THE WORLD BANK RE SEARCH OBSERVER VOLUME 11 NUMBER 2 AUGUST 1996 20867 Some Lessons from the East Asian Miracle Joseph E. Stiglitz In Search of Owners: Privatization and Corporate Governance in Transition Economies Cheryl W. Gray How Well Can Method Substitute for Data? Five Experiments in Poverty Analysis Martin Ravallion Deforestation and Forest Land Use: Theory, Evidence, and Policy Implications William F. Hyde, Gregory S. Amacher, and William Magrath Financial Markets, Public Policy, and the East Asian Miracle Joseph E. Stiglitz and Marilou Uy Credit Policies: Lessons from Japan and Korea Dimitri Vittas and Yoon Je Cho FILE COPY THE WORLD BANK RESEARCH OBSERVER EDITOR Moshe Syrquin COEDITORS Shantayanan Devarajan, Shahid Yusuf CONSULTING EDITOR Elinor Berg EDITORIAL BOARD Claire Liuksila (International Monetary Fund); Willem Buiter (Cambridge University); Angus Deaton (Princeton University); Howard Pack (Universiry of Pennsylvania); Martha Dr Melo, Gershon FIder, Gregory K. Ingram, George Psacharopoulos, William Tyler, Michael Walton (World Bank) The World Bank Research Observer is intended for anyone who has a professional interest in development. Observer articles are written to be accessible to nonspecialist readers; contributors examine key issues in development economics, survey the literature and the latest World Bank research, and debate issues of development policy. Articles are reviewed by an editorial board drawn from across the Bank and the international community of economists. Inconsistency with Bank policy is not grounds for rejection. The journal welcomes editorial comments and responses, which will be considered for publication to the extent that space permits. On occasion the Observer considers unsolicited contributions. Any reader interested in preparing such an article is invited to submit a proposal of nor more than two pages to the editor. Please direct all editorial correspon- dence to the Editor, The World Bank Research Observer, at the address in the copyright notice below. The views and interpretations expressed in this journal are those of the auithors and do nor necessarily represent the views and policies of the World Bank or of its Executive Directors or the countries they represent. The World Bank does not guarantee the accuracy of the data included in this publicarion and accepts no responsibility whatsoever for any consequences of their use. When maps are used, the boundaries, denominations, and other information do not imply on the part of the World Bank Group any judgment on the legal status of any territory or the endorsement or acceptance of such boundaries. The World Bank Research Observer is published twice a year in February and August. Single copies may be purchased for US$12.95. Subscription rates are as follows: Individuals Institutions 1-year subscription US$25 US$40 2-year subscription US$45 US$75 3-year subscription US$60 US$100 Orders should be sent to: World Bank Publications, Box 7247-7956, Philadelphia, Pennsylvania 19170-7956, U.S.A. Readers with mailing addresses in developing countries and in socialist economies in transition are eligible to receive complementary subscriptions oni request. Written confirmation is required every two years to renew such subscriptions. O 1996 The International Bank for Reconstruction and Development / THE WORILD BANK 181 8 H Street, N.W., Washington, D.C. 20433, U.S.A. All rights reserved Manufactured in the United States of America ISSN 0257-3032 Material in this journal is copyrighted. Requests for permission to reproduce portions of it should be sent to the Office of the Publisher at the address in the copyright notice above. The World Bank encourages dissemination of its work and will normally give permission promptly and, when the reproduction is for noncommercial purposes, without asking a fee. Permission to copy portions for classroom use is granted through the Copyright Clearance Center, Suite 910,222 Rosewood Drive, Danvers, Massachusetts 01923, U.S.A. The World Bank Research Observer is indexed by the Journal of Economic Literature, the Standard Periodical Directory, ARI/INFORM, the Index to International Statistics, the Public Affairs Information Service, Inc., the Social Sciences Citation Index, and, online, by the Economic Literature Index and DIALOC. It is available in microform through University Microfilms, Inc., 300 North Zeeb Road, Ann Arbor, Michigan 48106, U.S.A. THE WORLD BANK RE-SEARCH OBSERVER VOLUME 11 NUMBER 2 AUGUST 1996 Some Lessons from the East Asian Miracle Joseph E. Stiglitz 151 In Search of Owners: Privatization and Corporate Governance in Transition Economies Cheryl W. Gray 179 How Well Can Method Substitute for Data? Five Experiments in Poverty Analysis Martin Ravallion 199 Deforestation and Forest Land Use: Theory, Evidence, and Policy Implications William F. Hyde, Gregory S. Amacher, and William Magrath 223 Financial Markets, Public Policy, and the East Asian Miracle Joseph E. Stiglitz and Marilou Uy 249 Credit Policies: Lessons from Japan and Korea Dimitri Vittas and Yoon Je Cho 277 SOME LESSONS FROM THE EAST ASIAN MIRACLE Joseph E. Stiglitz The rapid economic growth of eight East Asian economies, often called the "East Asian miracle," raises two questions: What policies and other factors contributed to that growth? And can other developing countries replicate those policies to stimulate equally rapid growth? This article, based on case studies, econometric data, and economic theory, offers a list of the ingredients that contributed to that success. But it is the combination of these ingredients, many of which involve government inter- ventions acting together, that accounts for East Asia's success. T- he remarkable success of the economies of East Asia raises the question: to what can that success be attributed? In most of the eight economies that are part of the "East Asian miracle"-Hong Kong, Indonesia, Japan, the Republic of Korea, Malaysia, Singapore, Taiwan (China), and Thai- land-government undertook major responsibility for the promotion of eco- nomic growth. Which policies contributed to the success of these economies, and why? Ascertaining what would have happened in the absence of the speci- fied policy is often difficult. That the government subsidized a sector that grew rapidly does not imply that the growth should be attributed to the government's action. The sector might have grown without government intervention. This article is an interpretive essay based on case studies, econometric data, and economic theory. In formulating a coherent explanation of East Asia's ex- perience, I do not present a formula or a simple recipe, but rather a list of ingre- dients. Because these ingredients are interactive, and because they were intro- duced in conjunction with other policies, the government's approach has to be evaluated as a package. Indeed, East Asia's success was based on a combination of factors, particularly the high savings rate interacting with high levels of hu- man capital accumulation, in a stable, market-oriented environment-but one with active government intervention-that was conducive to the transfer of technology. ' Each of the economies is unique; each differs in its history and culture. Some, such as Singapore and Hong Kong, are small city-states. Others are large. Many The World Bank Research Observer, vol. 11, no. 2 (August 1996), pp. 151-77 ( 1996 The International Bank for Reconstruction and Development / THE WORLD BANK 151 are racially and culturally homogeneous; some, such as Malaysia, are culturally diverse. But it seems implausible to attribute the success of each of these coun- tries to special factors; the task instead is to discover the common threads. Moreover, the unique factors typically refer to certain cultural aspects, such as a Confucian heritage, that are suspect: not that long ago, the Confucian heri- tage, with its emphasis on traditional values, was cited as an explanation for why these countries had not grown. To be sure, cultural factors may play an important role: the stress on education has contributed much to the success of these countries. Statistical Explanations Having expressed reservations about the usefulness of cultural explanations, it is important to note as well the limitations to the standard statistical tech- niques. For almost four decades (Solow 1957), the standard approach has been to ask to what extent this growth can be explained by increases in inputs, that is, human and physical capital, and expenditures for the acquisition of technology. In this approach, the "miracle" is the amount of this growth that cannot be so explained (the residual). Several such studies have argued that the East Asian experience can largely be explained by rapid increases in inputs-high levels of investment and heavy expenditures for education. Krugman (1994) and Young (1993, 1995) argue that essentially all of the growth in Singapore can be so explained. Others, such as Kim and Lau (1993) and World Bank (1993), find more evidence of a positive residual, but even then it is not unusually high. There are a variety of technical reasons why the applicability of this methodol- ogy to at least several of the East Asian countries should be suspect. Whatever their flaws, however, these studies still offer an important lesson: policies that increase the accumulation of physical and human capital are likely to lead to mnore rapid growth. The real problem is what these studies leave unanswered. The unique and changing circumstances of each country and the multitude of programs involved, each with a number of potentially important features, imply that a statistical study would be relevant only for addressing the broadest ques- tions: were savings rates unusually high, or were financial restraints associated with faster rates of economic growth? Such studies do not identify those fea- tures that facilitated what in retrospect was a remarkable transformation of the economy. To understand