RayDebraj_What's New in Development Economics

8/6/2019 RayDebraj_What's New in Development Economics

1/14

WHAT'S NEW IN DEVELOPMENT ECONOMICS?Debraj Ray*

1 IntroductionThis essay is meant to describe the current fron-tiers of development economics, as I see them. I

might as well throw my hands up at the beginningand say there are too many frontiers. In recentyears, the subject has made excellent use of eco-nomic theory, econometric methods, sociology,anthropology, political science and demographyand has burgeoned into one of the liveliest areas ofresearch in all the social sciences. And about timetoo: the study of economic development is probablythe most challenging in all of economics, and pro-vided we are patient about getting to "the bottomline" and the "policy implications," it can haveenormous payoffs.

Fortunately, considerations of space allow me touse brevity as an excuse for selectivity. So ratherthan attempt an exhaustive review of several areas,I would like to concentrate on a few methodologi-cal points around which recent literature appears tohave clustered. More than anything else, I want tobring out for you a certain way of thinking aboutdevelopment that has become increasingly influen-tial over the last couple of decades, one that ischanging and will continue to change the face ofresearch in this discipline.

The main trend I would like to try and documentis a move-welcome, in my opinion-away from atraditional preoccupation with the notion of conver-gence. This is the basic notion that given certainparameters, say savings or fertility rates, economiesinevitably move towards some steady state. If theseparameters are the same across economies, then inthe long run all economies converge to one another.I review this approach very briefly in Section 2. Ithen explain why this view leads to (a) a limiteddepth in the way we ask development questions,and (b) a certain type of policy bias. In Section 3, Idiscuss the first of two types of theories that take usaway from the determinism inherent in the conver-

gence idea. This is an approach that is based on thenotion of multiple equilibria-several dramaticallydifferent outcomes can occur given the same funda-mentals. In Section 4 I return to equilibria that aredetermined fundamentally by historical conditions.That is, given a particular historical experience, theoutcome that results is fully pinned down, but theinfluence of that historical experience persiststhrough time in observed outcomes. In either case,there is no presumption of convergence or ahis-toricity. I will argue that this approach gives us dif-ferent insights, both in the way we ask questionsand with regard to policy.

While I am tempted by fashionable trends innomenclature, I hesitate to call this the "NewDevelopment Economics." If writers such as PaulRosenstein-Rodan or Albert Hirschman were toencounter such a phrase (and the subsequentaccompanying description), they would be scandal-ized. Much of recent thinking in development canbe traced to the insights of these two eminent writ-ers, though, to be sure, the retracing of their pathsbrings to light new insights and arguments.

2 A Traditional ViewOpen a book-almost any book--on the eco-

nomics of developing countries, and it will beginwith the usual litany of woes. Developing countries,notwithstanding the enormous strides they havemade in the last few decades, display fundamentaleconomic inadequacies in a wide range of indica-tors. Levels of physical capital per person are small.Nutrition levels are low. Other indicators of humancapital such as education-both at the primary andsecondary levels-are well below developed-coun-try benchmarks. So are access to sanitation, safewater and housing. Population growth rates arehigh, and so are infant mortality rates. One couldexpand this list indefinitely.

* New York University.Iam greatly indebted to Dilip Mookherjee for several conversations on this subject, and in particular, forallowing me to borrow freely from our joint introduction to A Reader in Development Economics, Lon-don: Blackwell (forthcoming), for the purpose of writing this essay. This essay will appear in MichaelSzenberg, ed. Shifting Paradigms, New Directions in Economics (Cambridge University Press)

Vol. 44, No.2 (Fall 2000) 3


2/14

Notice that some of these indicators-infant sur-vival rates or life expectancy, for instance-may beregarded as defining features of underdevelopment,so in this respect the list above may be viewed, notas a statement of correlations, but as a definition ofwhat we mean by development (or the lack of it).But other indicators, such as low quantities of phys-ical capital per capita, or population growth rates,are at least one step removed. These features don'tdefine underdevelopment. For instance, it is unclearwhether low fertility rates are intrinsically a featureof economic welfare or development: surely, manyfamilies in rich countries may take great pleasure inhaving a large number of offspring. Likewise, largeholdings of physical capital may well have aninstrumental value to play in the developmentprocess, but surely the mere existence of such hold-ings does not constitute a defining characteristic ofeconomic welfare.And indeed, that is how it should be. We do notmake a list of the features that go hand in hand withunderdevelopment simply to define the term. We doso because-implicitly or explicitly-we are look-ing for answers to the question: why are some coun-tries underdeveloped and others not?' One way ofgoing about answering this question is to look atempirical relationships between some measure ofdevelopment (say per-capita GOP) and other (pre-sumably exogenous) factors. For instance, onemight regress per-capita income on variables suchas the rate of savings (or investment) or populationgrowth rates (see, e.g., Mankiw, Romer and Weil[1992]).The background hypothesis of convergence,which goes back to Solow [1956]-but also has aparallel in the theory of optimal growth==has oftenbeen invoked to interpret empirical work of thissort. The basic idea of convergence is very simpleindeed. Suppose that all production is carried outusing capital and labor, and a constant fraction ofnational income is saved. Then countries with a lowendowment of capital relative to labor will have ahigh rate of return to capital (by the "law" of dimin-ishing returns). Consequently, a given addition tothe capital stock will have a larger impact on percapita income. It follows that, controlling for sav-ings rates, poorer countries will tend to grow fasterand hence will catch up, converge.To be sure, the savings rate is not the only factorthat qualifies the argument. Anything that systemat-ically affects the marginal addition to per-capita4

income must be controlled for, including sharpquantifiables such as population growth rates(which affect the denominator of per-capitaincome) or looser concepts such as "political cli-mate" or "corruption" (which might affect the rateof return to capital).

