significance level and dagger (dagger)denotes significance at the 10-percent level

1502 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 23: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

TABLE 6BmdashDEMAND PATTERNS WITHIN THE QUINOLONES SUBSEGMENTUNCONDITIONAL PRICE AND EXPENDITURE ELASTICITIES IN OTHER REGIONS

Demand patterns in the eastern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 594 014dagger 016 431 013 011 017 140(195) (008) (008) (201) (006) (007) (007) (031)

Foreign norfloxacin 329dagger 058 329dagger 264dagger 460 346 359 192(180) (083) (179) (153) (451) (138) (139) (082)

Foreign ofloxacin 012 010dagger 140 010 010 024 013 117(006) (006) (034) (031) (005) (030) (005) (019)

Domestic ciprofloxacin 018 001 001 172 009 008 011 117(008) (000) (002) (027) (002) (003) (002) (003)

Domestic norfloxacin 004 004 004 068 242 049 046 070(002) (004) (002) (020) (013) (005) (004) (010)

Domestic ofloxacin 004dagger 004 009 061 060 295 060 092(002) (001) (010) (028) (007) (022) (007) (017)

Domestic sparfloxacin 005 003 005 092 048 051 251 043(002) (001) (001) (017) (005) (006) (016) (010)

Demand patterns in the western region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 427 009dagger 011 286 008 006 011 126(129) (005) (005) (133) (004) (005) (004) (021)

Foreign norfloxacin 714dagger 008 710dagger 594dagger 995 742 770 299dagger

(369) (183) (368) (328) (1015) (297) (296) (178)Foreign ofloxacin 013 012dagger 145 009 012 026 013 119

(007) (006) (037) (029) (005) (033) (005) (021)Domestic ciprofloxacin 019 001 000 174 009 008 011 118

(008) (000) (001) (024) (002) (003) (002) (004)Domestic norfloxacin 004 004 004 067 243 050 047 069

(002) (004) (001) (018) (014) (005) (004) (010)Domestic ofloxacin 003 003 007 048 048 257 048 093

(002) (001) (008) (022) (005) (015) (004) (014)Domestic sparfloxacin 006 003 005 083 046 049 241 046

(002) (001) (001) (014) (004) (005) (012) (009)

Demand patterns in the southern region

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo pricesQuinolonesexpenditureCipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 560 013dagger 015 402 011 010 017 137(182) (007) (007) (187) (006) (006) (007) (029)

Foreign norfloxacin 533dagger 031 531dagger 431dagger 748 559 584 249dagger

(301) (134) (300) (245) (734) (219) (219) (128)Foreign ofloxacin 014 012dagger 148 011 012 029 016 120

(007) (007) (039) (033) (006) (035) (006) (022)Domestic ciprofloxacin 017 001 000 169 008 007 011 116

(008) (000) (001) (025) (002) (003) (002) (003)Domestic norfloxacin 004 003 003 059 225 043 040 073

(001) (003) (001) (019) (012) (004) (004) (009)Domestic ofloxacin 004dagger 004 009 062 060 296 060 091

(002) (001) (010) (026) (006) (019) (007) (017)Domestic sparfloxacin 008 005 008 138 073 075 323 016

(002) (001) (002) (018) (007) (008) (021) (014)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at 5-percent level anddagger (dagger) denotes significance at the 10-percent level

1499VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

foreign product is in stock) in aggregate datawe would expect to find precisely the substitu-tion patterns that we report in Table 6

Whether the particular explanation we provideabove is the correct one the high degree of sub-stitutability between domestic product groupsturns out to have important implications for thewelfare calculations We discuss these in moredetail below when we present the results of thecounterfactual welfare analysis Another elasticitywith important implications for the counterfactu-als is the price elasticity for the quinolone subseg-ment as a whole which indicates how likelyconsumers are to switch to other antibioticsgroups when faced with a price increase for quin-olones This elasticity is computed on the basis ofthe results in Table A3 and it is at 111 (stan-dard error 024) this is large in magnitudebutmdashas expectedmdashsmaller in absolute value thanthe own-price elasticities of the product groupswithin the quinolone subsegment

The results in Tables 6A and 6B are based onour preferred specification discussed in SectionII In Tables A4 to A6 in the Appendix weexperiment with some alternative specificationsTables A4(a)ndashA4(c) correspond to a specificationthat includes in addition to product-group-specificregional fixed effects product-group-specific (andfor the upper level antibiotics-segment-specific)seasonal effects We distinguish among three sea-sonsmdashthe summer monsoon and wintermdashandreport the unconditional demand elasticities for

the northern region for each of these seasons Asevident from the tables in the Appendix our elas-ticity estimates are robust to the inclusion of sea-sonal effects The demand elasticities in Table A5are based on estimation of the demand system byOLS Compared to the elasticities obtained by IVthe OLS elasticities are smaller in absolute valueimplying that welfare calculations based on theOLS estimates would produce larger welfare lossestimates Nevertheless some of the patterns re-garding the cross-price elasticities discussed ear-lier are also evident in the OLS results inparticular the cross-price elasticities between dif-ferent domestic product groups are all positivelarge and significant and in most instances largerthan the cross-price elasticities between drugs thatcontain the same molecule but are produced byfirms of different domesticforeign status Theclose substitutability of domestic products indi-cated by both the OLS and IV estimates seems tobe one of the most robust findings of the paper

B Cost and Markup Estimates

Table 7 displays the marginal costs markupsand profits implied by the price elasticity esti-mates of Tables 6A and 6B for each of the sevenproduct groups Given that our regional effectsimply different price elasticities for each regionour marginal cost and markup estimates alsodiffer by region Given however that based on

TABLE 7mdashUPPER AND LOWER BOUNDS FOR MARGINAL COST MARKUP AND ANNUAL PROFIT BY PRODUCT GROUPS WITHIN

THE QUINOLONE SUBSEGMENT

Product groupLower boundfor MC (Rs)

Upper boundfor markup

Upper boundfor profit(Rs mill)

Upper boundfor MC (Rs)

Lower boundfor markup

Lower boundfor profit

(Rs)

Foreign ciprofloxacin 83 19 269 103 0 00(123) (012) (1655)

Foreign norfloxacin NA NA NA 53 0 00Foreign ofloxacin 323 70 1061 1085 0 00

(2316) (021) (3185)Domestic ciprofloxacin 47 59 17019 112 0 00

(114) (010) (29858)Domestic norfloxacin 52 43 2807 90 0 00

(020) (002) (1532)Domestic ofloxacin 587 34 1612 901 0 00

(218) (002) (1280)Domestic sparfloxacin 495 37 1985 788 0 00

(157) (002) (1100)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level Estimated lower bound forforeign norfloxacinrsquos marginal cost is negative since the estimated price elasticity is less than one in absolute value

1500 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Tables 6A and 6B the price elasticities do notseem to differ substantially across regions wereport for ease of exposition only the nationalaverages for marginal costs and markups inTable 7

Table 7 has two parts In the left part (firstthree columns) we report the lower bound formarginal cost and the corresponding upperbound for markup and upper bound for totalannual profit for each product group Thesenumbers are based on the price elasticities weobtained from estimating the two-level AIDSdemand system Since we do not have a reliableestimate for the price elasticity of foreign nor-floxacin (the point estimate is negative but lessthan one in absolute value and insignificant)we cannot compute the lower bound for mar-ginal cost in this case It is important to note thatthese estimates do not reflect either the actualmarginal cost or the actual markup for thesedrugs both because the existence of price reg-ulation implies that the unconstrained first-orderconditions are unlikely to hold each period andbecause our aggregation across firms of thesame domesticforeign status supplying thesame molecule makes the interpretation of theseestimates problematic In particular the fact thatwe ignore competition among firms within eachproduct group implies that our estimates willtend to overstate market power These numberswill prove useful however in the counterfac-tual simulations as they can give us a sense ofhow large the maximum profit gains for multi-national firms and the maximum profit lossesfor domestic firms are likely to be under patentenforcement The right part of the table (lastthree columns) reports the upper bound for mar-ginal cost which we obtain by simply taking themarginal cost to be equal to the observed priceWe do not report standard errors in this casesince the numbers are based on actual data Themarkups corresponding to these marginal costupper bounds are of course zero We conductthe counterfactual simulations using both thelower and upper bounds for marginal costs

The (upper bounds for) markups on the leftside of the table are generally plausible Thedomestic ciprofloxacin product group that dom-inates the quinolone subsegment and for whichwe documented high prices a high marketshare and a relatively low elasticity of demandenjoys one of the highest markup upper bounds(60 percent) and accounts for nearly 70 percent

of all profits derived within the subsegmentForeign ciprofloxacin on the other hand whichfaces the stiffest competition from domesticfirms and for which we estimated a highly elas-tic demand has the lowest markup upper bound(19 percent)

