Information and the Change in the Paradigm in Author(s): Joseph E. Stiglitz Source: The , Vol. 92, No. 3, (Jun., 2002), pp. 460-501 Published by: American Economic Association Stable URL: http://www.jstor.org/stable/3083351 Accessed: 22/07/2008 15:44

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=aea.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that promotes the discovery and use of these resources. For more information about JSTOR, please contact [email protected].

http://www.jstor.org Informationand the Change in the Paradigmin Economicst

By JOSEPHE. STIGLITZ*

The research for which , named after its Chairmanof the Board, Gary , and I are being recognized is has declined to but a shadow of its former part of a larger research programwhich today self. But even in its heyday, it was marredby embraces a great numberof researchersaround poverty, periods of high , and the world. In this article, I want to set the massive racial discrimination.Yet the economic particular work which was cited within this theories we were taught paid little attention to broaderagenda, and that agenda within the still poverty, said that all marketscleared-including broaderperspective of the history of economic the labormarket, so that unemploymentmust be thought. I hope to show that information eco- nothing more than a phantasm-and claimed nomics representsa fundamentalchange in the that the profit motive ensured that there could prevailing paradigmwithin economics. not be economic discrimination(, Informationeconomics has alreadyhad a pro- 1971). As a graduatestudent, I was determined found effect on how we think about economic to try to create models with assumptions-and policy and is likely to have an even greater conclusions-closer to those that accordedwith influence in the future. Many of the major pol- the world I saw, with all of its imperfections. icy debates over the past two decades have My first visits to the developing world in centered around the related issues of the effi- 1967, and a more extensive stay in Kenya in ciency of the and the appro- 1969, made an indelible impression on me. priate relationshipbetween the market and the Models of perfect markets, as badly flawed as government. The argument of they might seem for Europeor America,seemed (1776) that free markets lead to efficient out- truly inappropriatefor these countries. While comes, "as if by an ,"has played many of the key assumptionsthat went into the a centralrole in these debates:It suggested that competitiveequilibrium model seemed not to fit we could, by and large, rely on marketswithout these economieswell, I was particularlystruck by government intervention (or, at most, with a the imperfectionsof information,the absence of limited role for government). The set of ideas markets, and the pervasiveness and persist- that I will present here undermined Smith's ence of seemingly dysfunctional institutions, theory and the view of the role of government such as . I had seen cyclical that rested on it. They have suggested that the unemployment-sometimes quite large-and reason that the hand may be invisible is that it is the hardshipit brought as I grew up, but I had simply not there-or at least that if is there, it is not seen the massive unemployment that palsied. characterized African cities, unemployment When I began the study of economics some that could not be explained either by unions 41 years ago, I was struck by the incongruity or minimum wage laws (which, even when between the models that I was taught and the they existed, were regularly circumvented). world that I had seen growing up in Gary, Again, there was a massive discrepancy be- Indiana. Founded in 1906 by U.S. Steel, and tween the models we had been taught and what I saw. In contrast, the ideas and models I will dis- t This article is a revised version of the lectureJoseph E. cuss here have proved useful not only in ad- on Stiglitz delivered in Stockholm, Sweden December 8, dressing broad philosophical questions, such as 2001, when he received the Bank of Sweden Prize in Eco- the role of the but also in nomic Sciences in Memory of Alfred Nobel. The article is appropriate state, copyright ? The Nobel Foundation2001 and is published analyzingconcrete policy issues. For example, I here with the permission of the Nobel Foundation. believe that some of the huge mistakes which * Graduate School of Business, Uris Hall, Columbia have been made in policy in the last decade, in University, 3022 Broadway, New York, NY 10027. for instance the managementof the East Asian 460 VOL. 92 NO. 3 STIGLI7Z: AND PARADIGMCHANGE 461 crisis or the transitionof the former communist want to trace out some of their origins. To a countries to the market, might have been large extent, these ideas evolved from attempts avoided had there been a better understanding to answer specific policy questions or to explain of issues-such as financial structure, bank- specific phenomena to which the standardthe- ruptcy,and corporategovernance-to which the ory provided an inadequate explanation. But new information economics has called atten- any discipline has a life of its own, a prevailing tion. And the so-called "Washington consen- paradigm, with assumptions and conventions. sus"' policies, which have predominatedin the Much of the work was motivatedby an attempt policy advice of the internationalfinancial in- to explore the limits of that paradigm-to see stitutions over the past quarter century, have how the standardmodels could embrace prob- been based on market fundamentalistpolicies lems of information imperfections (which which ignored the information-theoreticcon- turned out to be not very well). cerns; this explains, at least partly, their wide- For more than 100 years, formal modeling in spread failures. Information affects decision- economics had focused on models in which making in every context-not just inside firms information was assumed to be perfect. Of and . More recently, as I discuss course, everyone recognized that information below, I have turned my attention to some as- was in fact imperfect, but the hope, following pects of what might be called the political econ- Marshall's dictum "Natura non facit saltum," omy of information: the role of information was that economies in which information was in political processes and collective decision- not too imperfect would look very much like making. There are asymmetries of information economies in which information was perfect. between those governing and those governed, One of the main results of our research was to and just as participants in markets strive to show that this was not true; that even a small overcome asymmetriesof information,we need amountof informationimperfection could have to look for ways by which the asymmetries of a profound effect on the nature of the informationin political processes can be limited equilibrium. and their consequences mitigated. The creators of the neoclassical model, the reigningeconomic paradigm of the twentiethcen- I. The HistoricalSetting tury, ignored the warningsof nineteenth-century and still earlier masters about how information I do not want here to review in detail the concernsmight alter their analyses-perhaps be- models that were constructedexploring the role cause they could not see how to embrace them of information;in recent years, there has been a in their seemingly precise models, perhaps be- number of survey articles and interpretivees- cause doing so would have led to uncomfortable says, even several books in this area.2 I do conclusions about the efficiency of markets.For want to highlight some of the dramaticimpacts instance, Smith, in anticipatinglater discussions that information economics has had on how of adverse selection, wrote that as firms raise economics is approachedtoday, how it has pro- interestrates, the best borrowersdrop out of the vided explanations for phenomena that were market.3If lenders knew perfectly the risks as- previously unexplained, how it has altered our sociated with each borrower,this would matter views about how the economy functions, and, little; each borrower would be charged an ap- perhaps most importantly,how it has led to a propriaterisk premium.It is because lenders do rethinking of the appropriaterole for govern- ment in our society. In describing the ideas, I 3 "If the legal rate ... was fixed so high ... the greaterpart of the which was to be lent, would be lent to ' See John Williamson (1990) for a description and prodigals and profectors, who alone would be willing to Stiglitz (1999c) for a critique. give this higher .Sober who will for the 2 people, give Review articles include Stiglitz (1975b, 1985d, 1987a, use of money no more than a part of what they are likely to 1988b, 1992a, 2000d) and John G. Riley (2001). Book- make by the use of it, would not venture into the competi- length references include, among others, Drew Fudenberg tion" (Smith, 1776). See also Jean-Charles-LeonardSimonde and (1991), Jack Hirshleifer and Riley (1992), de Sismondi (1815), John S. Mill (1848), and Alfred and Oliver D. Hart (1995). Marshall (1890), as cited in Stiglitz (1987a). 462 THE AMERICANECONOMIC REVIEW JUNE 2002 not know the default probabilitiesof borrowers model. Some authors, like George J. Stigler perfectly that this process of adverse selection (1961), Nobel laureatein 1982, while recogniz- has such importantconsequences. ing the importanceof information,argued that I have already noted that something was once the real costs of information were taken wrong-indeed seriously wrong-with the into account, the standardresults of economics competitive equilibrium models which repre- would still hold. Informationwas just a trans- sented the prevailingparadigm when we went to actions cost. In the approachof many Chicago graduate school. The paradigm seemed to say School , informationeconomics was thatunemployment did not exist, and that issues like any otherbranch of ; one of efficiency and could be neatly sepa- simply analyzed the special factors determining rated, so that economists could set aside prob- the demand for and supply of information,just lems of inequality and poverty as they went as one might analyze the factors affecting the about their business of designing more efficient marketfor wheat. For the more mathematically economic systems. But beyond these question- inclined, informationcould be incorporatedinto able conclusions there were also a host of em- production functions by inserting an I for the pirical puzzles-facts that were hard to input "information,"where I itself could be reconcile with the standardtheory, institutional produced by inputs, like labor. Our analysis arrangements left unexplained. In microeco- showed that this approachwas wrong, as were nomics, there were puzzles, such the conclusions derived from it. as why firms appear not to take actions which Practical economists who could not ignore minimize their liabilities; security market the bouts of unemploymentwhich had plagued paradoxes,4 such as why asset prices are so capitalismsince its inception talked of the "neo- volatile (RobertJ. Shiller, 2000) and why equity classical synthesis":If Keynesian interventions plays such a limited role in the financingof new were used to ensure that the economy remained (Colin Mayer, 1990); and other im- at ,the story went, the standard portantbehavioral questions, such as why firms neoclassical propositions would once again be respond to risks in ways markedly different true. But while the (Paul from those predicted by the theory. In macro- A. Samuelson [1947], Nobel laureate in 1970) economics, the cyclical movements of many of had enormous intellectual influence, by the the key aggregate variables proved difficult to 1970's and 1980's it had come under attack reconcile with the standardtheory. For exam- from two sides. One side attackedthe underpin- ple, if labor-supplycurves are highly inelastic, nings of , its microfoun- as most evidence suggests is the case (especially dations. Why would rational actors fail to for primaryworkers), then falls in employment achieve equilibrium-with unemploymentper- during cyclical downturnsshould be accompa- sisting-in the way that John MaynardKeynes nied by large declines in the real consumption (1936) had suggested? This form of the argu- wage. This does not appear to happen. And if ment effectively denied the existence of the the perfect market assumptions were even ap- phenomena that Keynes was attemptingto ex- proximatelysatisfied, the distress caused by cy- plain. Worse still, from this perspective some clical movements in the economy would be saw the unemploymentthat did exist as largely much less than seems to be the case.5 reflecting an interference (e.g., by government There were, to be sure, some Ptolemaic at- in setting minimum wages, or by trade unions tempts to defend and elaborate on the old using their monopoly power to set too-high wages) with the free workings of the market. The implication was that unemploymentwould 4 be eliminated if marketswere made more There was so many of these that the Journal of Eco- flex- nomic Perspectives ran a regular column with each issue ible, that is, if unions and governmentinterven- highlighting these paradoxes. For a discussion of other tions were eliminated. Even if wages fell by a paradoxes, see Stiglitz (1973b, 1982d, 1989g). thirdin the Great should have, 5 Robert E. Jr. who won the Depression, they Lucas, (1987), in this view, fallen even more. in 1995, uses the perfect marketsmodel with a representa- tive to try to argue that these cyclical fluctuationsin There was however an alternativeperspective fact have a relatively small welfare costs. (articulatedmore fully in Bruce C. Greenwald VOL. 92 NO. 3 STIGLITZ:INFORMATION ECONOMICS AND PARADIGMCHANGE 463 and Stiglitz, 1987a, 1988b) which asked why social returnto education might decline. From we shouldn't believe that massive unemploy- this perspective, education was performing a ment was just the tip of an iceberg of more markedly different function than it did in the pervasive marketefficiencies that are harderto traditionaleconomics literature,where it simply detect. If markets seemed to function so badly added to and improved produc- some of the time, they must be underperforming tivity.6This analysis had importantimplications in more subtle ways much of the time. The for Kenya's decision about how much to invest economics of information bolstered this view. in higher education. The problem with Fields' Indeed, given the nature of the debt contracts, work was that it did not provide a full equilib- falling prices in the Depression led to bank- rium analysis: wages were fixed, rather than ruptcy and economic disruptions,actually exac- competitively determined. erbating the economic downturn. Had there This omission led me to ask what the market been more wage and price flexibility, matters equilibriumwould look like if wages were set might have been even worse. equal to mean marginalproducts conditional on In a later section, I shall explain how it was the information that was available (Stiglitz, not just the discrepanciesbetween the standard 1975c). And this in turnforced me to ask: what competitive model and its predictions which were the incentives and mechanisms for em- led to its being questioned, but the model's ployers and employees to acquire or transmit lack of robustness-even slight departures information?Within a group of otherwise sim- from the underlying assumption of perfect ilar job applicants(who thereforeface the same information had large consequences. But be- wage), the employer has an incentive to identify fore turning to those issues, it may be useful who is the most able, to find some way of to describe some of the specific questions sorting or among them, if he could which underlay the beginnings of my research keep that informationprivate. But often he can- program in this area. not; and if others find out about a worker's true ability, the wage will be bid up, and the em- II. Some Motivating Ideas ployer will be unableto appropriatethe returnto the information.At the very beginning of this A. Education as a Screening Device researchprogram we had thus identified one of the key issues in information economics: the Key to my thinking on these issues was the difficulty of appropriatingthe returnsto creat- time between 1969 and 1971 I spent at the ing information. Institute for Development Studies at the Uni- On the other hand, if the employee knew his versity of Nairobi with the supportof the Rock- own ability (thatis, if there were asymmetriesof efeller Foundation. The newly independent informationbetween the employee and the em- Kenyan government was asking questions that ployer), then a differentset of incentives were at had not been raised by its former colonial mas- play. Someone who knows his abilities are ters, as it attempted to forge policies which above average has an incentive to convince his would promoteits growth and development.For potential employer of that, but a worker at the example, how much should the governmentin- bottom of the ability distributionhas an equally vest in education? It was clear that a better strongincentive to keep the informationprivate. educationgot people betterjobs-the credential Here was a second principle that was to be put one at the head of the job queue. Gary S. explored in subsequent years: there are incen- Fields, a young scholar working at the Institute tives on the part of individuals for information of Development Studies there, developed a sim- not to be revealed, for secrecy, or, in moder ple model (publishedin 1972) suggesting, how- ever, that the private returnsto education-the enhanced probability of getting a good job- 6 See, e.g., Theodore W. Schultz (1960), who won the might differ from the social return. Indeed, it Nobel Prize in 1979, and Jacob Mincer (1974). At the time, was that as more there was other ongoing work criticizing the human-capital possible people got educated, formulation, which focused on the role of education in the private returnsgot higher (it was even more socialization and providing credentials; see, for example, necessary to get the credential)even though the Samuel Bowles and HerbertGintis (1976). 464 THEAMERICAN ECONOMIC REVIEW JUNE 2002 parlance,for a lack of transparency.This raised turnover,and thereforehigher turnovercosts for questions: How did the forces for secrecy and the firm.9 It was not until some years later than for informationdisclosure get balanced? What we were able to explain more fully-based on was the equilibriumthat emerged? I will post- limitations of information-why it was that pone until the next section a descriptionof that firms have to bear these turnovercosts (Richard equilibrium. J. Arott and Stiglitz, 1985; Arott et al., 1988). Another explanation for efficiency wages B. Theory was related to the work I was beginning on asymmetricinformation. Any managerwill tell That summer in Kenya I began three other you that paying higher wages attracts better research projects related to informationimper- workers-this is just an applicationof the gen- fections. At the time I was working in Kenya, eral notion of adverse selection, which played there was heavy urbanunemployment. My col- a central role in earlier insurance literature leagues at the Institute for Development Stud- (KennethJ. Arrow, 1965). Firms in a marketdo ies, Michael Todaro and John Harris, had not passively have to accept the "marketwage." formulated a simple model of labor migration Even in competitive markets, firms could, if from the rural to the urban sector which ac- they wanted, offer higher wages than others; counted for the unemployment.7 High urban indeed, it might pay a firm to offer a higher wages attractedworkers, who were willing to wage, to attractmore able workers. Again, the risk unemployment for the chance at those efficiency wage theory explained the existence higher wages. Here was a simple, general- of unemployment in equilibrium. It was thus equilibriummodel of unemployment,but again clear that the notion that underlay much of there was one missing piece: an explanationof traditional competitive equilibrium analysis- high urban wages, well in excess of the legal that marketshad to clear-was simply not true minimum wage. It did not seem as if either if informationwere imperfect. government or unions were forcing employers The formulationof the efficiency wage the- to pay these high wages. One needed an equi- ory that has received the most attentionover the libriumtheory of wage determination.I recalled years, however, has focused on problems of discussions I had once had in Cambridgewith incentives. Many firms claim that paying high Harvey Leibenstein, who had postulatedthat in wages induces their workers to work harder. very poor countries,because of nutrition,higher The problem that and I (1984) wages led to higher productivity (Leibenstein, faced was to try to make sense of this claim. If 1957). The key insight was thatimperfections in all workersare identical, then if it benefitedone informationand contractingmight also rational- firm to pay a high wage, it would likewise ize a dependence of productivityon wages.8 In benefit all firms. But if a worker was fired for that case, firms might find it profitableto pay a shirking, and there were full employment, he higher wage than the minimum necessary to hire labor; such wages I referredto as efficiency wages. With efficiency wages, unemployment 9 In Nairobi, in 1969, I wrote a long, comprehensive could exist in equilibrium.I explored four ex- analysis of efficiency wages, entitled "AlternativeTheories for of Wage Determination and Unemployment in LDC's." planations why productivity might depend Given the custom of writing relatively short papers, focus- on wages (other than through nutrition). The ing on one issue at a time, ratherthan publishing the paper simplest was that lower wages lead to higher as a whole, I had to breakthe paperdown into several parts. Each of these had a long gestation period. The labor turn- over paper was published as Stiglitz (1974a); the adverse selection model as Stiglitz (1982a, 1992d [a revision of a 7 See Michael P. Todaro (1969) and John R. Harrisand 1976 unpublished paper]). I elaborated on the nutritional Todaro (1970). I developed these ideas further in Stiglitz efficiency wage theory in Stiglitz (1976). Various versions (1969b). of these ideas have subsequently been elaborated on in a 8 Others were independently coming to the same in- large number of papers, including Andrew W. Weiss sight, in particular,Edmund S. Phelps (1968). Phelps and (1980), Stiglitz (1982f, 1986b, 1987a, 1987g), Akerlof and Sidney G. Winter (1970) also realized that the same Yellen (1986), Andr6s Rodriguez and Stiglitz (1991a, b), issues applied to product markets, in their theory of Raaj K. Sah and Stiglitz (1992), Barry J. Nalebuff et al. customer markets. (1993), and PatrickRey and Stiglitz (1996). VOL. 92 NO. 3 STIGLITZ:INFORMATION ECONOMICS AND PARADIGMCHANGE 465 could immediately get anotherjob at the same natives? The worker could rent the . He wage. The high wage would thus provide no would have full incentives but then he would incentive. Only if there were unemployment have to bear all the risk of fluctuationsin output; would the worker pay a price for shirking. We and beside, he often did not have the requisite showed that in equilibriumthere had to be un- capital to pay the rent ahead of time and access employment: unemploymentwas the discipline to credit was limited (for reasons to be ex- device that forced workers to work hard (see plained below). He could work as wage labor, Rey and Stiglitz [1996] for an alternativegeneral- but then the landlord would have to monitor equilibriumformulation). The model had strong him, to ensure that he worked. Sharecropping policy implications, some of which I shall de- represented a compromise between balancing scribe below. Our work illustrated the use of concerns about risk sharingand incentives. The highly simplified models to help clarify think- underlying information problem was that the ing aboutquite complicatedmatters. In practice, input of the worker could not be observed, but of course, workers are not identical, so prob- only his output, which was not perfectly corre- lems of adverse selection become intertwined lated with his input. The sharecroppingcontract with those of incentives. For example, being could be thoughtof as a combinationof a rental fired usually does convey information-there is contractplus an insurancecontract, in which the typically a stigma. landlord "rebates"part of the rent if crops turn There was a fourth version of the efficiency out badly. There is not full insurance (which wage, where productivitywas related to morale would be equivalentto a wage contract)because effects, perceptions about how fairly they were such insurance would attenuate all incentives. being treated. While I briefly discussed this The adverse effect of insuranceon incentives to version in my earlier work (see in particular avoid the insured-against contingency is re- Stiglitz, 1974d), it was not until almost 20 years ferred to as .10 later that the idea was fully developed in the In Stiglitz (1974b) I analyzed the equilibrium importantwork of Akerlof and Yellen (1990). sharecroppingcontract. In that paper, I recog- nized the similarity of the incentive problems I C. Sharecroppingand the General explored to those facing moder corporations, Theory of Incentives e.g., in providingincentives to their managers- a type of problemlater to be called the principal- This work on the economics of incentives in agent problem (Stephen A. Ross, 1973). There labor markets was closely related to the third followed a large literatureon optimal and equi- research project that I began in Kenya. In tra- libriumincentive schemes, in labor, capital, and ditional economic theory,while considerablelip insurance markets." An important principle was paid to incentives, there was little was that contracts had to be based on observ- serious attention to issues of incentives, moti- ables, whether they be inputs, processes, or vation, and monitoring. With perfect informa- outcomes. Many of the results obtained earlier tion, individualsare paid to performa particular in the work on adverse selection had their par- service. If they performthe service they receive the contractedamount; and if not, they do not. o0This like adverse in the With imperfectinformation, firms have to mo- term, selection, originates insurance literature. Insurance firms recognized that the tivate and monitor, rewarding workers for ob- greaterthe insurancecoverage, the less incentive there was served good performance and punishing them for the insured to take care; if a property was insured for for bad. My interest in these issues was first more than 100 percent of its , there was even an aroused about a incentive to have an accident (a fire). Not taking appropriate by thinking sharecropping, care was to be common form of land in thought "immoral";hence the name. Arrow's tenancy developing work in moral hazard (Arrow, 1963, 1965) was among the countries.Under sharecropping,the workersur- most importantprecursors, as it was in the economics of rendershalf of the adverse selection. (sometimes two-thirds) pro- 11 duce to the landlordin returnfor the use of his For a classic reference see Hartand Bengt Holmstrom land. At first this seemed a (1987). In addition, see Stiglitz (1975a, 1982c), Kevin J. blush, highly inef- Michael C. Jensen and ficient Murphy (1985), Murphy (1990), arrangement,equivalent to a 50-percent Joseph G. Haubrich (1994), and Brian J. Hall and Jeffrey tax on workers' labor. But what were the alter- B. Liebman (1998). 466 THEAMERICAN ECONOMIC REVIEW JUNE 2002 allel in this area of "adverse incentives." For ticipants in the economy better or worse engi- instance, Arnott and I (1988a, 1990) analyzed neers. Each was solving a maximization equilibria which entail partial insurance as a problem, with full information: households way of mitigating the adverse incentive effects maximizingutility subjectto budget constraints, (just as partial insurance characterizedequilib- firms maximizing profits (market value), and rium with adverse selection). the two interactingin competitive product, la- bor, and capital markets. One of the peculiar D. EquilibriumWage and Price Distributions implicationswas that there never were disagree- ments about what the firm should do. Alterna- The fourth strand of my research looked at tive management teams would presumably the issue of wage differentialsfrom a different come up with the same solution to the maximi- perspective.My earlierwork had suggested that zation problems. Another peculiar implication firms that faced higher turnover might pay was for the meaning of risk: When a firm said higher wages to mitigate the problem. But one that a project was risky, that (should have) of the reasons that individualsquit is to obtain a meant that it was highly correlated with the higher-payingjob, so the turnoverrate in turn , not that it had a high chance of dependson the wage distribution.The challenge failure (Stiglitz, 1989g). I have already de- was to formulate an equilibrium model that scribed some of the other peculiar implications incorporatedboth of these observations,that is, of the model: the fact that there was no unem- where the wage distributionitself which moti- ployment or credit ,that it focused on vated the search was explained as part of the only a limited subset of the informationprob- equilibrium. lems facing society, that it seemed not to ad- More generally, efficiency wage theory said dress issues such as incentives and motivation. that firms might pay a higher wage than neces- But much of the research in the profession sary to obtain workers; but the level of the was directed not at these big gaps, but at seem- efficiency wage might vary across firms. For ingly more technical issues-at the mathemati- example, firmswith higherturnover costs, or for cal structures. The underlying mathematics which worker inefficiency could lead to large requiredassumptions of convexity and continu- losses of capital, or for which monitoring was ity, and with these assumptionsone could prove more difficult, might find it desirable to pay the existence of equilibrium and its (Pareto) higher wages. The implication was that similar efficiency (see Gerard Debreu, 1959; Arrow, labor might receive quite different compensa- 1964). The standardproofs of these fundamen- tion in differentjobs. The distributionof wages tal theorems of welfare economics did not even might not, in general, be explicable solely in list in their enumeratedassumptions those con- terms of differences in abilities. cerning information:the perfect informationas- I was to returnto these four themes repeat- sumptionwas so ingrainedit did not have to be edly in my research over the following three explicitly stated. The economic assumptions to decades. which the proofs of efficiency called attention concernedthe absence of externalitiesand pub- III. From the Competitive Paradigm to the lic . The market failures approachto the Information Paradigm economics of the public sector (Francis M. Bator, 1958) discussed alternative approaches In the previous section, I described how the by which these market failures could be cor- disparitiesbetween the models economists used rected, but these market failures were highly and the world that I saw, especially in Kenya, circumscribedby assumption. had motivated a search for an alternativepara- There was, moreover, a curious disjunction digm. But there was anothermotivation, driven between the language economists used to ex- more by the internal logic and structureof the plain marketsand the models they constructed. competitive model itself. They talked about the informationefficiency of The competitive model virtually made eco- the marketeconomy, though they focused on a nomics a branch of engineering (no aspersions single informationproblem, that of .But on that noble profession intended), and the par- there are a myriad of other information prob- VOL.92 NO. 3 STIGLITZ:INFORMATION ECONOMICS AND PARADIGMCHANGE 467 lems faced by consumers and firms every day, Stiglitz, 1984) showed that there was a funda- concerning,for instance,the prices and qualities mental nonconcavity in the value of informa- of the various objects that are for sale in the tion, that is, under quite general conditions, it market, the quality and efforts of the workers never paid to buy just a little bit of information. they hire, or the potential returnsto investment Arnott and Stiglitz (1988a) showed that such projects. In the standardparadigm, the compet- problemswere pervasive in even the simplest of itive general-equilibrium model (for which moral hazardproblems (where individualshad a Kenneth J. Arrow and GerardDebreu received choice of alternativeactions, e.g. the amountof Nobel Prizes in 1972 and 1983, respectively), risk to undertake).While we had not repealed there were no shocks, no unanticipatedevents: the law of , we had shown At the beginning of time, the full equilibrium its domain to be more limited than had previ- was solved, and everythingfrom then on was an ously been realized. unfolding over time of what had been planned Michael Rothschildand I (1976) showed that in each of the contingencies. In the real world, under natural formulations of what might be the critical question was: how, and how well, meant by a competitive market with imperfect do markets handle fundamental problems of information,equilibrium often did not exist 3- information? even when there was an arbitrarily small There were other aspects of the standardpar- amount of information imperfection.14 While adigm that seemed hardto accept. It arguedthat subsequent research has looked for alternative institutions did not matter-markets could see definitions of equilibrium (e.g., Riley, 1979), through them, and equilibriumwas simply de- we remain unconvinced;most of these alterna- terminedby the laws of supply and demand. It tives violate the natural meaning of "competi- said that the distributionof wealth did not mat- tion," that each participantin the market is so ter, so long as there were well-defined property small that he believes that he will have no effect rights (Ronald H. Coase [1960], who won the on the behavior of others (Rothschild and Nobel Prize in 1991). And it said that (by and Stiglitz, 1997). large) history did not matter-knowing prefer- The new informationparadigm went further ences and technology and initial endowments, in underminingthe foundations of competitive one could describe the time path of the equilibrium analysis, the basic "laws" of eco- economy. nomics. For example, we have shown how, Work on the economics of information be- gan by questioning each of these underlying premises. Consider, to begin with, the con- Stiglitz [1977], Steven Salop [1977], and Stiglitz [1979a, b, vexity assumptions which corresponded to 1989f]), though the basis of imperfect was long-standingprinciples of diminishingreturns. markedly different from that originally envisioned by With information (and the costs of Edward H. Chamberlin(1933). imperfect 13 Nonconvexities rise to these were no naturallygive discontinuities, acquiringit) assumptions longer and discontinuities to problems of existence, but the non- plausible. It was not just that the cost of acquir- existence problemthat Rothschildand I had uncovered was ing information could be viewed as fixed of a different, and more fundamentalnature. The problem costs.12 My work with Roy Radner(Radner and was in partthat a single action of an individual-a choice of one insurancepolicy over another-discretely changed be- liefs, e.g., about his type; and that a slight change in the actions of, an insurance available a new 12 say firm-making In the natural "spaces," indifference curves and iso- insurancepolicy-could lead to discrete changes in actions, profit curves were ill behaved. The nonconvexities which and thereby beliefs. Partha Dasgupta and naturallyarose implied, in turn, that equilibriummight be (1986) have explored mixed strategy equilibria in game- characterized by randomization (Stiglitz, 1975b), or that theoretic formulations,but these seem less convincing than Pareto-efficienttax and policies might be char- the imperfectcompetition resolutions of the existence prob- acterizedby randomization(see Stiglitz [1982g], Arnottand lems describedbelow. I explored other problems of nonex- Stiglitz [1988a], and Dagobert L. Brito et al. [1995]). Even istence in the context of moral hazard problems in work small fixed costs (of search, of finding out about character- with RichardArott (1987, 199 b). 14 istics of different , of obtaining information This had a particularly inconvenient implication: about relevant technology) imply that markets will not be when there was a continuum of types, such as in the A. perfectly competitive;they will be better describedby mod- Michael Spence (1973, 1974) models, a full equilibrium els of (see Avinash K. Dixit and never existed. 468 THE AMERICANECONOMIC REVIEW JUNE 2002 when prices affect "quality"- either because of markets, e.g., for risk, has profound implica- incentive or selection effects- equilibriummay tions for how other markets function. The fact be characterizedby demand not equaling sup- that workers and firms cannot buy insurance ply; firms will not pay lower wages to workers, against many of the risks which they face af- even when they can obtain such workers, be- fects labor and capital markets; it leads, for cause doing so will raise their labor costs. Con- instance, to labor contracts in which the em- trary to the , we have shown ployer provides some insurance.But the design that the market will be characterizedby wage of these more complicated, but still imperfect and price distributions,even when there is no and incomplete, contractsaffects the efficiency, exogenous source of "noise" in the economy, and overall performance,of the economy. and even when all firms and workers are (oth- Perhapsmost importantly,under the standard erwise) identical. Contraryto standardcompet- paradigm, markets are Pareto efficient, except itive results, we have shown that in equilibrium, when one of a limited numberof marketfailures firms may charge a price in excess of the mar- occurs. Under the imperfect informationpara- ginal costs, or workers may be paid a wage in digm, marketsare almost never Paretoefficient. excess of their reservation wage, so that the While information economics thus under- incentive to maintaina reputationis maintained mined these long-standing principles of eco- (see also Benjamin Klein and Keith B. Leffler, nomics, it also provided explanationsfor many 1981; Shapiro, 1983). Contraryto the efficient phenomena that had long been unexplained. markets hypothesis (Eugene F. Fama, 1970), Before turning to these applications, I want which holds that stock prices convey all the to present a somewhat a more systematic ac- relevant informationfrom the informed to the count of the principles of the economics of uninformed,Sanford J. Grossmanand I (1976, information. 1980a) showed that, when informationis costly to collect, stock prices necessarily aggregate A. Some Problems in Constructingan information imperfectly (to induce people to AlternativeParadigm gather information,there must be an "equilib- rium amountof disequilibrium").Each of these The fact that informationis imperfectwas, of cornerstones of the competitive paradigm was course, well recognized by all economists. The rejected,or was shown to hold only undermuch reason that models with imperfect information more restrictive conditions. were not developed earlier was that it was not The most fundamental reason that markets obvious how to do so: While there is a single with imperfect informationdiffer from those in way in which informationis perfect, there are which informationis complete is that, with im- an infinitenumber of ways in which information perfect information, market actions or choices can be imperfect. One of the keys to success convey information.Market participantsknow was formulatingsimple models in which the set this and respond accordingly. For example, of relevantinformation could be fully specified- firms provide guaranteesnot only because they and so the precise ways in which information are better able to absorb the risk of product was imperfectcould also be fully specified. But failure but to convey information about their there was a danger in this methodology, as confidence in their products.A person takes an useful as it was: In these overly simplistic mod- insurancepolicy with a large deductibleto con- els, full revelation of information was some- vey to the insurerhis belief that the likelihood times possible. In the real world, of course, this of his havingan accidentis low. Informationmay never happens, which is why in some of the also be concealed: A firm may not assign an later work (e.g., Grossman and Stiglitz, 1976, employeeto a highly visiblejob, becauseit knows 1980a), we worked with models with an infinite that the assignmentwill be interpretedas an indi- number of states. Similarly there may well be cation thatthe employee is good, makingit more ways of fully resolving incentive problems in likely thata rival will try to hire the personaway. simple models, which collapse when models are One of the early insights (Akerlof, 1970) was made more realistic, for example by combining that, with imperfect information,markets may selection and incentive problems (Stiglitz and be thin or absent. The absence of particular Weiss, 1986). VOL. 92 NO. 3 STIGLITZ:INFORMATION ECONOMICS AND PARADIGMCHANGE 469