Thus the convergence hypothesis, properly inter-preted, does not really mean that all countries doactually converge. But it does mean that a failure toobserve convergence must be traced to one oranother of the so-called exogenous factors thatwe've just described. This has two important-andunfortunate-implications for the way we thinkabout development.First, it limits our search for deep explanations.It is not uncommon to find economists "explaining"inter-country variation by stating that one country ismore corrupt than another, or more democratic, oris imbued with some particularly hardworking cul-tural ethic. One might even hang one's hat on thefollowing sort of theory: different societies havesome intrinsic difference in their willingness-orability-to save, or to procreate. Therefore such-and-such country is poor or underdevelopedbecause it is populated by people who habituallysave very little, or procreate a lot.At some level these "explanations" are perfectlyvalid. But they are not very deep. We would like tohave a theory which-while not belittling or down-playing the role of social, cultural and political fac-tors--does not simply stop there. We would like toknow, for instance, whether low incomes provoke,in turn, low savings rates so that we have a genuinechicken-and-egg problem. The same is true ofdemographics-might underdevelopment be acause of high population growth rates, just as highpopulation growth rates themselves retard thedevelopment process? More boldly, we might seeka theory of corruption that views corruption just asmuch an outcome as a cause.Now simply asserting that "nothing is trulyexogenous" doesn't take us very far. The question iswhether one can study these interactions in a waythat yields new insights. In what follows, I will tryand argue that this can be (and is being) done.The second problem with the convergenceapproach is that it generates a particular set of atti-tudes towards economic policy. By stressing therole of factors such as savings, population growthor levels of corruption that might actually be symp-toms rather than causes of underdevelopment, they

THE AMERICAN ECONOMIST


3/14

promote superficial (and sometimes wrong) policyinterventions. If these factors are a result of under-development rather than simply its cause-they areunlikely to be prone to manipulation by simple-minded policy tinkering. And even if the policies areeffective, such approaches can lead to misjudgmenton the required duration of necessary interventions.For instance, suppose we believe that Bangladeshigrowth rates are low because Bangladeshi societysomehow promotes high fertility (the outcome, let ussay, of religious or cultural attitudes). If the fertilityrate is truly believed to be exogenous as a conse-quence, a policy of lowering fertility (say, throughmonetary incentives) will certainly have an effect ongrowth rates. But the incentives would have to beoffered indefinitely. In contrast, an interactiveapproach to the fertility-growth problem may sug-gest permanent effects of one-time interventions, anissue we shall return to below in more detail.

In the two sections of the paper that follow, lout-line theories that go beyond the convergence idea.In these theories, societies that are fundamentallysimilar in all respects might behave differently, andpersistently so. I shall discuss two reasons for thispersistent difference. The first is based on thenotion of underdevelopment as a self-fulfilling fail-ure of expectations. According to this approach(Section 3), economies exhibit multiple equilibria.Some societies may be stuck in the "bad" equilibri-um, exhibiting shortfalls in familiar developmentindicators. Simultaneously, such societies may dis-play low savings rates or "cultures of corruption,"but this latter set of features cannot be relatedcausally to the former.

The second set of theories (Section 4) is based onthe notion of underdevelopment as a persistent out-come of certain historical configurations. Onceagain, two blueprints of two societies may be thesame, but differences in certain initial conditionscause persistent differences in subsequent trajecto-ries. In particular, we will focus on differences ininitial economic inequality, though all sorts of otherinitial conditions could profitably be considered.

3 Underdevelopment and Expectations3.1 Multiple Equilibria

Paul Rosenstein-Rodan [1943] and AlbertHirschman [1958] argued that economic develop-

Vol. 44, No.2 (Fall 2000)

ment could be thought of as a massive coordinationfailure, in which several investments do not occursimply because other complementary investmentsare not made, and similarly, these latter investmentsare not forthcoming simply because the former aremissing. Thus one might conceive of two equilibriaunder the very same fundamental conditions, one inwhich active investment is taking place, with eachindustry's efforts motivated and justified by theexpansion of other industries, and another equilibri-um involving persistent stagnation, in which theinactivity of one industry seeps into another. Thisserves as a potential explanation of why similareconomies may behave very differently.

The work of these two writers brings out theessential feature that is needed for "multiple equi-libria" to arise, at least for multiple equilibria thatcan be ranked by some welfare criterion such asPareto-dominance. This is the basic idea of comple-mentarily: a particular form of externality in whichthe taking of an action by an agent increases themarginal benefit to other agents from taking thesame (or similar) action. As examples, consider thetwo main sources of coordination failure discussedby Rosenstein-Rodan and Hirschman.

I. Inter-Industry Links. The expansion of a par-ticular production sector will have both direct andindirect implications for other sectors through theselinks. For instance, the development of a trans-portation network, such as railways, will facilitatethe export of certain types of products, and therebyencourage their production. This is an example ofwhat might be called a supply link: one that worksby lowering the cost of inputs to another sector. Atthe same time, the expansion of railways will raisethe demand for railway inputs, such as steel. This isan example of a demand link.

Supply and demand links may, in tum, be director indirect. For instance, it is possible for railwaysto have a direct demand link to the coal industry (atleast in the days when steam engines were in oper-ation), as also an indirect demand link to the coalindustry (via steel, for instance). The entire produc-tive sector of an economy is covered by a web ofsuch links.

Figure 1 provides a (vastly oversimplified) pic-ture of what these links might look like. As an illus-tration of complementarily, suppose that the"action" in question is as follows: the magnitude ofinvestment in a particular industry. Then a comple-mentarily exists if the links are "positive." as in the

5


4/14

examples given above. For instance, an investmentexpansion in railways increases the incentive toinvest in steel. In such cases, it is possible that thevery same economy may be plunged into a low levelof activity for no other reason than the fact that sec-toral depressions are self-reinforcing. At the sametime, there may exist another (self-fulfilling) levelof economic activity that is better for all concerned.2. Demand complementarities. A different set of

connections is also emphasized in this early litera-ture. This is the possibility that an expansion insome industries will serve to raise income, and inthis way, generate demand for the final goods pro-duced by other industries. Once again, there is apotential complementarily here, at least across theproducers of non-inferior goods. An expansionaryinvestment in some subset of sectors will increasethe incentives of other sectors to follow suit,because there is now a greater demand for theirproducts.As usual, this complementarily raises the possi-bility of multiple equilibria. Each entrepreneurwould invest if he were to believe that demandwould be high, and if all entrepreneurs harboredsuch optimism, demand would indeed be high-

these expectations would be self-fulfilling. But pes-simism may also be self-fulfilling, because lack ofinvestment would lower demand in general for allproducts.

The argument here is that an enhanced level ofeconomic activity generates greater nationalincome, and the generation of national income cre-ates additional demand to justify that activity.