C Counterfactual Estimates of the Impact onPrices and Welfare

With estimates of the key demand and costparameters in hand we turn to the counterfac-tuals We consider the five separate scenarioslisted in the previous section All of the scenar-ios involve the withdrawal of one or more of thedomestic product groups from the market Ta-ble 8 displays our estimates of the consumerwelfare losses that result under the differentscenarios The losses are expressed in billion ofrupees per year All numbers presented in Ta-ble 8 and subsequent tables are based on usingthe lower bounds for marginal cost and upperbounds for markup in the simulations As dis-cussed earlier these numbers are the more in-teresting to work with since they give us anupper bound for the changes in the profits ofdomestic and foreign firms that would resultfrom patent enforcement In the Appendix wealso present results based on using the upperbounds for marginal costs in the simulations inwhich case the pre-TRIPS profits of domesticand foreign firms are zero In all cases marginalcosts are assumed to be constant in outputWhile naturally the profit implications differdepending on whether one uses the upper orlower bounds for marginal costs (firm profits arezero if one assumes the upper bound of mar-ginal cost in which case price equals marginalcost) the estimated consumer welfare losses aresimilar in the two cases We discuss these re-sults in more detail at the end of this section

The first column presents our estimates of theconsumer welfare losses attributable to the pureloss of product variety effect where we fix theprices of all remaining products as well as theoverall expenditure on quinolones while with-drawing one or more of the domestic productgroups Note that in our initial specification ofthe demand system had we not allowed for thepossibility that consumers might differentiatebetween domestic and foreign products evenwhen they contain the same molecule this

1501VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

particular component of the loss of consumerwelfare would not have arisen

The estimates reported in the second columnincorporate the expenditure-switching effect ontop of the loss of product variety Here basedupon the price elasticity estimates from thehigher-level AIDS system we adjust (down-ward) the expenditures allocated to the quino-lone subsegment as the composite price ofquinolones effectively increases as a consequenceof the higher virtual prices of the domestic productgroups that are withdrawn from the market Be-cause the estimates in this column are generatedassuming that the prices of the products that re-main in the market are not adjusted upward theyprovide a sense of what consumer losses would beif the introduction of product patents were coupledwith strict price regulation aimed at maintainingprices at prepatent levels Alternatively they canbe thought of as the relevant welfare numbers ifintense competition among firms within the re-maining product groups kept the prices of theproducts that were still offered in the market closeto the firmsrsquo marginal costs

The last column of Table 8 displays the esti-mated consumer welfare losses when bothcross-segment expenditure-switching and with-in-segment upward price adjustments are takeninto account

If we compare the results across the firstsecond and third columns all the counterfac-tual scenarios produce qualitatively similar pat-terns patterns that are consistent with what we

would expect Starting from the initial loss ofwelfare attributable to the loss of product vari-ety the option of switching expenditures outof the quinolone subsegment to other subseg-ments mitigates some of the initial welfareloss But if we then incorporate the upwardprice adjustments that result in response to thereduced competition the welfare losses aremagnified

Of particular interest from a policy perspec-tive are the relative magnitudes of these threeeffects which are similar under all the counter-factual scenarios though absolute levels varyconsiderably First despite the fact that the de-mand for quinolones is quite sensitive to thecomposite price of quinolonesmdashthe upper levelprice elasticity is 111mdashthe cross-subsegmentexpenditure-switching effects are in all thecases small (in absolute value terms) relative tothe other two effects For instance under thescenario where all the domestic quinolone prod-uct groups are withdrawn from the market theoverall consumer welfare loss of Rs 1781 bil-lion per year can be decomposed into an initialloss of Rs 1176 billion (66 percent) attribut-able to the loss of product variety a slightreduction in this initial loss of Rs 041 billion(2 percent) from Rs 1176 billion to Rs1135 billion because of expenditure switchingand a subsequent additional loss of Rs 646billion (36 percent) from Rs 1135 billion toRs 1781 billion because of the reduced com-petition and consequent price increases

TABLE 8mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure loss ofvariety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 732(087) (089) (146)

Only ofloxacin 008 008 023(008) (008) (010)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1253(177) (180) (415)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1058(142) (145) (331)

All four domestic quinolones products 1176dagger 1135dagger 1781(643) (634) (1270)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent significance level and dagger (dagger)denotes significance at the 10-percent level

1502 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 24: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

foreign product is in stock) in aggregate datawe would expect to find precisely the substitu-tion patterns that we report in Table 6

Whether the particular explanation we provideabove is the correct one the high degree of sub-stitutability between domestic product groupsturns out to have important implications for thewelfare calculations We discuss these in moredetail below when we present the results of thecounterfactual welfare analysis Another elasticitywith important implications for the counterfactu-als is the price elasticity for the quinolone subseg-ment as a whole which indicates how likelyconsumers are to switch to other antibioticsgroups when faced with a price increase for quin-olones This elasticity is computed on the basis ofthe results in Table A3 and it is at 111 (stan-dard error 024) this is large in magnitudebutmdashas expectedmdashsmaller in absolute value thanthe own-price elasticities of the product groupswithin the quinolone subsegment

The results in Tables 6A and 6B are based onour preferred specification discussed in SectionII In Tables A4 to A6 in the Appendix weexperiment with some alternative specificationsTables A4(a)ndashA4(c) correspond to a specificationthat includes in addition to product-group-specificregional fixed effects product-group-specific (andfor the upper level antibiotics-segment-specific)seasonal effects We distinguish among three sea-sonsmdashthe summer monsoon and wintermdashandreport the unconditional demand elasticities for

the northern region for each of these seasons Asevident from the tables in the Appendix our elas-ticity estimates are robust to the inclusion of sea-sonal effects The demand elasticities in Table A5are based on estimation of the demand system byOLS Compared to the elasticities obtained by IVthe OLS elasticities are smaller in absolute valueimplying that welfare calculations based on theOLS estimates would produce larger welfare lossestimates Nevertheless some of the patterns re-garding the cross-price elasticities discussed ear-lier are also evident in the OLS results inparticular the cross-price elasticities between dif-ferent domestic product groups are all positivelarge and significant and in most instances largerthan the cross-price elasticities between drugs thatcontain the same molecule but are produced byfirms of different domesticforeign status Theclose substitutability of domestic products indi-cated by both the OLS and IV estimates seems tobe one of the most robust findings of the paper

B Cost and Markup Estimates

Table 7 displays the marginal costs markupsand profits implied by the price elasticity esti-mates of Tables 6A and 6B for each of the sevenproduct groups Given that our regional effectsimply different price elasticities for each regionour marginal cost and markup estimates alsodiffer by region Given however that based on

TABLE 7mdashUPPER AND LOWER BOUNDS FOR MARGINAL COST MARKUP AND ANNUAL PROFIT BY PRODUCT GROUPS WITHIN

THE QUINOLONE SUBSEGMENT

Product groupLower boundfor MC (Rs)

Upper boundfor markup

Upper boundfor profit(Rs mill)

Upper boundfor MC (Rs)

Lower boundfor markup

Lower boundfor profit

(Rs)

Foreign ciprofloxacin 83 19 269 103 0 00(123) (012) (1655)

Foreign norfloxacin NA NA NA 53 0 00Foreign ofloxacin 323 70 1061 1085 0 00

(2316) (021) (3185)Domestic ciprofloxacin 47 59 17019 112 0 00

(114) (010) (29858)Domestic norfloxacin 52 43 2807 90 0 00

(020) (002) (1532)Domestic ofloxacin 587 34 1612 901 0 00

(218) (002) (1280)Domestic sparfloxacin 495 37 1985 788 0 00

(157) (002) (1100)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level Estimated lower bound forforeign norfloxacinrsquos marginal cost is negative since the estimated price elasticity is less than one in absolute value

1500 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Tables 6A and 6B the price elasticities do notseem to differ substantially across regions wereport for ease of exposition only the nationalaverages for marginal costs and markups inTable 7

Table 7 has two parts In the left part (firstthree columns) we report the lower bound formarginal cost and the corresponding upperbound for markup and upper bound for totalannual profit for each product group Thesenumbers are based on the price elasticities weobtained from estimating the two-level AIDSdemand system Since we do not have a reliableestimate for the price elasticity of foreign nor-floxacin (the point estimate is negative but lessthan one in absolute value and insignificant)we cannot compute the lower bound for mar-ginal cost in this case It is important to note thatthese estimates do not reflect either the actualmarginal cost or the actual markup for thesedrugs both because the existence of price reg-ulation implies that the unconstrained first-orderconditions are unlikely to hold each period andbecause our aggregation across firms of thesame domesticforeign status supplying thesame molecule makes the interpretation of theseestimates problematic In particular the fact thatwe ignore competition among firms within eachproduct group implies that our estimates willtend to overstate market power These numberswill prove useful however in the counterfac-tual simulations as they can give us a sense ofhow large the maximum profit gains for multi-national firms and the maximum profit lossesfor domestic firms are likely to be under patentenforcement The right part of the table (lastthree columns) reports the upper bound for mar-ginal cost which we obtain by simply taking themarginal cost to be equal to the observed priceWe do not report standard errors in this casesince the numbers are based on actual data Themarkups corresponding to these marginal costupper bounds are of course zero We conductthe counterfactual simulations using both thelower and upper bounds for marginal costs