Perhaps the hardest problem in building the the mix of customers, in such a way that the new paradigmwas modeling equilibrium.It was profits of the new entrantactually became neg- important to think about both sides of the ative. One had to rethink all the conclusions market-employers and employees, insurance from first premises. company and the insured, lender and borrower. We made progress in our analyses because Each had to be modeled as "rational,"in some we began with highly simplified models of par- sense, making inferences on the basis of avail- ticularmarkets, that allowed us to think through able information and behaving accordingly. I carefully each of the assumptions and conclu- wanted to model competitive behavior, where sions. From the analysis of particularmarkets each actor in the economy was small, and be- (whether the insurance market, the education lieved he was small-and so his actions could market, the labor market, or the land tenancy/ not or would not affect the equilibrium(though sharecroppingmarket), we attemptedto identify others' inferences about himself might be af- general principles, to explore how these princi- fected). Finally, one had to think carefully about ples operated in each of the other markets. In what was the feasible set of actions: what might doing so, we identified particular features, each side do to extractor convey informationto particular informational assumptions, which others. seemed to be more relevant in one market or As we shall see, the variety of results ob- another.The natureof competition in the labor tained (and much of the confusion in the early market is different from that in the insurance literature)arose partly from a failure to be as marketor the capital market,though these mar- clear as one might have been about the assump- kets have much in common. This interplay, tions. For instance, the standardadverse selec- between looking at the ways in which such tion model had the quality of the good offered in marketsare similar and dissimilar,proved to be the market(say of used cars, or riskiness of the a fruitful research strategy.'5 insured)depending on price. The car buyer (the seller of insurance) knows the statistical rela- B. Sources of InformationAsymmetries tionship between price and quality, and this affects his demand. The market equilibrium is Information imperfections are pervasive in the price at which demand equals supply. But the economy: indeed, it is hardto imagine what that is an equilibriumif and only if there is no a world with perfect informationwould be like. way by which the seller of a good car can Much of the researchI describe here focuses on convey that information to the buyer-so that asymmetries of information, that fact that dif- he can earn a quality premium-and if there is ferent people know different things. Workers no way by which the buyer can sort out good know more about their own abilities than the cars from bad cars. Typically, there are such firm does; the person buying insurance knows ways, and it is the attempt to elicit that infor- more about his health, e.g., whether he smokes mation which has profoundeffects on how mar- and drinks immoderately, than the insurance kets function. To develop a new paradigm,we had to break out from long-established pre- 15 mises, to ask what should be taken as assump- Some earlier work, especially in general-equilibrium tions and what should be derived from the theory, by (1960, 1972), Jacob Marschak and Radner (1972), and Radner (1972), among others, had analysis. Marketclearing could not be taken as recognized the importanceof problems of information,and an assumption;neither could the premise that a had even identified some of the ways that limited informa- firm sells a good at a particular price to all tion affected the natureof the marketequilibrium (e.g., one could have contractsthat comers. One could not begin the analysis even only were contingent on states of that in nature that were observable by both sides to the contract). by assuming competitive equilibrium But the attempt to modify the abstract theory of general there would be zero profits. In the standard equilibriumto incorporateproblems of informationimper- theory, if there were positive profits, a firm fects proved, in the end, less fruitful than the alternative might enter, bidding away existing customers. approach of beginning with highly simplified, quite con- In the new the to bid new crete models. Arrow (1963, 1965, 1973, 1974, 1978), while theory, attempt away a within the was customers key figure general-equilibriumapproach, by slightly lowering prices might one of the first to identify the importanceof adverse selec- lead to marked changes in their behavior or in tion and moral hazard effects. 470 THE AMERICANECONOMIC REVIEW JUNE 2002 firm. Similarly,the owner of a car knows more lem of the "winner's curse" (Wilson, 1969; aboutthe car thanpotential buyers; the ownerof a Edward Capen et al., 1971). The government firm knows more about the firm that a potential (or other owners of large tractsto be developed) investor;the borrowerknows more aboutthe risk- should take this into account in its leasing strat- iness of his projectthan the lenderdoes; and so on. egy. And the biddersin the initial leases too will An essential featureof a decentralizedmarket take this into account:part of the value of win- economy is that differentpeople know different ning in the initial auction is the informationrent things, and in some sense, economists had long that will accrue in later rounds. been thinking of markets with information While early work in the economics of infor- asymmetries.But the earlier literaturehad nei- mation dealt with how marketsovercame prob- ther thought about how these were created, or lems of information asymmetries, later work what their consequences might be. While such turnedto how actors in marketscreate informa- information asymmetries inevitably arise, the tion problems, for example in an attempt to extent to which they do so and their conse- exploit marketpower. An example is managers quences depend on how the market is struc- of firms who attempt to entrench themselves, tured, and the recognition that they will arise and reduce competition in the marketfor man- affects marketbehavior. For instance, even if an agers, by taking actions to increase information individual has no more information about his asymmetry ( and Robert W. ability than potentialemployers, the moment he Vishny, 1989; Aaron S. Edlin and Stiglitz, goes to work for a specific employer, an infor- 1995). This is an example of the general prob- mation asymmetry has been created-the em- lem of corporate governance, to which I will ployer may now know more about the returnlater. Similarly, the presence of informa- individual's ability than others do. A conse- tion imperfectionsgive rise to marketpower in quence is that the "used labor"market may not product markets.Firms can exploit this market work well. Other employers will be reserved in power through"sales" and other ways of differ- bidding for the worker's services, knowing that entiating among individuals who have different they will succeed in luring him away from his search costs (Salop, 1977; Salop and Stiglitz, current employer only if they bid too much. 1977, 1982; Stiglitz, 1979a). The price disper- This impedimentto labor mobility gives market sions which exist in the market are created by power to the first employer, which he will be the market-they are not just the failure of temptedto exercise. But then, because a worker markets to arbitrage fully price differences knows he will tend to be locked into a job, he caused by shocks that affect different markets will be more risk averse in accepting an offer. differently. The terms of the initial contractthus have to be designed to reflect the diminutionof the work- C. Overcoming Information Asymmetries er's bargainingpower that occurs the moment he accepts a job. I now want to discuss briefly the ways by To take anotherexample, it is naturalthat in which informationasymmetries are dealt with, the process of oil exploration, a company may how they can be (partially)overcome. obtain information relevant to the likelihood that there will be oil in a neighboringtract-an 1. Incentives for Gathering and Disclosing informational exterality (see Stiglitz, 1975d; Information.-There are two key issues: what Jeffrey J. Leitzinger and Stiglitz, 1984). The are the incentivesfor obtaininginformation, and existence of this asymmetricinformation affects what are the mechanisms. My brief discussion the nature of the bidding for oil rights on the of the analysis of education as a screening de- neighboringtract. Bidding when there is known vice suggested the fundamentalincentive: More to be asymmetriesof informationwill be mark- able individuals (lower risk individuals, firms edly different from that where such asymme- with betterproducts) will receive a higher wage tries do not exist (Robert B. Wilson, 1977). (will have to pay a lower premium,will receive Those who are uninformed will presume that a higher price for their products) if they can they will win only if they bid too much- establish that they are more productive (lower information asymmetries exacerbate the prob- risk, higher quality). VOL. 92 NO. 3 STIGLITZ:INFORMATION ECONOMICS AND PARADIGMCHANGE 471