this transformation, answers must be found to the following: First, why were saving rates so high? Elsewhere, such saving rates had only been attained under the compulsion of strong government force, as in the Com- munist countries. Although studies suggest that these saving rates may be ex- plained in part by economic factors-such as high growth rates that spurred high saving rates, as consumption lagged increases in incomes (Carroll, Weil, and Summers 1993; Stiglitz and Uy 1996)-government actions also played an 152 The World Bank Research Observer, vol. 11, no. 2 (August 1996) important role in mobilizing savings (although this "virtuous cycle" between growth and savings was important as well). Second, how was it possible to invest efficiently at such a rapid pace? To be sure, if life consisted of nothing more than adding homogeneous capital to a homogeneous production process, East Asia's success would hardly be remark- able. But in that case other countries that attempted to invest rapidly would have had far more success than they have had. Third, how was it possible to reduce the technology gap so quickly? Clearly, more was entailed than just buying technology. To encourage the transfer of technology from foreign investors, the East Asian economies made enormous investments in human capital, educating large numbers of skilled engineers able to absorb and adapt the most advanced technology. And the East Asian econo- mies were willing to accept foreign investment and create an economic atmo- sphere conducive to its entry.2 Moreover, they combined these efforts with an emphasis on the most technologically advanced investment. And finally how did East Asia ensure that the benefits of rapid growth were spread widely among the population? Previous theories suggested that rapid growth was associated with rapid capital accumulation, which in turn was asso- ciated with high degrees of inequality; and that growth would in fact be accom- panied by an increase in the degree of inequality (Kuznets 1955). Not only did this assumption prove to be false, but there are reasons to believe that govern- ment policies that promoted greater equality contributed in no small measure to the remarkable growth of these countries. Metaphors of Economic Growth Several metaphors are used to describe the process of economic growth. These metaphors undoubtedly influence how we think about the subject. An Engine Metaphor Perhaps the most popular metaphor is that which refers to the engine of growth-as if there were a motor driving the performance of the economy. Capital accumulation is often given credit for being the engine of growth. And the coun- tries of East Asia certainly have accumulated capital at an impressive rate. Some- times the concept of capital accumulation is broadened to include human capi- tal-the improvement in the skills of the labor force. And sometimes these two are given credit not only for their direct contribution, but also for the technical progress that might not have occurred in their absence. Once one identifies the engine of growth, one tries to make the engine stron- ger. Thus, if capital accumulation is the engine, the task is to increase capital accumulation. The role of government is to rev up the engine to encourage a higher rate of capital accumulation. The engine metaphor has some important Joseph E. Stiglitz 153 limitations: it encourages a search for particular factors that account for growth, although it may in fact be the system as a whole, including the interactions among the parts, that accounts for growth. If human capital accumulation is inadequate, even rapid physical capital accumulation may be ineffective. But if both are required, which one is the engine? A Chemical Metaphor I prefer two other metaphors for thinking about the growth process. One is borrowed from chemistry: the government as catalyst. The government can be a catalyst for growth without necessarily providing a great deal of resources. In- deed, that is the remarkable property of catalysts-having set off a chemical reaction, they are themselves not used up in the process. At the very least this metaphor warns that the effect of some government policy should not be mea- sured simply by asking about the magnitude of the subsidy or what fraction of the funds was provided by the government. More concretely, investments in human capital and infrastructure, both physical and institutional, can increase the private return to investment and thereby promote growth. A Biological Metaphor The second metaphor, "adaptive systems," is borrowed from biological ter- minology. Species that survive adapt to changes in their surroundings. More advanced species survive in part because of their ability to learn. Thus the most important characteristic for survival is not a particular policy, but the ability to respond to changes in the environment and to learn from past mistakes. Adapt- ability is often said to distinguish private sector enterprises from government bureaucracy. But government, because of its monopoly powers, can survive even if it does not adapt well or quickly. The East Asian economies demonstrated that government too can be highly adaptive. When changes in the environment made previously adopted policies inappropriate, these governments changed course, and they learned quickly from their mistakes. As their economies grew and became more complex, the state's role clearly had to change: there was neither the need nor the capacity for active intervention on the scale previously assumed. And officials recognized the importance of adopting policies to pro- mote higher levels of technology and higher value-added industries. A Metaphor from Physics One metaphor has been omitted from this list-the one that has in fact domi- nated the economics profession for almost a century-the economy as an equi- librium system. The omission is deliberate: it is a metaphor that provides little insight into the dramatic changes that occurred in these societies. The equilib- rium metaphor suggests that individuals had (perhaps rational) expectations 154 The World Bank Research Observer, vol. 11, no. 2 (August 1996) concerning future rates of return; given those expectations, they determined their saving rate; meanwhile, profit-maximizing firms scoured the world look- ing for the best products and technologies to employ, given the costs of adjust- ment. In this metaphor, too, government played at most an ancillary role. This metaphor leaves unasked-and unanswered-such fundamental questions as, What set the East Asian countries apart from other countries? Why is their experience so different? Complementing Markets Rather Than Replacing Them Before the East Asia miracle there were two dominant paradigms for devel- opment, one focused on markets, the other on government and planning. The first had its intellectual roots in Adam Smith's "invisible hand": markets lead to efficient outcomes. All that government needs to do to promote growth is get out of the way. The basic slogan is "get the prices right." With the right prices, everyone will have an incentive to make the right resource allocations. Under- mining this particular religion was the disturbing observation that countries that seemed to get the prices right-to follow all the advice of the visiting preachers of the free market-too often failed to grow. To be sure, like medieval medicine, there was always the allegation that the patient had not followed the doctor's orders precisely, and it was this that accounted for the failure of the remedy. At the opposite side were those who had little faith in the market and who looked to government to ensure through the planning process that resources were deployed in a way that promoted economic growth. The lack of success of those countries that followed this paradigm has led to the virtual extinction of this school of thought. Ironically, almost none of the successful industrial countries followed either of these extreme strategies. They are mixed economies in which government plays an important role. The appropriate question to be asked is not whether government should play a role, but what role and how can it be performed most effectively. At the same time that the success of the East Asian economies and the col- lapse of the socialist economies called into question the standard paradigm, ad- vances in economic theory called into question the intellectual foundations of these two approaches. In the mid-1950s Arrow and Debreu (1954) identified several conditions that must be satisfied if markets are to yield efficient out- comes. These include, first, the absence of externalities (external economies or diseconomies that affect the activity in question) and of public goods (com- modities or services that, once provided, can be obtained without payment by others); second, the presence of perfect competition; and, third, a complete set of markets, including markets extending infinitely far into the future and cover- ing all risks. A market failure is said to occur where these conditions are not satisfied. This approach identified specific interventions by the government to Joseph E. Stiglitz 155 correct each market failure, for instance, pollution taxes to correct for environ- mental damage. Government had a well-defined, highly circumscribed role. It was not until thirty years later, however, that the full limits of the market mechanism became well understood. Hidden in Arrow and Debreu's frame- work were strong assumptions about information and technology. In their model information need not be perfect, but it could not change as a result of actions taken within the economy. Greenwald and Stiglitz (1986) showed that when- ever information was imperfect or markets were incomplete, government could devise interventions that filled in for these imperfections and that could make everyone better off. Because information was never perfect and markets never complete, these results completely undermined