The (upper bounds for) markups on the leftside of the table are generally plausible Thedomestic ciprofloxacin product group that dom-inates the quinolone subsegment and for whichwe documented high prices a high marketshare and a relatively low elasticity of demandenjoys one of the highest markup upper bounds(60 percent) and accounts for nearly 70 percent

of all profits derived within the subsegmentForeign ciprofloxacin on the other hand whichfaces the stiffest competition from domesticfirms and for which we estimated a highly elas-tic demand has the lowest markup upper bound(19 percent)

C Counterfactual Estimates of the Impact onPrices and Welfare

With estimates of the key demand and costparameters in hand we turn to the counterfac-tuals We consider the five separate scenarioslisted in the previous section All of the scenar-ios involve the withdrawal of one or more of thedomestic product groups from the market Ta-ble 8 displays our estimates of the consumerwelfare losses that result under the differentscenarios The losses are expressed in billion ofrupees per year All numbers presented in Ta-ble 8 and subsequent tables are based on usingthe lower bounds for marginal cost and upperbounds for markup in the simulations As dis-cussed earlier these numbers are the more in-teresting to work with since they give us anupper bound for the changes in the profits ofdomestic and foreign firms that would resultfrom patent enforcement In the Appendix wealso present results based on using the upperbounds for marginal costs in the simulations inwhich case the pre-TRIPS profits of domesticand foreign firms are zero In all cases marginalcosts are assumed to be constant in outputWhile naturally the profit implications differdepending on whether one uses the upper orlower bounds for marginal costs (firm profits arezero if one assumes the upper bound of mar-ginal cost in which case price equals marginalcost) the estimated consumer welfare losses aresimilar in the two cases We discuss these re-sults in more detail at the end of this section

The first column presents our estimates of theconsumer welfare losses attributable to the pureloss of product variety effect where we fix theprices of all remaining products as well as theoverall expenditure on quinolones while with-drawing one or more of the domestic productgroups Note that in our initial specification ofthe demand system had we not allowed for thepossibility that consumers might differentiatebetween domestic and foreign products evenwhen they contain the same molecule this

1501VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

particular component of the loss of consumerwelfare would not have arisen

The estimates reported in the second columnincorporate the expenditure-switching effect ontop of the loss of product variety Here basedupon the price elasticity estimates from thehigher-level AIDS system we adjust (down-ward) the expenditures allocated to the quino-lone subsegment as the composite price ofquinolones effectively increases as a consequenceof the higher virtual prices of the domestic productgroups that are withdrawn from the market Be-cause the estimates in this column are generatedassuming that the prices of the products that re-main in the market are not adjusted upward theyprovide a sense of what consumer losses would beif the introduction of product patents were coupledwith strict price regulation aimed at maintainingprices at prepatent levels Alternatively they canbe thought of as the relevant welfare numbers ifintense competition among firms within the re-maining product groups kept the prices of theproducts that were still offered in the market closeto the firmsrsquo marginal costs

The last column of Table 8 displays the esti-mated consumer welfare losses when bothcross-segment expenditure-switching and with-in-segment upward price adjustments are takeninto account

If we compare the results across the firstsecond and third columns all the counterfac-tual scenarios produce qualitatively similar pat-terns patterns that are consistent with what we

would expect Starting from the initial loss ofwelfare attributable to the loss of product vari-ety the option of switching expenditures outof the quinolone subsegment to other subseg-ments mitigates some of the initial welfareloss But if we then incorporate the upwardprice adjustments that result in response to thereduced competition the welfare losses aremagnified

Of particular interest from a policy perspec-tive are the relative magnitudes of these threeeffects which are similar under all the counter-factual scenarios though absolute levels varyconsiderably First despite the fact that the de-mand for quinolones is quite sensitive to thecomposite price of quinolonesmdashthe upper levelprice elasticity is 111mdashthe cross-subsegmentexpenditure-switching effects are in all thecases small (in absolute value terms) relative tothe other two effects For instance under thescenario where all the domestic quinolone prod-uct groups are withdrawn from the market theoverall consumer welfare loss of Rs 1781 bil-lion per year can be decomposed into an initialloss of Rs 1176 billion (66 percent) attribut-able to the loss of product variety a slightreduction in this initial loss of Rs 041 billion(2 percent) from Rs 1176 billion to Rs1135 billion because of expenditure switchingand a subsequent additional loss of Rs 646billion (36 percent) from Rs 1135 billion toRs 1781 billion because of the reduced com-petition and consequent price increases

TABLE 8mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure loss ofvariety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 732(087) (089) (146)

Only ofloxacin 008 008 023(008) (008) (010)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1253(177) (180) (415)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1058(142) (145) (331)

All four domestic quinolones products 1176dagger 1135dagger 1781(643) (634) (1270)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent significance level and dagger (dagger)denotes significance at the 10-percent level

1502 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 25: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

Tables 6A and 6B the price elasticities do notseem to differ substantially across regions wereport for ease of exposition only the nationalaverages for marginal costs and markups inTable 7

Table 7 has two parts In the left part (firstthree columns) we report the lower bound formarginal cost and the corresponding upperbound for markup and upper bound for totalannual profit for each product group Thesenumbers are based on the price elasticities weobtained from estimating the two-level AIDSdemand system Since we do not have a reliableestimate for the price elasticity of foreign nor-floxacin (the point estimate is negative but lessthan one in absolute value and insignificant)we cannot compute the lower bound for mar-ginal cost in this case It is important to note thatthese estimates do not reflect either the actualmarginal cost or the actual markup for thesedrugs both because the existence of price reg-ulation implies that the unconstrained first-orderconditions are unlikely to hold each period andbecause our aggregation across firms of thesame domesticforeign status supplying thesame molecule makes the interpretation of theseestimates problematic In particular the fact thatwe ignore competition among firms within eachproduct group implies that our estimates willtend to overstate market power These numberswill prove useful however in the counterfac-tual simulations as they can give us a sense ofhow large the maximum profit gains for multi-national firms and the maximum profit lossesfor domestic firms are likely to be under patentenforcement The right part of the table (lastthree columns) reports the upper bound for mar-ginal cost which we obtain by simply taking themarginal cost to be equal to the observed priceWe do not report standard errors in this casesince the numbers are based on actual data Themarkups corresponding to these marginal costupper bounds are of course zero We conductthe counterfactual simulations using both thelower and upper bounds for marginal costs

The (upper bounds for) markups on the leftside of the table are generally plausible Thedomestic ciprofloxacin product group that dom-inates the quinolone subsegment and for whichwe documented high prices a high marketshare and a relatively low elasticity of demandenjoys one of the highest markup upper bounds(60 percent) and accounts for nearly 70 percent

of all profits derived within the subsegmentForeign ciprofloxacin on the other hand whichfaces the stiffest competition from domesticfirms and for which we estimated a highly elas-tic demand has the lowest markup upper bound(19 percent)

C Counterfactual Estimates of the Impact onPrices and Welfare

With estimates of the key demand and costparameters in hand we turn to the counterfac-tuals We consider the five separate scenarioslisted in the previous section All of the scenar-ios involve the withdrawal of one or more of thedomestic product groups from the market Ta-ble 8 displays our estimates of the consumerwelfare losses that result under the differentscenarios The losses are expressed in billion ofrupees per year All numbers presented in Ta-ble 8 and subsequent tables are based on usingthe lower bounds for marginal cost and upperbounds for markup in the simulations As dis-cussed earlier these numbers are the more in-teresting to work with since they give us anupper bound for the changes in the profits ofdomestic and foreign firms that would resultfrom patent enforcement In the Appendix wealso present results based on using the upperbounds for marginal costs in the simulations inwhich case the pre-TRIPS profits of domesticand foreign firms are zero In all cases marginalcosts are assumed to be constant in outputWhile naturally the profit implications differdepending on whether one uses the upper orlower bounds for marginal costs (firm profits arezero if one assumes the upper bound of mar-ginal cost in which case price equals marginalcost) the estimated consumer welfare losses aresimilar in the two cases We discuss these re-sults in more detail at the end of this section

The first column presents our estimates of theconsumer welfare losses attributable to the pureloss of product variety effect where we fix theprices of all remaining products as well as theoverall expenditure on quinolones while with-drawing one or more of the domestic productgroups Note that in our initial specification ofthe demand system had we not allowed for thepossibility that consumers might differentiatebetween domestic and foreign products evenwhen they contain the same molecule this