Notice that such "indirect" complementarities(not via specific interindustry links, but through theeconomy as a whole) do not need to work throughdemand alone. Suppose that the expansion of somesectors contributes to the generation of a skilled,reliable, educated workforce. Then the supply of alabor pool of high quality will stimulate the devel-opment of other industries. This is a complernentar-ily that works by facilitating production, not byraising the demand for products.Complementarities lead to a view of the worldthat is essentially non-deterministic. In its purestform, the theory says nothing about which equilib-rium will prevail.' Interpreted with care and someimagination, it also acts as a critique of the conver-gence-based methodology in the previous section.For instance, it is possible for the same economy to

IronRailways

1 ~y_ ./Coal

Steel Exports_ _ . .

r ~I Shipping//ining" VLachinery ~ ConsumerGoods

FIGURE 1. Inter-Industry Links.

6 THE AMERICAN ECONOMIST


5/14

be in a low-income/high-fertility trap, or in a high-income/low-fertility growth phase. In this view, fer-tility rates are not casually responsible for income,nor are fertility rates some exogenous social char-acteristic impeding inter-country convergence.

Once complementarities-and their implicationsfor equilibrium multiplicity--enter our way ofthinking, they seem to pop up everywhere. TheRosenstein-Rodan view of demand complernentari-ties was given new life in a paper by Murplhy,Shleifer and Vishny [1989]. Since then, there hasbeen an explosion of interest in demand comple-mentaries (though the equally important study ofcomplementarily via direct inter-industry links hasbeen surprisingly dormant). See, for example,Rodriguez [1996], Ciccone and Matsuyama [1996],and other papers in a recent issue of the Journal ofDevelopment Economics [1996] dedicated especial-ly to this topic.

To be sure, there is no need to restrict the analy-sis to cross-industry interactions. As Arthur [1983]and David [1985] have argued, it is possible toshow how the presence of complementarities canstifle the arrival of new technologies and new stan-dards. As discussed by Acemoglu and Zilibotti[1997] and others, complementarities can beinvoked to explain low financial depth in develop-ing countries. Complementaries make an appear-ance in the theory of economic growth, as in thepioneering work of Romer r 1986], Lucas [1990]and others. Complementarities can be used tounderstand spatial trends in crime, corruption orlarge scale defaults on debt. See Ray [1998, Ch. 4]for an introductory discussion and several otherapplications of the idea.

3.2 Policy ImplicationsThe methodology of this section has three strik-ing implications for policy.First, a policy is to be viewed as a device for

moving the economy out of one equilibrium intoanother. This is, conceptually, completely differentfrom the viewpoint implicit in the previous section:that there is a single long run outcome towardswhich all economies must go, but this long run out-come may display underdevelopment because theunderlying parameters of the economy are not right.Policy amounts to a sustained tweaking of theseparameters. In contrast, the multiple equilibria con-

Vol. 44, No.2 (Fall 2000)

text views policy as a way of pulling the economyout of one equilibrium into another.

This sort of view lies at the heart of argumentsput forward by Rosenstein-Rodan and Hirschmanin the 1950s, arguments that led to a vigorousdebate between "balanced versus unbalancedgrowth" (see Ray [1998] for a description of thetwo approaches). It is a pity that these argumentshave not received the full attention that theydeserve from a modem perspective; in the next sub-section, I attempt to explain why.

Second, according to the multiple equilibriaview, a policy need not be permanent or persistent,precisely because the desired end-state is also anequilibrium in the absence of the policy. Indeed,after a temporary phase in which the old, bad equi-librium is artificially ruled out with the impositionof the policy, leaving only the good equilibrium, thepolicy may be removed in the expectation that thenew state of affairs will hold on its own. Severalsocio-economic phenomena conform to this view. Itmay be an equilibrium response for citizens of asociety to employ slaves, to bum their wives, todemand dowry, to have many children, to throwgarbage in or otherwise soil public places, to notobserve codes of orderly conduct, or to bribe and bebribed, provided everybody else is doing the samething. The very same individuals may refrain fromall these activities if no one else engages in them.Consider, for instance, a policy in which one oranother of these activities is made unlawful, withattendant penalties for breaking the law. It is to beexpected that after some time has passed, the law(while still on the statute books in name) will notneed to be enforced anymore, assuming that initial-ly it has been implemented well. Social pressuresmay suffice. The same is true of many economic sit-uations, such as those studied by Rosenstein-Rodanand Hirschman.

Finally, while freed of certain responsibilities ofpersistent implementation, an equilibrium-tippingpolicy will need to be artfully chosen and closelyimplemented in the transition, or it can fail badly.Take, for example, the notion of compulsory prima-ry education. The reason that primary educationmay need to be compulsorily imposed in a societyis that its benefits are unclear in a world where laborpower is needed for current output and no one elseis particularly educated (this was even true of West-ern Europe, by the way; see, e.g., Eckstein andZilcha [1994 n . Yet, in a world where everyone else

7


6/14

is educating their children, it would be dangerousfor a single family not to do so. So this is a classicexample of the multiple equilibrium story. Now, ifthe policy of imposing primary education is notproperly implemented, the outcome may be muchworse than it ever was. Resources would be com-mitted to schools. Yet children may not be sent tothem. Worse still, children (such as young girls)may be selectively removed from school for thepurposes of child labor," The point is that at the timeof imposition of the policy, it is (still) not optimal todo as the policy says, so if there are resourcesexpended on the policy (such as schools) and otherresources spent on avoiding the policy (say, collec-tive "coaxing" of the village headmaster to keep afalse attendance register), the resulting outcomemay be worse than it was to start with.'3.3 Transitions

I end this section with some comments on thepersistence of particular equilibrium outcomes inthe mutiple-equilibria framework. I believe that it isour imperfect understanding of these issues thathinder a more careful study of issues such as bal-anced versus unbalanced growth.How does an economy "move from one equilib-rium to another"? I place this phrase in quotes

because it is imprecise: so-called transitions fromone "equilibrium" to another must themselves beviewed as the equilibria of some encompassingintertemporal process. Unfortunately, when embed-ded in an intertemporal setting, the multiple equi-libria or coordination-game paradigm is not ofmuch use in this regard beyond the demonstrationthat multiplicities may exist. In some sense, itavoids altogether any answer to the question: whyis one society less developed than another, and whatcan be done about it? For this would require a the-ory of where the pessimistic beliefs originally camefrom. The paradigm is at a loss for explaining his-torical inertia: repeat a story of multiple equilibriastory and numerous dynamic equilibria emerge,including those in which the society jumps betweenthe bad and good equilibria in all sorts of deftlycoordinated ways. We lack good economic theorythat actually identifies the "stickiness" of equilibria.A small literature-too small, in my opinion-exists on this topic. See, for instance, Krugman[1991], Matsuyama [1991], and Adsera and Ray[1998] in the development literature. There is also a8

corresponding smattering of literature amongmacroeconomists studying business-cycle modelsbased on coordination failure (see, e.g., Chamleyand Gale [1994] and Cooper [1999]).

