Do Bilateral Investment Treaties Increase Foreign Direct Investment to Developing Countries 2005...

World Development Vol. 33, No. 10, pp. 1567–1585, 2005� 2005 Elsevier Ltd. All rights reserved

Printed in Great Britain
0305-750X/$ - see front matter
doi:10.1016/j.worlddev.2005.07.001www.elsevier.com/locate/worlddev

Do Bilateral Investment Treaties Increase Foreign

Direct Investment to Developing Countries?

ERIC NEUMAYER and LAURA SPESS *

London School of Economics and Political Science, UK

Summary. — Foreign investors are often skeptical toward the quality of the domestic institutionsand the enforceability of the law in developing countries. Bilateral investment treaties (BITs) guar-antee certain standards of treatment that can be enforced via binding investor-to-state dispute set-tlement outside the domestic juridical system. Developing countries accept restrictions on theirsovereignty in the hope that the protection from political and other risks leads to an increase inforeign direct investment (FDI), which is also the stated purpose of BITs. We provide the first rig-orous quantitative evidence that a higher number of BITs raises the FDI that flows to a developingcountry. This result is very robust to changes in model specification, estimation technique, and sam-ple size. There is also some limited evidence that BITs might function as substitutes for gooddomestic institutional quality, but this result is not robust to different specifications of institutionalquality.
� 2005 Elsevier Ltd. All rights reserved.
Key words — foreign direct investment, bilateral investment treaties, institutional quality, protec-tion, risk

* We thank three anonymous referees as well as Anne

van Aaken, Ken Shadlen, and Susan Rose-Ackerman

for many helpful comments. Eric Neumayer acknowl-

edges financial support from the Leverhulme Trust.

1. INTRODUCTION

Developing countries sign bilateral invest-ment treaties (BITs) in order to attract moreforeign direct investment (FDI). In recent dec-ades, BITs have become ‘‘the most importantinternational legal mechanism for the encour-agement and governance’’ of FDI (Elkins, Guz-man, & Simmons, 2004, p. 0). The preambles ofthe thousands of existing BITs state that thepurpose of BITs is to promote the flow ofFDI and, undoubtedly, BITs are so popular be-cause policy makers in developing countries be-lieve that signing them will increase FDI. Butdo these treaties fulfil their stated purpose andattract more FDI to developing countries thatsubmit to the obligations of a BIT? Despitethe large and increasing number of BITs con-cluded, there exists very little evidence answer-ing this question. Most existing scholarships,typically written with a legal perspective, sim-ply restrict themselves to an analysis of theBIT practice of one country or certain similarprovisions in a range of BITs (Vandevelde,1996, p. 545). This omission is strange giventhat the question is of great importance todeveloping countries. They invest time and

156

other scarce resources to negotiate, conclude,sign, and ratify BITs. Such treaties represent anontrivial interference with the host countries’sovereignty as they provide protections to for-eign investors that are enforceable via bindinginvestor-to-state dispute settlement. While themotivations driving developing countries to in-cur these costs may be varied (see Elkins et al.,2004; Guzman, 1998; Neumayer, 2005), thecosts might be justified if the ultimate outcomeis an increase in the inward flow of FDI. 1 Butis this what actually occurs?In the absence of hard, quantitative evidence,

some observers have been rather pessimistic to-ward the effect of BITs on FDI location. Sorn-arajah (1986, p. 82), for example, suggests that‘‘in reality attracting foreign investment de-pends more on the political and economic cli-mate for its existence rather than on thecreation of a legal structure for its protection.’’An expert group meeting sponsored by the

7

1568 WORLD DEVELOPMENT

United Nations Conference on Trade andDevelopment (UNCTAD) in 1997 reportedlyheld a similar position (Raghavan, 1997). Sup-portive of this view is that some major hosts ofFDI such as Brazil or Mexico for a long timewere reluctant to sign BITs. As UNCTAD(1998, p. 141) has put it in a review of BITsfrom almost a decade ago: ‘‘There are manyexamples of countries with large FDI inflowsand few, if any, BITs.’’ And yet, most develop-ing countries have signed a great many BITs bynow. Is there evidence that those that havesigned more BITs have also managed to attractmore FDI?Two studies analyze this issue over the period

1980–2000 (Hallward-Driemeier, 2003; Tobin& Rose-Ackerman, 2005) and one over the per-iod 1991–2000 (Salacuse & Sullivan, 2005). Thefirst study by Hallward-Driemeier (2003) doesnot find any statistically significant effect. Thesecond study by Tobin and Rose-Ackerman(2005) finds a negative effect at high levels ofrisk and a positive effect only at low levels ofrisk, with the majority of developing countriesfalling into the high-risk category. The thirdstudy by Salacuse and Sullivan (2005) finds apositive effect only for United States BITs,but not for BITs from other countries of theOrganization of Economic Co-operation andDevelopment (OECD). The existing evidencegoes against expectation and would suggestthat the enormous amount of effort developingcountries have spent on BITs has basically beenwasted. One of the problems of existing studiesis that they infer results from a rather restrictedsample of countries (31 and 63, respectively) orare based on cross-sectional regressions. Incontrast, we employ a much larger panel overthe period 1970–2001, covering up to 119 coun-tries. Importantly, we find a positive effect ofBITs on FDI inflows that is consistent and ro-bust across various model specifications. Theeffect is sometimes conditional on institutionalquality, but is always positive and statisticallysignificantly different from zero at all levels ofinstitutional quality. To our knowledge, weprovide the first hard evidence that there is apayoff to developing countries’ willingness toincur the costs of negotiating BITs and to suc-cumb to the restrictions on sovereignty con-tained therein.Having demonstrated that BITs successfully

increase the flow of FDI coming to a country,another important question that we addressis whether BITs function as substitutes orcomplements to good institutional quality.

Naturally, one would expect them to be substi-tutes, that is, they provide security and certainstandards of treatment to foreign investorswhere domestic institutions fail to deliver thesame security and standards. However, some,like Hallward-Driemeier (2003) argue that BITsmight only be seen as credible in an environ-ment of good institutional quality. This wouldimply that BITs are most effective in countrieswhere they are least needed. Our results providesome limited evidence that BITs might functionas substitutes to poor institutional quality,which would suggest that they are most effec-tive where such quality is low and that theyare most successful where they are neededmost. However, this result is not robust to dif-ferent specifications of institutional quality.This article is structured as follows: Section 2

briefly describes the well-known fact of increas-ing importance of foreign investment todeveloping countries, illustrates the growth ofBITs, and analyzes the role of their main provi-sions for the promotion of FDI. We then re-view the three existing empirical studies anddiscuss their shortcomings, which we aim toovercome in our own analysis. After presentingour research design, we report results and testthe sensitivity of results to important changesin model specification. The final section con-cludes.

2. BITS AND FDI

The flow of FDI has dramatically increasedin the past several decades to become a majorforce in the worldwide allocation of funds andtechnology. Prior to 1970, world trade gener-ally grew at a greater pace than that of FDI,but in the decades since then, the flow of FDIhas grown at more than twice the pace of thegrowth of worldwide exports. By the early1990s, the sales of worldwide exports wouldbe eclipsed by the sales of foreign affiliates ofmultinational firms (Dunning, 1998). Not onlyhas the flow of FDI increased worldwide, butthe importance of FDI as a source of funds todeveloping countries in particular has also sig-nificantly increased. Private international flowsof financial resources have become increasinglyimportant to developing countries. In the1980s, tight budgets, the debt crisis, and anoverall decreased interest in providing tradi-tional development aid led to a decline inofficial development assistance from the devel-oped world. When capital flows to developing

BILATERAL INVESTMENT TREATIES 1569

countries began to rise again in the latter partof that decade, the flows were increasingly com-posed of FDI (Zebregs, 1998). Only very re-cently have aid flows slightly increased againin the wake of the so-called Monterrey Consen-sus. However, in 2003, FDI was the largestcomponent of the net resource flows to devel-oping countries, and this is bound to remainthe case for some time to come (UNCTAD,2003a). Although the developed countries re-main both the dominating source and the majorrecipient of FDI, their dominance has de-creased over time with developing countries in2003 receiving almost 31% of FDI as opposedto only about 20% in the 1980s (UNCTAD,2004). Indeed, FDI inflows per unit of GDPare much higher in many developing countriesthan in developed ones (UNCTAD, 2004). Itwas during this same period that BITs wereintroduced and eventually proliferated and inlight of the importance of FDI, particularly todeveloping countries, the extent to which thesetwo phenomena are causally related warrantscareful scrutiny.The first BITs appeared at the end of the