1501VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

particular component of the loss of consumerwelfare would not have arisen

The estimates reported in the second columnincorporate the expenditure-switching effect ontop of the loss of product variety Here basedupon the price elasticity estimates from thehigher-level AIDS system we adjust (down-ward) the expenditures allocated to the quino-lone subsegment as the composite price ofquinolones effectively increases as a consequenceof the higher virtual prices of the domestic productgroups that are withdrawn from the market Be-cause the estimates in this column are generatedassuming that the prices of the products that re-main in the market are not adjusted upward theyprovide a sense of what consumer losses would beif the introduction of product patents were coupledwith strict price regulation aimed at maintainingprices at prepatent levels Alternatively they canbe thought of as the relevant welfare numbers ifintense competition among firms within the re-maining product groups kept the prices of theproducts that were still offered in the market closeto the firmsrsquo marginal costs

The last column of Table 8 displays the esti-mated consumer welfare losses when bothcross-segment expenditure-switching and with-in-segment upward price adjustments are takeninto account

If we compare the results across the firstsecond and third columns all the counterfac-tual scenarios produce qualitatively similar pat-terns patterns that are consistent with what we

would expect Starting from the initial loss ofwelfare attributable to the loss of product vari-ety the option of switching expenditures outof the quinolone subsegment to other subseg-ments mitigates some of the initial welfareloss But if we then incorporate the upwardprice adjustments that result in response to thereduced competition the welfare losses aremagnified

Of particular interest from a policy perspec-tive are the relative magnitudes of these threeeffects which are similar under all the counter-factual scenarios though absolute levels varyconsiderably First despite the fact that the de-mand for quinolones is quite sensitive to thecomposite price of quinolonesmdashthe upper levelprice elasticity is 111mdashthe cross-subsegmentexpenditure-switching effects are in all thecases small (in absolute value terms) relative tothe other two effects For instance under thescenario where all the domestic quinolone prod-uct groups are withdrawn from the market theoverall consumer welfare loss of Rs 1781 bil-lion per year can be decomposed into an initialloss of Rs 1176 billion (66 percent) attribut-able to the loss of product variety a slightreduction in this initial loss of Rs 041 billion(2 percent) from Rs 1176 billion to Rs1135 billion because of expenditure switchingand a subsequent additional loss of Rs 646billion (36 percent) from Rs 1135 billion toRs 1781 billion because of the reduced com-petition and consequent price increases

TABLE 8mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure loss ofvariety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 732(087) (089) (146)

Only ofloxacin 008 008 023(008) (008) (010)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1253(177) (180) (415)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1058(142) (145) (331)

All four domestic quinolones products 1176dagger 1135dagger 1781(643) (634) (1270)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent significance level and dagger (dagger)denotes significance at the 10-percent level

1502 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 26: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

particular component of the loss of consumerwelfare would not have arisen

The estimates reported in the second columnincorporate the expenditure-switching effect ontop of the loss of product variety Here basedupon the price elasticity estimates from thehigher-level AIDS system we adjust (down-ward) the expenditures allocated to the quino-lone subsegment as the composite price ofquinolones effectively increases as a consequenceof the higher virtual prices of the domestic productgroups that are withdrawn from the market Be-cause the estimates in this column are generatedassuming that the prices of the products that re-main in the market are not adjusted upward theyprovide a sense of what consumer losses would beif the introduction of product patents were coupledwith strict price regulation aimed at maintainingprices at prepatent levels Alternatively they canbe thought of as the relevant welfare numbers ifintense competition among firms within the re-maining product groups kept the prices of theproducts that were still offered in the market closeto the firmsrsquo marginal costs

The last column of Table 8 displays the esti-mated consumer welfare losses when bothcross-segment expenditure-switching and with-in-segment upward price adjustments are takeninto account

If we compare the results across the firstsecond and third columns all the counterfac-tual scenarios produce qualitatively similar pat-terns patterns that are consistent with what we

would expect Starting from the initial loss ofwelfare attributable to the loss of product vari-ety the option of switching expenditures outof the quinolone subsegment to other subseg-ments mitigates some of the initial welfareloss But if we then incorporate the upwardprice adjustments that result in response to thereduced competition the welfare losses aremagnified

Of particular interest from a policy perspec-tive are the relative magnitudes of these threeeffects which are similar under all the counter-factual scenarios though absolute levels varyconsiderably First despite the fact that the de-mand for quinolones is quite sensitive to thecomposite price of quinolonesmdashthe upper levelprice elasticity is 111mdashthe cross-subsegmentexpenditure-switching effects are in all thecases small (in absolute value terms) relative tothe other two effects For instance under thescenario where all the domestic quinolone prod-uct groups are withdrawn from the market theoverall consumer welfare loss of Rs 1781 bil-lion per year can be decomposed into an initialloss of Rs 1176 billion (66 percent) attribut-able to the loss of product variety a slightreduction in this initial loss of Rs 041 billion(2 percent) from Rs 1176 billion to Rs1135 billion because of expenditure switchingand a subsequent additional loss of Rs 646billion (36 percent) from Rs 1135 billion toRs 1781 billion because of the reduced com-petition and consequent price increases

TABLE 8mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure loss ofvariety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 732(087) (089) (146)

Only ofloxacin 008 008 023(008) (008) (010)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1253(177) (180) (415)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1058(142) (145) (331)

All four domestic quinolones products 1176dagger 1135dagger 1781(643) (634) (1270)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent significance level and dagger (dagger)denotes significance at the 10-percent level

1502 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 27: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

The basic claim made by proponents ofTRIPS is that any adverse impacts on consumerwelfare from the introduction of a product patentin a particular market will be mitigated by theavailability of close therapeutic substitutes Therelatively minor role that cross-subsegment ex-penditure switching appears to play suggests thatfor this claim to be valid there need to be unpat-ented (ie patent-expired) substitutes availablewithin fairly narrowly defined therapeutic catego-ries Since the extent to which this is true will varyacross therapeutic segments the impact of TRIPSis likely to be correspondingly variegated a pointemphasized by Maskus (2000 p 163)

Price regulation and compulsory licensing aretwo of the most widely mentioned post-TRIPSpolicy options available to governments of devel-oping economies There is an ongoing debateabout how much leeway governments shouldhave to introduce these options and about therelative efficacy of the two options in limitingprice increases The magnitude and importance ofthe welfare losses we estimate from the loss ofproduct variety suggest that there may be an in-dependent role for compulsory licensing in addi-tion to or in lieu of price regulation for the solepurpose of mitigating the loss of product variety

Turning next to a comparison of the consumerwelfare losses under the different scenarios themost striking result is that the estimated loss ofconsumer welfare (Rs 1781 billion) from thesimultaneous withdrawal of all four domesticproduct groupsmdashthe scenario that most closelyresembles what is likely to happen underTRIPSmdashis more than two times the sum of theestimated losses from the four separate scenariosin each of which only one of the domestic productgroups is withdrawn37 What this very clearlyindicates is that past studies that have estimatedthe aggregate effects of patent protection by add-ing up the losses estimated separately in each ofa number of patentable markets may have sub-stantially underestimated the magnitude of the

consumer welfare losses from the introduction ofpharmaceutical product patents

The result that the simultaneous withdrawal ofall domestic products magnifies the scale of thewelfare losses is driven by our estimates of highpositive cross-price elasticities between domesticproducts As noted earlier these elasticities implythat such products are close substitutes to oneanother Hence when all four domestic productsdisappear from the market the resulting consumerloss is substantial In contrast the welfare lossesassociated with the withdrawal of a single domes-tic product or a subset of domestic products aremore modest with domestic product groupswithin the quinolone subsegment being relativelygood substitutes if only one of them is withdrawnconsumers switch to the others and this limits anywelfare losses

We should note that if as we speculated abovethe high degree of substitutability between domes-tic products stems in part from the differentialreach of the distribution networks of domestic andforeign firms these estimates may overstate thewelfare loss from the simultaneous withdrawal ofall domestic products That is because with Indiabecoming TRIPS compliant foreign subsidiariesmay well choose to expand their product portfo-lios in India and simultaneously expand their dis-tribution networks in India most likely throughjoint marketing ventures with Indian firms Mediaaccounts and interviews with industry sources in-dicate that such initiatives are increasing in num-ber In this case the welfare loss from thereduction in variety would be a purely transitionalphenomenon Over time foreign products wouldbe more readily available in local pharmaciesthroughout India and this would compensate forthe reduction in the number of domestic productsAlternatively if Indian consumers insist on buy-ing products produced by Indian firms foreignmultinationals could use licensing to recover thewelfare loss associated with the loss of varietyNote however that even under this scenario thecomponent of consumer welfare loss due to up-ward price adjustment remains And a crude cal-culation based on the estimates in the last row ofTable 8 suggests that this is likely to be significantIn particular if we subtract from our estimate ofthe overall consumer welfare loss (Rs 1781 bil-lion) the component attributable to the reductionin variety taking into account expenditure switch-ing (Rs 1135 billion) we are still left with anestimated welfare loss of Rs 646 billion Given

37 For ease of exposition Table 8 reports only a subset ofthe scenarios we have investigated The consumer welfareloss associated with the withdrawal of domestic norfloxacinonly (not reported in Table 8) is approximately Rs 01billion the welfare loss associated with the withdrawal ofsparfloxacin only is close to zero Hence the sum of theestimated welfare losses from the scenarios in which onlyone domestic product is withdrawn is (732 023 01 0) Rs 765 billion per year