the standard theoretical basis for relying on the market mechanism. Similarly the standard models ignored changes in technology; for a variety of reasons markets may underinvest in re- search and development (see, for example, Stiglitz 1987,1988, and Arrow 1962). Because developing economies have underdeveloped (missing) markets and im- perfect information and because the development process is associated with ac- quiring new technology (new information), these reservations about the adequacy of market mechanisms may be particularly relevant to developing countries (Stiglitz 1989). The modern theory of market failures recognizes, however, that government interventions may not actually improve matters. Theories of regulatory capture and rent-seeking imply that government interventions may contribute to ineffi- cient resource allocation, and whatever their weaknesses, these theories have sufficient plausibility to suggest that governments need to exercise caution. How the government intervenes may matter a great deal. The fundamental mistake of the countries of the former Soviet Union and those developing countries that tried to rely on planning was that they sought to correct market failures by replacing the market. The governments of East Asia, by contrast, recognized the limitations of markets but confined the government's role to * Policies that actively sought to ensure macroeconomic stability. * Making markets work more effectively by, for instance, regulating financial markets. * Creating markets where they did not exist. * Helping to direct investment to ensure that resources were deployed in ways that would enhance economic growth and stability. * Creating an atmosphere conducive to private investment and ensured political stability. In short, rather than replacing markets, these governments promoted and used them. Such interventions had to be carefully balanced; if they were too heavy-handed, they might have squelched the market. This agenda required gov- ernment to design interventions in a way that reduced the likelihood of rent- seeking behavior and that increased its ability to adapt to changing circum- 156 The World Bank Research Observer, vol. 11, no. 2 (August 1996) stances. One such mechanism was a performance-based reward structure that provided strong growth-oriented incentives and served as a basis for awarding government subsidies. This structure was relatively free from corruption and helped to direct resources to areas that produced high economic returns. An- other essential step was to design a civil service system based on merit, which compensated employees well and built in provisions that reduced the dangers of corruption. In this discussion, the interventions are organized around four major themes: industrial policies, cooperation and competition, equality, and export-led growth. Some of the most important actions to promote economic growth were directed to the financial market, and these interventions are the subject of the accompa- nying article in this journal by Stiglitz and Uy. Industrial Policies Industrial policies are directed at developing and encouraging certain sectors. What were these industrial policies? Why were they adopted? And did they work, either by directing resources to desired areas or, more broadly, in pro- moting economic growth? What Policies Were Pursued? Most countries shared three objectives: developing technological capabili- ties; promoting exports; and building the domestic capacity to manufacture a range of intermediate goods (such as plastics and steel). Support for par- ticular industries and imports of the necessary foreign technology took sev- eral forms. First, the support for education-particularly engincering and science education-provided an intellectual infrastructure that facilitated technological transfer. Second, the decision to discourage (through financial market regulations) the allocation of capital to areas such as real estate meant that more capital was available for areas with higher technological benefits, such as plants and equipment. Third, as discussed later, the government en- couraged exports. Fourth, in some industries, particularly those with many firms, government promoted technology programs, including science centers that offered services ranging from identifying new products to providing re- search and development for firms that had no facilities of their own. Taiwan (China) and Malaysia developed industrial parks for high-technology indus- tries, both to allow firms to capture some of the diffuse externalities associ- ated with these industries as well as to lower the barriers to entry. (Diffuse externalities arise when the actions of one firm benefit-or confer costs upon-many firms, rather than, say, just one upstream firm or one down- stream firm.) And finally, the government provided explicit and implicit sub- sidies (through cheap credit) to industries it wished to support. joseph E. Stiglitz 157 An important element in the expansion of certain industries was a receptivity to direct foreign investment. The East Asian economies not only resisted xeno- phobic aversions to foreign investments, but they also induced capital inflows by providing sound macroeconomic management, a stable political environment, and well-managed labor markets with educated workers. In many cases govern- ments took explicit steps to ensure that a transfer of technological and human capital would accompany these inflows. Foreign investment increased the pace of expansion, reducing the constraints imposed by limitations on the availability of capital, domestic entrepreneurship, and technological know-how. Why Were Industrial Policies Adopted? Market failures are likely to be particularly significant in developing coun- tries for several reasons.3 Understanding these market failures helps explain the policies that were adopted and the reasons they were so effective. WEAK AND NONEXISTENT MARKETS. In the early stages of development, mar- kets often do not exist or work well, so prices may not provide good signals for resource allocation. In East Asia capital markets were particularly weak, lead- ing government to create institutions to promote savings (the postal savings banks) and to extend long-term credit (the development banks). Governments also tried to develop the financial infrastructure by helping to establish bond and equity markets (Stiglitz and Uy 1996). Having promoted savings, governments had to decide how to allocate these funds. If there had been well-established market institutions for allocating long- term capital, governments could have made use of those institutions. But be- cause the governments had to decide how to allocate resources, it was natural to direct the funds to projects that would yield the highest level of social welfare. TECHNOLOGICAL SPILLOVERS. Private markets have inadequate incentives for investing in the production and acquisition of technology, largely because it is difficult to appropriate the returns to knowledge. Developing countries typi- cally operate at a level of technology far below that of industrial countries; development is, to a large extent, the process of acquiring and adapting existing technologies. Patent protection ensures that the seller can command some pay- ment for new technology, but it does not provide much protection for a firm that transfers and adapts an existing technology. Adopting and adapting new technologies involves a risk. If successes are quickly imitated, then firms face a "heads I lose, tails you win" situation: when they succeed, there is little profit because of the force of competition; when they fail, they lose money. MARKETING SPILLOVERS. Still another kind of valuable information concerns marketing. Knowing where there is a market for a product is not information that can be kept secret. If a firm spends money to discover that Americans like 158 The World Bank Research Observer, vol. II. no. 2 (August 1996) madras shirts, then any manufacturer of madras shirts can take advantage of that information. The converse is that the products of a country establish a reputation. Thus, Japan's reputation for high quality benefits all Japanese pro- ducers. Such marketing spillovers have led governments to adopt programs aimed at promoting the country's products. (In Hong Kong these programs are financed by a special tax. In Singapore they are directed by the powerful Economic De- velopment Board.) Spillovers have also resulted in an array of programs to im- prove the countries' reputation. Most notable in this respect is the recent effort by Taiwan (China) to encourage its domestic firms to obtain brand recognition. RETURNS TO SCALE: a problematic explanation. Not all of the arguments ad- vanced as rationales for industrial policies are persuasive, however. One that seemed particularly influential in Japan held that government intervention was required to rationalize industry. It was argued that without government sup- port, firms would be too small, and the large number of such firms would re- duce the profitability of all firms in a sector. (Thus, the Japanese government not only condoned the increased concentration in the steel industry in the late 1960s but, in one of its most famous mistakes, tried to discourage Honda-at the time a successful manufacturer of motorcycles-from entering the automo- bile market.) This argument is unpersuasive because if there truly were increas- ing returns to scale, then a single firm would benefit by increasing its produc- tion; in time its costs would be lowered, and it would then be able to undercut its rivals. Natural economic forces lead to the rationalization of industries without government intervention. A slight variant of the argument about returns to scale does have some valid- ity. Increasing returns combined with a shortage of capital may stunt small firms. They cannot expand to take advantage of increasing returns either because they cannot get access to capital or