1950s. Some trace their history back to the trea-ties of friendship, commerce, and navigation(FCN) concluded by the United States overcenturies (Salacuse, 1990). The FCN treatieshad the expansion of international trade andthe improvement of US foreign relations astheir prime purpose, even though some invest-ment provisions were later added (Guzman,1998). BITs on the other hand are more clearly

0

500

1000

1500

2000

2500

1960

1963

1966

1969

1972

1975

1978

1

Num

ber

of B

ITs

BITs Signed Per Year

Figure 1. Proliferation of BITs 1960–

focused on foreign investment protection. Ger-many, having lost almost all of her foreigninvestment during the Second World War,signed the very first BIT with Pakistan in1959. After that, it took almost two decades be-fore BITs gained momentum. By the end of the1960s, there were 75 treaties, which rose to 167by the end of the 1970s and to 389 by the end ofthe 1980s. The number of BITs worldwidebegan to grow rapidly in the 1990s, and by2002, there were 2,181 BITs worldwide (UNC-TAD, 2003a). Figure 1 shows the number ofBITs signed per year and the cumulative num-ber of BITs worldwide.In order to explain the popularity of BITs, it

is necessary to understand how they fit into thelarger regime of state–foreign investor rela-tions. Prior to the advent of BITs, the only pro-tection for foreign investors was the customaryinternational legal rule of minimum standard oftreatment and the so-called Hull rule. The min-imum standard of treatment rule provides onlyvery minimal protection, as the name alreadysuggests, while the Hull rule dealt exclusivelywith cases of expropriation and therefore pro-vided no general protection against discrimina-tory treatment. It grew out of a dispute betweenMexico and the United States in the 1930s overproperties expropriated by the government ofMexico. In one of a series of diplomatic notesto the Mexican Minister of Foreign Affairs,the US Secretary of State Cordell Hull statedthat ‘‘no government is entitled to expropri-ate private property, for whatever purpose,

981

1984

1987

1990

1993

1996

1999

2002

YearBITs Signed Cumulative

2003. Source: UNCTAD (2003b).


without provision for prompt, adequate, andeffective payment therefore’’ (Guzman, 1998,p. 641). Subsequently, the rule of ‘‘prompt, ade-quate, and effective’’ compensation would bethe standard known as the Hull rule. However,it is disputed whether the Hull rule representscustomary international law. Developing coun-tries challenged its validity as part of their de-mands for a New International EconomicOrder with some success: Resolution 1803 ofthe UN General Assembly merely requires‘‘appropriate compensation’’ for expropriation(Ginsburg, 2004). Guzman (1998, p. 641) sug-gests that by the mid-1970s ‘‘the Hull rule hadceased to be a rule of customary internationallaw,’’ if ever it had been one. The fact that therewere several spectacular expropriations in the1960s and 1970s taking place without whatinvestors regarded as adequate compensationsupports this view. 2 This raises a doubt onwhether the Hull rule ever represented custom-ary international law, for which conformingstate practice is a requirement.Surprisingly, even as many developing coun-

tries resisted the Hull rule, many of the samecountries began to sign on to BITs that incorpo-rate similar and indeed more far-reaching provi-sions. Guzman (1998) and Elkins et al. (2004)suggest that this seemingly contradictory behav-ior is explained by the increased competitionamong developing countries for FDI from devel-oped countries. Collectively, they would be bet-ter off refusing to sign any BIT and retaining asmuch control over their assets as possible, whichexplains their resistance to multilateral invest-ment treaties at fora such as the UNCTADwhere they can collectively express and organizetheir interests. In a classic example of the pris-oner’s dilemma, however, the individual countrybenefits from being able to provide crediblecommitments to investors. In the context of lim-ited multilateral or customary protection forinvestors, the individual country gains a compet-itive advantage as an investment location bysubmitting to a BIT. Further, when a less devel-oped country’s neighbor or economic competi-tor signs such an agreement, in order to remaincompetitive they must sign one as well. Some-what at odds with this explanation is that the lat-est trend is for developing countries to sign BITsamong themselves. This has been a rather recentdevelopment, however, and the vast majority ofexisting BITs are concluded between a devel-oped and a developing country.The basic provisions of a BIT typically guar-

antee certain standards of treatment for the for-

eign investor (see Dolzer & Stevens, 1995;UNCTAD, 1998). By entering into a BIT, sig-natories agree to grant certain relative standardtreatments such as national treatment (foreigninvestors may not be treated any worse thannational investors, but may be treated betterand, in fact, often are) and most-favored coun-try treatment (privileges granted to one foreigninvestor must be granted to all foreign inves-tors). They also agree to guarantee certainabsolute standards of treatment such as fairand equitable treatment for foreign investorsin accordance with international standardsafter the investment has taken place. BITs typ-ically ban discriminatory treatment against for-eign investors and include guarantees ofcompensation for expropriated property orfunds, and free transfer and repatriation of cap-ital and profits. Further, the BIT parties agreeto submit to binding dispute settlement, shoulda dispute concerning these provisions arise(UNCTAD, 1998). Ostensibly, these provisionsshould secure some of the basic requirementsfor credible protection of property and contractrights that foreign investors look for in hostcountries. They should also protect foreigninvestors against political and other risks highlyprevalent in many developing countries. Farfrom being neutral, foreign investors are oftengranted higher security and better treatmentthan domestic investors (Vandevelde, 1998).The basic provisions of BITs are all direct

answers to the fundamental ‘‘holdup’’ or ‘‘dy-namic inconsistency’’ problem that faces devel-oping countries attempting to attract FDI. Thedynamic inconsistency problem arises from thefact that although host countries have an incen-tive to promise fair and equitable treatmentbeforehand in order to attract foreign invest-ment, once that investment is established andinvestors have sunk significant costs, the hostcountry’s incentive is to exploit or even expro-priate the assets of foreign investors. Eventhose host countries that are willing to foregotaking advantage in these circumstances willfind it very difficult to credibly commit to theirposition. Many developing countries haveadopted domestic legal changes over the lastdecade or so with a view toward encouraginga greater FDI inflow (UNCTAD, 2004). How-ever, these domestic legal rules cannot substi-tute for the commitment device offered byentering into a legally binding bilateral treaty.BITs, and their binding investor-to-state dis-pute settlement provision in particular, aremeant to overcome the dilemma facing host


countries who are willing to denounce exploit-ing foreign investors after the investment hasalready been undertaken. Interestingly, at thesame time as BITs flourished in the 1980s and1990s, outright expropriations of foreign inves-tors, which were common during the 1960s and1970s, practically ceased to take place (Minor,1994).The extent of interference with domestic reg-

ulatory sovereignty that developing countriessuccumb to in signing BITs is enormous. Infact, virtually any public policy regulation canpotentially be challenged through the disputesettlement mechanism as long as it affects for-eign investors. Often, foreign investors neednot have exhausted domestic legal remediesand can thus bypass or avoid national legal sys-tems, reaching straight for international arbi-tration, where they can freely choose one ofthe three panelists, where their consensus isneeded for one other panelist, and where theycan expect that the rules laid out in the BITsare fully applied (Peterson, 2004). This con-trasts with domestic courts, where investorshave no say on the composition of judges andwhere domestic rules might trump BIT provi-sions. BITs have been criticized for not con-forming to a truly liberal economic model byfailing to ban distorting government policiessuch as protective tariffs or tax incentives forforeign investors (Vandevelde, 2000). However,even critics such as Vandevelde (2000, p. 499)admit that ‘‘BITs seriously restrict the abilityof host states to regulate foreign invest-ment.’’ 3 In concluding BITs, developing coun-tries are therefore ‘‘trading sovereignty forcredibility’’ (Elkins et al., 2004, p. 4).BITs are therefore an important instrument

of protection to foreign investors, for whichthere is currently not much legal alternative.Only few regional free trade agreements con-tain investment protection provisions like theNorth American Free Trade Agreement (NAF-TA). The World Trade Organization’s (WTO)Agreement on Trade-Related InvestmentMeasures (TRIMs Agreement) imposes onlyrudimentary disciplines on the regulation offoreign investment that are by far not as com-prehensive as and fall much short of provisionscontained in BITs.Of course, not all BITs are identical in their