As an example of the various approaches, theAdsera-Ray paper embeds a coordination game intoa real-time model of "intersectoral choice" (thechoices corresponding to the actions of the staticcoordination game). Now agents may switch sec-tors (more than once, if they so desire), and theirreturns are added over time, by applying a discountfactor. The objective of the paper is to give meaningto the notion of inertia, to the idea that historicalpredominance of a "sector" might impede thedevelopment of a Pareto-superior "sector." Themain result is that if externalities manifest them-selves with a lag (which may be arbitrarily small),and if there are no congestion costs in intersectoralmigration, then initial conditions do pin down equi-libria-there is inertia. The paper suggests aresearch program in which the study of laggedexternalities may be fruitful, as also the study ofmoving costs (a topic given more emphasis in theKrugman and Matsuyama papers).

There is much work to be done in the area ofintertemporal persistence of equilibrium. In partic-ular, only after we have a theory of "inter-equilibri-um transition" can we get to the serious details ofpolicy interventions.

4 Underdevelopment and History4.1 Historical Legacies

Underdevelopment-viewed as a coordinationfailure-directs our attention to the beliefs orexpectations of the economic agents which shore upone or another of several equilibria. In particular,one might ask-and we did ask this above-howthe formation of such expectations may be signifi-cantly conditioned by history. But history may dic-tate much more than expectation-formation; it mayactually pin down the values of certain tangiblevariables and influence future developments. Putanother way, historical legacies may actually selectamong different sets of equilibria (quite apart fromthe possible multiplicities in each set).Once again, variations in historical legacies--orinitial conditions-are not to be thought of as vari-ations in the fundamental makeup of the economy.



7/14

For instance, two economies may have the sametechnological possibilities and individual prefer-ences, but differ, perhaps, in the size of the initialcapital stock. The capital stock is the legacy; tech-nology and preferences represent the fundamentals.Can the former have persistent effects even if thelatter are all the same? As we have seen, the con-vergence hypothesis says no.Historical legacies need not be limited to anation's inheritance of capital stock or GOP from itsancestors. Factors as diverse as legal structure, tra-ditions, or group reputations may serve as initialconditions (see, e.g., the review in Ray [1998] orHoff and Stiglitz [1999]). But of all these, perhapsthe darkest shadow is cast by historically giveninequalities in the distribution of asset ownership.With imperfect capital markets, the poor are limitedin their access to credit necessary for productionand investment (this includes investment not onlyin projects but also in themselves, via education ornutrition). Hence increased inequality can exertnegative effects on both levels and growth rates ofper capita income. High initial inequalities may alsocreate conditions for self-perpetuation, generating alock-in effect with economic stagnation. The verysame fundamental economy would perform differ-ently were initial inequality to be altered.

4.2 InequalityOne may think of the literature that addressesthis sort of question as studying the functional roleof inequality, as opposed to the intrinsic merits anddemerits of unequal treatment. The question is:what effects does inequality have on other variablesof interest, such as aggregate output, employment,efficiency or growth? The relevant literatureincludes Dasgupta and Ray [1986, 1987], Baland

and Ray r 19911. Banerjee and Newman [1993],Galor and Zeira [1993], Lundqvist [1993], Ray andStrenfert [1993], Bowles and Gintis [1994, 1995],Hoff (1994), Hoff and Lyon [1995], Legros andNewman [1996], Aghion and Bolton [1997],Mookherjee [1997], Piketty [1997] and others.Some of the current literature based on dynamicmodels, finds its roots-paradoxically enough-ina paper that did not depart from the convergenceidea (Loury [1981D . Nevertheless, this pioneeringpaper did pin down the crucial interaction betweenlimited capital markets and dynamic inefficiency.The inefficiency of limited access to capital is aVol. 44, No.2 (Fall 2000)

theme that is common to several of the papers,though they depart significantly from the Lourymodel in other aspects.

As an illustration, consider the simplest versionof the Galor-Zeira [1993] model. It shows how theconvergence prediction of the neoclassical growthmodel can be overturned by dropping the assump-tions of a convex technology and perfect capitalmarkets. With setup costs in the acquisition of cer-tain occupations or skills, and borrowing con-straints for poor agents, the initial distribution ofwealth will influence the aggregate skill composi-tion of the economy and total output, resulting inreinforcement of those very same initial conditions.Poor families will not find it worthwhile to invest inthe education of their children, locking theirdescendants into a poverty trap. High initialinequalities thus tend to perpetuate themselves.Moreover, countries with a historically higherpoverty rate will have a persistently lower per capi-ta income.The demonstration of history-dependence in thesimple version of the Galor-Zeira model can be crit-icized. Even in the presence of indivisibilities ininvestment, substantial stochastic perturbationsmight restore ergodicity, by simply permitting dif-ferent wealth levels to communicate (though possi-bly with very small probability). For instance, in thepresence of random elements reflecting luck, a poorfamily may tip over the required threshold and jointhe ranks of the prosperous, just as wealthy familiesmay encounter a string of failures and temporarilydrift into poverty.A rebuttal to this criticism would argue thatunder the conditions of the Galor-Zeira model,those in poverty would remain locked there for along period of time; the problem would appear inthe guise of a low degree of wealth mobility. In partthis is a signal that ergodicity (and convergence,more generally) is itself a problematic concept, atopic that would take me somewhat afield of mycurrent program. But in part, it points to a secondinadequacy of these simple models, which is thatthey are not interactive across agents. The economyisjust several copies of isolated agents (or families)running in parallel. Then inequality has no aggre-gate effects that are not simply trivial sums of indi-vidual effects. The model (in this simple form)misses the interdependence in the evolution of for-tunes of different families in a given society, whichmay strengthen the tendency towards lock-in.

9


8/14

In contrast, the more complicated interactivemodels-such as those in the later part of the Galor-Zeira paper, and in several of the other papers citedabove--do not allow us to conclude anything fromthe behavior of a single family or dynasty of fami-lies. The joint behavior of all families affect impor-tant economic variables such as commodity prices,wage rates, or the rate of interest, and these in tumfeed back on the decision-making of individuals.