because the only form of capital to which they have access is credit, which imposes too high a risk. In this case, government intervention can lower the costs of capital and increase economic efficiency. Increasing returns, especially when combined with capital market imperfec- tions, provide the foundation for strategic trade policy. Historically, arguments for government trade interventions focused on industries with learning by do- ing. If today's production lowers future marginal costs, that creates a form of increasing returns akin to the more familiar static increasing returns. A firm that expands production lowers its future production costs and undercuts its rivals. The infant industry argument holds that protection is important so that the young firm can gain the experience required to lower its production costs and allow it to become viable. Critics of this argument claim that if the firm is to be profitable in the long run, it should incur any necessary losses today. But this assumption is based on the premise that capital markets are perfect. With im- perfect capital markets, a firm may not be able to sustain the losses that would enable it to produce at a level at which it would eventually become profitable. Joseph E. Stnglitz 159 Moreover, if the firm is unable to appropriate all the returns to its learning, then social returns to production will exceed private returns (Dasgupta and Stiglitz 1988). In addition, dominant firms in industrial countries are likely to take ad- vantage of the lack of competition that prevails when learning is important by raising prices and increasing their profits. Government policies may be directed at trying to appropriate some of these rents (the excess profits that result from a dominant competitive position). COORDINATION FAILURES. The widespread absence of markets in developing countrics means that prices cannot perform their coordination rolc. Govern- ment may thus have to assume a more active role in performing this function. The traditional examples relate to the development of downstream and upstream industries: developing a steel-manufacturing industry does not pay unless there is a steel-using industry; and developing a steel-using industry does not pay if there is no steel-manufacturing industry. If both wait, nothing happens. Accord- ing to this view, the government has an important function in coordinating the two activities. Such coordination failures, it is argued, are likely to be most important when the returns to scale are large. For instance, if manufacturing steel is deemed to be desirable, it is necessary to build a large steel plant and a large steel-using industry. Other market failures, such as the absence of risk markets, interact with this failure: large risks are likely to accompany such large- scale investments, and the market provides no mechanism by which these risks can be divested. Moreover, no single entrepreneur could amass the capital re- quired, and the imperfections of the capital market mean that it cannot supply the funds required. Developing countries are less likely than industrial countries to have the organizations capable of undertaking these large investments in a single sector, let alone the capacity to undertake the investments in both the upstream and downstream firms. Thus coordination problems may be larger in developing countries, and the capacity to deal with them may be smaller. The earlier arguments for coordination failures (Rosenstein-Rodan 1943 and Murphy, Shleifer, and Vishny 1989) were rightly criticized as unpersuasive (Stiglitz 1994a). Such a problem could easily be addressed through trade-one of the solutions devised by the East Asian countries (without benefit of the theo- retical literature). It is possible to develop a steel-using industry simply by im- porting steel and to develop steel producers without steel users simply by ex- porting steel. In the early stages of rapid growth, the subsectors responsible for the takeoff in many, if not most, of the East Asian countries-textiles, footwear, sporting goods, toys-were not those in which economies of scale or coordination prob- lems seemed important. But there was a more subtle form of returns to scale in which government intervention did matter and which affected growth even in these areas: the availability of a wide range of intermediate-often fairly com- plex-goods, tailored for the producers of final goods. The sellers of these inter- mediate goods do not capture all of the benefits that their greater availability 160 The World Bank Research Observer, vol. 11, no. 2 (August 1996) provides. The improved two-way flow of information between the producer and the user, which permits better coordination in the development of the interme- diate and final goods, is a benefit of proximity. That explains why importing the intermediate good does not serve as a perfect substitute for domestic production and also provides a rationale for government intervention. In Malaysia it is claimed that the local auto manufacturer has provided important spillovers to the intermediate goods firms that produce parts and that these firms, in turn, have benefited producers of other final goods. STRATEGIC NEGOTIATIONS. In negotiations with other countries or companies, the governments of East Asia have often recognized-and taken advantage of- the nature of the market environment. The outcome of any bargaining depends on the strength of competition on both sides. By reducing competition among buyers of technology and trying to increase competition among sellers, the gov- ernments succeeded in appropriating more of the surplus associated with the transfer of technology than otherwise could have been captured. In Japan, for instance, a single firm was sometimes given the right to negotiate a licensing agreement; it might then be compelled to share the technology with other firms in the industry. Did These Policies Work? Industrial policies have been widely criticized, on the (somewhat contradic- tory) grounds that they were ineffective or distortionary. The first criticism sug- gests that industrial policies are more form than substance. Critics cite statistics such as the small percentage of loans made by the development banks. These statistics are unconvincing, however: the consequences of, say, a loan by the Industrial Bank of Japan may be far greater than the actual dollars lent, because of either its signaling or risk-sharing effect (Stiglitz and Uy 1996). Government policies that increase the equity of a firm can have immense effects through the power of leveraging. Beyond that, there was a wide range of instruments for effecting industrial policies; it is the cumulative effect of all of these that mat- ters. The criticism is more properly directed at those who have suggested that Japan's Ministry of International Trade and Industry totally controlled the allo- cation of resources. This assumption is wrong on two counts. First, firms made most of the decisions about resource allocation-influenced, to be sure, by gov- ernment policies, but not directly controlled by them. None of the East Asian countries is a command-and-control economy. Second, the view that govern- ment makes decisions on its own seems misguided. Consultation between busi- ness and government was extensive (and many of the top leaders of business were former government employees). The charge that industrial policy was distortionary, however, is of more con- cern. Even if there is a rationale for government intervention, this view alleges that government does not do a good job at picking winners. Instances of mis- Joseph E. Stzglitz 161 takes by the government are typically cited. In some cases the government dis- couraged a firm (Honda, for example), when in retrospect it clearly should not have; and in others the government encouraged some industry (such as petro- chemicals), when in retrospect it probably should not have. There are four responses to this criticism. First, good decisionmaking by the government necessarily involves making mistakes: a policy that supported only sure winners would have taken no risks. The relatively few mistakes speak well for the government's ability to pick winners. Second, the government was not heavy-handed. Although it made mistakes of judgment, it did not force its opin- ions on others when they were willing to risk their own capital. This is one of the strengths of decentralized decisionmaking: it ensures that mistaken views will not dominate. Third, to a large extent, government policies were not directed at picking winners in the narrow sense of the term. Several governments decided to sup- port export-oriented industries. In a sense, that was choosing a winning devel- opment strategy; it did not necessarily entail micromanaging. Even when the government identified an industry for support, the banks seem to have had dis- cretion to select which firms or projects within that industry to support. Fourth, industrial policies were focused not so much on picking winners as on identifying market failures-instances where investors could not capture large potential spillovers. Concern about such spillovers helps explain the government's encouragement of high technology industries. Training provides another example. Firms would benefit from a trained labor force, but, because workers can leave for a better job once they are trained, firms have inadequate incentives to pro- ceed with training. Yet a skilled work force is essential for economic growth, so government undertook to improve the quality of the labor force by emphasizing education. Moreover, the criticism of industrial policies as misguided attempts to pick winners ignores the broader range of government actions, such as its role in spearheading the expansion of certain manufacturing sectors. "Picking winners" seems to imply culling from a fixed pool of applicants to find those with the highest long-run social returns. East Asian governments have instead performed an entrepreneurial role. Entrepreneurship requires combining technological and marketing knowledge, a vision of the future, a willingness to take risks, and an ability to raise capital. In early stages of development, these ingredients are typically in short supply. The governments in East Asia stepped in to fill the gap-but in a way that promoted rather than thwarted the development of pri- vate entrepreneurship. Government was also effective in monitoring the recipients of its support and ensuring that they did not siphon off funds for private use. Other government policies, such as those that led to more equity financing, reduced the magnitude of the monitoring problem; that is, they resulted in firms having more appropri- ate incentives. Still other policies, such as those that enhanced the stability of the banking system, led to more effective monitoring by financial institutions. 