We noted earlier that while some individuals sharecropping, which arise when workers do have an incentive to disclose information,some not own the land that they till. This pro- have an incentive not to have the information blem could be overcome if individuals could disclosed. Was it possible that in market equi- borrow to buy their land. But capital market librium, only some of the informationwould be imperfections-limitations on the ability to bor- revealed?One of the early importantresults was row, which themselves arise from information that, if the more able can costlessly establish imperfections-explain why this "solution" that they are more able, then the marketwill be does not work. fully revealing, even though those who are be- There is another important consequence: if low averagewould preferthat no informationbe marketswere fully informationallyefficient-that revealed. In the simplest models, I described a is, if information disseminated instantaneously process of unraveling: If the most able could and perfectlythroughout the economy-then no establish his ability, he would; but then all but one would have any incentive to gather infor- the most able would be grouped together, re- mation, so long as there was any cost of doing ceiving the mean marginal product of that so. Hence marketscannot be fully information- group; and the most able of that group would ally efficient (Grossman and Stiglitz, 1976, have an incentive to reveal his ability. And so 1980a). on down the line, until there was full revelation. (I jokingly referredto this as "Walras'Law of 2. Mechanismsfor Elimination of Reducing Sorting"-if all but one group sorts itself out InformationAsymmetries. In simple models from the others, then the last group is also where (for example) individuals know their identified.) own abilities there might seem an easy way to What happens if those who are more able resolve the problem of informationasymmetry: cannot credibly convince potentialemployers of Let each person tell his true characteristic.Un- their ability?The other side of the markethas an fortunately,individuals do not necessarily have incentive too to gather information. An em- the incentive to tell the truth. Talk is cheap. ployer that can find a worker that is better than Other methods must be used to convey infor- is recognized by others will have found a bar- mation credibly. gain, because the worker's wage will be deter- The simplest way by which that could be mined by what others think of him. The done was an exam. Models of competitive equi- problem, as we noted, is that if what the em- librium (Arrow, 1973; Stiglitz, 1974a) with ex- ployer knows becomes known to others, the ams make two general points. First, in worker's wage will be bid up, and the employer equilibrium the gains of the more able were will be unable to appropriatethe returnson his largely at the expense of the less able. Estab- investment in informationacquisition. lishing that an individual is of higher ability The fact that competition makes it difficult provides that person with higher wages, but for the screener to appropriatethe returnsfrom simultaneously establishes that others are of screening has an importantimplication: In mar- lower ability. Hence the private returns to ex- kets where, for one reason or another,the more penditureson educational screening exceed the able cannot fully convey their attributes,invest- social returns.It was clear that there were im- ment in screening requires imperfect competi- portant associated with informa- tion in screening. The economy, in effect, has to tion, a theme which was to recur in later work. choose between two different imperfections: Second, and a more striking result, there could imperfections of information or imperfections exist multiple equilibria-one in which infor- of competition.Of course, in the end, there will mation was fully revealed (the marketidentified be both forms of imperfection,and no particular the high and low ability people) and anotherin reason that these imperfections will be "bal- which it was not (called a pooling equilibrium). anced" optimally (Stiglitz, 1975b; Dwight The pooling equilibrium Pareto-dominatedthe Jaffee and Stiglitz, 1990). This is but one of equilibrium with full revelation. This work, many examples of the interplaybetween market done some 30 years ago, establishedtwo results imperfections. Earlier, for instance, we dis- of great policy import, which remarkablyhave cussed the incentive problems associated with not been fully absorbedinto policy discussions 472 THEAMERICAN ECONOMIC REVIEW JUNE 2002 even today. First, markets do not provide ap- locate itself, if it only wants to get healthy propriateincentives for informationdisclosure. applicants, depends on other elements of its There is, in principle, a role for government. strategy, such as the premium charged. Or the And second, expenditureson informationmay company may decide to throw in a member- be too great (see also Hirshleifer, 1971). ship in a health club, but charge a higher premium. Those who value a health club- 3. Conveying Information Through Ac- because they will use it-willingly pay the tions. But much of the information firms higher premium. But these individuals are glean about their employees, banks about their likely to be healthier. borrowers,or insurance companies about their There are a host of other actions which insured,comes not from examinationsbut from convey information. The quality of the guar- making inferencesbased on their behavior.This antee offered by a firm can convey informa- is a commonplacein life-but it was not in our tion about the quality of the product; only economic models. As I have already noted, the firms that believe that their product is reliable early discussions of adverse selection in insur- will be willing to offer a good guarantee. The ance markets recognized that as an insurance guarantee is desirable not just because it re- company raised its premiums, those who were duces risk, but because it conveys informa- least likely to have an accidentmight decide not tion. The number of years of schooling may to purchase the insurance; the willingness to convey information about the ability of an purchase insurance at a particularprice con- individual. More able individuals may go to veyed information to the insurance company. school longer, in which case the increase in George Akerlof recognized that this phenome- wages associated with an increase in school- non is far more general:the owner's willingness ing may not be a consequence of the human to sell a used car, for instance, conveyed infor- capital that has been added, but rather simply mation about the car's quality. be a result of the sorting that occurs. The size Bruce C. Greenwald(1979, 1986) took these of the deductible that an individual chooses ideas one importantstep further,showing how in an insurance policy may convey infor- adverse selection applied to labor and capital mation about his view about the likelihood of markets (see also Greenwald et al., 1984; an accident or the size of the accidents he Stewart C. Myers and Nicholas S. Majluf, anticipates-on average, those who are less 1984). For example, the willingness of insiders likely to have an accident may be more will- in a firm to sell stock at a particular price ing to accept high deductibles. The willing- conveys information about their view of what ness of an entrepreneurto hold large fractions the stock is really worth. Akerlof's insight that of his wealth in a firm (or to retain large the result of these informationasymmetries was fractions of the shares of the firm) conveys that markets would be thin or absent helped information about his beliefs in the firm's explain why labor and capital marketsoften did future performance. If a firm promotes an not function well. It providedpart of the expla- individual to a particularjob, it may convey nation for why firms raised so little of their information about the firm's assessment of his funds throughequity (Mayer, 1990). Stigler was ability. wrong: imperfectinformation was not just like a The fact that these actions may convey infor- transactionscost. mation affects behavior. In some cases, the ac- There is a much richer set of actions which tion will be designed to obfuscate, to limit convey information beyond those on which informationdisclosure. The firmthat knows that traditional adverse selection models have fo- others are looking at who it promotes, and will cused. An insurance company wants to attract compete more vigorously for those workers, healthy applicants. It might realize that by may affect the willingness of the firm to pro- locating itself on the fifth floor of a walk-up mote some individualsor assign them to partic- building, only those with a strong heart would ular jobs (Michael Waldman, 1984). In others, apply. The willingness or ability to walk up the action will be designed to convey informa- five floors conveys information. More subtly, tion in a credible way to alter beliefs. The fact it might recognize that how far up it needs to that customers will treat a firm that issues a VOL.92 NO. 3 STIGLITZ:INFORMATION ECONOMICS AND PARADIGMCHANGE 473 better guaranteeas if its productis better-and result in those with different characteristicsin thereforebe willing to pay a higher price-may effect identifyingthemselves throughtheir self- affect the guaranteethat the firm is willing to selection. The particularmechanism which we issue. Knowing that selling his shares will con- explored in our insurance model illustrates vey a negative signal concerning his views of how self-selection mechanisms work. People the futureprospects of his firm, an entrepreneur who know they are less likely to have an may retain more of the shares of the firm; he accident will be more willing to accept an will be less diversifiedthan he otherwise would insurance policy with a high deductible, so have been (and accordingly, he may act in a that an insurance company that offered two more risk-averse manner). policies, one at a high premium and no de- A simple lesson emerges: Some individuals ductible, one with a low premium and high wish to convey information; some individuals deductible, would be able to sort out who wish not to have informationconveyed (either were high risk and who low. It is an easy because such information might lead others to matter to construct choices which thus sepa- think less well of them, or because conveying rate people into classes. information may interfere with their ability to It was clear that information was conveyed appropriaterents). In either case, the fact that because the actions were costly, and more actions convey informationleads people to alter costly for some than others. The attempt to their behavior, and changes how marketsfunc- convey informationhad to distort behavior.Our tion. This is why information imperfections analysis also made it clear that it was not just have such profound effects. information asymmetries, but information im- Once one recognizes that actions convey perfections more generally, that were relevant. information, two results follow. First, in mak- Even if those buying insurance did not know ing decisions about what to do, individuals their accident probabilities(or know them with will not only think about what they like (as in greater accuracy than the insurance company), traditional economics) but how it will affect so long as those with higher accident probabil- others' beliefs about them. If I choose to go to ities on average differed in some way reflected school longer, it may lead others to believe in their preferences and actions, self-selection that I am more able. I may therefore decide to mechanisms could and would be employed to stay in school longer, not because I value sort. what is being taught, but because I value how Yet another set of issues arise from the fact it changes others' beliefs concerning my that actions may not be costlessly observable. ability. This means, of course, that we have to The employer would like to know how hardhis rethink completely firm and worker is working; the lender would like to decision-making. know the actions which borrower will under- Secondly, we noted earlier that individuals take. These asymmetries of information about have an incentive to "lie"-the less able to say actions are as importantas the earlierdiscussed that they are more able. Similarly, if it becomes asymmetries. Just as in the adverse selection recognized that those who walk up to the fifth model, the seller of insurancemay try to over- floor to apply for insurance are more healthy, come the problemsposed by informationasym- then I might be willing to do so even if I am not metries by examination, so too in the moral so healthy, simply to fool the insurance com- hazard or adverse incentive model, he may try pany. Recognizing this, one needs to look for to monitor the actions of the insured. But ex- ways by which informationis conveyed in equi- aminationsand monitoringare costly, and while librium. The critical insight in how that could they yield some information,typically there re- occur was provided in a paper I wrote with mains a high level of residual informationim- (1976). If those who were perfection. One response to this problem is to more able, less risk prone, or more creditworthy try to induce desired behavior through the set- acted in some observable way (had different ting of contractterms. For example, borrowers' preferences)than those who were less able, less risk-taking behavior may be affected by the risk prone, or less creditworthy,then it might be charged by the lender (Stiglitz and possible to design a set of choices, which would Weiss, 1981). 474 THEAMERICAN ECONOMIC REVIEW JUNE 2002

D. Consequences for Market Equilibrium that a firm might make depends not only on what that firmdoes, but also on what other firms The law of supply and demandhad long been do, turned out, however, to be a difficult task. treated as a fundamentalprinciple of econom- The easiest situation to analyze was that of a ics. But there is in fact no law that requiresthe monopolist (Stiglitz, 1977). The monopolist insurance firm to sell to all who apply at the could construct a set of choices that would announcedpremium, or the lender to lend to all differentiateamong different types of individu- who apply at the announcedinterest rate, or the als, and analyze whether it was profit maximiz- employer to employ all those who apply at the ing for him to do so fully, or to (partially) posted wage. With perfect informationand per- "pool"-that is, offer a set of contractssuch that fect competition, any firm that charged a price several types might choose the same one. This higher than the others would lose all of its work laid the foundationsof a general theory of customers; and at the going price, one faced a price discrimination. Under standard theories of perfectly elastic supply of customers.In adverse monopoly, with perfect information, firms selection and incentive models, what mattered would have an incentive to price discriminate was not just the supply of customersor employ- perfectly (extracting the full consumer surplus ees or borrowers,but their "quality"-the risk- from each). If they did this, then monopoly iness of the insuredor the borrower,the returns would in fact be nondistortionary.Yet most on the investment,the productivityof the worker. models assumedno price discrimination(that is, Since "quality"may increase with price, it the monopolist offered the same price to all may be profitable(for example) to pay a higher customers), without explaining why they did wage than the "market-clearing"wage, whether not do so. The new work showed how, given the dependence on quality arises from adverse limited information,firms could price discrim- selection or adverse incentive effects (or, in the inate, but could do so only imperfectly. Subse- labor market, because of morale or nutritional quent work by a variety of authors (such as effects). The consequence, as we have noted, is William J. Adams and Yellen, 1976; Salop, that market equilibrium may be characterized 1977) explored ways by which a monopolist by demandnot equalingsupply in the traditional might find out relevant characteristics of his sense. In creditmarket equilibrium, the supplyof customers. Information economics thus pro- loans may be rationed(William R. Keeton, 1979; vided the first coherent theory of monopoly. Jonathan Eaton and Mark Gersovitz, 1981; The reason that analyzing monopoly was Stiglitzand Weiss, 1981). Or, in the labormarket, easy is that the monopolist could structurethe the wage ratemay be higherthan that at which the entire choice set facing his customers.The hard demandfor laborequals the supply(an efficiency question is to describethe full competitive equi- wage), leadingto unemployment.6 librium, e.g., a set of insurance contracts such Analyzing the choices which arise in full that no one can offer an alternative set that equilibrium,taking into account fully not only would be profitable.Each firm could control the the knowledge that the firms have, say, about choices that it offered, but not the choices of- their customers but also the knowledge that fered by others; and the decisions made by customers have about how firms will make in- customersdepended on the entire set of choices ferences about them from their behavior, and available. In our 1976 paper, Rothschild and I taking into account the fact that the inferences succeeded in analyzing this case. Three strikingresults emergedfrom this anal- ysis. The first I have alreadymentioned: Under the naturaldefinition 16 models with these effects is plausible conditions,given Constructingequilibrium of not exist. more difficult than might seem to be the case at first, since equilibrium, equilibrium might each agent's behavior depends on opportunitieselsewhere, There were two possible forms of equilibria: i.e., the behaviorof others. For example, the workersthat a pooling equilibria, in which the market is not firm attractsat a particularwage depend on the wage offers able to distinguish among the types, and sepa- of other firms. Shapiro and Stiglitz (1984), Rodriguez and in which it is. The different Stiglitz (1991a, b), and Rey and Stiglitz (1996), represent rating equilibria, attempts to come to terms with these general-equilibrium groups "separateout" by taking different ac- problems. tions. We showed in our context that there never VOL. 92 NO. 3 STIGLITZ:INFORMATION ECONOMICS AND PARADIGMCHANGE 475 could be a pooling equilibrium-if there were a insured,and lender and creditorare awareof the single contractthat everyone bought, there was informationalconsequences of their actions. In another contract that another firm could offer the case where, say, the insurance company or which would "break"the pooling equilibrium. employer takes the initiative in sorting out ap- On the other hand, there might not exist a sep- plicants, self-selection is an alternativeto exam- aratingequilibrium either, if the cost of separa- inations as a sorting device. In the case where tion was too great. Any putative separating the insured,or the employee, takes the initiative equilibrium could be broken by a profitable to identify himself as a more attractivecontrac- pooling contract, a contract which would be tual partner,then it is conventional to say he is bought by both low risk and high risk types.17 signaling (Spence, 1973). But of course, in Second, even small amountsof imperfections equilibriumboth sides are aware of the conse- of informationcan change the standardresults quences of alternative actions, and the differ- concerningthe existence and characterizationof ences between signaling and self-selection equilibrium. Equilibrium, for instance, never screening models lie in the technicalities of exists when the two types are very near each game theory, and in particularwhether the in- other. As we have seen, the competitive equi- formed or uninformedplayer moves first.18 librium model is simply not robust. Still, some of the seeming differences be- Third, we now can see how the fact that tween signaling and screening models arise be- actions convey informationaffects equilibrium. cause of a failure to specify a full equilibrium. In perfect information models, individuals We noted earlier that there might be many sep- would fully divest themselves of the risks which aratingcontracts, but a unique separatingequi- they face, and accordingly would act in a risk librium. We argued that if one considered any neutral manner. We explained why insurance other separating set of contracts, then (say, in markets would not work well-why most risk- the insurancemarket) a firm could come in and averse individualswould buy only partialinsur- offer an alternativeset of contractsand make a ance. The result was importantnot only for the profit. Then the original set of separatingcon- insights it provided into the workings of insur- tracts could not have been an equilibrium.The ance markets, but because there are important same is true in, say, the education signaling elements of insurancein many transactionsand model. There are many educational systems markets. The relationshipbetween the landlord which "separate"-that is, the more able choose and his tenant, or the employer and his em- to go to school longer, and the wages at each ployee, contains an insurancecomponent. level of education correspondto the productiv- In short, the general principle that actions ity of those who go to school for that length of convey information applies in many contexts. time. But all except one are not full equilibria. Further,limitations on the ability to divest one- Assume, for instance, there were two types of self of risk are importantin explaining a host of individuals, of low ability and of high ability. contractualrelationships. Then if the low-ability person has 12 years of schooling, then any education system in which E. Sorting, Screening, and Signaling the high-ability person went to school suffi- ciently long-say, more than 14 years-might In equilibrium,both buyers and sellers, em- separate. But the low-ability types would rec- ployers and employees, insurancecompany and

18 See, in particular,Stiglitz and Weiss (1983a, 1994) and Shiro Yabushita (1983). As we point out, in the real 17 Of course, insurance markets do exist in the real world, who moves first ought to be viewed as an endoge- world. I suspect that a major limitation of the applicability nous variable. In such a context, it appearsthat the screen- of Rothschild-Stiglitz (1976) is the assumption of perfect ing equilibria are more robust than the signaling competition. Factors such as search costs and uncertainty equilibrium. Assume, for instance, that there were some about how easy it is to get a company to pay a claim make signaling equilibriumthat differed from the screening equi- the assumption of less plausible. Self- librium, e.g., there were a pooling equilibrium, sustained selection is still relevant, but some version of monopolistic because of the out-of-equilibriumbeliefs of firms. Then competition,may be more relevantthan the model of perfect such an equilibrium could be broken by a prior or later competition. move of firms. 476 THE AMERICANECONOMIC REVIEW JUNE 2002 ognize that if they went to school for 11 years, The work on sharecroppingand on equilib- they would still be treatedas having low ability. rium with competitive insurance markets The unique equilibrium level of education for showed that with imperfect information, a far the low-ability person is that which maximizes richer set of contracts would be employed and his income (taking into account the produc- thus began a large literatureon the theory of tivity gains and costs of education).The unique contracting. In the simple sharecroppingcon- equilibrium level of education for the high- tractsof Stiglitz (1974b), the contractsinvolved ability type is the lowest level of educationsuch shares, fixed payments, and plot sizes. More that the low-ability type does not have the in- generally, optimal payment structures related centive to mimic the high-ability person's edu- payments to observables, such as inputs, pro- cational attainment. cesses, or outputs.19Further, because what goes The education system, of course, was partic- on in one market affects other parts of the ularly infelicitous for studying market equilib- economy, the credit, labor, and land marketsare rium. The structureof the education system is interlinked; one could not decentralize in the largely a matterof public choice, not of market way hypothesizedby the standardperfect infor- processes. Different countries have chosen mation model. ( and markedlydifferent systems. The minimumlevel Stiglitz, 1982, 1986a, b, 1989). of educationis typically not a matterof choice, These basic principleswere subsequentlyap- but set by the government. Within educational plied in a variety of other marketcontexts. The systems, examinationsplay as importanta role most obvious was the design of labor contracts as self-selection or signaling, though given a (Stiglitz, 1975a). Payments to workers can de- certain standardof testing, there is a process of pend not only on output,but on relative perfor- self-selection involved in deciding whether to mance, which may convey more relevant stay in school, or to try to pass the examination. informationthan absolute performance.For ex- For the same reason, the problems of existence ample, the fact that a particular company's which arise in the insurance market are not stock goes up when all other companies' stock relevantin the educationmarket-the "compet- goes up may say very little about the perfor- itive" supply side of the market is simply ab- mance of the manager. Nalebuff and Stiglitz sent. But when the signaling concepts are (1983a, b) analyzed the design of these relative translatedinto contexts in which there is a ro- performancecompensation schemes (contests). bust competitive market, the problems of exis- Creditmarkets too are characterizedby com- tence cannot be so easily ignored. In particular, plicated equilibrium contracts. Lenders may when there is a continuum of types, as in the specify not only an interestrate, but also impose Spence (1973) model, there never exists a other conditions (collateralrequirements, equity screening equilibrium. requirements)which would have both incentive and selection effects.20 Indeed, the simulta- F. EquilibriumContracts neous presence of both selection and incentive