While the following comments do run the risk ofsome mathematical abstraction, they permit me toquickly illustrate a number of the models in the rel-evant literature by adopting a framework fromMookherjee and Ray [2000].

Let H be some list of occupations, over which apopulation of unit size is distributed at any date t.The date t is to be interpreted as the lifetime of thegeneration alive at t.

For each A , A ', to be interpreted as occupationaldistributions (of successive generations), a wagefunction w = {w( h)} h< H is defined on H. Thesedefine the incomes earned by different occupations.

A wage function w on H in tum helps determinea cost function = Ix(h)} hEH' also defined on H. Thiscan be interpreted as the cost, payable in the currentdate, of acquiring skills necessary for occupation hfor members of the next generation.

Thus given a sequence IA , J ~ , of occupationaldistributions on H, we obtain a sequence Iw " x, l~,of wage and cost functions defined on H, whereeach wage function w, depends on the neighboringoccupational distributions ( A " A , + , ) , and each costfunction x, is determined in tum by this wage func-tion. We can then say that Iw, .r], ~ is generated by{ A , J ~ , .

Individuals only foresee the wage-cost sequence(the actual generation of this sequence is of littleimport to them). They care about their own income,and those of their descendants. For an individual i(or current representative of family i) with ho(i)given, the problem is to

00

maxI3'u(c)r=()

subject to the constraints

and

10

(3)

(1)

for all t. [Above, u is a single-period utility functionand the f3discount factor.] As in Loury [1981], thisformulation presumes that parents care about theutility (rather than just the consumption or incomelevels) of their descendants in a consistent fashion,so bequests or educational investments in childrenwill be nonpaternalistic, thus removing one poten-tial source of market imperfection. However, capi-tal markets are missing: investments must befinanced entirely from current income. The maxi-mization problem above will result in a sequence ofoccupational choices made by successive genera-tions, which we may denote by Ih,(i) }~) for eachfamily i.

Aggregate these occupational choices acrossfamilies by defining, for each t, A,(h) to be the mea-sure of individuals isuch that h (i) = h. [Of course,the distribution \ is exogenously given.] This gen-erates a sequence of occupational distributions: saythat {A, l~ ) is an aggregate response to {w" x,}~ (forgiven \).

An equilibrium (given the historical distributionA o ) is a sequence of succeeding occupational distri-butions, income and cost profiles l A " w, x, }~, suchthat (a) { w " x , J ~ , is generated by {A,}; : ; ) , and (b)IA , }~ , is an aggregate response to I w " x ,} ~ ,. In suchan equilibrium all families have perfect foresightconcerning the future evolution of the economy andthe returns to different occupations; their optimalresponses in tum justify their beliefs.

It is possible to embed several well-knownmodels-as well as the readings included in thisvolume-within this framework. Consider the fol-lowing examples:

[1] Mode Is of N oninteracting Agents. H is the setof all capital stocks, w(h) is independent of theoccupational distribution, and equals some produc-tion functionf(h), while x(h) = h. This is the frame-work (with uncertainty added) studied in Loury[1981], under the assumption thatfis a "standard"concave production function. Alternatively, onemight interpret H as some discrete set of skills. Thisis the first model studied in Galor-Zeira [1993](they also use a simpler paternalistic "warm-glow"formulation of the bequest motive).

[2] Entrepreneurship. H = {l, 2). I stands forworker; 2 stands for employer. x(h), the cost func-tion, is independent of the wage function: it is 0 ifh = 1, and is S, a setup cost for entrepreneurship, ifh = 2. To determine the wage function, suppose that

(2)



9/14

there is a production function F defined on theamount of employed labor. Each entrepreneurchooses L to

max F(L) - w(I)L.where w( I) is the wage rate for labor. In equilibri-um, L is just the employment per capitalist, whichis A(2)/A(1). So w(1) is given by

, (A(2)F A(l) =w(l),

while w(2) is the resulting profit:

w(2) = F (A(.2) _ F ' ( A ( 2 ) A(2) .A(1) A(1) A(1)This is essentially the Banerjee and Newman[1993] model (see also the second part of Galor-Zeira [1993]). Like Galor and Zeira, they employ awarm-glow model of bequests, and assume a fun-damental indivisibility in the occupational structure(i.e., there are two discretely different occupations).The evolution of wealth and of occupational deci-sions is, however, fundamentally interdependentacross different families. The resulting dynamicsare complicated. Banerjee and Newman manage todescribe the nature of this dynamic in a number ofspecial cases, and show how distinct occupationalstructures and related production systems (such asthe factory system rather than independent cottageproduction) may evolve in the long run, dependingon historical conditions.

Further developments of a related model with adivisible investment technology and random shockswere subsequently explored by Piketty [1997], whoshowed that the interactive nature of the wealthdynamic may still result in multiple long run steadystates from different historical conditions. In thissense historical lock-in can persist even in the pres-ence of wealth mobility at the level of individualfamilies, and the presence of a convex technology.

[3] Demand Effects. H is a finite set of com-modities. A person with occupation h can produceone unit of the specialized commodity h. Again,take x(h) as independent of other variables.

Let p = {p(h) ) " < 1 1 be a price vector on H. Givenincome y, a consumer generates a demand vectorc(p, y) on H.

Vol. 44, No.2 (Fall 2000)

An equilibrium price vector will equate supplyand demand. But the demand by occupants of occu-pation h is just c(p, p(h))A(h), so that equilibriumprices must be given by the solution to the systemI(p,p(h))A(h) = A

hEllBy constant returns to scale, take p(h) = w(h) for allh. A model of this kind is studied by Mani [1998].

[4 ] Labor Skills. This is the approach followedby Lundqvist [1993]. H = (I, 2). 1 stands forunskilled worker; 2 stands for skilled worker. Theproduction function F( a"a2) defines output pro-duced by a, and a2 units of unskilled and skilledlabor respectively. This determines the wage pat-tern:

w( h) = F h (a(l), a(2for h = 1, 2. The function x(h) defining the cost oftraining for different occupations in tum dependson the wage function: it is 0 if h = I, and is aw(2)if h = 2. The idea is that to acquire skill a workerneeds to be trained by a units of currently skilledworkers, who need to be paid their opportunity costof not working in the production sector and earningthe wage w(2). Skilled workers in the economy thusdivide themselves between the production andtraining sectors, depending on the demand in thetwo sectors. Unskilled workers work only in theproduction sector. In equilibrium, the occupationaldistributions at successive dates will determine theallocation of skilled workers in the following man-ner. Let A and A' denote the occupational distribu-tions for succeeding generations. Then notice that

a(1) = A(l),while

a(2) = A(2}--aA' (2),so that the wage function is ultimately related to thesuccessive occupational distributions:

w(h) = F h (A,(1), A(2) - aA'(2for h = 1,2.