provisions. Some developed country investorssuch as the United States insist on some limitedrights of its investors to establish investment inhost countries in the first place, whereas inves-tor’s rights in most BITs are restricted to fair

and equitable treatment after the investmenthas already taken place and provide no rightof entrance. United States BITs often prohibitcertain performance requirements such as localcontent, export, and employment requirementsbeyond the requirements contained in theWTO’s TRIMs Agreement, whereas BIT pro-grams of other developed countries do not con-tain such provisions (Vandevelde, 1998).Conversely, some nondeveloped countries suchas China and Eastern European countries havesuccessfully managed to restrict the compulsorydispute settlement provisions to disputes con-cerning expropriation or the compensationthereof (Peters, 1996, p. 107). However, byand large, BITs tend to be rather similar intheir provisions.BITs are also unlikely to be identical in their

effect on incoming FDI flows. In principle, BITprovisions only protect investors from the sig-natory states to whom binding commitmentsare made. 4 One would therefore expect thatsigning a BIT with a major capital exportersuch as Germany or the United States has a lar-ger impact on FDI inflows than signing a BITwith minor capital exporters such as New Zea-land or Portugal. However, the signing of BITssends out a signal to potential investors that thedeveloping country is generally serious aboutthe protection of foreign investment. Theencouragement of FDI flows therefore neednot be restricted to investors from developedcountries that are BIT partners of the develop-ing country. Instead, BITs can have positivespill-over effects. How important is the signal-ing effect, which benefits investors from allcountries, compared to the commitment effect,which only relates to investors from BIT part-ner countries, is difficult to say. Our researchdesign, described in detail below, does not re-strict the effect of BIT signature on FDI toinvestment from partner countries and ac-counts for differences in the size of potentialFDI to which the developing country has madebinding commitments by weighting BITs withthe relative importance of the developed coun-try partner as a capital exporter.

3. REVIEW OF OTHER STUDIES

It is most astonishing that despite the risingnumber of BITs, there are only three other seri-ous studies examining the effect of such treatieson the location of FDI. 5 The first studyhas been undertaken by Hallward-Driemeier


(2003), looking at the bilateral flow of FDIfrom 20 OECD countries to 31 developingcountries over the period 1980–2000. Her re-search design is dyadic, consisting of up to537 country pairs. Using fixed-effects estima-tions, she finds that the existence of a BIT be-tween two countries does not increase theflow of FDI from the developed to the develop-ing signatory country. This is true whether thedependent variable is measured as absoluteflows, flows divided by host country’s GDP,or the share of the source countries’ FDI out-flow. Interacting the BIT variable with variousmeasures of institutional quality, she finds a po-sitive coefficient of the interaction term that isoften statistically significant. This would sug-gest that, contrary to theoretical expectations,BITs are complements to good institutionalquality and therefore do not perform their ori-ginal function, namely to provide guarantees toforeign investors in the absence of good domes-tic institutional quality.In the second study, Tobin and Rose-Acker-

man (2005) analyze the impact of BITs on gen-eral nondyadic FDI inflows, also in a panelfrom 1980 to 2000, but with data averaged overfive-year periods, covering 63 countries. Whileboth studies draw upon data provided by Inter-national Country Risk Guide (ICRG), Hall-ward-Driemeier (2003) uses individualcomponents of institutional quality, whereasTobin and Rose-Ackerman (2005) use theaggregate political risk measure, which includesmany more components than institutionalquality, including some that are not directly re-lated to political risk (such as, among others,religious and ethnic tensions, armed conflict,and socioeconomic conditions such as unem-ployment and poverty). In a fixed-effects model,Tobin and Rose-Ackerman find that a highernumber of BITs either in total or signed witha high income country lowers the FDI a coun-try receives as a share of global FDI flows athigh levels of risk and raises the FDI only atlow levels of risk. In an additional dyadic anal-ysis of 54 countries, they fail to find any statis-tically significant effect of BITs signed with theUnited States on FDI flows from the UnitedStates to developing countries, either condition-ally on the level of political risk or uncondition-ally.The third study provides three cross-sectional

analyses of FDI inflows to up to 99 developingcountries in the years 1998, 1999, and 2000,respectively, as well as a fixed-effects estimationof the bilateral flow of FDI from the United

States to 31 developing countries over the per-iod 1991–2000. Salacuse and Sullivan (2005)find the signature of a BIT with the UnitedStates to be associated with higher FDI inflowsin both types of estimations, whereas the num-ber of BITs with other OECD countries isalways statistically insignificant.The three studies suffer from a number of

shortcomings that we try to improve on inour own study. Hallward-Driemeier’s (2003)model presumes that a BIT will only have an ef-fect on the flow of FDI from one developedcountry, namely the signatory, to the develop-ing country. However, this presumption ne-glects the signaling effect of BITs (Elkinset al., 2004, p. 21). As pointed out in the pre-ceding section, in concluding a BIT, the devel-oping country commits to protect foreigninvestments, explicitly only the FDI from thesignatory developed country, but implicitly italso signals its willingness to protect all foreigninvestment. There are therefore likely to bepositive spill-over effects from signing a BIT.Hallward-Driemeier’s modeling cannot capturethe potential of BITs to attract more FDI fromother developed nonsignatory countries as well,and consequently may underestimate the effectthat signing a BIT has on the inward flow ofFDI. In addition to not capturing this poten-tially important spill-over effect, the dyadic de-sign also has another major disadvantage. Dataon bilateral FDI flows are very sparse, and con-sequently, the size of the sample is significantlylimited by this choice. A sample of 31 develop-ing countries is everything but representative.Similar arguments apply to Salacuse and Sulli-van’s (2005) fixed-effects analysis. 6 Our ownstudy draws from a much larger and more rep-resentative sample.Where Tobin and Rose-Ackerman (2005) do

not use a dyadic research design, the paucity ofbilateral FDI flow data does not impose a bind-ing constraint on sample size. Nevertheless, forno clear reason, their sample consists of only 63countries. In comparison, our own sample isboth deeper and wider. It covers the period1970, the first year for which UNCTAD pro-vides FDI data, to 2001, the last year for whichwe have available data. It also covers up to 119developing countries, which amounts to a muchmore representative sample. The countries in-cluded in our sample are listed in AppendixA. Salacuse and Sullivan’s (2005) cross-sec-tional analysis also has the advantage of a largesample size. However, by definition, this type ofanalysis cannot control for country-specific


unobserved heterogeneity, which is likely to beimportant, nor does it exploit the full informa-tion available from looking at FDI flows over alonger time period.

4. RESEARCH DESIGN

(a) Dependent variable

As our main measure of FDI attractiveness,we use the absolute amount of FDI going toa developing country, converted to constantUS$ of 1996 with the help of the US GDPdeflator. The definition of a developing countryfollows World Bank classification. We useabsolute FDI flows because if one were to useFDI inflow as a percentage of host country’sGDP instead, the measure would capturechanges in the relative importance of foreigninvestment to the host country, but not changesin inflows directly. Quite possibly, the world-wide increase in the rate of the conclusion ofBITs is partly responsible for the increase inoverall FDI going to developing countries.However, there is always the danger that onefinds a statistically significant relationship be-tween two upward trending variables that isspurious. We deal with this potential problemin two ways: First, we employ year dummiesto account for any year-to-year variation intotal FDI flows unaccounted for by our explan-atory variables, which should mitigate potentialspuriousness of any significant results. Second,as an alternate dependent variable to the abso-lute amount of FDI, we use the FDI inflow acountry receives relative to the sum of FDIgoing to developing countries. Since this vari-able is not trending over time, no year-specifictime dummies are included in these sets of esti-mations. FDI inflow as a share of developingcountry FDI as the dependent variable capturesthe relative attractiveness of developing coun-tries as hosts for FDI and explicitly allows forcompetition among them for a fixed sized cakeof FDI to be divided. 7 Ideally, one would liketo disaggregate FDI flows according to eco-nomic sectors. Unfortunately, no comprehen-sive information is available for a large panelof countries.We take the natural log of the dependent var-

iable to reduce the skewness of its distribution.This increases the model fit substantially. To doso, we need to recode a small number of nega-tive FDI flows. Negative FDI flows essentiallyimply ‘‘instances of reverse investment or disin-

vestment’’ (UNCTAD, 2001, p. 292). In ouranalysis, we set negative FDI flows equal to po-sitive FDI flows of US$1. If instead one were todiscard all negative flows, then results arehardly affected.