The work with Rothschildwas relatedto ear- 19 lier work that I had done on incentives (such as In Stiglitz (1974b) the contractswere highly linear. In the work on in that both lines of principle, generalizing payment structures to nonlinear sharecropping) functions was simple. Though even here, there were subtle- work entailed an "equilibrium in contracts." ties, e.g., whether individuals exerted their efforts before The contracts that had characterizedeconomic they knew the realizationof the state of nature,and whether relations in the standard competitive model there were bounds on the penalties that could be imposed, in were I will a the event of bad outcomes (James A. Mirrlees [1975b]; extraordinarilysimple: pay you Mirrlees The literature has not certain amount if do such and such. If Stiglitz [1975a]; [1976]). you you fully resolved the reason that contracts are often much did not perform as promised, the pay was not simpler than the theory would have predicted (e.g., pay- given. But with perfect information,individuals ments are linear functions of output), and do not adjust to simply would not sign contracts that they did changes in circumstances(see, e.g., Franklin Allen, 1985; not intend to fulfill. Insurance contracts were Douglas Gale, 1991). 20 See, for instance, Stiglitz and Weiss (1983b, 1986, similarly simple: A payment occurred if and 1987). Even with these additional instrumentsthere could only if particularspecified events occurred. still be nonmarket-clearingequilibria. VOL. 92 NO. 3 STIGLITZ:INFORMATION ECONOMICS AND PARADIGMCHANGE 477 effects is important in credit markets. In the epsilon more would lose no customers and thus absence of the former, it might be possible to would choose to increase its price. Similarly, it increase the collateral requirement and raise would pay all other firms to increase their interest rates, still ensuring that the borrower prices. But at the higher price, it would again undertookthe safe project. pay each to increase price, and so on until the As another application, "contracting"- price charged at every firm is the monopoly including provisions that help information be price, even though search costs are small. This conveyed and risks be shared-have been showed convincingly that the competitive price shown to play an importantrole in explaining was not the equilibrium.But in some cases, not macroeconomic rigidities. See, for instance, even the monopoly price was an equilibrium.In Costas Azariadis and Stiglitz (1983), the papers general, Salop and Stiglitz (1977, 1982, 1987) of the symposium in the 1983 QuarterlyJour- and Stiglitz (1979b, 1985c, 1987b, 1989c) nal of Economics, the survey article by Sherwin showed that in situationswhere there were even Rosen (1985), Arott et al. (1988), and Lars small search costs, marketsmight be character- Werin and Hans Wijkander(1992). Moreover, ized by a price distribution.The standardwis- problems of asymmetries of information can dom that said that not everyone had to be help explain the perpetuationof seemingly in- informed to ensure that the market acted per- efficient contracts. (Stiglitz, 1992b). fectly competitive was simply not, in general, true (see Stiglitz, 1989c, for a survey). G. EquilibriumWage and Price Distributions IV. Efficiency of the Market Equilibrium and One of the most obvious differences between the Role of the State the predictionsof the model with perfect infor- mation and what we see in everyday life is the The fundamental theorems of neoclassical conclusion that the same good sells for the same welfare economics state that competitive econ- price everywhere. In reality, we all spend a omies will lead, as if by an invisible hand, to a considerableamount of time shopping for good (Pareto-) efficient allocation of resources, and buys. The differences in prices representmore that every Pareto-efficient resource allocation thanjust differences in quality or service. There can be achieved through a competitive mecha- are real price differences. Since Stigler's classic nism, provided only that the appropriatelump- paper (1961), there has been a large literature sum redistributions are undertaken. These exploring optimal search behavior. However theorems provide both the rationale for the re- Stigler, and most of the search literature,took liance on free markets, and for the belief that the price or wage distributionas given. They did issues of distribution can be separated from not ask how the distribution might arise and issues of efficiency, allowing the the whether, given the search costs, it could be freedom to push for reforms which increase sustained. efficiency, regardless of their seeming impact As I began to analyze these models, I found on distribution. (If society does not like the that there could be a nondegenerateequilibrium distributional consequences of a policy, it wage or price distribution even if all agents should simply redistributeincome.) were identical, e.g., faced the same searchcosts. The economics of information showed that Early on, it had become clear that even small neither of these theorems was particularlyrele- searchcosts could make a large differenceto the vant to real economies. To be sure, economists behavior of productand labor markets.Peter A. over the preceding three decades had identified Diamond (1971) had independently made this important market failures-such as the exter- point in a highly influentialpaper, which serves nalities associated with pollution-which re- to illustratepowerfully the lack of robustnessof quired government intervention.But the scope the competitive equilibriumtheory. Assume for for market failures was limited, and thus the example, as in the standardtheory, that all firms arenas in which government intervention was were chargingthe competitive price, but there is requiredwere correspondinglylimited. an epsilon cost of searching,of going to another Early work, already referredto, had laid the store. Then any firm which charged half an foundations for the idea that economies with 478 THE AMERICANECONOMIC REVIEW JUNE 2002 informationimperfections would not be Pareto While it was perhapsnot surprisingthat mar- efficient, even taking into account the costs of kets might not provide appropriateincentives obtaining information.There were interventions for the acquisition and dissemination of infor- in the marketthat could make all parties better mation, the marketfailures associated with im- off. We had shown, for instance, that incentives perfect informationmay be far more profound. for the disclosure and acquisitionof information The intuition can be seen most simply in the were far from perfect. On the one hand, imper- case of models with moral hazard. There, the fect appropriabilitymeant that there might be premiumcharged is associatedwith the average insufficient incentives to gather information; risk and, therefore, the average care, taken by but on the other, the fact that much of the gains seemingly similar individuals. The moral haz- were "rents,"gains by some at the expense of ard problem arises because the level of care others, suggested that there might be excessive cannot be observed. Each individualignores the expenditureson information.A traditionalargu- effect of his actions on the premium;but when ment for unfettered capital markets was that they all take less care, the premium increases. there are strong incentives to gather informa- The lack of care by each exerts a negative tion; discovering that some stock was more on others. The essential insight of valuable than others thoughtwould be rewarded Greenwald and Stiglitz (1986)22 was to rec- by a capital gain. This price discovery function ognize that such externality-like effects are of capital markets was often advertised as one pervasive whenever information is imperfect of its strengths. But while the individual who or markets incomplete-that is always-and discoveredthe informationa nanosecondbefore as a result, markets are essentially never anyone else might be betteroff, was society as a constrained Pareto efficient. In short, market whole better off? If having the informationa failures are pervasive. Arnott et al. (1994) nanosecondearlier did not lead to a changein real provide a simple exposition of this point us- decisions (e.g., concerning investment),then it ing the standard self-selection and incentive was largelyredistributive, with the gains of those compatibility constraints. obtainingthe informationoccurring at the expense An importantimplication is that efficient al- of others(Stiglitz, 1989c). locations cannot in general be decentralizedvia There are potentially other inefficiencies as- competitive markets.The notion that one could sociated with informationacquisition. Informa- decentralize decision-making to obtain (Pa- tion can have adverse effects on volatility reto-) efficient resource allocation is one of the (Stiglitz, 1989i). And information can lead to fundamental ideas in economics. Greenwald the destructionof markets,in ways which lead and Stiglitz (1986) showed that that was not to adverse effects on welfare. For example, in- possible in general. A simple example illus- dividuals may sometimes have incentives to trates what is at issue. An insurance company create information asymmetries in insurance cannot monitor the extent of smoking, which markets,which leads to the destructionof those marketsand a lowering of overall welfare. Wel- fare might be increasedif the acquisitionof this kind of information could be proscribed. Re- plications, see Arrow (1972), Phelps (1972), and Stiglitz such issues have become sources of real (1973a, 1974d). See also Stiglitz (1984a). cently, 22 Greenwald and Stiglitz (1986) focus on models with policy concern in the arena of genetic testing. adverse selection and incentive problems. Greenwald and Even when information is available, there are Stiglitz (1988a) showed that similar results hold in the issues concerningits use, with the use of certain context of search and other models with imperfect informa- kinds of information either a discrimi- tion. Earlier work, with Shapiro (1983) had shown, in the having context of a that in an intent or in circumstances in specific model, equilibria economy natory effect, with an agency or principal-agentproblem were not (con- which such directdiscrimination itself would be strained) Pareto efficient. Later work, with Arott (1990), prohibited.21 explored in more detail the market failures that arise with moral hazard. Earlier work had shown that with imperfect risk markets, themselves explicable by imperfections of information,market equilibrium was Pareto inefficient. See 21 See, e.g., Rothschild and Stiglitz (1982, 1997). For David M. G. Newbery and Stiglitz (1982, 1984) and Stiglitz models of statistical discriminationand some of their im- (1972a, 1981, 1982b). VOL. 92 NO. 3 STIGLITZ:INFORMATION ECONOMICS AND PARADIGMCHANGE 479 has an adverseeffect on health. The government same informational constraints, welfare could cannot monitor smoking any better than the be improved (Stiglitz, 1989a). insurancecompany, but it can impose , not There was another rear-guard argument, only on cigarettes, but also on other commodi- which ultimately holds up no better. It is that ties which are complements to smoking (and marketfailures-absent or imperfectmarkets- subsidies on substituteswhich have less adverse give rise to nonmarketinstitutions. For exam- effects). See Arott and Stiglitz (1991a) and ple, the absence of death insurancegave rise to Stiglitz (1989a, 1998b). burial societies. Families provide insurance to A related result from the new information their members against a host of risks for which economics is that issues of efficiency and equity they either cannot buy insurance, or for which cannot easily be delinked. For example, with the insurance premium is viewed as too high. imperfect information, a key source of market But in what I call the functionalistfallacy, it is failure is agency problems, such as those which easy to go from the observationthat an institu- arise when the owner of land is different from tion arises to fulfill a function to the conclusion the person working the land. The extent of that actually, in equilibrium,it serves that func- agency problems depends on the distributionof tion. Those who succumbed to this fallacy wealth, as we noted earlier in our discussion of seemed to argue that there was no need for sharecropping.Moreover, the notion that one governmentintervention because these nonmar- could separate out issues of equity and effi- ket institutions would "solve" the market fail- ciency also rested on the ability to engage in ure, or at least do as well as any government. lump sum redistributions. But as Mirrlees Richard Arott and I (1991a) showed that, to (1971) had pointed out, with imperfectinforma- the contrary,nonmarket institutions could actu- tion, this was not possible; all redistributive ally make mattersworse. Insuranceprovided by taxation must be distortionary. But this fact the family could crowd out market insurance, implies that interventionsin the market which for example. Insurancecompanies would recog- change the before-tax distribution of income nize that the insured would take less risk be- could be desirable, because they lessened the cause they had obtained insurancefrom others, burden on redistributive taxation (Stiglitz, and accordingly cut back on the amount of 1998a). Again, the conclusion: The second wel- insurance that they offered. But since the non- fare theorem, effectively asserting the ability to market (family) institutions did a poor job of separate issues of distribution and efficiency, divesting risk, welfare could be decreased. was not true. The Arott-Stiglitz analysis reemphasized In effect, the Arrow-Debreumodel had iden- the basic point made at the end of the last tified the single set of assumptionsunder which subsection: it was only under very special cir- markets were efficient. There had to be per- cumstances that markets could be shown to be fect information;more accurately, information efficient. Why then should we expect an equi- could not be endogenous, it could not change librium involving nonmarket institutions and either as a result of the actions of any individual markets to be efficient? or firm or through investments in information. But in the world we live in, a model which V. FurtherApplications of the New Paradigm assumes that information is fixed seems irrelevant. Of all the marketfailures, the extended peri- As the theoreticalcase that marketsin which ods of underutilizationof resources-especially information is imperfect were not efficient be- human resources-is of the greatest moment. came increasinglyclear, several new arguments The consequences of unemployment are exac- were put forward against governmentinterven- erbatedin turnby capital marketimperfections, tion. One we have already dealt with: that the which imply that even if the futureprospects of government too faces informational imper- an unemployed individual are good, he cannot fections. Our analysis had shown that the in- borrowenough to sustain his standardof living. centives and constraints facing government We referredearlier to the dissatisfactionwith differed from those facing the private sector, so traditionalKeynesian explanations, in particu- that even when government faced exactly the lar, the lack of .This dissatis- 480 THE AMERICANECONOMIC REVIEW JUNE 2002 faction gave rise to two schools of thought.One matters. Both because of the cost of bankrupt- sought to use the old perfect marketparadigm, cies and limitationsin the design of managerial relying heavily on representativeagent models. incentive schemes, firms act in a risk-averse While informationwas not perfect, expectations manner-with risk being more than just corre- were rational. But the representative agent lation with the business cycle (Greenwald and model, by construction,ruled out the informa- Stiglitz, 1990a; Stiglitz, 1989g). Moreover, be- tion asymmetrieswhich are at the heart of mac- cause of the potential for , not roeconomic problems. If one begins with a only does the firm's net worth matter, but so model that assumes that marketsclear, it is hard does its asset structure,including its liquidity.23 to see how one can get much insight into un- While there are many implications of the employment (the failure of the labor marketto theory of the risk-averse firm facing credit ra- clear). tioning, some of which are elaboratedupon in The constructionof a macroeconomicmodel the next section, one example should suffice to which embraces the consequences of imperfec- highlight the importanceof these ideas. In tra- tions of informationin labor, product,and cap- ditional neoclassical investment theory, invest- ital markets has become one of my major ment depends on the real interest rate and the preoccupations over the past 15 years. Given firm's perception of expected returns. The the complexity of each of these markets, firm's cash flow or its net worth should make no creating a general-equilibriummodel-simple difference. The earliest econometric studies of enough to be taughtto graduatestudents or used investment, by Edwin Kuh and John R. Meyer by policy makers-has not proven to be an easy (1957), suggested however that this was not the task. At the heartof that model lies a new theory case. Nevertheless these variables were ex- of the firm, for which the theory of asymmetric cluded from econometric analyses of invest- informationprovides the foundations.The mod- ment for two decades following the work of ern in turn rests on three RobertE. Hall and Dale W. Jorgenson(1967). It pillars, the theory of corporatefinance, the the- was not until work on asymmetricinformation ory of corporategovernance, and the theory of had restoredtheoretical respectability that it be- organizationaldesign. came acceptableto introducefinancial variables into investment regressions. When that was A. Theory of the Firm done, it was shown that-especially for small- and medium-sized enterprises-these variables Under the older, perfect information theory are crucial. (For a survey of the vast empirical ( and Merton H. Miller, literaturesee R. Glenn Hubbard,1998). 1958, 1961; see also Stiglitz, 1969a, 1974c, In the traditionaltheory, firms simply maxi- 1988d), it made no difference whether firms mized the expected present discounted value of raised capital by debt or equity, in the absence profits (which equaled marketvalue); with per- of tax distortions.But informationis at the core fect information,how that was to be done was of finance. The informationrequired to imple- simply an engineering problem. Disagreements ment equity contracts is greater than for debt about what the firm should do were of little contracts (Robert J. Townsend, 1979; Green- moment. In that context, corporate governance- wald and Stiglitz, 1992). Most importantly,the how firm decisions were made-mattered little willingness to hold (or to sell) shares conveys as well. But again, in reality, corporategover- information (Hayne E. Leland and David H. nance matters a great deal. There are disagree- Pyle, 1977; Ross, 1977; Stiglitz, 1982c; Green- ments about what the firm should do-partly wald et al., 1984; Myers and Majluf, 1984; motivated by differences in judgments, partly Thomas F. Hellman and Stiglitz, 2000; for em- motivatedby differences in objectives (Stiglitz, pirical verification see, e.g., Paul Asquith and David W. so that how firms Mullins, Jr., 1986), 23 raise does make a difference. In The very concept of liquidity-and the distinction capital practice, between lack of and on firms on debt (as to liquidity insolvency-rests infor- rely heavily opposed equity) mation asymmetries.If there were perfect information,any finance (Mayer, 1990), and bankruptcy,result- firm that was liquid would be able to obtain finance, and ing from the failure to meet debt obligations, thus would not face a liquidity problem. VOL. 92 NO. 3 STIGLITZ:INFORMATION ECONOMICS AND PARADIGMCHANGE 481