It is precisely the dependence of the wage andtraining cost functions on the occupational distribu-

11


10/14

tion that generates new insights. There are threeconsequences that merit particular emphasis.

First, even if there is perfect equality to startwith, the subsequent evolution of inequality isinevitable. To illustrate this, suppose all individualsin a particular generation have equal wealth. Is itpossible for all of them to make the same choices?The answer is, in general, no. If all of them chooseto leave their descendants unskilled, then the returnto skilled labor will become enormously high,encouraging some fraction of the population to edu-cate their children. Similarly, it is not possible forall parents to educate their children, if unskilledlabor is also necessary in production. Thus identicalagents are forced to take nonidentical actions, pre-cisely because of the interdependence of decisionsmade by different families. This means, of course,that in the next generation some inequality mustemerge.

Indeed, following this logic, it is possible toshow that every steady state of the system describedabove must involve inequality. The evolution ofunequal treatment is not precipitated by randomfactors such as bad luck; it is part of the inner logicof the economic system.

Second, this inequality, in tum, leads to a lack ofefficiency. Individuals cannot simply compensatefor their unequal positions by taking recourse to acredit market. In the models studied here, there isn'ta credit market; or if there is one, it is imperfect. Itis this imperfection that underlies the inefficiencyof inequality. Individuals with low wealth may beunable to take advantage of profitable opportunitiesopen to them, be these in the form of skill acquisi-tion, certain occupational advantages, or remunera-tive investment opportunities.

However, I should note that lack of efficiency inthis sense (in the sense of the inability to takeadvantage of productive opportunities) does notnecessarily imply that there are other equilibria thatare Pareto-superior. This has policy implicationsthat I note below.

Third, there may be several steady states, in thesense that many wealth distributions (and associat-ed levels of national output and prices) may all beself-reinforcing. One must be careful not to inter-pret these as "multiple equilibria"-given the initialhistorical conditions, this is perfectly consistentwith the idea that the economy follows a uniquepath." As we shall see, the policy implications aredifferent in each case.12

Finally, as we have already noted, inequality fun-damentally affects the working of equilibriumprices-broadly defined-and in so doing it affectsthe dynamic fate of individuals in a way that cannotbe disentangled by simple stochastic perturbationsof individual outcomes. Thus it is perfectly possiblethat a particular regime displays full mobility ofindividual dynasties, while at the same time thereare many such regimes (depending on history).4.3 Policy Revisited

The policy implications of history-dependencehave sharper political edges than the ones impliedby multiple equilibria. If multiple equilibria arePareto-ranked, then an equilibrium-tipping policywill-at least in the long run-benefit all the agentsin the economy. To be sure, there are serious prob-lems of implementation. Nevertheless, if the policyworks, it will benefit all concerned,' and this knowl-edge can serve to dilute opposition to the policy.

These conditions are significantly harder to meetin situations of history-dependence, especiallythose in which the relevant historical variable isasset inequality. The sharpest expression of thispossibility is in the static model of inequality andundernutrition studied in Daegupta and Ray [1986].Under some mild conditions, every competitiveequilibrium in that model is Pareto-efficient, eventhough they may display undernutrition, low output(compared to other equilibria), and involuntaryunemployment. It follows that-in the Dasgupta-Ray model-every policy designed to reduceundernutrition and unemployment must hurt somesegment of the population. In this scenario, theroots of opposition are not very far underground.

Admittedly, the Dasgupta-Ray model is anextreme illustration. Inefficiency appears as soon aswe tum to a dynamic formulation. But here too, wemust step carefully. As already discussed, the inef-ficiency of a particular equilibrium simply points tothe fact that there are allocations that make allagents better off. But there is no guarantee that suchallocations can themselves serve as equilibria. Thereason I highlight this concern is that, if we wish tocontinue the view of economic policy as anephemeral device, the final outcome of the policyintervention must itself be an equilibrium. But if thelatter equilibrium is not Pareto-superior to the for-mer, there may be serious opposition to the policy.



11/14

Leaving aside issues of opposition, let us take acloser look at the nature of these policies. In con-trast to the policies for eliminating bad equilibria ina multiple-equilibrium framework, the objectivehere is to change initial conditions, thereafter per-mitting equilibrium outcomes to be generated inaccordance with the changed history.

The models discussed above generally have theproperty that steady state equilibria in which thedistribution of wealth is relatively equal are "better"from the point of view of (productive and alloca-tive) efficiency, output and employment, and possi-bly the rate of growth. This suggests that a redis-tributive change in initial asset inequalities in favorof those who are relatively deprived will be benefi-cial from the point of view of other important eco-nomic indicators. But I must emphasize that theprevious two sentences are not necessarily connect-ed by infallible logic. The former assertion con-cerns steady states. For the latter assertion to bevalid, distributions which are originally more equalmust lead to steady states that are more equal aswell. But there is no guarantee that this is true-atpresent we know too little about the out-of-steady-state behavior of these models to tell with any cer-tainty. What we do know that future research willhave to study carefully-and in more detail-thesubtle and often complex connections between ini-tial conditions and final steady states.In addition, the first assertion may be wrong aswell. It is sometimes true that extremely poor soci-eties may gain in functional efficiency if there issome inequality (see, for example, Ray r 1998,Chapters 7 and 8] or Matsuyama [1999]). We arethen caught in a genuine tradeoff between efficien-cy and equality. A more complex phenomenon isthe possibility of "wrong" responses to small orhalf-hearted changes, as discussed for employmentin Dasgupta and Ray [1987]. In these cases, a smalldegree of redistribution may be worse than noredistribution at all.

The discussion so far may give the impressionthat we can say very little about the policy pre-scriptions of these models. This is not entirely true.Remember, the most important policy prescriptionis that in many cases, one-time interventions canhave persistent, permanent effects. Where I havetried to be careful is in cautioning against a cavalierapproach to such one-time interventions, arguingthat there is still much to be done in connecting ini-tial conditions to final steady state outcomes.