1503VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 28: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

the size of the welfare loss due to upward priceadjustment policymakers may be tempted to con-tinue the use of price controls and other domesticregulations Such policies however would put alimit not only on prices but also on the incentivesof foreign producers to expand their operations inthe Indian market so that the welfare loss due tothe reduction of product variety could become apermanent phenomenon

Table 9 documents our estimates of the priceincreases that would result under the variouscounterfactual scenarios The table reports theprice increases for the product groups foreignor domestic that would remain in the marketunder each of our scenarios The groups that arewithdrawn from the market are indicated by theshaded areas For the foreign products thatwould remain in the market we estimate priceincreases between 100 percent and 400 percentWhile these numbers are based on simulationsand thus not observed we can obtain a roughidea about their plausibility by comparing themto the prices of the same products observed incountries ldquosimilarrdquo to India in terms of demo-graphics but which have less competitive do-mestic markets Pakistan is a natural candidateFor the drug ciprofloxacin for example wepredict that the price of the (patented) foreignproducts in India would be approximately fivetimes higher than it is now (see first column ofTable 9 last row the relevant scenario here is

one where all domestic products are withdrawnfrom the market since this is the situation thatmost closely resembles that of Pakistan) Lan-jouw (1998 p 39 Table 2) reports that the priceof ciprofloxacin in Pakistan is about seven timesthe price of the same drug in India The twonumbers are of similar order of magnitude al-though our estimate is on the low side Thesecomparisons give us confidence that the empir-ical framework we use as a basis for conductingcounterfactual simulations in India captures themain features of this market

Table 10 presents our estimates of the netimpact of the withdrawal of one or more do-mestic product groups on the collective profitsof domestic Indian firms in the quinolone sub-segment Under the scenario where all the do-mestic product groups are withdrawn from themarket the net impact equals the gross impactand is simply the loss of the profits initiallyenjoyed by domestic firms Rs 234 billion peryear In the other cases the foregone profitsof those domestic firms whose products arewithdrawn from the market are partly orwholly offset by the increased profits of thosedomestic firms that remain in the market andbenefit from the reduced competition FromTable 10 it can be seen that this latter resultarises when domestic ofloxacin is withdrawnfrom the market in which case consumersswitch to other domestic drugs within quino-

TABLE 9mdashCOUNTERFACTUAL ESTIMATES OF DRUG PRICE CHANGES AFTER PRODUCT WITHDRAWAL DUE TO INTRODUCTION OF

PHARMACEUTICAL PATENTS

Counterfactual scenarios withdrawal ofone or more domestic product groups

Changes in prices with cross-segment expenditure switching andwithin-segment price adjustment ( of original prices)

Foreign product groups Domestic product groups

Cipro Norflo Oflo Cipro Norflo Oflo Spar

Only ciprofloxacin 1894 3147 982 1486 1414 1641(018) NA (021) (009) (008) (008)

Only ofloxacin 1004 3147 1029 1024 1083 1107(001) NA (008) (001) (001) (001)

Ciprofloxacin ofloxacin and norfloxacin 2476 6278 1544 2963(042) NA (039) (070)

Ciprofloxacin ofloxacin and sparfloxacin 2552 6278 1583 2501 (040) NA (037) (064)

All four domestic quinolones products 3964 6278 3184dagger (334) NA (173)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level We fix foreign norflorsquos price at the numbers shown because an estimate for the marginalcost lower bound is not available for this drug We tried many different values for foreign norflorsquos counterfactual prices andthe results are remarkably robust

1504 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 29: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

lones increasing the profits of the domesticfirms selling those drugs38

Critics of the Indian governmentrsquos stance onTRIPS frequently assert that it is motivated lessby concerns about consumer welfare than it isby a desire to protect the domestic pharmaceu-tical industry Whether or not that is the casethe estimates presented in Table 10 indicate thatthe loss of domestic producer surplus is unlikelyto be the biggest consequence of TRIPS-induced patent protection First as just men-tioned there are scenarios under which thecollective profits of domestic firms would actu-ally go up though there always is a segmentthat would be adversely affected39 Secondeven when collective profits go down a com-parison with Table 8 indicates that the loss ofconsumer welfare is much greater in every in-stance And under the scenario where the col-

lective loss of profits is the greatest and thereare no winners among the domestic Indianfirms the loss incurred by producersmdashRs 23billion on an annualized basismdashpales in com-parison to the decrease in consumer welfarereported in Table 8 under the same scenariomdashRs 1781 billion annuallymdashespecially if onetakes into account that the profit loss of Rs 23billion was derived by using an upper bound forthe domestic firmsrsquo Pre-TRIPS markups andhence clearly overstates the profit loss that thedomestic sector would actually incur40

Table A6 in the Appendix reports resultsfrom counterfactual simulations in which weused the upper bound for marginal costs Thepre-TRIPS domestic firm profits are of coursezero in this case since the upper bound wasderived by setting marginal cost equal to priceInterestingly the results in Table A6 that reportthe loss in consumer welfare indicate that theconsumer loss would be slightly higher thanbefore This is due to the fact that because of thehigher marginal cost estimates the price in-creases are now higher compared to the casewhere the lower bound of marginal cost wasused The new consumer welfare loss estimateis however roughly of the same order of

38 To be consistent with Table 8 which reported con-sumer welfare losses as positive numbers Table 10 reportsforegone profits as positive numbers Thus if the collectiveprofits of domestic Indian firms actually increase negativenumbers are reported

39 This may in part explain why the Indian pharmaceu-tical industry has been divided in its reaction to TRIPS TheOrganization of Pharmaceutical Producers of India whichincludes among its members most of the leading Indianfirms as well the subsidiaries of foreign MNCs is openlysupportive of strengthening Indiarsquos intellectual propertyrights regime (httpwwwindiaoppicom) Other industryassociations such as the Indian Drug Manufacturers Asso-ciation with memberships drawn from smaller firms tend tobe more critical of TRIPS

40 There are other factors as well that might serve tomitigate the losses experienced by Indian firms among themthe possibility of joint ventures with or contract manufacturingfor multinationals Such collaborations are increasing in fre-quency in the Indian pharmaceutical industry

TABLE 10mdashCOUNTERFACTUAL ESTIMATES OF FOREGONE PROFITS OF DOMESTIC PRODUCERS FROM PRODUCT WITHDRAWAL

DUE TO THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 095 109 040(016) (008) (031)

Only ofloxacin 004 003dagger 014(001) (002) (004)

Ciprofloxacin ofloxacin and norfloxacin 124 140 042(017) (011) (060)

Ciprofloxacin ofloxacin and sparfloxacin 111 126 048(017) (011) (050)

All four domestic quinolones products 234 234 234(019) (019) (019)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1505VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 30: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

magnitude as before so we focus the rest of ourdiscussion on the results we obtained by usingthe lower bound of marginal cost

Adding up the estimates of consumer welfarelosses from Table 8 and producer losses fromTable 10 we get estimates of the total welfarelosses to the Indian economy These are re-ported in Table 11 At the upper bound weestimate that in the absence of any price regu-lation or compulsory licensing the total annualwelfare losses to the Indian economy from thewithdrawal of just four domestic product groupsin the quinolone subsegment would be on theorder of Rs 2016 billion which translates atthe then-prevailing exchange rate into a figureof US$450 million

Given that in practice the simultaneous en-forcement of intellectual property rights andelimination of price controls is unlikely a morerealistic estimate of the welfare loss can beobtained by assuming that price regulation willprevent upward price adjustments as a result ofproduct withdrawals This gives us a mid-rangeestimate of Rs 137 billion or about US$305million per year for the scenario involving with-drawal of all four domestic quinolone productgroups Of this amount foregone profits of do-mestic producers constitute roughly Rs 23 bil-lion or US$50 million (circa 16 percent of thetotal welfare loss) The overwhelming portionof the total welfare loss therefore derives fromthe loss of consumer welfare

Lastly if we assume that the welfare lossesdue to the reduction in variety are a purely

transitional phenomenon or that they could beneutralized through expanded use of licensingand subtract these from our upper bound esti-mates we obtain a lower bound estimate of Rs65 billion ( 2016 137) or $144 millionannually Though only about 30 percent of ourupper bound estimate in absolute terms thislower bound estimate is still very large repre-senting about 24 percent of antibiotic sales in2000

Finally Table 12 presents our estimates ofthe profit gains realized by foreign producers asa result of patent introduction These estimatesindicate that the total profit gains to foreignproducers would be only about Rs 24 billionor approximately US$53 million per year Moreimportantly the US$53 million per year esti-mate corresponds to the rather unrealistic casewhere there is no price regulation so that mul-tinationals are free to adjust their prices upwardin response to the reduced competition In thepresence of price regulation that would keepprices fixed at their pre-patent-enforcementlevel (column 2 in Table 12) the profit gain forforeign multinationals becomes only Rs 088billion or US$196 million per year To putthese numbers in perspective sales of Ciproalone the main patented ciprofloxacin productof Bayer were roughly US$16 billion in 2000(Scott Hensley ldquoCipro Loses Share in Tradi-tional Market as Doctors Seek Other Cures inShortagerdquo Wall Street Journal November 72001) Assuming a 40-percent markup (themarkup usually quoted for the pharmaceutical