1972b; Grossman and Stiglitz, 1977, 1980b). be based. (Nalebuff and Stiglitz, 1983a, b). But Managerscan take actions which advance their there is another importantaspect of organiza- at the expense of that of shareholders, tion design. Even if individuals are well inten- and majority shareholders can advance their tioned, with limited information, mistakes get interests at the expense of minority sharehold- made. To err is human. Raaj K. Sah and I, in a ers. The owners not only could not monitortheir series of papers (1985, 1986, 1988a, b, 1991) workers and managers,because of asymmetries explored the consequences of alternativeorga- of information, they typically did not even nizational design and decision-making struc- know what these people who were supposed to tures for organizationalmistakes: for instance, be acting on their behalf should do. That there whether good projects get rejected or bad were importantconsequences for the theory of projects get accepted. We suggested that, in a the firm of the separation of ownership and variety of circumstances, decentralized poly- control had earlier been noted by Adolph A. archical organizationalstructures have distinct Berle and GardinerC. Means (1932), but it was advantages(see also Sah, 1991; Stiglitz, 1989d). not until informationeconomics that we had a These papersare just beginningto spawn a body coherent way of thinking about the implica- of research; see, for example, Bauke Visser tions (Jensen and Willima H. Meckling, 1976; (1998), Amar Bhid6 (2001), and Michael Chris- Stiglitz, 1985a). tensen and ThorbjomKnudsen (2002). Some who still held to the view that firms would maximize their marketvalue argued that B. (the threat of) takeovers would ensure compe- tition in the market for managers and hence With these points made, we can returnto the promote stock market value maximization. If importantarea of macroeconomics.The central the firm were not maximizing its stock market macroeconomic issue is unemployment. The value, then it would pay someone to buy the models I describedearlier explained why unem- firm, and change its actions so that its value ployment could exist in equilibrium.But much would increase. Early on in this debate, I raised of macroeconomics is concerned with dynam- questions on theoreticalgrounds about the effi- ics, with explaining why sometimes the econ- cacy of the takeover mechanism (Stiglitz, omy seems to amplify rather than absorb 1972b). The most forceful set of arguments shocks, and why the effects of shocks may long were subsequently put forward by Grossman persist. In joint work with Bruce Greenwaldand and Hart (1980), who observed that any small Andy Weiss, I have shown how theories of shareholder who believed that the takeover asymmetricinformation can help provide expla- would subsequently increase market value nations of these phenomena. (For an early sur- would not be willing to sell his shares. The vey, see Greenwaldand Stiglitz [1987a, 1988b, subsequentwork by Shleifer and Vishny (1989) 1993b] and Stiglitz [1988b, 1992a].) The im- and Edlin and Stiglitz (1995), referredto earlier, perfections of capital markets-the phenomena showed how existing managers could take ac- of credit and equity rationing which arise be- tions to reduce the effectiveness of competition cause of information asymmetries-are key. for management,i.e., the threatof takeovers,by They lead to risk-averse behavior of firms and increasing asymmetriesof information. to households and firms being affected by cash So far, we have discussed two of the three flow constraints. pillars of the modem theory of the firm: corpo- Standard interpretationsof Keynesian eco- rate finance and corporate governance. The nomics emphasizedthe importanceof wage and third is organizational design. In a world with price rigidities, but without a convincing ex- perfect information,organizational design too is planation of how those rigidities arise. For of little moment. In practice, it is of central instance, some theories had shown the im- concern to businesses. For example, as we have portance of costs of adjustment of prices already discussed, an organizationaldesign that (Akerlof and Yellen, 1985; N. Gregory Man- has alternative units performing comparable kiw, 1985). Still at issue, though, is why firms tasks can enable a firm to glean informationon tend to adjustquantities rather than prices, even the basis of which better incentive systems can though the costs of adjusting quantities seem 482 THE AMERICANECONOMIC REVIEW JUNE 2002 greater than those of prices. The Greenwald- there was a resolution, the firm's access to Stiglitz theory of adjustment(1989) providedan credit would be impaired,and for good reason. explanation based on capital market imperfec- Moreover,without "clearowners" those in con- tions arising from informationimperfections. In trol would in general not have incentives to brief, it argued that the risks created by infor- maximize the firm's value. mationalimperfections are generally greaterfor Even when the shocks were not large enough price and wage adjustmentsthan from quantity to lead to bankruptcy, they had impacts on adjustments. Risk-averse firms would make firms' ability and willingness to take risks. smalleradjustments to those variablesfor which Since all productionis risky, shocks affect ag- the consequences of adjustment were more gregate supply, as well as the demand for in- uncertain. vestment.Because firmnet worth would only be But even though wages and prices were not restored over time, the effects of a shock per- perfectly flexible, neither were they perfectly sisted. By the same token, there were hysteresis rigid, and indeed in the , they effects associated with policy: An increase in fell by a considerable amount. There had been interestrates which depleted firm net worth had large fluctuationsin earlierperiods, and in other impacts even after the interest rates were re- countries,in which therehad been a high degree duced. Firms that were bankruptedwith high of wage and price flexibility. Greenwald and I interest rates remain so. If firms were credit (1987a, b, 1988b, c, d, 1989, 1990b, 1993a, b, rationed,then reductionsin liquidity could have 1995) argued that other marketfailures, in par- particularlymarked effects (Stiglitz and Weiss, ticular,the imperfectionsof capital marketsand 1992). Every aspect of macroeconomic behav- incompleteness in contracting,were needed to ior is affected: The theories helped explain, for explain key observed macroeconomicphenom- instance, the seemingly anomalous cyclical be- ena. In debt contracts, which are typically not havior of inventories (the procyclical move- indexed for changes in prices, whenever prices ments in inventories, counter to the idea of fell below the level expected (or in variable production smoothing, result from cash con- interest rate contracts, whenever real interest straints and the resulting high shadow price of rates rose above the level expected) there were money in ); or of pricing (in reces- transfersfrom debtors to creditors.In these cir- sions, when the "shadow price" of capital is cumstances, excessive downwardprice flexibil- high, firms do not find it profitableto invest in ity (not just price rigidities) could give rise to acquiringnew customers by cutting prices). In problems; (1933) and Stiglitz short, our analysis emphasized the supply-side (1999d) emphasize the consequences of differ- effects of shocks, the interrelationshipsbetween ences in the speed of adjustmentof different supply and demand side effects, and the impor- prices. These (and other) redistributivechanges tance of finance in propagatingfluctuations. had large real effects, and could not be insured Earlier,I describedhow the informationpar- againstbecause of imperfectionsin capital mar- adigm explained credit rationing.A second im- kets. Large shocks could lead to bankruptcy, portant strand in our macroeconomic research and with bankruptcy(especially when it results explored the link between credit rationing and in firm liquidation) there was a loss of organi- macroeconomic activity (Alan S. Blinder and zational and informational capital.24 Even if Stiglitz, 1983), explained the role of banks as such large changes could be forestalled, until risk-aversefirms, as informationinstitutions in- volved in screening and monitoring, in deter- mining the supply of credit (Greenwald and and 24 In traditional economic theories Stiglitz, 1990b, 1991, 2002; Stiglitz Weiss, bankruptcyplayed describedthe macroeconomic of little role, partly because control (who made decisions) did 1990), impacts not matter,and so the change in control that was consequent changes in financial regulations, and analyzed to bankruptcywas of little moment, partly because with the implications for under a perfect information,there would be little reason for lenders variety of regimes, including dollarization to lend to someone, rather than extending funds through These differed in re- equity (especially if there were significant probabilitiesof, (Stiglitz, 2001d). many and costs to, bankruptcy).For an insightful discussion about spects from the traditional theories, such as control rights see Hart (1995). those based on the transactions demand for VOL.92 NO. 3 STIGLITZ:INFORMATION ECONOMICS AND PARADIGMCHANGE 483 money, the microfoundations of which were failures are more prevalent in less developed increasingly being discredited as money be- countries, and these market failures are often came increasingly interest bearing (the interest associated with information problems-the rate was not the opportunity cost of holding very problems that inspired much of the re- money) and as credit, not money, was increas- search described in this paper (see Stiglitz, ingly being used for transactions.We also ex- 1985b, 1986a, 1988a, 1989e, h, 1991a, 1997a; plained the importanceof credit linkages (e.g., Bravermanet al., 1993). While these perspec- not only between banks and firms but among tives help explain the failures of policies based firms themselves) and their role in transmitting on assuming perfect or well-functioning mar- shocks throughoutthe economy. A large body kets, they also direct attentionto policies which of empirical work has subsequentlyverified the might remedy or reduce the consequences of importance of credit constraintsfor macroeco- informationalimperfections (, 1999). nomic activity, especially investment (see Kuh One of the most importantdeterminants of and Meyer (1957), Charles W. Calomiris and the pace of growth is the acquisition of knowl- Hubbard(1990), and Hubbard(1990). edge. For developed countries, this requiresin- vestment in research; for less developed C. Growth and Development25 countries, efforts at closing the knowledge gap between themselves and more developed coun- While most of the macroeconomic analysis tries. Knowledge is, of course, a particularform focused on exploring the implications of imper- of information,and many of the issues that are fections of credit markets for cyclical fluctua- centralto the economics of informationare also tions, another strand of our research program key to understanding research-such as the focused on growth. The importance of capital problems of appropriability,the fixed costs as- markets for growth had long been recognized; sociated with investments in research (which without capital markets firms have to rely on give rise to imperfections in competition), and retainedearnings. But how firms raise capital is the nature of information. It was importantfor their growth.In particular,"equity thus naturalthat I turned to explore the impli- rationing"-especially importantin developing cations in a series of papers that looked at both countries, where informational problems are equilibrium in the research industry and the even greater-impedes firms' willingness to in- consequences for .26While it vest and undertakerisks, and thus slows growth. is not possible to summarizebriefly the results, Changes in which enable firms one conclusion does stand out: Market econo- to bear more risk (e.g., by reducing the size of mies in which research and innovation play an macroeconomic fluctuations,or which enhance important role are not well described by the firms' equity base, by suppressinginterest rates, standardcompetitive model, and that the market which result in firm's having larger profits) en- equilibrium, without government intervention, hance economic growth. Conversely, policies, is not in general efficient. such as those associated with IMF interven- tions, in which interest rates are raised to very D. Theory of Taxation27 high levels, discourage the use of debt, forcing firms to rely more heavily on retainedearnings. One of the functions of government is to The most challenging problems for growth redistributeincome. Even if it did not actively lie in economic development.Typically, market 26 There were, of course, several precursorsto what has come to be called . See in par- 25 For discussions of growth, see Greenwaldet al. (1990) ticular, the collection of essays in Karl Shell (1967) and and Stiglitz (1990, 1992c, 1994a, b). The somewhat sepa- Anthony B. Atkinson and Stiglitz (1969). For later work, rate topic of development is analyzed in Stiglitz (1985b, see, in particular,Dasgupta and Stiglitz (1980a, b, 1981, 1986a, 1988a, 1989b,e,h, 1991a, 1993, 1995, 1996, 1988), Dasguptaet al. (1982), and Stiglitz (1987c, d, 1990). 1997a,b, 1998b, 1999b,c, 2000c, 2001a, b), Sah and 27 The discussion of this section draws upon Mirrlees Stiglitz (1989a, b), Karla Hoff and Stiglitz (1990, 1998, (1971, 1975a), Atkinson and Stiglitz (1976), Stiglitz (1982e, 2001), Nicholas Stern and Stiglitz (1997), and Stiglitz and 1987f), Arnott and Stiglitz (1986), and Brito et al. (1990, Shahid Yusuf (2000). 1991, 1995). 484 THE AMERICANECONOMIC REVIEW JUNE 2002 wish to redistribute,the governmenthas to raise occasion to provide a complete description of revenues to finance public goods, and there is a the results, two are worth noting: What had concern that the revenue be raised in an equita- been thought of as optimal commodity tax ble manner,e.g., thatthose who are more able to structures(Frank P. Ramsey, 1927) were shown contributedo so. But governmenthas a problem to be part of a Pareto-efficienttax system only of identifying these individuals,just as (for ex- under highly restricted conditions, e.g., that ample) a monopolist may find it difficult to there was no income tax (see Atkinson and identify those who are willing to pay more for Stiglitz, 1976; Sah and Stiglitz, 1992; Stiglitz, its product. Importantly, the self-selection 1998a). On the other hand, it was shown that in mechanisms for information revelation that a central benchmarkcase, it was not optimal to Rothschild and I had explored in our competi- tax interest income. tive insurance model or that I had explored in my paper on discriminatingmonopoly can be E. Theory of Regulation and Privatization applied here. (The problem of the government, maximizing social "profit,"i.e., welfare, subject The governmentfaces the problemsposed by to the informationconstraints, is closely analo- information asymmetries in regulation as well gous to that of the monopolist, maximizing pri- as in taxation. Over the past quartercentury, a vate profit subject to information constraints. huge literature has developed making use of For this reason, Mirrlees' (1971) paper on op- self-selection mechanisms (see, for example, timaltaxation, though not couchedin information- David E. M. Sappington and Stiglitz [1987a]; theoretic terms, was an importantprecursor to Jean-JacquesLaffont and Tirole [1993]), allow- the work described here.) ing far better and more effective systems of The critical question for the design of a tax regulation than had existed in the past. An ex- system thus becomes what is observable. In ample of a sector in which governmentregula- older theories, in which information was per- tion is of particularimportance is banking; we fect, lump-sum taxes and redistributionsmade noted earlier that information problems are at sense. If ability is not directly observable, the the heart of credit markets, and it is thus not governmenthad to rely on other observables- surprisingthat market failures be more perva- like income-to make inferences; but, as in all sive, and the role of the government more im- such models, market participants,as they rec- portant in those markets (Stiglitz, 1994d). ognize that inferences are being made, alter Regulatory design needs to take into account their behavior. In Mirrlees (1971) only income explicitly the limitations in information (see, was assumed observable. But in different cir- e.g., Hellman et al., 2000; PatrickHonahan and cumstances, either more or less information Stiglitz, 2001; Stiglitz, 2001c; Greenwald and might be available. It might be possible to ob- Stiglitz, 2002). serve hours worked, in which case wages would The 1980's saw a strong movement towards be observable. It might be possible to observe privatizing state enterprises, even in areas in the quantity of each good purchased by any which there was a naturalmonopoly, in which particularindividual or it might be possible to case governmentownership would be replaced observe only the aggregate quantity of goods with government regulation. While it was ap- produced. parentthat there were frequentlyproblems with For each information structure, there is a government ownership, the theories of imper- Pareto-efficienttax structure,that is, a tax struc- fect informationalso made it clear that even the ture such that no group can be made better off without making some other group worse off. The choice among such tax structuresdepends Pareto frontierthat would be chosen by a governmentwith on the social welfare function, including atti- a utilitarian social welfare function. Some of the critical tudes towardsinequality.28 While this is not the properties,e.g., the zero marginaltax rate at the top, were, however, characteristicsof any Pareto-efficient tax struc- ture,though thatparticular property was not robust-that is, it depended strongly on his assumptionthat relative wages 28 In that sense, Mirrlees' work confounded the two between individuals of different abilities were fixed (see stages of the analysis. He described the point along the Stiglitz, 2002a). VOL. 92 NO. 3 STIGLI7Z:INFORMATION ECONOMICS AND PARADIGMCHANGE 485 best designed regulatory systems would work ures and the theories put forward here. Our imperfectly. This naturally raised the question work had emphasized the importanceof main- of under what circumstanceswe could be sure taining the credit supply and the risks of (espe- that privatizationwould enhance economic wel- cially poorly managed) bankruptcy. Poorly fare. As Herbert A. Simon (1991), winner of designed policies could lead to an unnecessarily the 1978 Nobel Prize, had emphasized, both large reductionin credit availabilityand unnec- public and private sectors face informationand essary large increases in bankruptcy,both lead- incentive problems; there was no compelling ing to large adverseeffects on aggregatesupply, theoretical argumentfor why large private or- exacerbatingthe economic downturn.But this is ganizations would solve these incentive prob- precisely what the IMF did: by raising interest lems better than public organizations. In work rates to extremelyhigh levels in countrieswhere with Sappington (1987b), I showed that the firms were already highly leveraged, it forced conditions under which privatization would massive bankruptcy,and the economies were necessarily be welfare enhancing were ex- thus plunged into deep . Capital was tremely restrictive, closely akin to those under not attracted to the country, but rather fled. which competitive marketswould yield Pareto- Thus, the policies even failed in their stated efficient outcomes (see Stiglitz [1991b, 1994c] purpose, which was to stabilize the exchange for an elaborationand applications). rate. There were strong hysteresis effects asso- ciated with these policies: when the interest VI. Some Policy Debates rates were subsequentlylowered, firms that had been forced into bankruptcy did not become The perspectives provided by the new infor- "unbankrupt,"and the firms that had seen their mation paradigm not only shaped theoretical net worth depleted did not see an immediate approachesto policy, but in innumerablecon- restoration. There were alternative policies crete issues also led to markedlydifferent policy available, debt standstillsfollowed by corporate stances from those wedded to the old paradigm. financial restructurings, for example; while Perhaps most noted were the controversies these might not have avoided a downturn,they concerning development strategies, where the would have made it shorterand more shallow. Washingtonconsensus policies, based on mar- Malaysia, whose economic policies conformed ket fundamentalism-the simplistic view of much more closely to those that our theories competitive markets with perfect information, would have suggested, not only recoveredmore inappropriate even for developed countries, quickly, but was left with less of a legacy of but particularly inappropriatefor developing debt to impairits futuregrowth, than did neigh- countries-had prevailed since the early 1980's boring Thailand,which conformedmore closely within the internationaleconomic institutions. to the IMF's recommendation.(For discussions Elsewhere, I have documented the failures of of bankruptcyreform motivated by these expe- these policies in development (Stiglitz, 1999c), riences see Marcus Miller and Stiglitz, 1999; as well as in managingthe transitionfrom com- Stiglitz, 2000e.) munism to a marketeconomy (see, for instance, On anotherfront, the transitionfrom commu- Athar Hussein et al., 2000; Stiglitz [2000a, nism to a marketeconomy representsone of the 2001e]) and in crisis managementand preven- most important economic experiments of all tion (Stiglitz, 2000b). Ideas matter,and it is not time, and the failure (so far) in Russia, and the surprising that policies based on models that successes in China, shed considerable light on departas far from reality as those underlyingthe many of the issues which I have been discuss- so often led to failure. ing. The full dimension of Russia's failure is This point was brought home perhaps most hard to fathom. , with its central forcefully by the managementof the East Asia planning(requiring more informationgathering, crisis which began in Thailandon July 2, 1997. processing, and dissemination capacity than While I have written extensively on the many could be managedwith any technology), its lack dimensions of the failed responses (Jason Fur- of incentives, and its system rife with distor- man and Stiglitz, 1998; Stiglitz, 1999e), here I tions, was viewed as highly inefficient. The want to note the close link between these fail- movement to a market, it was assumed, would 486 THE AMERICANECONOMIC REVIEW JUNE 2002 bring enormous increases in incomes. Instead, ceived of development as nothing more than incomes plummeted, a decline confirmed not increasing the stock of capital and reducing only by GDP statistics and household surveys, economic distortions. But development repre- but also by social indicators. The numbers in sents a far more fundamentaltransformation of poverty soared,from 2 percentto upwardsof 40 society, including a change in "preferences"and percent, depending on the measureused. While attitudes,an acceptanceof change, and an aban- there were many dimensions to these failures, donment of many traditionalways of thinking one stands out: the privatizationstrategy, which (Stiglitz, 1995, 1999c). This perspective has paid little attention to the issues of corporate strong policy implications. For instance, some governance which we stressed earlier. Empiri- policies are more conducive to effecting a de- cal work (Stiglitz, 2001e) confirms that coun- velopment transformation.Many of the policies tries that privatized rapidly but lacked "good" of the IMF-including the mannerin which in corporate governance did not grow more rap- interacted with governments, basing loans on idly. Rather than providing a basis for wealth conditionality-were counterproductive.A fun- creation, privatizationled to asset strippingand damental change in development strategy oc- wealth destruction (Hussein et al., 2000; curred at the World Bank in the years I was Stiglitz, 2000a). there, one which embraced this more compre- hensive approachto development. By contrast, VII. BeyondInformation Economics policies which have ignored social conse- quences have frequently been disastrous. The We have seen how the competitive paradigm IMF policies in Indonesia, including the elimi- that dominatedeconomic thinking for two cen- nation of food and fuel subsidies for the very turies was not robust, did not explain key eco- poor as the country was plunging into depres- nomic phenomena,and led to misguided policy sion, predictably led to riots. The economic prescriptions. The research over the past 30 consequences are still being felt. years on informationeconomics that I have just In some ways, as I developed these perspec- described has focused, however, on only one tives, I was returningto a theme I had raised 30 aspect of my dissatisfactionwith that paradigm. years ago, during my work on the efficiency It is not easy to change views of the world, and wage theory in Kenya. In that work I had it seemed to me the most effective way of suggested psychological factors-morale, re- attackingthe paradigmwas to keep within the flecting a sense that one is receiving a fair standardframework as much as possible. I only wage-could affect efforts, an alternative,and varied one assumption-the assumption con- in some cases more persuasive reason for the cerning perfect information-and in ways efficiency wage theory. It is curious how econ- which seemed to me highly plausible. omists have almost studiously ignored factors, There were other deficiencies in the theory, which are not only the center of day-to-daylife, some of which were closely connected. The but even of business school education.Surely, if standard theory assumed that technology and marketswere efficient, such attentionwould not preferenceswere fixed. But changes in technol- be given to such matters,to issues of corporate ogy, R & D, are at the heart of capitalism. The culture and intrinsic rewards, unless they were new information economics-extended to in- of some considerable importance.And if such corporatechanges in knowledge-at last began issues are of importancewithin a firm, they are to addresssystematically these foundationsof a equally importantwithin a society. marketeconomy. Finally, I have become convinced that the As I thought about the problems of de- dynamics of change may not be well described velopment, I similarly became increasingly by equilibrium models that have long been at convinced of the inappropriateness of the the center of economic analysis. Information assumptionof fixed preferences,and of the im- economics has alertedus to the fact that history portance of embedding economic analysis in a matters; there are importanthysteresis effects. broadersocial and political context. I have crit- Random events-the Black Plague, to take an icized the Washington consensus development extreme example-have consequences that are strategies partly on the grounds that they per- irreversible.Dynamics may be better described VOL.92 NO. 3 STIGLITZ:INFORMATION ECONOMICS AND PARADIGMCHANGE 487 by evolutionaryprocesses and models, than by was political, a point which became clear when equilibrium processes. And while it may be it was noted that partialdisclosures could be of difficult to describe fully these evolutionary only limited value. Indeed, they could possibly processes, this much is already clear: there is be counterproductive,as capital would be in- no reason to believe that they are, in any duced to move throughchannels involving less general sense, "optimal." (I discussed these disclosure, channels like off-shore bankingcen- issues briefly in Stiglitz [1975b, 1992e, ters, which were also less well regulated.When 1994c] and Sah and Stiglitz [1991]; some of demands for transparencywent beyond East the problems are associated with capital mar- Asia to Westernhedge funds and - ket imperfections.) ing centers, suddenly the advocates of more Many of the same themes that emerged from transparencybecame less enthralled,and began our simpler work in informationeconomics ap- praising the advantages of partial secrecy in plied here. For instance, in the information- enhancingincentives to gatherinformation. The theoretic models discussed above we showed and the Treasurythen opposed the that multiple equilibria (some of which Pareto- OECD initiative to combat money laundering dominated others) could easily arise. So, too, through greater transparencyof offshore bank- here (Stiglitz, 1995). This in turn has several ing centers-these institutionsserved particular important consequences, beyond the observa- political and economic interests-until it be- tion already made that history matters. First, it came clear that terroristsmight be using them to means that one cannot simply predict where the help finance their operations.At that point, the economy will be by knowing preferences,tech- balance of American interests changed, and the nology, and initial endowments. There can a Treasurychanged its position. high level of indeterminacy(see, e.g., Stiglitz, Political processes inevitably entail asymme- 1973c) Second, as in Darwinian ecological tries of information (for a more extensive dis- models, the majordeterminant of one's environ- cussion, see Patrick D. Moynihan, 1998; ment is the behavior of others, and their behav- Stiglitz, 2002b): our political leaders are sup- ior may in turn depend on their beliefs about posed to know more about threats to defense, others' behavior (Hoff and Stiglitz, 2001). about our economic situation, etc., than ordi- Third, government interventioncan sometimes nary citizens. There has been a delegation of move the economy from one equilibrium to responsibility for day-to-day decision-making, another;and having done that, continued inter- just as there is within a firm. The problem is to vention might not be required. provide incentives for those so entrusted to act on behalf of those who they are supposed VIII. The PoliticalEconomy of Information to be serving-the standard principal-agent problem. Democracy-contestability in politi- Information affects political processes as cal processes-provides a check on abuses of well as economic ones. First, we have already the powers that come from delegationjust as it noted the distributiveconsequences of informa- does in economic processes; but just as we tion disclosures. Not surprisingly,then, the "in- recognize that the takeovermechanism provides formation rules of the game," both for the an imperfect check on management,so too we economy and for political processes, can be- should recognize that the electoral process pro- come a subject of intense political debate. The vides an imperfect check on politicians. As in United States and the IMF argued strongly that the theory of the firm where the current man- lack of transparencywas at the root of the 1997 agement has an incentive to increase asymme- , and said that the East Asian tries of informationin order to enhance market countries had to become more transparent.The power, so too in public life. And as disclosure attention to quantitative data on capital flows requirements-greater transparency-can af- and loans by the IMF and the U.S. Treasury fect the effectiveness of the takeover mecha- could be taken as conceding the inappropriate- nism and the overall quality of corporate ness of the competitive paradigm (in which governance, so too these factors can affect po- prices convey all the relevant information);but litical contestability and the quality of public the more appropriateway of viewing the debate governance. 488 THE AMERICANECONOMIC REVIEW JUNE 2002