(b) Explanatory variables

Our main explanatory variable is the cumula-tive number of BITs a developing country hassigned with OECD countries, weighted by theshare of outward FDI flow the OECD countryaccounts for relative to total world outwardFDI flow. 8 The weighting is to account for dif-ferences in the size of potential FDI a develop-ing country makes protection commitments forvia signing a BIT. We exclude BITs signed be-tween developing countries since FDI flows be-tween developing countries are rare. Tobin andRose-Ackerman (2005) do not weight thecumulative BIT variable by the share of out-ward FDI flow of the developed country part-ner, but our results are similar if we includethe unweighted BIT variable instead. 9 Theyalso take the natural log of the cumulativeBIT variable. We keep the variable in its levelform, not least because the log of zero (BITs)is undefined.Our control variables are very similar to the

ones used by Hallward-Driemeier (2003) andTobin and Rose-Ackerman (2005). They arealso among the ones more consistently foundto be determinants of FDI flows (Chakrabarti,2001). We include the natural log of per capitaincome, the log of total population size, and theeconomic growth rate as indicators of marketsize and market potential (data from WorldBank, 2003a, 2003b). 10 Developing countries,which have concluded a free trade agreementwith a developed country, might receive moreFDI as it is easier to export goods back intothe developed or other countries. Such agree-ments sometimes also contain provisions onpolicies that might be beneficial to foreigninvestors. We account for this with a dummyvariable indicating whether a country is a mem-ber of the WTO as well as a variable countingthe number of bilateral trade agreements adeveloping country has concluded with theUnited States, the European Community/Euro-pean Union, or Japan, based on informationcontained in WTO (2004) and EU (2004). 11

The inflation rate is a proxy variable for macro-economic stability. In sensitivity analysis, wealso included trade openness and the secondaryenrollment ratio, but these two variables are


not included in the main analysis due to loss ofobservations following their inclusion. Data aretaken from World Bank (2003a). We employ ameasure of natural resource intensity to controlfor the fact that, all other things being equal,large natural resources are a major attractorto foreign investors. Our measure is equal tothe sum of rents from mineral resource and fos-sil fuel energy depletion divided by gross na-tional income, as reported in World Bank(2003b). Rents are estimated as (P � AC) * R,that is, as price minus average cost multipliedby the amount of resource extracted, anamount known as total Hoteling rent in thenatural resource economics literature.There is a long tradition of studies analyzing

the effect of political stability and institutionalquality on FDI inflows (see, e.g., Alesina & Per-otti, 1996; Globerman & Shapiro, 2002; Perry,2000; Schneider & Frey, 1985; Wheeler &Mody, 1992). We include five different mea-sures of institutional quality in separate estima-tions together with interaction effects with theBIT variable. First, we use the political con-straints (POLCON) index developed by Henisz(2000). Henisz has designed his index as anindicator of the ability of political institutionsto make credible commitments to an existingpolicy regime, which he argues is the most rele-vant political variable of interest to investors.Building on a simple spatial model of politicalinteraction, the index makes use of the struc-ture of the government in a given country andthe political views represented by the differentlevels of that government (i.e., the executiveand the lower and upper legislative chambers).It measures the extent to which political actorsare constrained in their choice of future policiesby the existence of other political actors withveto power who will have to consent. Usinginformation on party composition of the execu-tive and the legislative branches allows takinginto account how alignment across branchesof the government and the extent of preferenceheterogeneity within each legislative branch im-pacts the feasibility of policy change. Scoresrange from 0, which indicates that the executivehas total political discretion and could changeexisting policies at any point of time, to 1,which indicates that a change of existing poli-cies is totally infeasible. Of course, in practice,agreement is always feasible, so the maximumscore is less than 1.The remaining measures of institutional qual-

ity are all compiled from the ICRG, publishedby the Political Risk Services (PRS) Group

(www.prsgroup.com). They are the investmentprofile index, the index of government stability,an index of law and order as well as ICRG’scomposite political risk index. These data areavailable from 1984 onwards and have the wid-est country coverage of all the sources for coun-try risk ratings. The composite political riskindex incorporates the other indices. The indexvaries from 0 (high risk) to 100 (low risk). Thefull composite political index or a comparablemeasure is more commonly used in studiesinvestigating the determinants of FDI and BITsthan the individual subcomponents (see, e.g.,Alesina & Perotti, 1996; Tobin & Rose-Acker-man, 2005; UNCTAD, 1998; Wheeler & Mody,1992). However, as mentioned already, it alsoincludes many items that are not strictly speak-ing relevant to institutional quality such as indi-cators of socioeconomic conditions, measuresof conflict and ethnic tensions, and measuresof military and religious involvement in thepolitical process—see Appendix B for a de-tailed description of this variable. The compos-ite index is therefore strictly speaking notpurely a measure of institutional quality. Forthis reason, we also employ the three subcom-ponents most relevant to investors and mostclosely connected to BIT provisions. Theinvestment profile index varies from 0 to 12,12 representing very low risk and 0 indicatingvery high risk. The index is made of ratingson three separate elements, each receiving equalweight: contract viability (risk of expropria-tion), profit repatriation, and payment delays.Similar to the investment profile index, the gov-ernment stability index varies from 0 (high risk)to 12 (low risk) and is composed of three ele-ments that receive equal weight: governmentunity, legislative strength, and popular support.The law and order index runs from 0 (worst) to6 (best) and consists of a law component mea-suring the strength and impartiality of the legalsystem and an order component measuring theextent of popular observance of the law.

(c) Estimation technique

We estimate both random-effects and fixed-effects models, in which case we can employ ro-bust standard errors. We suspect that there arefactors making a country attractive to foreigninvestors that are not captured by our explana-tory variables and that are (approximately)time invariant, such as colonial history, culture,language, climate, geographical distance to thecenters of the Western developed world, legal

http://www.prsgroup.com


restrictions on inward FDI, etc. Hausman tests,which test the random-effects assumption thatthese time-invariant factors are uncorrelatedwith the explanatory variables, by and large re-ject the assumption. We therefore focus on thefixed-effects estimation results. To mitigate po-tential reverse causality problems, we lag allexplanatory variables by one period. Ideally,one would like to tackle this problem morecomprehensively with the help of instrumentalvariable regression. However, practically allexplanatory variables are potentially subjectto reverse causality, and it would be simplyimpossible to find adequate and valid instru-ments. Tables 1a and 1b provide summarydescriptive variable information together witha bivariate correlation matrix. Variance infla-tion analysis did not suggest reason for concernwith multicollinearity problems. As in anyregressions analysis, there is of course alwaysthe possibility of omitted variable bias. Forexample, we cannot account for over-timechanges in domestic legislation encouraging ordiscouraging FDI other than what is capturedby BITs as there is no comprehensive informa-tion available. However, we see no reason whythis or any other potentially omitted variableshould be systematically correlated with ourexplanatory variables to an extent that our re-sults would be significantly biased.

5. RESULTS

Table 2 presents random-effects estimationresults for the logged amount of FDI in US$

Table 1a. Descriptive statis

Variable Obs. Mean

ln FDI flow 2,767 3.92ln FDI flow share 2,767 �7.06BITs 2,767 23.97ln GDP p.c. 2,767 7.93In population 2,767 15.72Economic growth 2,767 0.01Inflation 2,767 65.78Resource rents 2,767 5.55Bilateral trade agreements 2,767 0.07WTO membership 2,767 0.67POLCON 2,767 0.18Composite political risk 1,331 58.71Investment profile 1,351 6.03Government stability 1,350 6.88Law and order 1,349 3.09

of 1996 flowing to a country as the dependentvariable. Column I starts with POLCON asthe measure of institutional quality. Most vari-ables test in accordance with theoretical expec-tations. Countries with a higher cumulativenumber of BITs, richer countries with fast-growing economies, and larger populations re-ceive more FDI. So do countries that are moreintensive in natural resource extraction, thatare members of the WTO and have a highernumber of trade agreements with developedcountries. A higher inflation rate deters FDI.The POLCON variable is statistically insignifi-cant. How to interpret this result? With interac-tion terms included, one cannot interpret thecoefficients on the individual components inthe conventional way. Instead, the coefficienton POLCON in a model with a significantinteraction term BITs * POLCON is the effectof POLCON on FDI when the BIT variableis zero (see Braumoeller, 2004, for a nice expo-sition). It follows from our estimations thatinstitutional quality as measured by POLCONhas no effect on FDI in the absence of BITs.The interaction term is marginally significant,however, suggesting that institutional qualityas measured by POLCON and BITs are com-plements since the positive effect of cumulativeBIT signature is higher when the POLCONindex is high as well.In column II, we replace Henisz’s (2000) pol-

icy constraints variable with the ICRG’s com-posite political risk index (where higher valueson the index mean lower risk). Note the signif-icant drop in the number of observations. Thisis not so much due to a loss of countries

tical variable information

Std. dev. Min Max

2.52 �4.69 10.782.45 �16.98 �1.1726.98 0 99.340.83 5.64 9.721.88 10.62 20.990.07 �0.42 0.78684.46 �31.52 26,762.029.76 0 66.600.26 0 20.47 0 10.20 0 0.6714.27 11.50 94.171.91 1 11.132.32 1 121.27 0 6