TABLE 11mdashCOUNTERFACTUAL ESTIMATES OF TOTAL WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 594 601 772(094) (087) (120)

Only ofloxacin 004 005 009(008) (009) (010)

Ciprofloxacin ofloxacin and norfloxacin 876 880 1295(184) (177) (361)

Ciprofloxacin ofloxacin and sparfloxacin 725 728 1107(146) (142) (288)

All four domestic quinolones products 1410 1370 2016(657) (648) (1285)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level

1506 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 31: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

industry) this translates to annual profits ofroughly US$640 million per year for BayerrsquosCipro alone

IV Conclusion

The results of our analysis suggest thatconcerns about the potentially adverse wel-fare effects of TRIPS in developing countriesmay have some basis Specifically we esti-mate that in the quinolone subsegment of thesystemic antibacterials segment alone patentenforcement would result in a large welfareloss for the Indian economy The estimatedloss ranges from $144 million to an upperbound of $450 million annually dependingon the way policies are implemented the ex-tent of price regulation and the degree towhich foreign multinationals respond topatent protection by expanding their distribu-tion networks or using licensing more exten-sively Of this amount only a small fractionaccounts for the forgone profits of domestic(Indian) pharmaceutical firms Hence we donot find much support for the claim thatTRIPS would have detrimental effects on theIndian pharmaceutical industry In fact undersome scenarios we find that the profits ofdomestic firms may even increase this hap-pens because when certain domestic productsbecome unavailable as a result of patent en-forcement consumers substitute toward other

domestic products containing different mole-cules rather than foreign products containingthe same molecule This differential effect ofTRIPS on domestic firmsrsquo profits may partlyexplain the divided position of the Indianpharmaceutical industry regarding TRIPS

With respect to the subsidiaries of foreignmultinationals we estimate the profit gains ofthese firms to be approximately US$53 millionper year when patents are enforced This is inthe absence of compulsory licensing or priceregulation With price regulation that wouldkeep the prices of drugs supplied by multina-tional subsidiaries at their pre-TRIPS level theprofit gains drop to only US$196 million peryear While we certainly do not attempt to drawany conclusions about the relationship betweenintellectual property rights protection and re-search and innovation we note that this numberrepresents a very small fraction of the annualsales of big pharmaceutical firms in this sub-segment

By far the biggest effects of TRIPS concernthe Indian consumers for whom we estimatesubstantial welfare losses The losses increasein the number of domestic products that areaffected by TRIPS The worst-case scenarioinvolves simultaneous withdrawal of all do-mestic product groups in the quinolone sub-segment In contrast when only one domesticproduct or a subset of domestic products arewithdrawn the consumer losses are modest

TABLE 12mdashCOUNTERFACTUAL ESTIMATES OF PROFIT GAINS OF FOREIGN PRODUCERS FROM PRODUCT WITHDRAWAL DUE TO

THE INTRODUCTION OF PHARMACEUTICAL PATENTS (RS BILL PER YEAR)

Counterfactual scenarios Withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 017 014 035(003) (006) (017)

Only ofloxacin 002 002 001(001) (001) (002)

Ciprofloxacin ofloxacin and norfloxacin 036 028 071dagger

(005) (011) (040)Ciprofloxacin ofloxacin and sparfloxacin 037 030 079

(005) (010) (038)All four domestic quinolones products 117 088 243dagger

(031) (041) (144)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

1507VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 32: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

This pattern is driven by the empirical findingthat domestic products are viewed by Indianconsumers as close substitutes accordinglythe existence of some degree of domesticcompetition has a big impact on consumerwell-being

Finally our decomposition of the total con-sumer loss into a ldquoproduct-varietyrdquo effect anldquoexpenditure-switchingrdquo effect and a ldquoprice-adjustmentrdquo effect has interesting policy im-plications We find that a substantial fractionof the total welfare loss is attributable to theloss of variety which we interpret as primar-ily capturing an ldquoease of accessrdquo effect be-cause the retail coverage of domestic firms inIndia is substantially more extensive than thatof foreign multinationals drugs produced bydomestic firms are more readily available toIndian consumers than drugs sold by foreignproducers This suggests a potentially inde-pendent role of compulsory licensing in addi-tion to or in lieu of price regulation for thesole purpose of mitigating the loss of productvariety effect Even if one considers this ef-fect to be only a transitional phenomenon thatwill diminish in importance as foreign firmsrespond to TRIPS enforcement by expandingtheir product portfolios and distribution net-works or by using licensing more exten-sively the welfare loss due to upward priceadjustment remains substantial The price-adjustment component of welfare loss couldpotentially be mitigated by appropriate pricecontrols or other regulations In this casehowever the incentives of multinationals toexpand their operations in the Indian marketwould become questionable and the welfare

loss attributable to the loss of product varietycould become a permanent effect

In general our simulations indicate that froma consumer welfare point of view the issue ofproduct availability is as important as the issueof affordability In this sense our analysis sug-gests that policymakers should evaluate TRIPS-related policies not only in terms of theireffects on drug prices but also in terms oftheir impact on product availability This ob-servation is more relevant the more likely it isthat there will be tension between policiesdesigned to address these two sets of effectsIntellectual property rights enforcement with-out price regulation is likely to bolster foreignfirmsrsquo incentives to market their products indeveloping countries and use licensing moreextensively than in the past but it brings withit the potential of substantial price increasesof patented products Accompanying priceregulation can prevent patent holders fromexploiting their market power but not withoutdiminishing the incentives of such firms toexpand their operations in the developingworld A combination of policies that wouldcompletely neutralize TRIPSrsquo adverse effectson consumer welfare is hence unlikely

Lastly we find that expenditure switchingacross subsegments has a limited role in con-taining consumer welfare loss The claim ofTRIPS proponents that any adverse effects aris-ing from the introduction of a patent in a par-ticular market would be mitigated by theavailability of close therapeutic substitutes isthus valid only if there are patent-expired sub-stitutes available within fairly narrowly definedtherapeutic categories

1508 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 33: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

APPENDIX ADDITIONAL TABLES AND ROBUSTNESS CHECKS

TABLE A1mdashSPECTRUM OF ACTIVITY OF VARIOUS FAMILIES OF ANTI-BACTERIAL DRUGS

Organism Tet

racy

clin

es

Chl

oram

peni

cols

Am

pici

llin

amox

ycill

in

Cep

halo

spor

ins

Tri

met

hopr

imco

mbi

natio

ns

Mac

rolid

es

Oth

erpe

nici

llins

Am

inog

lyco

side

s

Fluo

roqu

inol

ones

Gram-positive cocciStaphylococcus aureus

Non-penicillinase producing x x x x xPenicillinase producing x x x x x

Streptococcus bovisSerious infections x x xUncomplicated urinary tract infection x x x

Streptococcus pneumoniae x x x x x xGram-negative cocci

Neisseria meningitidis x x xNeisseria gonorrhaoeae

Non-beta-lactamase producing x x x x xBeta-lactamase producing x x x x

Gram-negative bacilliAcinetobacter spp x x x xBrucella spp x x x xCampylobacter jejuni x x x xEnterobacter spp x x x xEscherichia coli

Uncomplicated urinary tract infection x x x x xSystemic infection x x x x

Francisella tularensis x x x xHaemophilus influenzae

Meningitis x x x x xOther infections x x x x

Klebisiella pneumonia x x x x x xLegionella spp x x x xProteus mirabillis x x x xOther proteus spp x x x x xProvidencia spp x x x x xPseudomonas aeruginosa x x x xSalmonella spp x x x xSerratia marcescens x x x x xShigella spp x x xYersinia pestis x x x x

Anaerobic bacteriaAnaerobic streptococci x x x x xBacteroides spp

Oropharyngeal strains x x x x x xGastrointestinal strains x x x x x

Clostridium spp x x x

Notes An ldquoxrdquo in a cell indicates that at least one member of the family of drugs indicated in the column heading is listedas the antimicrobial drug of choice or as an alternative agent for the treatment of the bacterial infection indicated in the rowheading

Source Table 15-1 pp 225ndash26 Principles and Practice of Infectious Diseases (2000)

1509VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 34: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

TABLE A2mdashCOEFFICIENT ESTIMATES FROM THE LOWER-LEVEL AIDS SYSTEM

Product group ConstantOwn-pricecoefficients

Cross-price coefficients

Coefficienton

quinolonesexpenditure

Easternregion

dummy

Westernregion

dummy

Southernregion

dummy

Samemoleculedifferent

status

Differentmolecule

foreigngroup

Differentmoleculedomestic

group

Foreignciprofloxacin

0013 0120 0115 0003dagger 0004 0010 0027 0031 0007(002) (005) (005) (000) (000) (001) (001) (001) (000)