In the context of political processes, where One of the lessons of the economics of infor- "exit"options are limited,one needs to be partic- mation is that these problems cannot be fully ularly concernedabout abuses. If a firm is mis- resolved, but that laws and institutions can de- managed-if the managers attempt to enrich cidedly improve matters. Right-to-know laws, themselves at the expense of shareholdersand for example, which require increased transpar- customersand entrenchthemselves against com- ency, have been part of governance in Sweden petition, the damage is limited-customers, at for 200 years; they have become an importantif least, can switch. But in political processes, imperfect check on government abuses in the switchingis not so easy. If all individualswere as United States over the past quartercentury. In selfish as economistshave traditionallymodeled the past five years, there has become a growing them, matterswould indeed be bleak, for-as I internationalacceptance of such laws; Thailand have put it elsewhere-ensuring the public good has gone so far as to include such laws in its is itself a public good. But there is a wealth of new constitution. Regrettably, these principles evidencethat the economists'traditional model of of transparencyhave yet to be endorsed by the the individualis too narrow-and that indeed in- internationaleconomic institutions. trinsicrewards, e.g., of publicservice, can be even more effective than extrinsicrewards, e.g., mon- IX. Concluding Remarks etarycompensation (which is not to say thatcom- pensationis not of some importance).This public In this article I have traced the replace- spiritedness(even if blendedwith a modicumof ment of one paradigm with another. The de- self-interest)is manifestedin a variety of civil ficiencies of the neoclassical paradigm-the society organizations,through which individuals failed predictions,the phenomenathat were left voluntarilywork together to advancetheir percep- unexplained-made it inevitable that it would tion of the collective interests. be challenged. One might ask, though, how can There are strong incentives on the part of we explain the persistence of this paradigmfor those in government to reduce transparency. so long? Despite its deficiencies, the competi- More transparency reduces their scope for tive paradigm did provide insights into many action-it not only exposes mistakes, but also economic phenomena. There are some mar- corruption(as the expression goes, "sunshineis kets in which the issues which we have dis- the strongest antiseptic").Government officials cussed are not important-the marketfor wheat may try to enhance their power by trying to or corn-though even there, pervasive govern- advance specious arguments for secrecy, and ment interventions make the reigning com- then saying, in effect, to justify their otherwise petitive paradigm of limited relevance. The inexplicable or self-serving behavior, "trust underlyingforces of demandand supply are still me ... if you only knew what I knew." important, though in the new paradigm, they There is a furtherrationale for secrecy, from become only part of the analysis; they are not the point of view of politicians: Secrecy is an the whole analysis. But one cannot ignore the artificially created scarcity of information,and possibility that the survivalof the paradigmwas like most artificially created , it gives partly because the belief in that paradigm,and rise to rents, rents which in some countries are the policy prescriptionsthat were derived from appropriatedthrough outright corruption(sell- it, has served certain interests. ing information). In other contexts these rents As a social scientist, I have tried to follow the become part of a "gift exchange," as when analysis, whereverit might lead. My colleagues reporterstrade "puff pieces" and distortedcov- and I know that our ideas can be used or erage in exchange for privileged access to in- abused-or ignored. Understanding the com- formation. I was in the unfortunateposition of plex forces that shape our economy is of value watching this process work, and work quite in its own right; there is an innate curiosity effectively. Without unbiased information, the about how this system works. But, as Shake- effectiveness of the check that can be provided speare said, "All the world's a stage, and all the by the citizenry is limited; without good infor- men and women merely players."Each of us in mation, the contestability of the political pro- our own way, if only as a voter, is an actor in cesses can be undermined. this granddrama. And what we do is affected by VOL.92 NO. 3 STIGLITZ:INFORMATION ECONOMICS AND PARADIGMCHANGE 489 our perceptions of how this complex system Akerlof, George A. and Yellen, Janet L."The Fair works. Wage-Effort Hypothesis and Unemploy- I entered economics with the hope that it ment."Quarterly Journal of Economics,May might enable me to do something about unem- 1990, 105(2), pp. 255-83. ployment, poverty, and discrimination. As an Allen, Franklin. "On the Fixed Nature of Share- economic researcher,I have been lucky enough cropping Contracts." Economic Journal, to hit upon some ideas that I think do enhance March 1985, 95(377), pp. 30-48. our understandingof these phenomena. As an Arnott, Richard J.; Greenwald, Bruce C. and educator, I have had the opportunityto reduce Stiglitz, Joseph E. "Information and Economic some of the asymmetries of information,espe- Efficiency."Information Economics and Pol- cially concerningwhat the new informationpar- icy, March 1994, 6(1), pp. 77-88. adigm and other developments in modem Arnott, Richard J.; Hosios, Arthur J. and Stiglitz, economic science have to say about these phe- Joseph E. "Implicit Contracts,Labor Mobil- nomena, and to have had some first-rate stu- ity, and Unemployment." American Eco- dents who, themselves, have pushed the nomic Review, December 1988, 78(5), pp. research agenda forward. 1046-66. As an individual, I have however not been Arnott, Richard J. and Stiglitz, Joseph E. "Labor content just to let others translate these ideas Turnover,Wage Structures,and Moral Haz- into practice. I have had the good fortune to ard: The Inefficiency of Competitive Mar- be able to do so myself, as a public servant kets." Journal of Labor Economics, October both in the American government and at the 1985, 3(4), pp. 434-62. World Bank. We have the good fortune to live . "MoralHazard and OptimalCommod- in democracies, in which individuals can fight ity Taxation."Journal of , for their perception of what a better world February 1986, 29(1), pp. 1-24. might be like. We as academics have the good . "Equilibriumin CompetitiveInsurance fortune to be further protected by our aca- Marketswith Moral Hazard."Princeton Uni- demic freedom. With freedom comes respon- versity Discussion Paper No. 4, October sibility: the responsibility to use that freedom 1987. to do what we can to ensure that the world of . "Randomizationwith Asymmetric In- the future be one in which there is not only formation." RAND Journal of Economics, greater economic prosperity, but also more Autumn 1988a, 19(3), pp. 344-62. social justice. . "The Basic Analytics of Moral Haz- ard." Scandinavian Journal of Economics, REFERENCES September 1988b, 90(3), pp. 383-413. . "The Welfare Economics of Moral Adams, William J. and Yellen, Janet L. "Com- Hazard,"in H. Louberge, ed., Risk informa- modity Bundling and the Burden of Monop- tion and insurance: Essays in the memoryof oly." Quarterly Journal of Economics, Karl H. Borch. Norwell, MA: Kluwer, 1990, August 1976, 90(3), pp. 475-98. pp. 91-122. Akerlof, George A. "The Market for 'Lemons': . "Moral Hazard and NonmarketInsti- Quality Uncertaintyand the Market Mecha- tutions: Dysfunctional Crowding Out or Peer nism." QuarterlyJournal of Economics, Au- Monitoring?" American Economic Review, gust 1970, 84(3), pp. 488-500. March 1991a, 81(1), pp. 179-90. Akerlof,George A. and Yellen,Janet L. "A Near- . "Price Equilibrium, Efficiency, and Rational Model of the Business Cycle with Decentralizabilityin InsuranceMarkets." Na- Wage and Price Inertia." Quarterly Journal tional Bureau of Economic Research (Cam- of Economics, 1985, Supp., 100(5), pp. 823- bridge, MA) Working Paper No. 3642, 38. March 1991b. Akerlof, George A. and Yellen, Janet L., eds. Ef- Arrow, Kenneth J. "Uncertainty and the Welfare ficiency wages model of the labor market. Economics of Medical Care."American Eco- New York: Cambridge University Press, nomic Review, December 1963, 53(5), pp. 1986. 941-73. 490 THEAMERICAN ECONOMIC REVIEW JUNE 2002

. "The Role of Securitiesin the Optimal Bowles, Samuel and Gintis, Herbert. Schooling in Allocation of Risk-bearing."Review of Eco- capitalist America. New York: Basic Books, nomic Studies, April 1964, 31(2), pp. 91-96. 1976. . Aspects of the theory of risk-bearing Braverman, Avishay; Hoff, Karla and Stiglitz, Jo- (Yrjo Jahnsson lectures). Helsinki, Finland: seph E., eds. The economics of rural organi- Yrjo JahnssoninSaatio, 1965. zation: Theory, practice, and policy. New . "Some MathematicalModels of Race York: , 1993. in the Labor Market,"in A. H. Pascal, ed., Braverman, Avishay and Stiglitz, Joseph E. Racial discrimination in economic life. Lan- "Sharecropping and the Interlinking of ham, MD: Lexington Books, 1972, pp. 187- Agrarian Markets." American Economic Re- 204. view, September 1982, 72(4), pp. 695-715. . "HigherEducation as a Filter." Jour- . "Cost-Sharing Arrangements under nal of Public Economics, July 1973, 3(2), pp. Sharecropping: Moral Hazard, Incentive 193-216. Flexibility, and Risk." Journal of Agricul- . "Limited Knowledge and Economic tural Economics, August 1986a, 68(3), pp. Analysis." American Economic Review, 642-52. March 1974, 64(1), pp. 1-10. . "Landlords,Tenants and Technologi- . "Risk Allocation and Information: cal Innovations." Journal of Development Some Theoretical Development." Geneva Economics,October 1986b, 23(2), pp. 313-32. Papers on Risk and Insurance, June 1978, 8. "Credit Rationing, Tenancy, Produc- Asquith, Paul and Mullins, David W., Jr. "Equity tivity, and the Dynamics of Inequality,"in P. Issues and Offering Dilution."Journal of Fi- Bardhan, ed., The economic theory of agrar- nancial Economics, January-February 1986, ian institutions. Oxford University Press, 15(1-2), pp. 61-89. 1989, pp. 185-202. Atkinson, Anthony B. and Stiglitz, Joseph E. "A Brito, Dagobert L.; Hamilton, Jonathan H.; New View of Technological Change." Eco- Slutsky, Steven M. and Stiglitz, Joseph E. nomic Journal, September1969, 79(315), pp. "Pareto Efficient Tax Structures." Oxford 573-78. Economic Papers, January 1990, 42(1), pp. . "The Design of Tax Structure:Direct 61-77. Versus IndirectTaxation." Journal of Public . "Dynamic Optimal Income Taxation Economics, July-August 1976, 6(1-2), pp. With GovernmentCommitment." Journal of 55-75. Public Economics, February 1991. 44(1), pp. Azariadis, Costas and Stiglitz, Joseph E. "Implicit 15-35. Contractsand Fixed Price Equilibria."Quar- . "Randomizationin Optimal Income terly Journal of Economics, 1983, Supp., Tax Schedules." Journal of Public Econom- 98(3), pp. 1-22. ics, February1995, 56(2), pp. 189-223. Bator,Francis M. "TheAnatomy of MarketFail- Calomiris, Charles W. and Hubbard, R. Glenn. ure." Quarterly Journal of Economics, Au- "Firm Heterogeneity, Internal Finance, and gust 1958, 72(3), pp. 351-79. Credit Rationing." Economic Journal, March Becker, Gary. The economics of discrimination, 1990, 100(399), pp. 90-104. 2nd Ed. Chicago: Capen, Edward; Clapp, Robert and Campbell, Press, 1971. William."Competitive Bidding in High Risk Berle, Adolph A. and Means, Gardiner C. The Situations." Journal of Petroleum Technol- modern corporation and private property. ogy, June 1971, 23(1), pp. 641-53. New York: HarcourtBrace and World, 1932. Chamberlin, Edward H. The theory of monopo- Bhide,Amar. "Taking Care: Ambiguity, Pooling listic competition. Cambridge, MA: Harvard and ErrorControl." Working paper, Colum- University Press, 1933. bia Business School, November 2001. Christensen, Michael and Knudsen, Thorbjorn. Blinder, Alan S. and Stiglitz, Joseph E. "Money, "The Architecture of Economic Organiza- Credit Constraints,and Economic Activity." tion: Towarda GeneralFramework." Univer- American Economic Review, May 1983 (Pa- sity of Southern Denmark Working Paper pers and Proceedings), 73(2), pp. 297-302. No. 02-7, January2002. VOL. 92 NO. 3 STIGLITZ:INFORMATION ECONOMICS AND PARADIGMCHANGE 491

Coase,Ronald H. "The Problemof ." Fisher, Irving. "The Debt Theory of Journal of , October Great Depressions." Econometrica, October 1960, 3, pp. 1-44. 1933, 1(4), pp. 337-57. Dasgupta, Partha; Gilbert, Richard J. and Fudenberg, Drew and Tirole, Jean. Game theory. Stiglitz, Joseph E. "Invention and Innovation Cambridge,MA: MIT Press, 1991. underAlternative Market Structures: The Case Furman,Jason and Stiglitz,Joseph E. "Economic of Natural Resources."Review of Economic Crises: Evidence and Insights from East Studies,October 1982, 49(4), pp. 567-82. Asia." BrookingsPapers on EconomicActiv- Dasgupta,Partha and Maskin, Eric. "The Exis- ity, 1998, (2), pp. 1-114. tence of Equilibriumin Discontinuous Eco- Gale, Douglas. "Optimal Risk Sharing through nomic Games, I: Theory." Review of Renegotiationof Simple Contracts."Journal Economic Studies, January 1986, 53(1), pp. of Financial Intermediation,December 1991, 1-26. 1(4), pp. 283-306. Dasgupta, Partha and Stiglitz, Joseph E. "Indus- Greenwald, Bruce C. Adverse selection in the trial Structureand the Nature of Innovative labor market.New York:Garland Press, 1979. Activity." Economic Journal, June 1980a, . "Adverse Selection in the Labor Mar- 90(358), pp. 266-93. ket." Review of Economic Studies, July 1986, . "Uncertainty,Market Structure and the 53(3), pp. 325-47. Speed of R&D." Bell Journal of Economics, Greenwald,Bruce C.; Kohn, Meir and Stiglitz, Spring 1980b, 11(1), pp. 1-28. Joseph E. "Financial Market Imperfections . "Entry, Innovation, Exit: Toward a and Productivity Growth." Journal of Eco- Dynamic Theory of Oligopolistic Industrial nomic Behavior and Organization, June Structure."European Economic Review, Feb- 1990, 13(3), pp. 321-45. ruary 1981, 15(2), pp. 137-58. Greenwald, Bruce C. and Stiglitz, Joseph E. . "Learningby Doing, MarketStructure, "Externalitiesin Economies with Imperfect and Industrial and Trade Policies." Oxford Informationand IncompleteMarkets." Quar- Economic Papers, 1988, 40(2), pp. 246-68. terly Journal of Economics, May 1986, Debreu, Gerard. The theory of value. New Ha- 101(2), pp. 229-64. ven, CT: Press, 1959. . "Keynesian,New Keynesian and New Diamond, Peter A. "A Model of Price Adjust- ." OxfordEconomic Pa- ment." Journal of Economic Theory, June pers, March 1987a, 39(1), pp. 119-33. 1971, 3(2), pp. 156-68. . "Imperfect Information, Credit Mar- Dixit,Avinash K. and Stiglitz,Joseph E. "Monop- kets and Unemployment." European Eco- olistic Competitionand Optimal ProductDi- nomic Review, 1987b, 31(1-2), pp. 444-56. versity." American Economic Review, June . "ParetoInefficiency of Market Econ- 1977, 67(3), pp. 297-308. omies: Search and Efficiency Wage Models." Eaton, Jonathan and Gersovitz, Mark. "Debt AmericanEconomic Review, May 1988a (Pa- with Potential Repudiation:Theoretical and pers and Proceedings), 78(2), pp. 351-55. Empirical Analysis." Review of Economic . "Examining Alternative Macroeco- Studies, April 1981, 48(2), pp. 289-309. nomic Theories."Brookings Papers on Eco- Edlin, Aaron S. and Stiglitz, Joseph E. "Discour- nomic Activity, 1988b, (1), pp. 207-60. aging Rivals: Managerial Rent-Seeking and . "ImperfectInformation, Finance Con- Economic Inefficiencies." American Eco- straints and Business Fluctuations," in M. nomic Review, December 1995, 85(5), pp. Kohn and S. C. Tsiang, eds., Finance con- 1301-12. straints, expectations, and macroeconomics. Fama, Eugene F. "Efficient Capital Markets: a Oxford: Oxford University Press, 1988c, pp. Review and EmpiricalWork." Journal of Fi- 103-40. nance, May 1970, 25(2), pp. 383-417. . "Money, Imperfect Information and Fields, Gary S. "Private and Social Returns to Economic Fluctuations," in M. Kohn and Education to Labor Surplus Economies." S. C. Tsiang, eds., Finance constraints, ex- Eastern Africa Economic Review, June 1972, pectations and macroeconomics. Oxford: 4(1), pp. 41-62. Oxford University Press, 1988d, pp. 141-65. 492 THE AMERICANECONOMIC REVIEW JUNE 2002