Table 1b. Bivariate correlation matrix

I II III IV V VI VI VIII IX X XI XII XIII XIV XV

I: ln FDI flow share 1.00II: ln FDI flow 0.95 1.00III: BITs 0.41 0.28 1.00IV: ln GDP p.c. 0.49 0.48 0.31 1.00V: In population 0.43 0.45 0.21 �0.02 1.00VI: Economic growth 0.16 0.15 0.08 0.13 0.07 1.00VII: Inflation �0.07 �0.06 �0.05 �0.07 0.02 �0.16 1.00VIII: Resource rents 0.09 0.14 �0.12 0.13 0.07 �0.04 0.02 1.00IX: Bilateral trade agreements 0.21 0.15 0.30 0.30 0.05 0.03 �0.03 0.04 1.00X: WTO membership 0.07 0.02 0.05 �0.03 0.02 �0.02 0.02 �0.21 �0.09 1.00XI: POLCON 0.36 0.28 0.16 0.41 �0.02 0.08 �0.02 �0.13 0.08 0.19 1.00XII: Investment profile 0.38 0.23 0.26 0.33 0.04 0.14 �0.12 �0.08 0.22 0.18 0.29 1.00XIIII: Government stability 0.31 0.19 0.25 0.29 �0.02 0.14 �0.15 �0.09 0.20 0.20 0.24 0.49 1.00XIV: Law and order 0.35 0.17 0.37 0.28 0.07 0.15 �0.12 0.00 0.23 0.04 0.18 0.58 0.68 1.00XV: Composite political risk 0.40 0.28 0.28 0.38 0.03 0.15 �0.10 �0.05 0.24 0.03 0.22 0.58 0.35 0.46 1.00

1576WORLD

DEVELOPMENT


included in the sample, which drops from 120to 91, but due to the fact that we lose all obser-vations before 1984, the first year for which thisvariable has been coded. Despite the substan-tial reduction in sample size, the weighted

Table 2. Random-effects estimation results

I

BITs 0.015 0.(5.34)*** (3.7

ln GDP p.c. 0.548 0.(5.53)*** (1

In population 0.506 0.(10.27)*** (6.9

Economic growth 1.195 1.(2.52)** (2.

Inflation �0.0001 �0(2.07)** (1

Resource rents 0.025 0.(4.34)*** (3.6

Bilateral trade agreements 0.343 0.(1.83)* (1

WTO membership 0.212 �0(1.98)** (0

POLCON 0.350(1.17)

BITs * POLCON 0.012(1.78)*

Composite political risk 0.(1

BITs * composite political risk �0(1

Investment profile

BITs * investment profile

Government stability

BITs * government stability

Law and order

BITs * law and order

Observations 2,767 1,Countries 120Period 1970–2001 1984R-squared (overall) 0.46 0Hausman test 28.58 72

0.8659 0.0

Notes: Absolute t-values in parentheses. Year-specific timHausman test is asymptotically v2 distributed with p-value* Significant at the 0.1% level.** Significant at the 0.05% level.*** Significant at the 0.01% level.

sum of BITs variable is statistically significantwith the expected positive sign. Results on theother variables are also relatively consistent interms of sign and statistical significance of vari-ables, but the trade agreement variables are no

(logged FDI flows in 1996, in dollars)

II III IV V

031 0.012 0.020 0.0209)*** (2.17)** (3.66)*** (3.96)***

265 0.238 0.294 0.245.68)* (1.50) (1.86)* (1.64)*

594 0.626 0.610 0.6258)*** (7.24)*** (7.08)*** (7.96)***

683 1.553 1.602 1.59742)** (2.19)** (2.25)** (2.26)**

.0001 �0.0001 �0.0001 �0.0001.73)* (1.46) (1.65)* (1.64)030 0.023 0.022 0.0211)*** (2.76)*** (2.62)*** (2.59)***

278 0.160 0.344 0.288.34) (0.72) (1.56) (1.34).018 �0.024 0.066 0.137.12) (0.16) (0.44) (0.92)

011.81)*

.0002.73)*

0.091(2.16)**

0.001(1.08)

0.091(2.34)**

�0.001(0.77)

0.247(3.56)***

�0.001(0.87)

346 1,369 1,368 1,36791 91 91 91–2001 1984–2001 1984–2001 1984–2001.49 0.50 0.49 0.51.11 68.46 73.93 90.24000 0.0000 0.0000 0.0000

e dummies included, but coefficients not reported. Thes in brackets.


longer statistically significant. There is one fur-ther important further exception: The interac-tion term between institutional quality andour BIT variable is now statistically significantwith a negative coefficient sign. This would sug-gest that BITs function as substitutes for highdomestic institutional quality since the positive

Table 3. Fixed-effects estimation results

I

BITs 0.016 0.(4.23)*** (3.6

ln GDP p.c. 1.916 3.(4.04)*** (4.3

In population �1.344 �4(2.66)*** (5.2

Economic growth 1.134 2.(1.81)* (3.3

Inflation �0.0001 �0(3.16)*** (2.

Resource rents 0.030 0.(3.63)*** (2.9

Bilateral trade agreements 0.532 0.(2.17)** (0

WTO membership 0.218 �0(1.98)** (0

POLCON 0.233(0.71)

BITs * POLCON 0.011(1.29)

Composite political risk 0.(2.

BITs * composite political risk �0(2.

Investment profile




Law and order


Observations 2,767 1,Countries 120Period 1970–2001 1984R-squared (within) 0.22 0

Notes: Absolute t-values in parentheses. Year-specific timestandard errors.* Significant at the 0.1% level.** Significant at the 0.05% level.*** Significant at the 0.01% level.

effect of cumulative BIT signature is higherwhen the ICRG composite index is low, thatis, in high-risk environments. Importantly,while the positive effect of BITs on FDI de-creases as political risk is reduced, the effect al-ways remains positive, even at very low levels ofrisk. 12 In the remaining columns, we replace

(logged FDI flows in 1996, in dollars)

II III IV V

033 0.015 0.025 0.0208)*** (2.93)*** (4.52)*** (3.98)***

771 3.304 4.052 3.6917)*** (3.60)*** (4.47)*** (4.21)***

.942 �4.513 �5.176 �5.0331)*** (4.51)*** (5.19)*** (5.24)***

372 2.366 2.343 2.4645)*** (3.13)*** (3.12)*** (3.24)***

.0001 �0.0001 �0.0001 �0.000136)** (2.18)** (2.41)** (2.48)**

036 0.031 0.031 0.0286)*** (2.41)** (2.47)** (2.16)**

119 0.061 0.289 0.199.66) (0.32) (1.44) (1.02).081 �0.111 �0.047 0.027.52) (0.73) (0.31) (0.18)

01405)**

.000317)**

0.117(2.84)***

�0.000(0.01)

0.128(2.88)***

�0.001(2.12)**

0.290(4.13)***

�0.002(1.14)

346 1,369 1,368 1,36791 91 91 91–2001 1984–2001 1984–2001 1984–2001.30 0.30 0.30 0.30

dummies included, but coefficients not reported. Robust


the ICRG’s composite index with the selectedindividual subcomponents. In column III, welook at a country’s investment profile, in col-umn IV at a country’s governmental stabilityand in column V at its degree of law and order,respectively. Results are largely consistent withthe ones for the ICRG composite index. Impor-tantly, the weighted cumulative number of

Table 4. Fixed-effects estimation results (logged

I

BITs 0.013 0.(3.47)*** (4.0

ln GDP p.c. 3.008 1.(10.10)*** (3.1

In population �2.613 �2(11.45)*** (5.1

Economic growth 1.334 2.(2.15)** (3.1

Inflation �0.0001 �0(3.27)*** (2.