Foreignnorfloxacin

0006 0000 0005 0003dagger 0004 0001 0001 0000 0000(001) (000) (000) (000) (000) (000) (000) (000) (000)

Foreignofloxacin

0047 0013 0008 0003dagger 0004 0005 0001 0003 0007(002) (001) (001) (000) (000) (001) (000) (000) (000)

Domesticciprofloxacin

0603 0298 0115 0004 0058 0102 0001 0044 0033(004) (005) (005) (000) (000) (002) (001) (001) (001)

Domesticnorfloxacin

0206 0177 0005 0004 0058 0038 0041 0034 0031(003) (002) (000) (000) (000) (001) (001) (001) (001)

Domesticofloxacin

0339 0191 0008 0004 0058 0008 0018 0034 0026(004) (001) (001) (000) (000) (002) (001) (001) (001)

Domesticsparfloxacin

0209 0186 0004 0058 0072 0006 0016 0028(003) (002) (000) (000) (001) (001) (000) (000)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent level and dagger (dagger) denotessignificance at the 10-percent level

TABLE A3mdashCOEFFICIENT ESTIMATES FROM THE UPPER-LEVEL AIDS SYSTEM

Product group Con

stan

t

Cross-price coefficients

Ant

ibio

tics

expe

nditu

re

Eas

tern

regi

ondu

mm

y

Wes

tern

regi

ondu

mm

y

Sout

hern

regi

ondu

mm

y

Tet

racy

clin

e

Chl

oram

phen

icol

Am

pici

llin

Cep

halo

spor

in

Tri

met

hopr

im

Mac

rolid

es

Oth

erpe

nici

llin

Qui

nolo

nes

Tetracycline 011 005 003 014dagger 013 007dagger 008dagger 002 005 003 003 000 000(009) (006) (002) (008) (006) (004) (004) (003) (005) (001) (001) (004) (004)

Chloramphenicol 002 003 004 015 009 000 009 000 014 002 002 001 001(005) (002) (001) (004) (003) (002) (002) (002) (004) (000) (001) (001) (001)

Ampicillin 027 014dagger 015 000 010 002 004 009 003 001 007 013 018(017) (008) (004) (014) (010) (006) (008) (007) (007) (001) (003) (005) (005)

Cephalosporin 022 013 009 010 026dagger 015dagger 008 016dagger 019 003 009 003 006(022) (006) (003) (010) (015) (008) (010) (009) (008) (001) (002) (005) (005)

Trimethoprim 036 007dagger 000 002 015dagger 004 016 003 022 001 001 006 005dagger

(013) (004) (002) (006) (008) (013) (006) (007) (005) (001) (001) (003) (003)Macrolides 014 008dagger 009 004 008 016 019dagger 008 005 000 002 004 003

(021) (004) (002) (008) (010) (006) (011) (008) (005) (001) (002) (003) (003)Other penicillin 017 002 000 009 016dagger 003 008 006 008dagger 001 002dagger 000 000

(018) (003) (002) (007) (009) (007) (008) (007) (005) (000) (001) (002) (003)Quinolones 041 005 014 003 019 022 005 008dagger 002 004 003 001 003

(011) (005) (004) (007) (008) (005) (005) (005) (006) (002) (002) (003) (003)

Notes Standard errors in parentheses Asterisk () denotes significance at the 5-percent confidence level and dagger (dagger)denotes significance at the 10-percent confidence level

1510 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 35: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

TABLE A4mdashOTHER SPECIFICATIONS FOR THE NORTHERN REGION

A4(a) Demand patterns with seasonal dummies Summer

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 606 014dagger 015 454 012 013dagger 016(191) (007) (008) (215) (007) (007) (007)

Foreign norfloxacin 610 012 609 527 747 610 619(455) (226) (455) (601) (1168) (462) (458)

Foreign ofloxacin 011dagger 011dagger 158 008 011 047 011(006) (006) (032) (032) (005) (029) (006)

Domestic ciprofloxacin 018 001dagger 000 172 007 007 010(008) (000) (001) (029) (003) (002) (003)

Domestic norfloxacin 003 003 003 058 204 036 033(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 006dagger 005 023dagger 089 079 367 077(003) (002) (014) (040) (011) (036) (012)

Domestic sparfloxacin 007 004 006 125 060 058 281(003) (002) (002) (020) (006) (007) (017)

A4(b) Demand patterns with seasonal dummies Monsoon

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 580 013dagger 015 432 011dagger 011 015(189) (007) (007) (207) (007) (007) (007)

Foreign norfloxacin 352 035 352 303 431 351 357(219) (104) (219) (217) (492) (171) (172)

Foreign ofloxacin 010dagger 009dagger 151 006 009 041dagger 010(005) (005) (026) (029) (004) (025) (004)

Domestic ciprofloxacin 019 001 000 172 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 003 061 210 038 035(001) (003) (001) (019) (011) (004) (004)

Domestic ofloxacin 005dagger 004 019dagger 072 065 320 064(002) (002) (011) (032) (008) (027) (009)

Domestic sparfloxacin 007 004 007 125 060 061 282(002) (002) (002) (021) (007) (007) (019)

A4(c) Demand patterns with seasonal dummies Winter

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 566 013dagger 014dagger 419 012dagger 011 014(180) (007) (007) (202) (007) (008) (007)

Foreign norfloxacin 437dagger 020 436dagger 378 533 434 441(254) (142) (253) (266) (610) (213) (213)

Foreign ofloxacin 010dagger 010dagger 155 007 010 044dagger 010(006) (005) (028) (030) (005) (026) (005)

Domestic ciprofloxacin 019 001 000 173 007 007 010(008) (000) (001) (029) (002) (003) (002)

Domestic norfloxacin 004 003 004 068 222 043 039(002) (003) (002) (020) (013) (005) (004)

Domestic ofloxacin 004dagger 004 018 068 061 308 061(003) (002) (012) (033) (010) (035) (010)

Domestic sparfloxacin 006 003 006 106 051 053 256(002) (001) (002) (020) (006) (007) (018)

Notes Standard errors in parentheses Elasticities evaluated at average revenue shares Asterisk () denotes significance at the5-percent confidence level and dagger (dagger) denotes significance at the 10-percent confidence level

1511VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 36: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

REFERENCES

Andrews Donald W K and Moshe Buchinsky2000 ldquoA Three-Step Method for Choosingthe Number of Bootstrap RepetitionsrdquoEconometrica 68(1) 23ndash51

Bale Harvey E Jr 2001 ldquoConsumption andTrade in Off-Patented Medicinesrdquo WorldHealth Organization Commission for Mac-roeconomics and Health Working Paper Se-ries WG 4 3

Barton John 2001 ldquoDifferentiated Pricing of

Patented Productsrdquo World Health Organiza-tion Commission for Macroeconomics andHealth Working Paper Series WG4 2

Berndt Ernst R 1994 Uniform PharmaceuticalPricing An Economic Analysis WashingtonDC AEI Press

Berry Steven James Levinsohn and Ariel Pakes1995 ldquoAutomobile Prices in Market Equilib-riumrdquo Econometrica 63(4) 841ndash90

Bresnahan Timothy F 1997 ldquoValuation of NewGoods under Perfect and Imperfect Compe-tition Commentrdquo In The Economics of New

TABLE A5mdashDEMAND PATTERNS WITH OLS COEFFICIENTS

Product group

Elasticity with respect to

Foreign groupsrsquo prices Domestic groupsrsquo prices

Cipro Norflo Oflo Cipro Norflo Oflo Sparflo

Foreign ciprofloxacin 276 005 006 139 000 001 005(082) (005) (005) (092) (007) (007) (007)

Foreign norfloxacin 166 124 164 059 225 102 110(175) (161) (175) (204) (673) (204) (207)

Foreign ofloxacin 004 004 099 007 002 002 003(005) (004) (026) (024) (005) (023) (005)

Domestic ciprofloxacin 006 000 001 156 007 007 010(004) (000) (001) (025) (003) (003) (002)

Domestic norfloxacin 001 001 001 046 218 040 039(002) (004) (001) (020) (011) (004) (004)

Domestic ofloxacin 002 001 000 081 073 316 071(003) (002) (010) (022) (007) (024) (007)

Domestic sparfloxacin 006 001 004dagger 125 063 064 276(002) (002) (002) (015) (006) (006) (017)

TABLE A6mdashCOUNTERFACTUAL ESTIMATES OF CONSUMER WELFARE LOSSES FROM PRODUCT WITHDRAWAL DUE TO THE

INTRODUCTION OF PHARMACEUTICAL PATENTS ASSUMING MC P (RS BILL PER YEAR)

Counterfactual scenarios withdrawal ofone or more domestic product groups

Pure lossof variety

Loss of variety and

Cross-segmentexpenditureswitching

Within-segment price-adjustmentand cross-segment expenditure

switching

Only ciprofloxacin 498 492 947(092) (094) (232)