Vol. 44, No.2 (Fall 2000)

5 Concluding RemarksAs mentioned in the introduction, my goal in this

article has been particular in nature, rather thancomprehensive. 1 wanted to write about innovativeapproaches in the theory of development economicsthat view underdevelopment not as a failure ofsome fundamental economic parameters, or socio-cultural values, but as an interacting "equilibrium"that hangs together, precipitated by expectationalinertia or by historical conditions.

Why is this view of the development process animportant one? There are three reasons why I feelthis view should be examined very seriously.[1] This point of view leads to a theory, or a set of

theories. in which economic "convergence" (ofincomes, wealth, levels of well-being) acrosscountries is not to be automatically had. Actu-ally, the intelligent layperson reading thesewords will find this reasoning a bit abstruse:why on earth would one expect convergence inthe first place? And why, indeed, should I finda theory interesting on the grounds that it doesnot predict convergence, when I knew that allalong? This is not a bad line of reasoning, butto appreciate why it is misguided, it is impor-tant to refer to a venerable tradition in econom-ics that has convergence as its very core pre-diction. The idea is based-roughly--on theargument that countries which are poor willhave higher marginal products of capital, andconsequently a higher rate of return to capital.This means that a dollar of extra savings willhave a higher payoff in poor countries, allow-ing it grow faster. The prediction: poorer coun-tries will tend to grow faster, so that over timerich and poor countries will come together, or"converge."

Of course. I have not examined the conver-gence hypothesis in detail, as my intention is tocover other views of development." But oneshould notice that convergence theories in theraw form described above have rarely beenfound acceptable, and there are several subtlevariants of the theory. Some of these variantsstill preserve the idea that "other things" beingequal, convergence in some conditional senseis still to be had. It's only if we start acceptingthe possibility that these "other things"-suchas savings or fertility rates=-often cannot bekept equal, that the notion of conditional con-

13


12/14

vergence starts losing its relevance and verydifferent views of development, not at all basedon the idea of convergence, must be sought.[2] The second reason why I find these theoriesimportant is that they do not rely on "funda-mental" differences across peoples or cultures.Thus we may worry about whether Confucian-ism is better than the Protestant ethic in pro-moting hard-headed, successful economicagents, and we might certainly decry Hindufatalism as deeply inimical to purposeful, eco-nomic self-advancement, but we have seenagain and again that when it comes down to theeconomic crunch and circumstances are right,both Confucian and Hindu will make the bestof available opportunities-and so will a hostof other religions and cultures besides. Onceagain, this is not the place to examine in detailfundamentalist explanations based on culturalor religious differences, but I simply don't findthem very convincing. This is not to say thatculture-like conditional convergence--doesnot playa role. But I also take the view that cul-ture, along with several other economic, socialand political institutions, are all part of somebroader interactive theory in which "firstcause" is to be found-if at all-in historicallegacies. And yes-if we do insist on recursinghistory backwards to find the "originalcause"-I would reply that there is no suchthing, that small initial "butterfly effects" haveenormous, magnified consequences.[3] The last reason why I wish to focus on thesetheories is that create a very different role forgovernment policy. Specifically, I have arguedthat these theories place a much greater weighton one-time, or temporary, interventions thantheories that are based on fundamentals. Forinstance, if we were to observe that Indian sav-ings rates are low compared to other East Asiancountries, and we were to believe that Hindufatalism is somehow responsible for this out-come, then a policy of encouraging savings(say, through tax breaks) will certainly have aneffect on growth rates. But there is no tellingwhen that policy can be taken away, or indeed,if it can be taken away at all. For in the absenceof the policy, the theory would tell us that sav-ings would revert to the old Hindu level. Incontrast, a theory that is based on an interactive"chicken-and-egg" approach would promote a

14

policy that attempts to push the chicken-eggcycle into a new equilibrium. Once that hap-pens, the policy can be removed. This is not tosay that once-and-for-all policies are the onlycorrect ones, but to appreciate that the interac-tive theories that we have discussed have verydifferent implications from the traditional ones.

I have discussed one of several frontiers indevelopment economics, but I believe this particu-lar frontier is particularly important. Because it ismore abstract than, say, an account of the latestresearch on labor markets, it is more a methodolog-ical frontier than anything else, and permeatesmuch of our thinking about various theories. Ofcourse, even the practitioners of traditional conver-gence theories are aware of the viewpointsexpressed here, and many are even sympathetic toit. But one hopes that future researchers willembrace this methodology not just from a distance,but in the essential way in which their models areconstructed.

Notes1. Perhaps the word "underdeveloped" does not

constitute politically correct usage, so that sev-eral publications-those by well-known inter-national organizations chief among them-usethe somewhat more hopeful and placatory pre-sent continuous developing." I won't be usingsuch niceties in this article, because it should beclear--or at least it is clear in my mind-thateconomic underdevelopment pins no derogato-ry social label on those who live in, or comefrom, such societies.2. See the literature on turnpike theory inspiredinitially by the work of von Neumarn [1945],followed by several writers-see MeKenzie[1976] for a survey.3. This can sometimes be embarrassing, becausesuch a theory is also unable to predict what willhappen today even if one of the equilibria hasbeen played repeatedly over the past 100 peri-ods. I return to this theme below.4. For a distinct but related view on child laborand multiple equilibria, see Basu and Van[1998]5. For an interesting theoretical discussion of theappearance of (policy-induced) worse equilib-



13/14

ria in the Murphy-Shleifer-Vishuy model, seeBond and Pande [1999].

6. To be sure, it is possible that multiple equilibriaand history-dependence coexist in the samemodel. That is, it may be true both that equilib-ria vary with history, and that there are severalequilibria for each history. This can be shownto happen, for instance, in the model studied byRomer [1986].

7. Of course, this is not true if the multiple equi-libria under consideration are not Pareto-ranked. Sometimes this is indeed the case, as inthe model of child labor studied by Basu andVan [1999].

8. On the convergence postulate and studies stem-ming from it, see, e.g., Barro and Xala-i-Martin[1995], Jones [1997] and Ray [1998].

ReferencesAcemoglu, D. and Zilibotti (1997). "Was

Prometheus Unbound by Chance? Risk, Diversi-fication and Growth" Journal of Political Econ-omy 105,709-751.