Resource rents 0.027 0.(3.47)*** (2.7

Bilateral trade agreements 0.539 0.(2.25)** (1

WTO membership 0.133 �0(1.26) (0

POLCON 0.184(0.57)


Composite political risk 0.(2.

BITs * composite political risk �0(2.

Investment profile




Law and order


Observations 2,767 1,Countries 120Period 1970–2001 1984R-squared (within) 0.09 0

Notes: Absolute t-values in parentheses. Robust standard e* Significant at the 0.1% level.** Significant at the 0.05% level.*** Significant at the 0.01% level.

signed BITs is always statistically significantwith the expected positive sign. However, theinteraction term is not statistically signifi-cant for these subcomponents of the ICRGcomposite index. The Hausman test fails to re-ject the random-effects assumption only in col-umn I, which underlines the importance ofcontrolling for country fixed effects. Table 3

FDI flows as share of developing country FDI)

II III IV V

035 0.018 0.027 0.0234)*** (3.33)*** (5.02)*** (4.65)***

916 1.754 2.144 2.0473)*** (3.03)*** (3.42)*** (3.52)***

.686 �2.654 �2.864 �2.8443)*** (5.66)*** (5.37)*** (5.98)***

268 2.348 2.225 2.1468)*** (3.06)*** (2.93)*** (2.81)***

.0001 �0.0001 �0.0001 �0.000133)** (2.10)** (2.38)** (2.53)**

032 0.027 0.028 0.0270)*** (2.18)** (2.24)** (2.18)**

216 0.113 0.343 0.302.29) (0.60) (1.82)* (1.65)*

.049 �0.075 �0.017 0.031.31) (0.49) (0.11) (0.20)

01300)**

.000304)**

0.108(2.69)***

�0.000(0.02)

0.104(2.59)***

�0.001(2.11)**

0.231(3.36)***

�0.002(1.20)

346 1,369 1,368 1,36791 91 91 91–2001 1984–2001 1984–2001 1984–2001.07 0.07 0.06 0.07

rrors.


therefore reports results from fixed-effects esti-mation.Fixed-effects estimation results are rather

similar to the ones from random-effects estima-tion with two important exceptions: First, theinteraction term between POLCON and theBIT variable is now insignificant, whereasthe interaction term with governmental stabil-ity is significant and negative, suggesting thatBITs and governmental stability function assubstitutes. As before, the positive effect ofBITs on FDI becomes smaller the more stablethe governments are, but never to an extentthat the effect would become negative. Second,the log of population size now switches signsand is statistically significant with a negativecoefficient. Keeping in mind that the fixed-effects estimation is based on the within-variation of the data in each country only,whereas the random-effects estimation is basedon both cross-country and within-variation,this can be interpreted to the effect that coun-tries with larger populations receive more FDIconditional on the other explanatory variables,but as a country’s population grows, it receivesless rather than more FDI conditional on theother explanatory variables and the country-specific fixed effects.Table 4 presents results for the logged coun-

try share of developing country FDI as thedependent variable. The reported results arebased on fixed-effects estimation, since Haus-man tests overwhelmingly rejected the ran-dom-effects assumption in all cases. Resultsare very similar to the ones reported abovefor the other dependent variable. In particular,a higher cumulative number of signed BITs isassociated with a higher share of FDI inflows,so is institutional quality with the exceptionof POLCON. As before, there is some limitedevidence that BITs and institutional qualityare substitutes for each other, but the interac-tion term is only statistically significant with anegative coefficient sign for the compositeICRG index and its governmental stability sub-component.

6. SENSITIVITY ANALYSIS

Lagging the explanatory variables by oneyear mitigates potential simultaneity bias, butthis lag length is somewhat arbitrary. The posi-tive impact of the BIT variable on FDI inflowsis maintained if the lag length is two, three, orfour years instead. 13 Tobin and Rose-Acker-

man (2005) use five-year period averages toavoid the impact of year-to-year variation.Maintaining a lag of one year, but averagingthe data over five-year periods does not dra-matically change our results as the fixed-effectsestimation results reported in Table 5 attestto, with the share of developing country FDIinflow as the dependent variable. With lessvariation in the data, we are not surprised tofind that some variables lose statistical signifi-cance. Importantly, however, the BIT variableremains statistically significant with the ex-pected positive sign throughout. The institu-tional variables and interaction terms also testalmost as before in terms of sign and whetherthey are statistically significantly differentfrom zero. The exception is the interaction be-tween the BIT variable and governmental sta-bility, which is only marginally insignificant,however.In a further nonreported sensitivity analysis,

we briefly explored the issue of whether BITsare all about a signaling effect. Maybe the onlything that matters is that a developing countryhas signed one BIT (perhaps with a major cap-ital exporter) and signing any more BITs has noadditional effect. When we added a dummy var-iable that is set to one in case a developingcountry has signed a BIT with any developedcountry, then this dummy variable was statisti-cally significant with the expected positive sign.However, as before, the weighted cumulativesum of BITs remained positive and statisticallysignificant as well with its magnitude onlyslightly reduced. This remains true if we replacethe dummy variable with one for BITs signedwith any of the six major capital exporters,namely, France, Germany, Japan, the Nether-lands, the United Kingdom, and the UnitedStates. We interpret this as further evidencethat BITs are likely to fulfil the dual functionof both signaling and commitment.Next, we checked whether our main results

are due to the presence of problematic coun-tries. In particular, we excluded all EasternEuropean and former Soviet Union countriesfrom the sample since, with the exception ofHungary, these countries do not seem to be in-cluded in the analysis by Tobin and Rose-Ack-erman (2005). However, results hardly change.We also excluded countries with a populationsize of less than one million from the sampleto eliminate the influence of very small coun-tries, but the results were hardly affected. Weeven restricted the list of countries to be exactlythe same as in Tobin and Rose-Ackerman

Table 5. Five-year period averages fixed-effects estimation results (logged FDI flows as share of developingcountry FDI)

I II III IV V

BITs 0.014 0.048 0.030 0.042 0.034(2.15)** (3.80)*** (2.91)*** (4.16)*** (3.68)***

ln GDP p.c. 3.036 1.948 1.792 2.099 1.957(6.47)*** (2.33)** (2.17)** (2.39)** (2.47)**

In population �2.601 �2.766 �2.702 �2.849 �2.944(6.92)*** (3.50)*** (3.63)*** (3.58)*** (4.25)***

Economic growth 4.337 3.488 3.300 3.564 4.014(2.14)** (1.21) (1.03) (1.11) (1.29)

Inflation �0.0001 �0.0001 �0.0001 �0.0001 �0.0001(1.95)* (0.97) (1.06) (1.21) (1.23)

Resource rents 0.029 0.039 0.042 0.045 0.045(1.94)* (1.55) (1.69)* (1.90)* (1.84)*

Bilateral trade agreements 1.112 0.482 0.336 0.634 0.530(2.26)** (1.33) (0.79) (1.50) (1.38)

WTO membership 0.242 �0.245 �0.231 �0.146 �0.099(1.17) (0.82) (0.77) (0.49) (0.34)

POLCON �0.066(0.10)


Composite political risk 0.020(1.72)*

BITs * composite political risk �0.000(1.79)*

Investment profile 0.138(1.64)*

BITs * investment profile �0.001(0.37)

Government stability 0.127(1.71)*

BITs * government stability �0.002(1.61)

Law and order 0.341(3.01)***

BITs * law and order �0.002(0.91)

Observations 637 317 314 314 314Countries 120 91 91 91 91Period 1970–2001 1984–2001 1984–2001 1984–2001 1984–2001R-squared (within) 0.19 0.19 0.20 0.19 0.21

Notes: Absolute t-values in parentheses. Robust standard errors.* Significant at the 0.1% level.** Significant at the 0.05% level.*** Significant at the 0.01% level.


(2005) and still found a positive effect of BITson FDI throughout, with one exception (invest-ment profile as measure of institutionalquality), in which case the BIT variable is mar-ginally insignificant with a positive coefficientsign. This holds true both in the annual andin the five-year period model specification.