. "Toward a Theory of Rigidities." "On Value Maximizationand Alterna- American Economic Review, May 1989 tive Objectives of the Firm." Journal of Fi- (Papers and Proceedings), 79(2), pp. 364- nance, May 1977, 32(2), pp. 389-402. 69. . "On the Impossibility of Information- . "Asymmetric Information and the ally Efficient Markets."American Economic New Theory of the Firm: Financial Con- Review, June 1980a, 70(3), pp. 393-408. straints and Risk Behavior."American Eco- . "StockholderUnanimity in the Making nomic Review, May 1990a (Papers and of Production and Financial Decisions." Proceedings), 80(2), pp. 160-65. QuarterlyJournal of Economics, May 1980b, . "MacroeconomicModels with Equity 94(3), pp. 543-66. and CreditRationing," in R. Glenn Hubbard, Hall, Brian J. and Liebman, Jeffrey B. "Are ed., Asymmetric information, corporate fi- CEO's Really Paid Like Bureaucrats?"Quar- nance, and investment. Chicago: University terly Journal of Economics, August 1998, of Chicago Press, 1990b, pp. 15-42. 113(3), pp. 653-91. . "Towarda Reformulationof Monetary Hall, Robert E. and Jorgenson, Dale W. "Tax Theory: Competitive Banking." Economic Policy and Investment Behavior."American and Social Review, October 1991, 23(1), pp. Economic Review, June 1967, 57(3), pp. 1-34. 391-414. ."Information, Finance and Markets: Harris, John R. and Todaro, Michael P. "Migra- The Architectureof Allocative Mechanisms." tion, Unemployment and Development: A Industrial and Corporate Change, 1992, Two-Sector Analysis." American Economic 1(1), pp. 37-68. Review, March 1970, 60(1), pp. 126-42. . "Financial Market Imperfections and Hart, Oliver D. Firms, contracts, and financial Business Cycles." QuarterlyJournal of Eco- structure. Oxford: Oxford University Press, nomics, February 1993a, 108(1), pp. 77- 1995. 114. Hart, Oliver D. and Holmstrom, Bengt. "The "New and Old Keynesians." Journal Theory of Contracts,"in T. Bewley, ed., Ad- of Economic Perspectives, Winter 1993b, vances of economic theory: Fifth WorldCon- 7(1), pp. 23-44. gress. Cambridge: Cambridge University . "Labor Market Adjustments and the Press, 1987, pp. 71-155. Persistence of Unemployment." American Haubrich, Joseph G. ", Perfor- Economic Review, May 1995 (Papers and mance Pay, and the Principal-Agent Prob- Proceedings), 85(2), pp. 219-25. lem." Journal of , April . Towardsa new paradigmfor monetary 1994, 102(2), pp. 258-76. economics. London: Cambridge University Hellman, Thomas F. and Stiglitz, Joseph E. Press, 2002 (forthcoming). "Credit and Equity Rationing in Markets with Greenwald, Bruce C.; Stiglitz, Joseph E. and Adverse Selection." EuropeanEconomic Re- Weiss, Andrew W. "InformationalImperfec- view, February2000, 44(2), pp. 281-304. tions in the Capital Markets and Macroeco- Hellman, Thomas F.; Murdock, Kevin C. and nomic Fluctuations." American Economic Stiglitz, Joseph E. "Liberalization, Moral Haz- Review, May 1984 (Papers and Proceed- ard in Banking and Prudential Regulation: ings), 74(2), pp. 194-99. Are Capital RequirementsEnough?" Ameri- Grossman,Sanford J. and Hart,Oliver D. "Take- can Economic Review, March 2000, 90(1), over Bids, the Free-Rider Problem, and pp. 147-65. the Theory of the Corporation."Bell Journal Hirshleifer, Jack. "The Private and Social Value of Economics, Spring 1980, 11(1), pp. 42- of Informationand the Reward to Inventive 64. Activity." American Economic Review, Sep- Grossman, Sanford J. and Stiglitz, Joseph E. tember 1971, 61(4), pp. 561-74. "Information and Competitive Price Sys- Hirshleifer, Jack and Riley, John G. The analyt- tems." American Economic Review, May ics of uncertainty and information. Cam- 1976 (Papers and Proceedings), 66(2), pp. bridge: CambridgeUniversity Press, 1992. 246-53. Hoff, Karla and Stiglitz, Joseph E. "Imper- VOL. 92 NO. 3 STIGLI7Z:INFORMATION ECONOMICS AND PARADIGMCHANGE 493

fect Informationand Rural Credit Markets- Journal of , October Puzzles and Policy Perspectives." World 1976, 3(4), pp. 305-60. Bank Economic Review, September 1990, Jensen, Michael C. and Murphy, Kevin J. "Per- 4(3), pp. 235-50. formance Pay and Top-ManagementIncen- . "Moneylenders and Bankers: Price- tives." Journal of Political Economy, April Increasing Subsidies in a Monopolistically 1990, 98(2), pp. 225-64. Competitive Market." Journal of Develop- Keeton,William R. Equilibriumcredit rationing. ment Economics, April 1998, 55(2), pp. 485- New York: GarlandPress, 1979. 518. Keynes, John Maynard. The general theory of ."Modern Economic Theory and De- employment,interest and money. New York: velopment," in G. Meier and J. E. Stiglitz, HarcourtBrace, 1936. eds., Frontiers of : Klein, Benjaminand Leffler,Keith B. "The Role The future in perspective. New York: Ox- of Market Forces in Assuring Contractual ford University Press, March 2001, pp. Performance."Journal of Political Economy, 389-485. August 1981, 89(4), pp. 615-41. Honahan,Patrick and Stiglitz,Joseph E. "Robust Kuh, Edwin and Meyer, John R. The investment Financial Restraint,"in G. Caprio, P. Hono- decision: An empirical study. Cambridge, han, and J. E. Stiglitz, eds., Financial liber- MA: HarvardUniversity Press, 1957. alization: How far, how fast? New York: Laffont,Jean-Jacques and Tirole,Jean. A theory Cambridge University Press, October 2001, of incentives in procurementand regulation. pp. 31-62. Cambridge,MA: MIT Press, 1993. Hubbard,R. Glenn,ed. Asymmetricinformation, Leibenstein,Harvey. "The Theory of Underem- corporatefinance, and investment.Chicago: ployment in BackwardEconomies." Journal University of Chicago Press, 1990. of Political Economy, April 1957, 65(2), pp. . "Capital-MarketImperfections and In- 91-103. vestment." Journal of Economic Literature, Leitzinger,Jeffrey J. and Stiglitz,Joseph E. "In- March 1998, 36(1), pp. 193-225. formationExternalities in Oil and Gas Leas- Hurwicz,Leonid. "Optimality and Informational ing." Contemporary Policy Issues, March Efficiency in Resource Allocation Process- 1984, (5), pp. 44-57. es," in K. J. Arrow; S. Karlin,and P. Suppes, Leland, Hayne E. and Pyle, David H. "Informa- eds., Mathematicalmethods in the social sci- tional Asymmetries, Financial Structure,and ences. Stanford, CA: Financial Intermediation."Journal of Fi- Press, 1960, pp. 27-46. nance, May 1977, 32(2), pp. 371-87. . "On Informationally Decentralized Lucas, Robert E., Jr. Models of business cycles. Systems," in C. B. McGuire and R. Radner, New York: Blackwell, 1987. eds., Decision and organization.Amsterdam: Mankiw, N. Gregory. "Small Menu Costs and North-Holland, 1972, pp. 297-336. Large Business Cycles: A Macroeconomic Hussein,Athar; Stern, Nicholas and Stiglitz, Jo- Model of Monopoly." Quarterly Journal of seph E. "Chinese Reforms from a Compara- Economics, May 1985, 10(2), pp. 529-37. tive Perspective," in P. J. Hammond and Marschak, Jacob and Radner, Roy. Economic G. D. Myles, eds., Incentives, organization, theory of teams. New Haven, CT: Yale Uni- and public economics: Papers in honour of versity Press, 1972. Sir . Oxford: Oxford Univer- Marshall,Alfred. Principles of economics. Lon- sity Press, 2000, pp. 243-77. don: Macmillan, 1890. Jaffee,Dwight and Stiglitz,Joseph E. "CreditRa- Mayer, Colin. "Financial Systems, Corporate Fi- tioning," in B. Friedman and F. Hahn, eds., nance, and Economic Development,"in R. G. Handbook of , Vol. 2. Hubbard,ed., Asymmetric information,cor- Amsterdam:North-Holland, 1990, pp. 837- porate finance and investment.Chicago: Uni- 88. versity of Chicago Press, 1990, pp. 307-32. Jensen, Michael C. and Meckling, William H. Mill, John S. Principles of political economy "Theory of the Firm: Managerial Behavior, with some of their applications to social phi- Agency Costs and Ownership Structure." losophy. London: John W. Parker, 1848. 494 THEAMERICAN ECONOMIC REVIEW JUNE 2002

Miller, Marcus and Stiglitz, Joseph E. "Bank- "Prizesand Incentives:Toward a Gen- ruptcy Protection Against Macroeconomic eral Theory of Compensation and Competi- Shocks: The Case for a 'Super Chapter11'." tion." Bell Journal of Economics, Spring Unpublished manuscript presented at the 1983b, 14(1), pp. 21-43. World Bank Conference on Capital Flows, Newbery, David M. G. and Stiglitz, Joseph E. Financial Crises, and Policies, April 1999. "The Choice of Techniquesand the Optimal- Mincer,Jacob. Schooling, experience and earn- ity of MarketEquilibrium with Rational Ex- ings. New York: Press, pectations." Journal of Political Economy, 1974. April 1982, 90(2), pp. 223-46. Mirrlees, James A. "An Exploration in the The- . "Pareto Inferior Trade." Review of ory of Optimum Income Taxation."Review Economic Studies, January 1984, 51(1), pp. of Economic Studies, April 1971, 38(2), pp. 1-12. 175-208. Phelps, Edmund S. "Money-Wage Dynamics . "Optimal Commodity Taxation in a and Labor-MarketEquilibrium." Journal of Two-Class Economy." Journal of Public Political Economy, July-August 1968, Pt. 2, Economics, February1975a, 4(1), pp. 27-33. 76(4), pp. 678-711. . "The Theory of Moral Hazard and "The StatisticalTheory of Racism and UnobservableBehaviour I." Mimeo, Nuffield Sexism." American Economic Review, Sep- College, 1975b. tember 1972, 62(4), pp. 659-61. . "The Optimal Structureof Incentives Phelps, Edmund S. and Winter, Sidney G. and Authority within an Organization."Bell "OptimalPrice Policy underAtomistic Com- Journal of Economics, Spring 1976, 7(1), pp. petition,"in E. Phelps, et al., eds., Microeco- 105-31. nomic foundations of employment and Modigliani, Franco and Miller, Merton H. "The theory. New York: Norton, 1970, Cost of Capital,Corporation Finance, and the pp. 309-37. Theory of Investment."American Economic Radner, Roy. "Existence of Equilibrium of Review, June 1958, 48(3), pp. 261-97. Plans, Prices, and Price Expectations in a . "Dividend Policy, Growth, and the Sequence of Markets."Econometrica, March Valuation of Shares." Journal of Business, 1972, 40(2), pp. 289-303. October 1961, 34(4), pp. 411-33. Radner, Roy and Stiglitz, Joseph E. "A Noncon- Moynihan, Patrick D. Secrecy: The American cavity in the Value of Information,"in M. experience.New Haven, CT: Yale University Boyer and R. Khilstrom,eds., Bayesian mod- Press, 1998. els in economic theory. New York: Elsevier, Murphy, Kevin J. "CorporatePerformance and 1984, pp. 33-52. Managerial Remuneration: An Empirical Ramsey, Frank P. "A Contribution to the Theory Analysis." Journal of Accounting and Eco- of Taxation." Economic Journal, March nomics, 1985, 7(1-3), pp. 11-42. 1927, 37(145), pp. 47-61. Myers, StewartC. and Majluf,Nicholas S. "Cor- Rey, Patrick and Stiglitz, Joseph E. "Moral porate Financing and Investment Decisions Hazard and Unemployment in Competitive When Firms Have InformationThat Inves- Equilibrium."Unpublished manuscript,Uni- tors Do Not Have." Journal of Financial versity of Toulouse, July 1996. Economics, June 1984, 13(2), pp. 187-221. Riley, John G. "Informational Equilibrium." Nalebuff, Barry J.; Rodriguez, Andres and Econometrica, March 1979, 47(2), pp. 331- Stiglitz, Joseph E. "EquilibriumUnemploy- 60. ment as a Worker Screening Device." Na- "Silver Signals: Twenty-Five Years of tional Bureau of Economic Research Screening and Signaling." Journal of Eco- (Cambridge,MA) Working Paper No. 4357, nomic Literature,June 2001, 39(2), pp. 432- May 1993. 78. Nalebuff, Barry J. and Stiglitz, Joseph E. "Infor- Rodriguez, Andres and Stiglitz, Joseph E. "Equi- mation, Competition and Markets."Ameri- librium Unemployment, Testing, and the can Economic Review, May 1983a (Papers Pure Theory of Selection." Unpublished and Proceedings), 73(2), pp. 278-83. manuscript presented at the NBER/CEPR VOL.92 NO. 3 STIGLITZ:INFORMATION ECONOMICS AND PARADIGMCHANGE 495

Conference on Unemployment and Wage on Reliability, December 1988b, 37(5), pp. Determination,Boston, October 1991a. 515-20. . "Unemploymentand EfficiencyWages: . "Sourcesof Technological Divergence The Adverse Selection Model." Unpublished between Developed and Less Developed manuscriptpresented at NBER/CEPR Con- Economies," in G. Calvo, R. Findlay, P. ference on Unemployment and Wage Deter- Kouri, and J. Braga de Macedo, eds., Debt, mination, Boston, October 199 b. stabilizations and development: Essays in Rosen,Sherwin. "Implicit Contracts: A Survey." memory of Carlos Diaz-Alejandro. Cam- Journal of Economic Literature, September bridge, MA: Blackwell (for WIDER of the 1985, 23(3), pp. 1144-75. UnitedNations University), 1989a, pp. 423-46. Ross, Stephen A. "The Economic Theory of . "Technological Learning, Social Agency: The PrincipalsProblem." American Learning and Technological Change," in S. Economic Review, May 1973 (Papers and Chakravarty,ed., The balance between indus- Proceedings), 63(2), pp. 134-39. try and agriculture in economic develop- . "The Determination of Financial ment: Proceedings of the Eighth World Structure: The Incentive-Signalling Ap- Congress of the InternationalEconomic As- proach." Bell Journal of Economics, Spring sociation, Delhi, India, volume 3, Manpower 1977, 8(1), pp. 23-40. and transfers. New York: St. Martin's, Rothschild, Michael and Stiglitz, Joseph E. 1989b, pp. 285-98. "Equilibrium in Competitive Insurance . "The Quality of Managersin Central- Markets: An Essay on the Economics of ized Versus Decentralized Organizations." Imperfect Information." Quarterly Journal Quarterly Journal of Economics, February of Economics, November 1976, 90(4), pp. 1991, 106(1), pp. 289-95. 629-49. . Peasants versus city-dwellers: Taxa- . "A Model of Employment Outcomes tion and the burden of economic develop- Illustratingthe Effect of the Structureof In- ment. Oxford: ClarendonPress, 1992. formation on the Level and Distribution of Salop,Steven. "The Noisy Monopolist:Informa- Income." Economic Letters, 1982, 10(3-4), tion, Price Dispersion and Price Discrimina- pp. 231-36. tion." Review of Economic Studies, October . "Competition and Insurance Twenty 1977, 44(3), pp. 393-406. Years Later." Geneva Papers on Risk and . "MonopolisticCompetition with Out- Insurance Theory, December 1997, 22(2), side Goods." Bell Journal of Economics, pp. 73-79. Spring 1979, 10(1), pp. 141-56. Sah, Raaj K. "Fallibility in Human Organiza- Salop, Steven and Stiglitz, Joseph E. "Bargains tions and Political Systems." Journal of Eco- and Ripoffs: A Model of Monopolistically nomic Perspectives, Spring 1991, 5(2), pp. Competitive Price Dispersions." Review of 67-88. Economic Studies, October 1977, 44(3), pp. Sah, Raaj K. and Stiglitz, Joseph E. "Human 493-510; reprintedin S. A. Lippmanand D. K. Fallibility and Economic Organization." Levine, eds., The economics of information. American Economic Review, May 1985 Aldershot,U.K.: Elgar, 1995, pp. 198-215. (Papers and Proceedings), 75(2), pp. 292- . "The Theory of Sales: A Simple 96. Model of EquilibriumPrice Dispersion with "The Architectureof Economic Sys- Identical Agents." American Economic Re- tems: Hierarchies and Polyarchies." Ameri- view, December 1982, 72(5), pp. 1121-30. can Economic Review, September 1986, . "Information,Welfare and ProductDi- 76(4), pp. 716-27. versity," in G. Feiwel, ed., Arrow and the . "Committees, Hierarchies and Poly- foundations of the theory of economicpolicy. archies." Economic Journal, June 1988a, New York: New York University Press, 98(391), pp. 451-70. 1987, pp. 328-40. . "Qualitative Properties of Profit- Samuelson, Paul A. Foundations of economic Maximizing K-out-of-N Systems Subject to analysis. Cambridge, MA: HarvardUniver- Two Kinds of Failure." IEEE Transactions sity Press, 1947. 496 THE AMERICANECONOMIC REVIEW JUNE 2002

Sappington, David E. M. and Stiglitz, Joseph E. tion." Journal of Political Economy, June "Information and Regulation," in E. Bailey, 1961, 69(3), pp. 213-25. ed., Public regulation. London: MIT Press, "Imperfectionsin the CapitalMarket." 1987a, pp. 3-43. Journal of Political Economy, May/June "Privatization,Information and Incen- 1967, 75(3), pp. 287-92. tives." Journal of Policy Analysis and Man- Stiglitz, Joseph E. "A Re-Examination of the agement, 1987b, 6(4), pp. 567-82. Modigliani-MillerTheorem." American Eco- Schultz, Theodore W. "Capital Formation by Ed- nomic Review, December 1969a, 59(5), pp. ucation."Journal of Political Economy, De- 784-93. cember 1960, 68(6), pp. 571-83. . "Rural-UrbanMigration, Surplus La- Shapiro, Carl. "Premiums for High Quality bour, and the Relationship between Urban Products as Returns to Reputations."Quar- and Rural Wages." East African Economic terly Journal of Economics, November 1983, Review, December 1969b, 1(2), pp. 1-27. 98(4), pp. 659-80. . "On the Optimalityof the Stock Mar- Shapiro, Carl and Stiglitz, Joseph E. "Equilib- ket Allocation of Investment." Quarterly rium Unemployment as a Worker Discipline Journal of Economics, February 1972a, Device." American Economic Review, June 86(1), pp. 25-60. 1984, 74(3), pp. 433-44. . "Some Aspects of the Pure Theory of Shell, Karl, ed. Essays on the theory of optimal CorporateFinance: Bankruptcies and Take- economic growth. Cambridge, MA: MIT Overs." Bell Journal of Economics, Autumn Press, 1967. 1972b, 3(2), pp. 458-82. Shiller,Robert J. Irrational exuberance.Prince- "Approachesto the Economics of Dis- ton, NJ: PrincetonUniversity Press, 2000. crimination." American Economic Review, Shleifer, Andrei and Vishny, Robert W. May 1973a (Papers and Proceedings), 62(2), "Management Entrenchment:The Case of pp. 287-95; reprinted in W. Darity and C. Manager-SpecificAssets." Journal of Finan- Boshamer, eds., Economics and discrimina- cial Economics, November 1989, 25(1), pp. tion. Aldershot, U.K.: Elgar, 1993, pp. 325- 123-39. 33. Simon, Herbert A. "Organizations and Markets." "Taxation,Corporate Financial Policy Journal of Economic Perspectives, Spring and the Cost of Capital."Journal of Public 1991, 5(2), pp. 25-44. Economics, February1973b, 2(1), pp. 1-34. Simonde de Sismondi, Jean-Charles-Leonard. Po- . "The Badly Behaved Economy with litical economy. New York: Kelley, 1966 the Well Behaved ," in (first published in 1815). J. A. Mirrlees and N. H. Ster, eds., Models Smith, Adam. An inquiry into the nature and of economic growth. London: MacMillan, causes of the wealth of nations. Chicago: 1973c, pp. 118-37. University of Chicago Press, 1977 (first pub- . "AlternativeTheories of Wage Deter- lished in 1776). mination and Unemployment in L.D.C.'s: Spence, A. Michael. "Job Market Signaling." The LaborTurnover Model." QuarterlyJour- Quarterly Journal of Economics, August nal of Economics, May 1974a, 88(2), pp. 1973, 87(3), pp. 355-74. 194-227. . Marketsignaling: Informationtransfer "Incentivesand Risk Sharingin Share- in hiring and related processes. Cambridge, cropping." Review of Economic Studies, MA: HarvardUniversity Press, 1974. April 1974b, 41(2), pp. 219-55. Stern, Nicholas and Stiglitz, Joseph E. "A . "On the Irrelevance of CorporateFi- Framework for a Development Strategy in nancial Policy." AmericanEconomic Review, a Market Economy," in E. Malinvaud and December 1974c, 64(6), pp. 851-66. A. K. Sen, eds., Development strategy and . "Theoriesof Discriminationand Eco- the management of the market economy. nomic Policy," in G. von Furstenberg, B. Oxford: Clarendon Press, 1997, pp. 253- Harrison,and A. R. Horowitz, eds., Patterns 95. of racial discrimination.Lanham, MD: Lex- Stigler, George J. "The Economics of Informa- ington Books, 1974d, pp. 5-26. VOL. 92 NO. 3 STIGLITZ:INFORMATION ECONOMICS AND PARADIGMCHANGE 497