Adsera, A. and D. Ray (1998). "History and Coor-dination Failure." Journal of Economic Growth3,267-276.

Aghion, P. and P. Bolton (1997). "A Theory ofTrickle-Down Growth and Development,"Review of Economic Studies 64, 151-172.

Arthur, W . B. (1983). "On Competing Technologiesand Historical Small Events: The Dynamics ofChoice Under Increasing Returns," mimeograph,Department of Economics, Stanford University.

Baland, J-M. and D. Ray (1991). "Why Does AssetInequality Affect Unemployment? A Study of theDemand Composition Problem," Journal ofDevelopment Economic, 35, 69-92.

Banerjee, A. V. and A. Newman (1993). "Occupa-tional Choice and the Process of Development,"Journal of Political Economy 101, 274-298.

Barro, R. J. and X. Sala-i-Martin (1995). EconomicGrowth. New York: McGrawHil1.Basu, K. and P. H. Van (1998). "The Economics of

Child Labor," American Economic Review 88,412-427.

Bond, P. and R. Pande (1999). "Implementing the'Big Push': the Limits of Policy Activism,"mimeograph, Department of Economics, Colum-bia University.

Vol. 44, No.2 (Fall 2000)

Bowles, S. and H. Gintis (1994. "Credit MarketImperfections and the Incidence of WorkerOwned Firms," Metroeconomica 45, 209-223.

Bowles, S. and H. Gintis (1995), "The Distributionof Wealth and the Assignment of ControlRights," mimeograph, Department of Econom-ics, University of Massachusetts, Amherst.

Charnley, C. and D. Gale (1994). "Information Rev-elation and Strategic Delay in a Model of Invest-ment," Econometrica 62, 1065-1085.

Ciccone, A. and K. Matsuyama (1996). "Start-upCosts and Pecuniary Externalities as Barriers toEconomic Development," .lournal of Develop-ment Economics 49,33-59.

Cooper, R. (1999). Coordination Games: Comple-mentarities and Macroeconomics. CambridgeUniversity Press.

Dasgupta, P. and D. Ray (1986. "Inequality as aDeterminant of Malnutrition and Unemploy-ment: Theory," Economic Journal 97,1011-1034.

Dasgupta, P . and D. Ray (1987). "Inequality as aDeterminant of Malnutrition and Unemploy-ment: Policy," EconomicJournal Tl , 177-188.

David, P. A. (1985). "Clio and the Economics ofQWERTY," American Economic Review 75,332-337.

Eckstein, Z. and I. Zilcha (1994), "The Effect ofCompulsory Schooling on Growth, Income Dis-tribution and Welfare," Journal of Public Eco-nomics 54,339-359.

Galor, O. and J. Zeira (1993). "Income Distributionand Macroeconomics," Review of EconomicStudies 60, 35-52.

Hirschman, A. O. (1958). The Strategy of Econom-ic Development. New Haven, CT: Yale Universi-ty Press.

Hoff, K. (1994). "The Second Theorem of the Sec-ond Best," Journal of Public Economics 45,223--242.

Hoff, K. And A. Lyon (1995), "Non-Leaky Buck-ets: Optimal Redistributive Taxation and AgencyCosts," Journal of Public Economics 58,365--390.

Hoff, K. And J. Stiglitz (1999). "Modern EconomicTheory and Development," in G. Meier and J.Stiglitz (eds), Pioneers in Development, OxfordUniversity Press, forthcoming.

Jones, C. (1998). Introduction to Economic Growth.W. W. Norton.

15


14/14

Krugman, P. (1991). "History versus Expectations,"Quarterly Journal of Economics 106,651-667.Legros, P. And A. Newman (1996). "Wealth

Effects, Distribution and the Theory of Organi-zation," Journal of Economic Theory 70,312-341.

Lowry, G. C. (1981). "Intergenerational Transfersand the Distribution of Earnings," Econometrica40, 843-867.

Lucas, R. E. (1988). "On the Mechanics of Eco-nomic Development," Journal of Monetary Eco-nomics 22. 3-42.

Lundqvist, L. (1993). "Economic Underdevelop-ment: The Case of a Missing Market for HumanCapital," Journal of Development Economics 40219-239.

Mani, A. (1998). "Income Distribution and theDemand Constraint," mimeograph, Departmentof Economics, Vanderbilt University.

Mankiw, N. G., Romer, P . And D. Weil (1992). "AContribution to the Empirics of EconomicGrowth," Quarterly Journal of Economics 107,407-438.

Matsuyama, K. (1991). "Increasing Returns, Indus-trialization, and Indeterminacy of Equilibrium,"Quarterly Journal of Economics 106,617-50.

Matsuyama, K. (1999). "The Rise of Mass Con-sumption Societies," mimeograph, Departmentof Economics, Northwestern University.McKenzie, L. (1976). "Turnpike Theory," Econo-metrica 44. 841-865.

16

Mookherjee, D. And D. Ray (2000). "PersistentInequality and Endogenous Investment Thresh-olds," mimeograph, Department of Economics,Boston University.

Murphy, K., A. Shleifer, And R. Vishny (1989a)."Industrialization and the Big Push," Journal ofPolitical Economy 17, 1003-1026.

Neumann, J. V. (1945--46). "A Model of GeneralEconomic Equilibrium," Review of EconomicStudies 13, 1-9.Pike tty, T. (1997), "The Dynamics of the Wealth

Distribution and the Interest Rate with CreditRationing," Review of Economic Studies 64,173-189.

Ray, D. (1998). Development Economics. Prince-ton: Princeton University Press.

Ray, D. And P. Streufert (1993). Udynamic Equilib-ria with Unemployment due to Undernourish-ment," Economic Theory 3, 61-85.

Rodriguez-Clare, A. (1996). "The Division ofLabor and Economic Development," Journal ofDevelopment Economics 49,3-32.

Romer, P. (1986). "Increasing Returns and Long-Run Growth," Journal of Political Economy 92,1002-1037.Rosenstein-Rodan, P. (1943) "Problems of Industri-alization of Eastern and Southeastern Europe,"Economic Journal 53, 202-211.

Solow, R. (1956). "A Contribution to the Theory ofEconomic Growth," Quarterly Journal of Eco-nomics 70, 65-94.


RayDebraj_What's New in Development Economics

Documents

Transcript of RayDebraj_What's New in Development Economics