Next, we wanted to test more formallywhether results are driven by a few influentialoutliers. Belsley, Kuh, and Welsch (1980) sug-gest that observations that have both highresiduals and a high leverage deserve specialattention. We excluded an observation if itsso-called DFITS is greater than twice the


square root of k/n, where k is the number ofindependent variables and n, the number ofobservations. DFITS is defined as the squareroot of hi/(1 � hi), where hi is an observation’sleverage, multiplied by its studentized residual.Applying this criterion leads to the exclusion ofup to 385 observations. Results are remarkablyconsistent, however.In a further sensitivity analysis, we also in-

cluded a measure of trade openness, whichhas a theoretically ambiguous effect on FDI(Taylor, 2000). On the one hand, countries thatare more open to trade can be more attractivehost countries if the main purpose of foreigninvestment is to export the goods or servicesproduced. On the other hand, high trade barri-ers could make it in a company’s best interest tolocate production within the host country inorder to circumvent the import barriers. Tradeopenness tested sometimes insignificant, some-times (marginally) significant with a negativecoefficient sign in the estimations, hardly affect-ing the results of the other variables. Since itsinclusion would reduce the sample size byabout 20 countries, we did not include this var-iable in the reported estimations. FollowingNoorbakhsh, Paloni, and Youssef (2001), wealso included the secondary enrollment ratioto account for human capital as a determinantof FDI in nonreported analysis. The number ofobservations dropped substantially, leaving re-sults for the other variables mainly unaffected.The enrollment ratio itself is never statisticallysignificant, even though it is positively signedin line with expectations.

7. CONCLUSION

Developing countries that sign more BITswith developed countries receive more FDI in-flows. The effect is robust to various samplesizes, model specifications, and whether or notFDI flows are normalized by the total flow ofFDI going to developing countries. There issome limited evidence that BITs function assubstitutes for institutional quality, as in afew estimations the interaction term betweenthe accumulated number of BIT variable andinstitutional quality is negative and statisticallysignificant. The message to developing coun-tries therefore is that succumbing to the obliga-tions of BITs does have the desired payoff ofhigher FDI inflows. To our knowledge, oursis the first study to provide robust empirical evi-dence that BITs fulfil their stated objective.

Those with particularly poor domestic institu-tional quality possibly stand most to gain fromBITs, but there is no robust and consistent evi-dence for this.Why do we come to different conclusions than

the three other relevant studies? Hallward-Driemeier’s (2003) study does not allow for asignaling effect and suffers from a small nonrep-resentative sample due to the dyadic researchdesign. Salacuse and Sullivan’s (2005) analysisis cross-sectional and therefore cannot detecthow a higher number of BITs raises the flowof FDI to signatory developing countries overtime. The difference to the results presented byTobin and Rose-Ackerman (2005) is more puz-zling. As we noted in the sensitivity analysis, ourresults uphold if we adopt their five-year period-averages approach and restrict the list of coun-tries to be exactly the same as in their analysis. Itis therefore difficult to know where the differ-ence comes from. One possibility is that we donot log the number of BITs, not least becausethe log of zero (BITs) is not defined. Whateverthe cause, we find from Tobin and Rose-Acker-man’s (2005) result, that each additional BITlowers (rather than raises) the flow of FDI todeveloping countries with high political risk asextremely difficult to believe. BITs might notraise FDI flows in contexts of high risk, butthere is no reason whatsoever to expect thatthey should lower FDI flows.Statistical significance is not equivalent to

substantive importance. We therefore need toknow how strong is the effect of the BIT vari-able on FDI flows. How much more FDI cana developing country expect if it aggressivelyengages in a program to sign BITs with devel-oped countries? To answer this question, welook at a one standard deviation increase inthe BIT variable (equivalent to an increase ofaround 27 in the weighted cumulative BIT var-iable running from 0 to 99). Since in someregressions the interaction effect between theBIT variable and institutional quality is statisti-cally significant, the overall effect of signingup to BITs sometimes depends on the level ofinstitutional quality, in which case, for simplic-ity, we fix institutional quality at its median. 14

Based on the estimations in Table 3, a countryexperiencing a one standard deviation increasein the BIT variable, is predicted to increase itsFDI inflow by between 43.7% and 93.2%.Based on the results from Table 4, such a coun-try is predicted to increase its share of FDI in-flow relative to the total inflow to developingcountries by between 42.0% and 104.1%.


Clearly, these are nonnegligible increases fol-lowing a substantial increase in BIT activity.But whether the demonstrated benefits of sign-ing up to BITs in the form of increased FDI in-flows are higher than the substantial costsdeveloping countries incur in negotiating,signing, concluding, and complying with the

obligations typically contained in suchtreaties is impossible to tell. What we do knowis that BITs fulfil their purpose, and thosedeveloping countries that have signed moreBITs with major capital exporting developedcountries are likely to have received moreFDI in return.

NOTES

1. For the purpose of this article, we presume that ahigher FDI inflow is beneficial to the host nation. Thispresumption can of course be contested (De Soysa &Oneal, 1999).

2. Expropriations without prompt, adequate, andeffective compensation also took place in many Com-munist countries after the Second World War.

3. When developed countries tried to extend theserestrictions in the context of a multilateral agreement oninvestment (MAI) in the mid-1990s to banning pre-investment restrictions, performance requirements, andto extending the measures of expropriation requiringcompensation to indirect and de facto (as opposed todirect) expropriation, most developing countries wererather happy when these negotiations imploded underthe mounting critique of civil society and internaldifferences among developed countries (Neumayer,2001). Many developed countries were also not tookeen on granting liberalized access to investors fromother developed countries the same way they normallyexpect developing countries to grant to investors fromtheir own countries.

4. Depending on the BIT, this can refer to the countryof the company’s incorporation, seat, registered office,or principal place of business or the nationality of theindividuals who have control over, or a substantialinterest in, the investing company (Salacuse & Sullivan,2005, p. 82).

5. A fourth study is provided by UNCTAD (1998).However, it is based on a purely cross-sectional stepwise(!) regression analysis with an unspecified number ofobservations from 1995. Not surprisingly, such ‘‘garbagecan’’ modeling leads to inconclusive results.

6. Their study also suffers from the absence of year-specific time dummies controlling for aggregate annualchanges in US FDI outflows, which could mean that theresults are spurious.

7. If we were to take the share a developing countryreceives relative to the sum of global FDI instead, thenresults are practically identical.

8. In future research, we would like to analyze BITs inmore detail. Not all BITs are the same, and one wouldlike to know whether it is certain elements in BITs thatmatter for FDI location more than others.

9. This is unsurprising given the high correlationbetween the weighted and unweighted cumulative countof BITs, which follows from the fact that by and largeminor capital exporting developed countries have signedfew BITs, whereas the opposite is the case for majorcapital exporters.

10. Note that replacing logged population with the logof total GDP leads to an identical result, which followsfrom the fact that the two variables plus the log of GDPper capita are not independent of each other.

11. We do not include the Lome Conventions or thefollow-on Cotonou Agreement between the EU and 77countries from Africa, the Caribbean, and the Pacific(ACP) since it is highly unlikely that these had a majorimpact on FDI.

12. Since the coefficient of the BIT variable representsthe effect of BITs when the value of institutional qualityis zero, one can re-scale the institutional quality variablesuch that zero represents the lowest level of risk. At thelowest observed level of risk in the sample, the positiveeffect of BITs is reduced from 0.030 (at the highestobserved level of risk in the sample) to 0.011.

13. All results that are not explicitly reported areavailable on request.

14. Strictly speaking, the coefficient of the BIT variablealways represents the effect of BITs at a zero value of theinstitutional quality measure, but if the interaction termis not statistically significantly different from zero, then,as a first approximation, it can be taken as the effect ofBITs at any value of institutional quality.


REFERENCES

Alesina, A., & Perotti, R. (1996). Income distribution,political instability, and investment. European Eco-nomic Review, 40, 1203–1228.

Belsley, D. A., Kuh, E., & Welsch, R. E. (1980).Regression diagnostics. New York: John Wiley.

Braumoeller, B. F. (2004). Hypothesis testing andmultiplicative interaction terms. International Orga-nization, 58, 807–820.

Chakrabarti, A. (2001). The determinants of foreigndirect investment: Sensitivity analyses of cross-coun-try regressions. Kyklos, 54, 89–114.

De Soysa, I., & Oneal, J. R. (1999). Boon or bane?Reassessing the productivity of foreign direct invest-ment. American Sociological Review, 64, 766–782.