162 The World Bank Research Observer, vol. 11, no. 2 (August 1996) Cooperation and Competition Popular discussions of the success of Japan and several other East Asian coun- tries have stressed the cooperative relations between government and business, between workers and employers, and between small and large businesses. Clearly, the extent of this cooperation (sometimes referred to as "Japan Inc."), has been exaggerated. Yet a variety of institutions and practices facilitate cooperation, and this kind of cooperation appears to have had beneficial effects. Adam Smith's "invisible hand" of perfect competition argues that because each individual, in pursuing his self-interest, is also maximizing the common welfare, cooperation is not necessary. But when market failures occur, it is not necessarily the case that the selfish pursuit of self-interest leads to efficient outcomes. The governments of East Asia recognized that the business community had superior information about investment decisions, but they also recognized that the overall information base could be improved. The establishment of formal and informal councils gave rival firms and industries a way to exchange infor- mation with each other and with the government. (This information exchange process is sometimes described as akin to indicative planning, but the analogy is, at best, an imperfect one.) These exchanges conveyed far more information than the traditional format used to display planned sectoral inputs and outputs. What made them more meaningful than such exchanges in other countries? Why would businesses, or government for that matter, tell the truth? To a large extent, good behavior is induced through long-term relationships and reputations. In the process of development, social sanctions become less effective in enforcing cooperative behavior, but establishing and maintaining alternative bases for cooperative relations may be difficult. The gains from co- operation are based on the perception that the future returns to cooperation exceed the short-run gains that might accrue from the pursuit of self-interest. But an environment of rapid change may heighten uncertainty about the value of the future relationship and the magnitude of the long-term gains from coop- eration. Moreover, future cooperative gains have to be discounted (meaning that, because they may not materialize in the future, they are worth less in today's terms). Typically uncertainty is greater and discount rates higher in developing countries. Further, concerns about the potential bankruptcy of one or the other firm, which could terminate the relationship, heighten the likelihood that coop- erative relationships will not materialize. Under these circumstances, future gains from cooperation must be greater to compensate firms for sacrificing the short- run gains from self-interested behavior. Encouraging Cooperation The Japanese government used both carrots and sticks to encourage coopera- tion and the exchange of truthful information. Although cultural characteristics are often credited with facilitating this harmonious result, other countries with Joseph E. Stiglitz 163 similar cultural backgrounds have not displayed the same sort of cooperative behavior seen in Japan. It seems far more likely that government actions were more important than culture in shaping these behavioral patterns. Of the institutions and mechanisms that facilitated cooperation, an impor- tant role was played by business councils set up to share reliable and timely information. Why did not some businesses try to "free ride," to obtain the infor- mation provided by others while providing no real information themselves? The answer is, in part, that they were in a longer-term relationship; a firm that "cheated" would be ostracized from the circle. The fact that the government was included in these circles was important: firms wanted to know what the government was thinking about specific projects or what policy changes were planned. Even if a firm's cooperative instincts went astray, self-interest was a strong incentive. Moreover, by paying attention to these councils, the govern- ment ensured that the gains from cooperation were even greater. The government's discretionary powers enabled it to reward cooperation and honesty, and there was at least a fear that the lack of cooperation and the ap- pearance of dishonesty would be punished. Government intervention in markets created rents that the government could then allocate to participants who be- haved cooperatively. For example, by restricting the formation of branch banks, a large franchise value was associated with the right to have a branch. Similarly, restricting credit meant that access to credit had value. And the Bank of Japan (the central bank) could, on a discretionary basis, provide banks with additional funds when needed. The relative stability of the East Asian governments increased the incentives for establishing long-term cooperative relations. At the same time, long-term relations enhanced the effectiveness of incentives (Stiglitz and Weiss 1983). Firms that performed well on one project could expect to be rewarded with another project. The East Asian governments also tried to create an environment conducive to close cooperation among businesses. In Japan, for instance, the government tried to encourage mergers. To the extent that these programs were successful (and there is considerable controversy about that), they reduced the difficulties of cooperation. The smaller the group, the easier cooperation is to attain. Here the government was walking a fine line; a small group could-and may-have led to collusion by restricting competition. In some circumstances, the government approved the formation of so-called recession cartels. These cartels were an explicit attempt to deal cooperatively (and collusively) with the problems that arise in a recession when there is excess capacity in a capital-intensive industry. Under certain conditions, as demand shifts down, prices drop and firms are unable to recover their capital costs. Recession cartels were a way to restrict competition to enable the industry in question to avoid the low prices that would damage all the firms. Whether the gains were worth the costs of reduced compe- tition, higher prices, and underutilized resources is not clear, however. Because of the strong incentives to cheat on such arrangements, cartels are seldom suc- 164 The World Bank Research Observer, vol. 11, no. 2 (August 1996) cessful without legal sanctions from the government. In some cases the Japanese government paid firms to destroy equipment, and in others, to seal equipment shut. Even these tactics were not always successful; some firms did not com- pletely dismantle their equipment. Labor markets were similarly designed to encourage cooperative behavior. The Japanese pattern of lifetime employment was important because it meant that employees had long-term relationships with employers, which facilitated cooperative behavior. The rapid increase in wages that came with age and expe- rience provided a strong incentive for workers to stay with their organizations. The average pay of each age cohort increased sharply, but differentiation within the age cohort remained smaller than in, say, the United States. Japan's preva- lent compensation scheme, in which a large part of the salary was paid as an annual bonus (based largely on profits of the previous year), also encouraged cooperation because workers had, in effect, an equity stake in the firm. This form of risk-sharing may be particularly important in early stages of develop- ment when capital markets are underdeveloped. Because wages are based on the group's performance, the individual has an incentive to monitor his peers to make sure that his co-workers are working hard (Arnott and Stiglitz 1991; Stiglitz 1990b). One might even go further. Basing salary on individual performance encourages self-interested, noncooperative behavior. Conversely, paying wages based on group performance signals the importance of cooperative behavior. Also important in Japan's labor market was the government-established Pro- ductivity Council, which dealt with the degree of inequality that could exist within a firm and limited salaries of top managers to no more than ten times the wages of the lowest-paid workers.4 This compressed wage structure enhanced the sense that top management was not taking advantage of workers and led to greater effort and lower labor turnover. Cooperative behavior between firms and their employees is particularly im- portant in facilitating technological change. Workers are often in the best posi- tion to identify improvements in efficiency, although such improvements do not always redound to the benefit of the workers. Because labor-saving innovations may result in less demand for labor and higher unemployment, employees are often reluctant to disclose such ideas. If, however, the firm provides a guarantee of lifetime employment, existing employees will see no conflict between their interests and