Only ofloxacin 008 008 408(009) (009) (083)

Ciprofloxacin ofloxacin and norfloxacin 752 740 1467(183) (188) (552)

Ciprofloxacin ofloxacin and sparfloxacin 614 603 1283(141) (146) (457)

All four domestic quinolones products 1176dagger 1135dagger 1946(628) (628) (1520)

Notes Standard errors in parentheses Asterisk () denotes significance at 5-percent confidence level and dagger (dagger) denotessignificance at 10-percent confidence level

1512 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 37: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

Goods ed Timothy F Bresnahan and RobertJ Gordon 237ndash47 Chicago University ofChicago Press

Caves Richard E Michael D Whinston andMark A Hurwitz 1991 ldquoPatent ExpirationEntry and Competition in the US Pharma-ceutical Industryrdquo Brookings Papers on Eco-nomic Activity Special Issue 1ndash48

Challu Pablo M 1991 ldquoThe Consequences ofPharmaceutical Product Patentingrdquo WorldCompetition 15(2) 65ndash126

Chin Judith C and Gene M Grossman 1990ldquoIntellectual Property Rights and North-South Traderdquo In The Political Economy ofInternational Trade Essays in Honor of Rob-ert E Baldwin ed Ronald W Jones andAnne O Krueger 90ndash107 Cambridge MABasil Blackwell

Christensen Laurits R Dale W Jorgenson andLawrence J Lau 1975 ldquoTranscendental Log-arithmic Utility Functionsrdquo American Eco-nomic Review 65(3) 367ndash83

Deardorff Alan V 1992 ldquoWelfare Effects ofGlobal Patent Protectionrdquo Economica59(233) 35ndash51

Deaton Angus S and John Muellbauer 1980aldquoAn Almost Ideal Demand Systemrdquo Ameri-can Economic Review 70(3) 312ndash26

Deaton Angus S and John Muellbauer 1980bEconomics and Consumer Behavior Cam-bridge Cambridge University Press

Diwan Ishac and Dani Rodrik 1991 ldquoPatentsAppropriate Technology and North-SouthTraderdquo Journal of International Economics30(1-2) 27ndash47

Efron Bradley and Robert J Tibshirani 1993An Introduction to the Bootstrap Mono-graphs on Statistics and Applied ProbabilityVol 57 New York Chapman and Hall

Ellison Sara F Iain M Cockburn Zvi Grilichesand Jerry A Hausman 1997 ldquoCharacteristicsof Demand for Pharmaceutical Products AnExamination of Four CephalosporinsrdquoRAND Journal of Economics 28(3) 426ndash46

Fink Carsten 2000 ldquoHow Stronger Patent Pro-tection in India Might Affect the Behavior ofTransnational Pharmaceutical IndustriesrdquoThe World Bank Policy Research WorkingPaper 2352

Frank Richard G and David S Salkever 1997ldquoGeneric Entry and the Pricing of Pharma-ceuticalsrdquo Journal of Economics and Man-agement Strategy 6(1) 75ndash90

Ganslandt Mattias Keith E Maskus and Eina VWong 2001 ldquoDeveloping and DistributingEssential Medicines to Poor Countries TheDEFEND Proposalrdquo World Economy 24(6)779ndash95

Goldberg Pinelopi K 1995 ldquoProduct Differen-tiation and Oligopoly in International Mar-kets The Case of the US AutomobileIndustryrdquo Econometrica 63(4) 891ndash951

Grossman Gene M and Edwin L-C Lai 2004ldquoInternational Protection of Intellectual Prop-ertyrdquo American Economic Review 94(5)1635ndash53

Hausman Jerry A 1994 ldquoValuation of NewGoods under Perfect and Imperfect Compe-titionrdquo National Bureau of Economic Re-search Working Paper 4970

Hausman Jerry A and Gregory K Leonard2002 ldquoThe Competitive Effects of a NewProduct Introduction A Case Studyrdquo Jour-nal of Industrial Economics 50(3) 237ndash63

Helpman Elhanan 1993 ldquoInnovation Imita-tion and Intellectual Property RightsrdquoEconometrica 61(6) 1247ndash80

Indian Credit Rating Agency 1999 Indian Phar-maceutical Industry New Delhi ICRA In-dustry Watch Series

Kremer Michael 1998 ldquoPatent Buyouts AMechanism for Encouraging InnovationrdquoQuarterly Journal of Economics 113(4)1137ndash67

Lanjouw Jean O 1998 ldquoThe Introduction ofPharmaceutical Product Patents in IndialsquoHeartless Exploitation of the Poor and Suf-feringrsquordquo National Bureau of Economic Re-search Working Paper 6366

Lanjouw Jean O and Iain M Cockburn 2001ldquoNew Pills for Poor People Empirical Evi-dence after GATTrdquo World Development29(2) 265ndash89

Lu Z John and William S Comanor 1998ldquoStrategic Pricing of New PharmaceuticalsrdquoReview of Economics and Statistics 80(1)108ndash18

Mandell Gerald l John E Bennett and RaphaelDolin 2000 Principles and Practice of Infec-tious Diseases 5th ed Philadelphia and Lon-don Churchill LivingstoneHarcourt Brace

Maskus Keith E 2000 Intellectual PropertyRights in the Global Economy WashingtonDC Institute for International Economics

Maskus Keith E and Denise E Konan 1994ldquoTrade-Related Intellectual Property Rights

1513VOL 96 NO 5 CHAUDHURI ET AL GLOBAL PATENT PROTECTION IN PHARMACEUTICALS

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006

Page 38: Estimating the Effects of Global Patent Protection in ...pages.stern.nyu.edu/~acollard/AIDSIndia.pdfcourse, to the extent that our estimates refer to antibiotics, they are not directly

Issues and Exploratory Resultsrdquo In Analyti-cal and Negotiating Issues in the GlobalTrading System ed Alan V Deardorff andRobert M Stern 401ndash46 Ann Arbor MIUniversity of Michigan Press

Mazzoleni Roberto and Richard R Nelson1998a ldquoThe Benefits and Costs of StrongPatent Protection A Contribution to the Cur-rent Debaterdquo Research Policy 27(3) 273ndash84

Mazzoleni Roberto and Richard R Nelson1998b ldquoEconomic Theories about the Bene-fits and Costs of Patentsrdquo Journal of Eco-nomic Issues 32(4) 1031ndash52

Nevo Aviv 2001 ldquoMeasuring Market Power inthe Ready-to-Eat Cereal Industryrdquo Econo-metrica 69(2) 307ndash42

Nogues Julio J 1993 ldquoSocial Costs and Bene-fits of Introducing Patent Protection for Phar-maceutical Drugs in Developing CountriesrdquoDeveloping Economies 31(1) 24ndash53

Nordhaus William D 1969 Invention Growthand Welfare A Theoretical Treatment ofTechnological Change Cambridge MAMIT Press

Pollack Andrew April 20 2001 ldquoNews Analy-sis Defensive Drug Industry Fuels Fight overPatentsrdquo New York Times

Quah Danny 2002 ldquoAlmost Efficient Innova-tion by Pricing Ideasrdquo London School ofEconomics Center for Economic Perfor-mance Working Paper 1219

Rozek Richard P and Ruth Berkowitz 1998ldquoThe Effects of Patent Protection on thePrices of Pharmaceutical Products Is Intel-

lectual Property Protection Raising the DrugBill in Developing Countriesrdquo Journal ofIntellectual Property 1(2) 179ndash243

Sachs Jeffrey D 2002 ldquoA New Global Effort toControl Malariardquo Science 298(5591) 122ndash24

Scherer Frederic M and Jayashree Watal 2002ldquoPost-TRIPS Options for Access to PatentedMedicines in Developing Nationsrdquo Journalof International Economic Law 5(4) 913ndash39

Stevenson Richard W November 15 2001ldquoMeasuring Success At Least the TalksDidnrsquot Collapserdquo New York Times

Subramanian Arvind 1995 ldquoPutting SomeNumbers on the TRIPS Pharmaceutical De-baterdquo International Journal of TechnologyManagement 10(2-3) 252ndash68

Trajtenberg Manuel 1989 ldquoThe Welfare Anal-ysis of Product Innovations with an Appli-cation to Computed Tomography ScannersrdquoJournal of Political Economy 97(2) 444ndash79

Wattal Jayashree 1996 ldquoIntroducing ProductPatents in the Indian Pharmaceutical Sec-tormdashImplications for Prices and WelfarerdquoWorld Competition 20(2) 5ndash21

Wattal Jayashree 2000 ldquoPharmaceutical Pat-ents Prices and Welfare Losses Policy Op-tions for India under the WTO TRIPSAgreementrdquo World Economy 23(5) 733ndash52

Wright Brian D 1983 ldquoThe Economics of In-vention Incentives Patents Prizes and Re-search Contractsrdquo American EconomicReview 73(4) 691ndash707

1514 THE AMERICAN ECONOMIC REVIEW DECEMBER 2006