Dolzer, R., & Stevens, M. (1995). Bilateral investmenttreaties. The Hague: Martinus Nijhoff Publishers.

Dunning, J. H. (1998). The changing geography offoreign direct investment. In N. Kumar (Ed.),Globalization, foreign direct investment and technol-ogy transfers: Impacts on and prospects for developingcountries. New York: Routledge.

Elkins, Z., Guzman, A., & Simmons, B. (2004). Com-peting for capital: The diffusion of bilateral investmenttreaties, 1960–2000. Working paper. University ofIllinois, University of California at Berkeley andHarvard University.

EU (2004). EC regional trade agreements. Brussels:European Union Directorate General Trade. Avail-able from: http://europa.eu.int/comm/trade/index_en.htm.

Ginsburg, T. (2004). International substitutes for domes-tic institutions: Bilateral investment treaties and gov-ernance. Working Paper. University of IllinoisCollege of Law.

Globerman, S., & Shapiro, D. (2002). Global foreigndirect investment flows: The role of governanceinfrastructure. World Development, 30, 1899–1919.

Guzman, A. (1998). Why LDCs sign treaties that hurtthem: Explaining the popularity of bilateral invest-ment treaties. Virginia Journal of International Law,38, 639–688.

Hallward-Driemeier, M. (2003). Do bilateral investmenttreaties attract FDI? Only a bit. . . and they could bite.Washington, DC: World Bank, World Bank PolicyResearch Paper WPS 3121.

Henisz, W. J. (2000). The institutional environment foreconomic growth. Economics and Politics, 12(1),1–31.

Minor, M. S. (1994). The demise of expropriation as aninstrument of LDC policy, 1980–92. Journal ofInternational Business Studies, 25, 177–188.

Neumayer, E. (2001). Greening trade and investment.London: Earthscan.

Neumayer, E. (2005). Self-interest, foreign need andgood governance: are bilateral investment treatyprograms similar to aid allocation? Mimeo, LondonSchool of Economics and Political Science.

Noorbakhsh, F., Paloni, A., & Youssef, A. (2001).Human capital and FDI inflows to developingcountries: New empirical evidence. World Develop-ment, 29, 1593–1610.

Perry, A. (2000). Effective legal systems and foreigndirect investment: In search of the evidence. Interna-tional and Comparative Law Quarterly, 49, 779–799.

Peters, P. (1996). Review of Dolzer and Stevens:Bilateral investment treaties. Netherlands Interna-tional Law Review, 43(1), 103–108.

Peterson, L. E. (2004). Bilateral investment treaties anddevelopment policy-making. Winnipeg: InternationalInstitute for Sustainable Development.

Raghavan, C. (1997). Bilateral investment treaties playonly a minor role in attracting FDI. Third WorldEconomics, 162, 1–15.

Salacuse, J. W. (1990). BIT by BIT: The growth ofbilateral investment treaties and their impact onforeign investment in developing countries. TheInternational Lawyer, 24(3), 655–675.

Salacuse, J. W., & Sullivan, N. P. (2005). Do BITs reallywork? An evaluation of bilateral investment treatiesand their grand bargain. Harvard International LawJournal, 46, 67–130.

Schneider, F., & Frey, B. (1985). Economic and politicaldeterminants of foreign direct investment. WorldDevelopment, 13, 161–175.

Sornarajah, M. (1986). State responsibility and bilateralinvestment treaties. Journal of World Trade Law, 20,79–98.

Taylor, C. J. (2000). The impact of host countrygovernment policy on US multinational investmentdecisions. The World Economy, 23, 635–647.

Tobin, J., & Rose-Ackerman, S. (2005). Foreign directinvestment and the business environment in developingcountries: The impact of bilateral investment treaties.Yale Law School Center for Law, Economics andPublic Policy Research Paper No. 293.

UNCTAD (1998). Bilateral investment treaties in themid-1990s. New York and Geneva: United Nations.

UNCTAD (2001). Bilateral investment treaties 1959–99.New York and Geneva: United Nations.

UNCTAD (2003a). World Investment Report 2003: FDIpolicies for development national and internationalperspectives. New York and Geneva: United Na-tions.

UNCTAD (2003b). World investment directory. NewYork and Geneva: United Nations.

UNCTAD (2004). World Investment Report 2004: Theshift toward services. New York and Geneva: UnitedNations.

Vandevelde, K. (1996). Review of Dolzer and Stevens:Bilateral investment treaties. American Journal ofInternational Law, 90(3), 545–547.

Vandevelde, K. (1998). The political economy of abilateral investment treaty. The American Journal ofInternational Law, 92(4), 621–641.

Vandevelde, K. (2000). The economics of bilateralinvestment treaties. Harvard International Law Jour-nal, 41(2), 469–502.

Wheeler, D., & Mody, A. (1992). International invest-ment location decisions: The case of US Firms.Journal of International Economics, 33, 57–76.

World Bank (2003a). World development indicators CDRom. Washington, DC: World Bank.

http://europa.eu.int/comm/trade/index_en.htm

http://europa.eu.int/comm/trade/index_en.htm


World Bank (2003b). Adjusted net savings data. Wash-ington, DC: World Bank, Available from: http://lnweb18.worldbank.org/ESSD/envext.nsf/44ByDoc-Name/GreenAccountingAdjustedNetSavings.

WTO (2004). Regional trade agreements. Geneva: WorldTrade Organization, Available from: www.wto.org.

Zebregs, H. (1998). Can the neoclassical model explainthe distribution of foreign direct investment acrossdeveloping countries? Washington, DC. InternationalMonetary Fund, IMF Working Paper WP/98/139.

Political risk components

Sequence Component Max.points

A Government stability 12B Socioeconomic conditions 12C Investment profile 12D Internal conflict 12E External conflict 12F Corruption 6G Military in politics 6H Religion in politics 6I Law and order 6J Ethnic tensions 6K Democratic accountability 6L Bureaucracy quality 4

Total 100

Source: www.prsgroup.com.

APPENDIX A. LIST OF COUNTRIESINCLUDED IN SAMPLE

Albania, Algeria, Angola, Antigua and Bar-buda, Argentina, Armenia, Azerbaijan, Bangla-desh, Barbados, Belarus, Belize, Benin, Bolivia,Botswana, Brazil, Bulgaria, Burkina Faso, Bur-undi, Cambodia, Cameroon, Cape Verde, Cen-tral African Republic, Chad, Chile, China,Colombia, Comoros, Congo (Dem. Rep.),Congo (Rep.), Costa Rica, Cote d’Ivoire, Croa-tia, Czech Republic, Dominica, DominicanRepublic, Ecuador, Egypt, El Salvador, Equa-torial Guinea, Estonia, Ethiopia, Fiji, Gabon,Gambia, Georgia, Ghana, Grenada, Guate-mala, Guinea, Guinea-Bissau, Guyana, Haiti,Honduras, Hungary, India, Indonesia, Iran, Ja-maica, Jordan, Kazakhstan, Kenya, Korea(Rep.), Kyrgyzstan, Latvia, Lebanon, Lesotho,Lithuania, Macedonia FYR, Madagascar, Ma-lawi, Malaysia, Mali, Mauritania, Mauritius,Mexico, Moldova, Morocco, Mozambique,Namibia, Nepal, Nicaragua, Niger, Nigeria,

Pakistan, Panama, Papua New Guinea, Para-guay, Peru, Philippines, Poland, Romania,Russian Federation, Rwanda, Sao Tome andPrincipe, Senegal, Seychelles, Sierra Leone,Slovak Republic, South Africa, Sri Lanka,St. Kitts and Nevis, St. Lucia, St. Vincent andthe Grenadines, Swaziland, Syria, Tanzania,Thailand, Togo, Trinidad and Tobago, Tuni-sia, Turkey, Uganda, Ukraine, Uruguay, Uzbe-kistan, Venezuela, Vietnam, Yemen, Zambia,Zimbabwe.

APPENDIX B. ICRG’S COMPOSITEPOLITICAL RISK INDEX

http://lnweb18.worldbank.org/ESSD/envext.nsf/44ByDocName/GreenAccountingAdjustedNetSavings



http://www.wto.org

http://www.prsgroup.com

Do Bilateral Investment Treaties Increase Foreign Direct Investment to Developing Countries 2005...

Documents

Transcript of Do Bilateral Investment Treaties Increase Foreign Direct Investment to Developing Countries 2005...