those of the firm. Moreover, when wages are based partially on firm profitability, interests coincide: if the productivity-enhancing innovation enhances profits in the long run, employees will share in the gain. Of course, when growth is rapid firms can more easily promise that labor-saving innova- tions will not result in reduced employment, which makes it more credible that all (existing) employees will benefit from such innovations. Cooperative behavior between firms and their banks was also evident in the operations of capital markets. In Japan each firm had a long-standing relation- ship with a single bank, and that bank played a large role in the affairs of the firm. Japanese banks, unlike American banks, are allowed to own shares in the Joseph E. Stiglitz 165 firms to which they lend, and when their client firms are in trouble, they step in. (The fact that the bank owns shares in the firm means that there is a greater coincidence of interest than there would be if the bank were simply a creditor; see Stiglitz 1985.) This pattern of active involvement between lenders and bor- rowers is seen in other countries of East Asia and was actively encouraged by governments. Another important aspect of business-government cooperation in Japan has been the attempt to reduce bankruptcies, which have been markedly less cyclical than those in the United States and other countries. This pattern reflects not only the country's better macroeconomic performance and a le- gal structure that encourages actions short of bankruptcy but also an active government policy directed at avoiding the economic disruption caused by bankruptcy. Combining Competition and Cooperation The East Asian countries succeeded (not always, but with a remarkable fre- quency) in harnessing the advantages of cooperation while retaining the advan- tages of competition. Cooperation to increase efficiency can easily be turned into collusion to raise prices and restrict output and entry. Worse still, discre- tionary powers needed for cooperation can give rise to rent-seeking and corrup- tion. Competition both enhanced efficiency and reduced the scope for abuses of discretionary powers. In fostering a competitive industrial structure, govern- ments looked not so much at the number of firms in an industry, but at the effectiveness of the competition; competition may be more effective with two evenly matched firms than with one firm competing with many small rivals (Nalebuff and Stiglitz 1983). By the same token, the process of identifying which workers to promote in Japanese firms may be more effective in encouraging competition than is the process in the United States. In Japan, where workers are less mobile, a cohort of workers hired together advances together. They all work hard; they all have to signal that they are committed to the firm; they all remain in the contest. In the United States decisions concerning who is on an upward career ladder often take place earlier. Under that system, incentives may be strong in the early stages of individual careers, but they may be greatly attenuated once these decisions are made. Those who know that they are not going to be "winners" have little reason to work hard. One method introduced to stimulate competition was the use of contests. Governments rewarded firms that performed well relative to others (such as in exports) by, for example, providing them with access to capital and for- eign exchange. In many instances, the value of the prize arose from the gov- ernment intervention: if the government had not created artificial scarcity of capital or foreign exchange, an increase in availability would have had no incentive effect. 166 The World Bank Research Observer, vol. 11, no. 2 (August 1996) Well-functioning contests are characterized by rules that establish a clear cri- terion for rewards, such as export performance; specify the nature of the reward (the allocation of credit or foreign exchange); and indicate who will evaluate performance. This system reduced the scope for abuse of bureaucratic discre- tion at the same time that it provided strong incentives. Ironically, licensing requirements put in place to restrict competition may give rise to more competitive behavior. At various times, the Japanese govern- ment imposed restrictions on the expansion of capacity in certain industries. It awarded licenses to expand capacity on the basis of firms' previous market shares. Thus performance-particularly growth-in one year may increase profits not only in that year, but also in subsequent years. Growth with Equality Although industrial policies attempt to direct resource allocations in ways that maximize growth, income distribution policies seek to promote greater equal- ity. Historically, the development process has been characterized by marked increases in inequality (the Kuznets curve). It was alleged that the massive amounts of capital accumulation required could only be attained through sig- nificant inequality; the poor simply could not save enough. Moreover, growth creates winners (the owners of those firms that do well), and losers (workers displaced from lagging industries, in particular agriculture). The economies of East Asia were able to achieve rapid growth without an increase in inequality. Indeed, active policies promoting equality probably enhanced growth (figure 1). In Korea, Japan, and Taiwan (China), land reforms-at least partially im- posed from the outside-were important in the initial stages of development. These had three effects: they increased rural productivity and income and re- sulted in increased savings; higher incomes provided the domestic demand that was important in these economies before export markets expanded; and the redistribution of income contributed to political stability, an important factor in creating a good environment for domestic and foreign investment. In later years policies to ensure more equitable distribution of income contin- ued to contribute to economic growth, with positive effects that more than off- set the possible negative effects of reduced capital accumulation upon which earlier discussions had focused. These policies continued to contribute to politi- cal stability. High and increasing wages reduced inequality, made workers not only more satisfied but also (by standard efficiency wage arguments) more pro- ductive, and promoted cooperative relations between workers and firms. Poli- cies that attempted to restrict real estate speculation (by limiting lending for that purpose) can be viewed both as part of industrial policy and as part of income distribution policy. While they directed funds into industry, they limited the increases in the prices of housing relative to what would otherwise have occurred. Such price increases would have led to demands for further wage in- Joseph E. Stiglitz 167 Figure 1. Income Inequality and Growth of Gross Domestic Product, 1970-93 GDP growth per capita (percent) 8 6 Singapore ® Botswana 6 _ Indonesia @> 'Hong Kong Thailand O O Chile 4 _ Sri Lanka O O Malaysia Pakistan @ Italy Brazil 0 Japan 0 O India (D Colombia 2 - Belgium 0 D D0 France ~ Clmi Australia D Nepal G Switzerland Bangladesh 0 0 Bolivia 0 Mexico 0 -Philippines ° ariai E) Mauritania Argentina ° ° Venezuela -2 e0 Ghana CD Kenya 0 C6te d'Ivoire O Peru -4 ( Zambia -6 . . i . 0 5 10 15 20 25 Income inequality' a. Income inequality is measured by the ratio of the income shares of the richest 20 percent and the poorest 20 percent of the population. Source: World Bank data. creases and would have had particularly adverse effects on the very poor, who often seem unable to obtain adequate housing under such conditions. Additionally, policies ensuring universal literacy both increased productivity and promoted greater equality. The emphasis on female education led to re- duced fertility, thus mitigating the adverse effects of population pressure felt in so many developing countries, and it directly increased the supply of educated labor. Most studies suggest that a worker's wage performance is more directly related to nonschool factors, such as home background, than to education in school. Education of women can be thought of as a roundabout but high-return way of enhancing labor force productivity. In Thailand a program to provide credit to the rural sector, although largely motivated by concerns about communist insurgency, seemed not only to have promoted equality but also to have yielded reasonably high economic returns. And in Malaysia, policies that would be regarded as affirmative action else- where were able not only to draw upon a reservoir of human talent that had not 168 The World Bank Research Observer, vol. 11, no. 2 (August 1996) been well used before, but to weld together a nation that had already demon- strated a potential for ethnic strife. There are positive relations between growth and equality. High rates of growth provided resources that could be used to promote equality, just as the high de- gree of equality helped sustain the high rates of growth. Although this may seem to be little more than common sense, until the experience of East Asia, "com- mon sense" suggested quite the contrary: growth produced inequality, and in- equality was necessary for growth. Export-Oriented Growth Why focus on exports? Should not countries simply produce the goods in which they have a comparative advantage, whether that happens to be products that are exported or substitutes for goods that are currently imported?5 Success in exporting provided policymakers with an objective way to award credit and foreign exchange.6 Two questions arise: why are exports a better measure of performance than profits? And second, do markets reward success in an appro- priate way without government intervention? In measuring performance to determine which firms to favor with credit and other scarce resources, governments faced an information problem. All governments face a similar problem, but in the context of development, the