. "Incentives, Risk and Information: . "Ownership, Control and Efficient Notes Toward a Theory of Hierarchy."Bell Markets: Some Paradoxes in the Theory of Journal of Economics, Autumn 1975a, 6(2), Capital Markets,"in K. D. Boyer and W. G. pp. 552-79. Shepherd,eds., Economic regulation: essays . "Informationand Economic Analy- in honor of James R. Nelson. East Lansing, sis," in J. M. Parkin and A. R. Nobay, eds., MI: Michigan State University Press, 1982d, Current economic problems. Cambridge: pp. 311-41. CambridgeUniversity Press, 1975b,pp. 27-52. . "Self-Selection and Pareto Efficient . "The Theory of Screening, Education Taxation." Journal of Public Economics, and the Distribution of Income." American March 1982e, 17(2), pp. 213-40. Economic Review, June 1975c, 65(3), pp. . "The Structureof Labor Markets and 283-300. Shadow Prices in L.D.C.'s," in R. Sabot, ed., . "The Efficiency of Market Prices in Migration and the labor market in develop- Long Run Allocations in the Oil Industry,"in ing countries. Boulder, CO: Westview, G. M. Brannon, ed., Studies in energy tax 1982f, pp. 13-64. policy. Cambridge, MA: Ballinger, 1975d, . "Utilitarianismand HorizontalEquity: pp. 55-99. The Case for Random Taxation."Journal of . "The Efficiency Wage Hypothesis, Public Economics, June 1982g, 18(1), pp. Surplus Labour, and the Distribution of In- 1-33. come in L.D.C.s." OxfordEconomic Papers, "Information,Screening and Welfare," July 1976, 28(2), pp. 185-207. in M. Boyer and R. Khilstrom,eds., Bayesian . "Monopoly, Non-linear Pricing and models in economic theory. New York: Imperfect Information:The Insurance Mar- Elsevier, 1984a, pp. 209-39. ket." Review of Economic Studies, October . "Price Rigidities and Market Struc- 1977, 44(3), pp. 407-30. ture." American Economic Review, May . "Equilibriumin ProductMarkets with 1984b (Papers and Proceedings), 74(2), pp. ImperfectInformation." American Economic 350-56. Review, May 1979a (Papers and Proceed- . "Credit Markets and the Control of ings), 69(2), pp. 339-45. Capital." Journal of Money, Credit, and "OnSearch and EquilibriumPrice Dis- Banking, May 1985a, 17(2), pp. 133-52. tributions,"in M. Boskin, ed., Economics and . "Economics of Information and the human welfare: Essays in honor of Tibor Theory of Economic Development."Revista Scitovsky. New York: Academic Press, de Econometria,April 1985b, 5(1), pp. 5-32. 1979b, pp. 203-36. . "Equilibrium Wage Distribution." "ParetoOptimality and Competition." Economic Journal, September 1985c, Journalof Finance,May 1981,36(2),pp. 235-51. 95(379), pp. 595-618. . "AlternativeTheories of Wage Deter- . "Informationand Economic Analysis: minationand Unemployment:The Efficiency A Perspective." Economic Journal, 1985d, Wage Model," in M. Gersovitz, C. F. Diaz- Supp., 95, pp. 21-41. Alejandro,G. Ranis, and M. R. Rosenzweig, . "The New Development Economics." eds., The theory and experience of economic World Development, 1986a, 14(2), pp. 257- development:Essays in honor of Sir Arthur 65. W. Lewis. London: Allen and Unwin, 1982a, "Theoriesof Wage Rigidities,"in J. L. pp. 78-106. Butkiewicz, K. J. Koford, and J. B. Miller, "The Inefficiency of the Stock Market eds., Keynes' economic legacy: Contempo- Equilibrium."Review of Economic Studies, rary economic theories. New York: Praeger, April 1982b, 49(2), pp. 241-61. 1986b, pp. 153-206. . "Informationand Capital Markets,"in . "The Causes and Consequences of the W. F. Sharpeand C. Cootner,eds., Financial Dependence of Quality on Prices."Journal of economics: Essays in honor of Paul Cootner. Economic Literature, March 1987a, 25(1), Upper Saddle River, NJ: Prentice Hall, pp. 1-48. 1982c, pp. 118-58. . "Competition and the Number of 498 THE AMERICANECONOMIC REVIEW JUNE 2002

Firms in a Market: Are Duopolies More ment." Oxford Review of Economic Policy, CompetitiveThan Atomistic Markets?"Jour- Winter 1989b, 5(4), pp. 55-68. nal of Political Economy, 1987b, 95(5), pp. . "ImperfectInformation in the Product 1041-61. Market,"in R. Schmalenseeand R. D. Willig, . "Design of Labor Contracts:Econom- eds., Handbook of , ics of Incentives and Risk-Sharing,"in H. Vol. 1. New York: Elsevier, 1989c, pp. 769- Nalbantian, ed., Incentives, cooperation and 847. risk sharing. Totowa, NJ: Rowman and "Incentives, Informationand Organi- Littlefield, 1987c, pp. 47-68. zational Design." Empirica, January 1989d, . "Learningto Learn, Localized Learn- 16(1), pp. 3-29. ing and Technological Progress,"in P. Das- . "Markets,Market Failures and Devel- gupta and P. Stoneman, eds., Economic opment." American Economic Review, May policy and technological performance. Cam- 1989e (Papers and Proceedings), 79(2), pp. bridge: CambridgeUniversity Press, 1987d, 197-203. pp. 125-53. "Monopolistic Competition and the . "On the Microeconomicsof Technical Capital Market," in G. Feiwel, ed., The Progress," in J. M. Katz, ed., Technology economics of imperfect competition and generation in Latin American manufacturing employment- and beyond. industries. New York: St. Martin's Press, New York: New York University Press, 1987e, pp. 56-77. 1989f, pp. 485-507. "Efficient and Optimal Taxation and . "MutualFunds, Capital Structure,and the New New Welfare Economics," in A. ," in S. Bhattacharya Auerbach and M. Feldstein, eds., Handbook and G. Constantinides, eds., Theory of of public economics. New York: Elsevier, valuation-frontiers of modern financial the- 1987f, pp. 991-1042. ory, Vol. 1. Totowa, NJ: Rowman and Little- . "The Wage-ProductivityHypothesis: field, 1989g, pp. 342-56. Its Economic Consequences and Policy Im- . "Rational Peasants, Efficient Institu- plications," in M. J. Boskin, ed., Modern tions and the Theory of Rural Organization," developments in public finance. Oxford: in P. Bardhan, ed., The economic theory of Blackwell, 1987g, pp. 130-65. agrarian institutions. Oxford: Clarendon . "EconomicOrganization, Information, Press, 1989h, pp. 18-29. and Development,"in H. Chenery and T. N. . "Using Tax Policy to CurbSpeculative Srinivasan, eds., Handbook of development Short-Term Trading." Journal of Financial economics. New York: Elsevier, 1988a, pp. Services Research, December 1989i, 3(2-3), 185-201. pp. 101-15. . "Money, Credit, and Business Fluctu- . "Growth Theory: Comments: Some ations." Economic Record, December 1988b, Retrospective Views on Growth Theory," in 64(187), pp. 62-72. P. Diamond, ed., Growth/productivity/unem- . "On the Relevance or Irrelevance of ployment: Essays to celebrate Bob Solow's Public Financial Policy," in K. J. Arrow and birthday.Cambridge, MA: MIT Press, 1990, M. J. Boskin, ed., The economics of public pp. 50-68. debt: Proceedings of a conference held by the . "Development Strategies: The Roles International Economic Association. New of the State and Private Sector," in S. York: St. Martin's, 1988c, pp. 41-76. Fischer, D. de Tray, and S. Shekhar, eds., . "Why Financial Structure Matters." Proceedings of the World Bank annual con- Journal of Economic Perspectives, Autumn ference on development economics 1990. 1988d, 2(4), pp. 121-26. Washington, DC: World Bank, 1991a, pp. . "On the Economic Role of the State," 430-33. in A. Heertje, ed., The economic role of the . "Some TheoreticalAspects of the Pri- state. Oxford, U.K.: Blackwell, 1989a, pp. vatization: Applications to Eastern Europe," 9-85. in Rivista di Politica , December . "Financial Markets and Develop- 1991b, 81(158), pp. 199-224; reprintedin M. VOL.92 NO. 3 STIGLITZ:INFORMATION ECONOMICS AND PARADIGMCHANGE 499

Baldassarri,L. Paganetto, and E. S. Phelps, eds., Social capability and long-term eco- eds., Privatizationprocesses in Eastern Eu- nomic growth. New York: St. Martin's, 1995, rope. Rome: St. Martin's, 1993, pp. 179- pp. 48-81. 204. . "Some Lessons from the East Asian . "CapitalMarkets and Economic Fluc- Miracle." World Bank Research Observer, tuations in CapitalistEconomies." European August 1996, 11(2), pp. 151-77. Economic Review, April 1992a, 36(2-3), pp. . "The Role of Government in Eco- 269-306. nomic Development," in M. Bruno and B. . "ContractTheory and Macroeconomic Pleskovic, eds., Annual WorldBank confer- Fluctuations,"in L. Werin and H. Wijkander, ence on development economics, 1996. eds., Contract economics. Cambridge, MA: Washington, DC: World Bank, 1997a, pp. Blackwell, 1992b, pp. 292-322. 11-23. . "ExplainingGrowth: Competition and . "TheRole of Governmentin the Econ- Finance."Rivista di Politica Economica, No- omies of Developing Countries,"in E. Mal- vember 1992c, 82(169), pp. 277-43. invaud and A. K. Sen, eds., Development . "Prices and Queues as Screening De- strategy and the management of the market vices in Competitive Markets," in D. Gale economy. Oxford:Clarendon, 1997b, pp. 61- and 0. Hart, eds., Economic analysis of mar- 109. kets and games: Essays in honor of Frank "ParetoEfficient Taxation and Expen- Hahn. Cambridge, MA: MIT Press, 1992d, diturePolicies, With Applicationsto the Tax- pp. 128-66. ation of Capital, Public Investment, and . "Notes on Evolutionary Economics: Externalities."Unpublished manuscript,pre- Imperfect Capital Markets, Organizational sented at conference in honor of Agnar Design, Long-run Efficiency." Unpublished Sandmo, January 1998a. manuscript presented at a Conference at . "Towardsa New Paradigmfor Devel- Osaka University, 1992e. opment: Strategies, Policies and Processes." . "Consequencesof Limited Risk Mar- Unpublished manuscript,9th Raul Prebisch kets and Imperfect Informationfor the De- Lecture delivered at the Palais des Nations, sign of Taxes and Transfers:An Overview," Geneva, UNCTAD, October 1998b. in K. Hoff, A. Braverman, and J. Stiglitz, . "Interest Rates, Risk, and Imperfect eds., The economics of rural organization: Markets: Puzzles and Policies." Oxford Re- Theory,practice, and policy. New York: Ox- view of Economic Policy, 1999a, 15(2), pp. ford University Press, 1993. 59-76. "Economic Growth Revisited." Indus- . "Knowledge for Development: Eco- trial and CorporateChange, 1994a, 3(1), pp. nomic Science, Economic Policy, and Eco- 65-110. nomic Advice," in B. Pleskovic and J. E. . "EndogenousGrowth and Cycles," in Stiglitz, eds., Proceedings from the annual Y. Shionoya and M. Perlman, eds., Innova- bank conference on developmenteconomics, tion in technology, industries, and institu- 1998. Washington DC: World Bank, 1999b, tions: Studies in Schumpeterianperspectives. pp. 9-58. Ann Arbor, MI: University of Michigan . "MoreInstruments and BroaderGoals: Press, 1994b, pp. 121-56. Moving Toward the Post-Washington Con- . Whithersocialism? Cambridge, MA: sensus." Revista de Economia Politica, MIT Press, 1994c. January-March 1999c, 19(1), pp. 94-120; . "The Role of the State in Financial reprinted in Ha-Joon Chang, ed., The rebel Markets," in M. Bruno and B. Pleskovic, within: at the World Bank. eds., Proceeding of the World Bank confer- London: Anthem, 2001, pp. 19-56. ence on development economics, 1993. . "Toward a General Theory of Wage Washington, DC: World Bank, 1994d, pp. and Price Rigidities and Economic Fluctua- 41-46. tions." American Economic Review, May . "Social AbsorptionCapability and In- 1999d (Papers and Proceedings), 89(2), pp. novation,"in Bon Ho Koo and D. H. Perkins, 75-80. 500 THE AMERICANECONOMIC REVIEW JUNE 2002

. "Respondingto Economic Crises: Pol- nomics in Europe. New York: Oxford Uni- icy Alternatives for Equitable Recovery and versity Press, 2001e, pp. 22-54. Development." Manchester School, 1999e, . "New Perspectiveson Public Finance: Spec. Iss., 67(5), pp. 409-27. Recent Achievements and Future Chal- . "WhitherReform? Ten Years of the lenges." Journal of Public Economics, 2002a Transition,"in B. Pleskovic and J. E. Stiglitz, (forthcoming). eds., Proceedings of the annual bank confer- . "On Liberty, the Right to Know and ence on development economics, 1999. Public Discourse: The Role of Transparency Washington, DC: World Bank, 2000a, pp. in Public Life," in M. Gibney, ed., Globaliz- 27-56. ing rights. Oxford, U.K.: Oxford University . "Capital Market Liberalization,Eco- Press, 2002b (forthcoming). nomic Growth, and Instability." World De- Stiglitz,Joseph E. and Weiss,Andrew W. "Credit velopment,June 2000b, 28(6), pp. 1075-86. Rationing in Markets with Imperfect Infor- . "Formal and Informal Institutions," mation." American Economic Review, June in P. Dasgupta and I. Serageldin, eds., So- 1981, 71(3), pp. 393-410. cial capital: A multifaceted perspective. ."AlternativeApproaches to the Anal- Washington, DC: World Bank, 2000c, pp. ysis of Markets with Asymmetric Informa- 59-68. tion." American Economic Review, March . "The Contributionsof the Economics 1983a, 73(1), pp. 246-49. of Information to Twentieth Century Eco- . "IncentiveEffects of Termination:Ap- nomics." Quarterly Journal of Economics, plications to the Credit and Labor Markets." November 2000d, 115(4), pp. 1441-78. American Economic Review, December . "Some ElementaryPrinciples of Bank- 1983b, 73(5), pp. 912-27. ruptcy," in Governance, equity and global . "CreditRationing and Collateral,"in J. markets(proceedings of annual bank confer- Edwards, J. Franks, C. Mayer, and S. ence for development economics in Europe, Schaefer, eds., Recent developments in cor- June 1999). Paris: La DocumentationFran- porate finance. New York: Cambridge Uni- caise, 2000e. versity Press, 1986, pp. 101-35. . "Challenges in the Analysis of the . "CreditRationing: Reply." American Role of Institutions in Economic Develop- Economic Review, March 1987, 77(1), pp. ment," in G. Kochendorfer-Luciusand B. 228-31. Pleskovic, eds., The institutionalfoundations . "Banks as Social Accountants and of a marketeconomy, Villa Borsig Workshop Screening Devices for the Allocation of Series 2000. Berlin: GermanFoundation for Credit." Greek Economic Review, Autumn InternationalDevelopment (DSE), 2001a, pp. 1990, Supp., 12, pp. 85-118. 15-28. . "Asymmetric Information in Credit . "From Miracle to Recovery: Lessons Markets and Its Implications for Macroeco- from Four Decades of East Asian Experi- nomics." Oxford Economic Papers, October ence," in Shahid Yusuf, ed., Rethinkingthe 1992, 44(4), pp. 694-724. East Asian miracle. Washington,DC: World . "Sorting Out the Differences Be- Bank, 2001b. tween Screening and Signaling Models," in . "Principlesof FinancialRegulation: A M. O. L. Bacharach,M. A. H. Dempster, and Dynamic Approach."World Bank Observer, J. L. Enos, eds., Mathematicalmodels in eco- Spring 2001c, 16(1), pp. 1-18. nomics. Oxford: Oxford University Press, . "Crisis y Restructuraci6nFinanciera: 1994. el Papel de la Banca Central." Cuestiones Stiglitz,Joseph E. and Yusuf, Shahid."Develop- Econdmicas, 2001d, 17(2), pp. 3-24. ment Issues: Settled and Open," in G. M. "Quis Custodiet Ipsos Custodes? Cor- Meier and J. E. Stiglitz, eds., Frontiers of porate Governance Failures in the Transi- development economic: The future in per- tion," in J. E. Stiglitz and P.-A. Muet, eds., spective. Oxford: Oxford University Press, Governance, equity, and global markets:the May 2000, pp. 227-68. annual bank conference on developmenteco- Todaro,Michael P. "A Model of Labor Migra- VOL. 92 NO. 3 STIGLITZ:INFORMATION ECONOMICS AND PARADIGMCHANGE 501

tion and Urban Unemployment in Less De- tract: Determinants,Properties and Implica- veloped Countries." American Economic tions, 1990 Nobel Symposium 77, Salts- Review, March 1969, 59(1), pp. 138-48. jobaden, Sweden.] Townsend, Robert J. "Optimal Contracts and Williamson, John. "What Washington Means Competitive Markets with Costly State Ver- by Policy Reform," in J. Williamson, ed., ifications."Journal of Economic Theory,Oc- Latin American adjustment: How much has tober 1979, 21(2), pp. 265-93. happened? Washington, DC: Institute of Visser, Bauke. "BinaryDecision Structuresand , April 1990, pp. the Required Detail of Information."Euro- 5-20. pean University InstituteWorking Paper No. Wilson, Robert B. "Competitive Bidding with 89/1, February 1998. Disparate Information." Management Sci- Waldman,Michael. "Job Assignments, Signal- ence, March 1969, 15(7), pp. 446-48. ing and Efficiency." RAND Journal of Eco- . "A Bidding Model of Perfect Compe- nomics, Summer 1984, 15(2), pp. 255-67. tition." Review of Economic Studies, 1977, Weiss, AndrewW. "Job Queues and Layoffs in 44(3), pp. 511-18. Labor Markets with Flexible Wages." Jour- World Bank, ed. Knowledge for development: nal of Political Economy, June 1980, 88(3), 1998/99 world developmentreport. Washing- pp. 526-38. ton, DC: World Bank, 1999. Werin,Lars and Wijkander,Hans, eds. Contract Yabushita,Shiro. "Theoryof Screening and the economics. Cambridge, MA and Oxford, Behavior of the Firm." American Economic U.K.: Blackwell, 1992. [Proceedingsof Con- Review, March 1983, 73(1), pp. 242-45.