information problem is particularly severe for two reasons. First, relatively few firms may be engaged in similar activities, so bases of comparison are limited. Second, a host of problems must be overcome-new supplier rela- tionships, new markets, and so on. In such circumstances, short-run profits may be imperfect indicators of long-term performance. Consider, for instance, the two sources of profits: those derived from exports, and those derived from domestic sales. The latter may reflect either the firm's efficiency or its monopoly position in the economy. The profits that result from imperfect competition in the domestic market accrue at least partially at the expense of consumers and should not be thought of as a social gain. By contrast, a firm that succeeds in the export market is more likely to be economically efficient. It can market a product at a lower price than can foreign rivals, or one better tailored for the world market. Export markets are more likely to be competitive. And even if they are not, it is of no concern: the profits of the firm are then at the expense of foreign consumers. Indeed, from the export- ing country's perspective, finding a niche within which some market power can be exercised is to be rewarded, not condemned. Other advantages are also associated with exports. Firms learn a great deal in international markets, benefiting from spillovers related to both marketing and production know-how. For instance, success in producing intermediate goods requires producing to standards that are typically higher than those that prevail within developing countries. This generates a demand for testing laboratories. Joseph E. Stiglitz 169 The recognition that standards are important and the knowledge about produc- ing goods of higher quality has implications across a broad range of products. Moreover, the contacts established through exporting may be of value when the firm decides to enter related markets. It will, for instance, know where to turn to acquire advanced technology. From a social perspective, success in exporting may be a better indicator of whether a firm merits additional funds than success in selling domestically, but banks have typically preferred lending to firms engaged in the domestic market, and for a simple reason. Banks do not care whether a firm makes social returns or private returns, as long as it can repay the loan. Banks are less informed about foreign markets and thus consider it riskier to lend for export projects than for the domestic market. It has been argued that the preferences East Asian governments gave to ex- ports were intended simply to offset the disadvantages of tariffs and other re- strictions on imports. From this perspective, government was not promoting exports but simply "getting the prices right." Upon closer examination (even without a detailed scrutiny of the statistics), this argument appears faulty on two grounds. First, it refers to averages of exports; but what is relevant is the effective subsidy on particular exports. If some exports are encouraged and oth- ers (perhaps unintentionally) discouraged, it is apparent that government has intervened in the allocation of resources. Second, the government actually en- gaged in a wide range of activities beyond direct subsidies to promote exports. Export Promotion Activities Four activities were very important in promoting export growth: the provi- sion of infrastructure; preferential access to capital and foreign exchange; the development of export markets; and licensing and other regulations designed to enhance the reputation of the country's exports. As noted previously, the close, long-term relationships between exporters and governments can be credited with making these mechanisms work. THE PROVISION OF INFRASTRUCTURE. Because poor infrastructure is an impor- tant barrier to trade, East Asian governments have invested in infrastructurc, including good port facilities and improved transportation systems to reduce the costs of shipping goods abroad. Transportation is not the only aspect of infra- structure that has received government attention. Singapore has been involved in efforts to provide an adequate supply of electricity and an effective telecom- munications system, both vital to the country's development as a financial center. PREFERENTIAL ACCESS TO CAPITAL AND FOREIGN EXCHANGE. Most of the coun- tries of East Asia engaged in some degree of financial restraint, that is, capital markets were controlled to give priority industries preferential access to capital and foreign exchange (Stiglitz and Uy 1996). Although in some cases govern- 170 The World Bank Research Observer, vol. 11, no. 2 (August 1996) ments provided subsidies (including lower interest rates) to encourage the ex- pansion of favored industries, most observers believe that the access to credit was far more important. Critics of this access raise the issue of fungibility: what if the government did provide credit and funds for investment in export-oriented industries? So long as money is fungible, large conglomerates could simply divert to other uses those funds that would have been allocated to exports. Consequently, financing ex- ports may have little incremental effect on exports. From this perspective, the allocation of capital to the export sector has no marginal effect. It has only an inframarginal effect on firms that are successful in exporting. This view, how- ever, does not take account of the process by which funds are allocated. If past export performance is used as one of the criteria for judging the creditworthi- ness of the borrower, firms have an incentive to increase exports. And firms that were successful exporters had demonstrated some set of abilities. If those abili- ties were correlated with other abilities that enhanced the likelihood of high marginal returns to investment, then the use of export performance may have been an efficient selection mechanism. DEVELOPING NEW EXPORT MARKETS. Information problems associated with the development of new export markets go beyond the problems of reputa- tion. One noted earlier was the "public good" nature of information. As in the case of other public goods, a strong case can be made for public provi- sion. And many of the East Asian economies have done just that. For in- stance, Singapore's Economic Development Board has actively worked on developing foreign markets and takes an active interest in what goods might be produced for export. Business executives are invited to join official trips abroad to persuade them that it is in their interests to enter into meaningful business relationships overseas. ENHANCING THE REPUTATION OF THE COUNTRY'S EXPORTS. In the 1950s and early 1960s, Japanese products had a reputation for being shoddy. American and European buyers had little information about individual Japanese produc- ers and were likely to make unfavorable inferences concerning any particular product. Because establishing a reputation is expensive for any firm seeking to export (particularly when consumers have strong negative prior beliefs), indi- vidual firms had little incentive to improve the quality of their products. The government conducted a concerted effort both to improve the quality of the products and to establish brand reputations for Japanese firms, so that they would have private incentives to maintain their reputation. Here is an example of an interaction between cooperative behavior and individual incentives. A simi- lar process is occurring in Taiwan (China), where the government is effectively providing subsidies for firms to establish brand recognition. In doing so, firms will have a private incentive to maintain high quality, with positive effects on the reputation of Taiwanese products in general. Joseph E. Stiglitz 171 Conclusion One of the reasons for attempting to delineate what East Asian governments did that resulted in such high growth rates is that other countries would like to replicate their success. If they did the same thing as the governments of East Asia, would they too grow at such rapid rates? To be sure, many countries did similar things, but often with adverse rather than positive effects. They created development banks, only to find that the development banks diverted scarce savings into projects with low returns and made investments that did more to line the pockets of politicians than to raise the welfare of the country. The East Asian miracle had many dimensions: rents were created, but they were used to encourage growth, not dissipated in rent-seeking. Government and businesses cooperated closely, but they collaborated without collusion. Many aspects of this transformation can be explained, and to the extent that they can be ex- plained, it is possible that what they did can be replicated. A high rate of saving leads to high growth; allocating resources on the basis of contests and other performance-based measures can both provide high-powered incentives and re- duce the scope for corruption; egalitarian policies, including active education policies, can contribute to a more stable political and economic environment and lead to faster growth through a more productive labor force. Governments that use markets and help create markets are likely to be more successful in promoting growth than governments that try to replace markets. What generalizations can be drawn from the findings of this article? To be sure, not all of these generalizations are held with the same degree of confi- dence. In some cases, there are alternative interpretations of the events and evi- dence. But a combination of theory and evidence supports these conclusions. Included in the discussion below are several interventions in the financial mar- ket, which, although mentioned only briefly, are amplified in the accompanying article by Stiglitz and Uy in this issue. Because governments in different coun- tries pursued somewhat different policies, not all the statements hold with equal validity in all countries; some may not even hold within all sectors of a given country. These conclusions are organized around six themes. • Making society function better. Economic growth required the maintenance of macroeconomic and political stability. Policies that sustained a more equitable distribution of income-and that supported basic education for women as well as men-contributed to economic progress by encouraging political stability and cooperative behavior within the private sector. The result was a better business climate for investment and more effective use of human resources. * Adaptability of government policies. Government policies adapted to changing economic circumstances, rather than remaining fixed. As the East Asian economies grew more complex, government had less need to assume an active role and found it more difficult to act effectively on a broad scale. 172 The World Bank Research Observer, vol. 11, no. 2 (August 1996) * Government and markets. Governments played an active role in creating market institutions, such as long-term development banks and capital markets to trade bonds and equities, and in establishing an institutional infrastructure that enabled markets to work more effectively. These institutions and markets helped ensure that the high volume of savings was invested efficiently. Governments also used their control of financial markets to help direct resources in ways that stimulated economic growth. This control was probably more important than direct subsidies or low interest rates. Credit was directed not only toward priority areas, but away from speculative real estate and consumer durables. Policies to improve government-business cooperation enabled govern- ments to design programs that served the needs of the business community, created a favorable business climate, and encouraged business to direct its energies in ways that contributed to high social returns. Sharing information enhanced the quality of decisionmaking. By using, directing, and supplementing markets rather than replacing them, the private sector remained the center of economic activity in most of the East Asian countries; when the private sector disagreed with the government, it was permitted to go ahead and risk its own capital. * Promoting accumulation of physical and human capital. The introduction of postal savings institutions and provident funds resulted in higher domestic savings. At the same time, measures that established prudential regulations (and in some cases, entry restrictions) enhanced the safety and soundness of financial institutions and promoted financial deepening. A variety of programs increased the returns to private investment and facilitated the development and transfer of technology; these included policies that promoted education and training, provided infrastructure, and, in most countries, established a receptivity to foreign investment. * Altering the allocation of resources. Governments in East Asia used industrial policies to affect the allocation of resources in ways that would stimulate economic growth. They took an entrepreneurial role in identifying industries in which research and development would have high payoffs. Support for industry, such as the establishment of research and science centers and quality control standards, was important both in attracting foreign investment and in encouraging domestic investors. Emphasizing industries with strong backward and forward links and large externalities may have helped long-term growth. In the short term, the lack of profitability does not provide a good measure of the potential long-run contribution to growth, precisely because it is the discrepancy between private and social returns that motivates government intervention. Governments actively encouraged firms to export. Exports provided a performance-based criterion for allocating credit, encouraged the adoption of international standards, and accelerated the diffusion of technology. Contests among exporters were used widely as incentive devices. The essential Josepb E. Stiglit: 173 ingredients of contests are rewards (here the allocation of credit), rules (measures of performance), and referees (who evaluate performance). In a world short of perfect competition, contests can provide strong incentives with limited risks, and, if the rules are well specified, reduce bureaucratic abuses. Government policies supporting investment. Mild financial repression had a positive effect on economic growth. The effects on national savings and on the efficiency with which scarce capital was allocated were likely positive; positive incentive effects may have been associated with the contest for scarce credit, and the increased equity of firms and banks (because of lower interest rates) enhanced their ability to bear risks. Equally important were other government programs that led to more effective risk-sharing within the economy. Risk-sharing reduced the effective cost of capital, thus stimulating investment. Government intervention in international eco- nomic relations (for instance, in bargaining for foreign technology, in impeding certain capital movements, and in insisting on certain transfers of technology as part of foreign investment) may have enhanced the national interest, promoted economic stability, and enhanced savings. No single policy ensured success, nor did the absence of any single ingredient ensure failure. There was a nexus of policies, varying from country to country, sharing the common themes that we have emphasized: governments intervened actively in the market, but used, complemented, regulated, and indeed created markets, rather than supplanted them. Governments created an environment in which markets could thrive. Governments promoted exports, education, and technology; encouraged cooperation between government and industry and be- tween firms and their workers; and at the same time encouraged competition. The real miracle of East Asia may be political more than economic: why did governments undertake these policies? Why did politicians or bureaucrats not subvert them for their own self-interest? Even here, the East Asian experience has many lessons, particularly the use of incentives and organizational design within the public sector to enhance efficiency and to reduce the likelihood of corruption. The recognition of institutional and individual fallibility gave rise to a flexibility and responsiveness that, in the end, must lie at the root of sustained success. Notes Joseph E. Stiglitz is chairman of President Clinton's Council of Economic Advisers, on leave from Stanford University, where he is professor of economics. This is a shortened version of a paper written as part of the World Bank project on The East Asian Miracle and Public Policv. Financial and technical support of the World Bank is gratefully ac- knowledged. The author is particularly indebted to Marilou Uy. He has also benefited from discussions with Nancy Birdsall, John Page, Richard Sabot, Howard Pack, Edward Campos, Masahiro Okuno, Masahiko Aoki, Daniel Okimoto, Lawrence Lau, Professor Gato, Professor Baba, and dozens of other government officials, academics, bankers, and 174 The World Bank Research Observer, vol. I], no. 2 (August 1996) industrialists who gave generously of their time during this research project. Research assistance from Thomas Hellman is also gratefully acknowledged. 1. In the literature on this subject, particular reference should be made to the work of Alam (1989), Aoki (1988), Wade (1990), Amsden (1989), Okimoto (1989), Lau (1990), Agrawal and others (1992), Johnson (1982), Pack and Westphal (1986), Itoh and others (1984), Komiya, Okuna, and Suzumura (1988), and Vogel (1991), as well as to the coun- try studies of the World Bank. The information theoretic foundation of the analyses pre- sented here is set forth in greater detail in Greenwald and Stiglitz (1986, 1988, 1992), Arnott, Greenwald, and Stiglitz (1993), and Stiglitz (1994b). The implications for govern- ment policy are discussed in greater length in Stiglitz (1990a, 1991a, 1991b). 2. The contrast between India and Singapore could not bring this point home more clearly: India, with a population 300 times that of Singapore and a gross domestic product ten times as large, has a cumulative foreign investment one-fifteenth that of Singapore's. 3. The discussion of this section focuses on standard market failures associated with externalities, missing markets, and competition. The Greenwald-Stiglitz theorems, which go beyond these standard market failures, establish that whenever information is incom- plete, a discrepancy may exist between social and private returns. An important applica- tion of this principle arises in the context of capital markets: the ratio of the private return to the supplier of capital to the social return may differ markedly (even in the absence of the traditional market failures). For instance, private lenders may be able to appropriate a larger fraction of the total returns to real estate lending than to other lending. For a fuller discussion of the implications, see Stiglitz and Uy (1996). 4. This should be contrasted with the United States, where, for instance, in recent years top executives often received 100 times the pay of recent hires. Within rapidly growing areas of China, the degree of inequality is even lower, with managers getting paid approxi- mately three times the amount received by workers. 5. Note that several of the countries went through an import substitution phase, during which they were very successful. It is questioned whether this phase was necessary, whether it helped (or hindered) the growth process, or whether it was primarily a consequence of the particular economic doctrines that were fashionable at the time. 6. The arguments here are not those provided by government officials at the time (or even subsequently). These focused on more immediate concerns: for instance, in the post- war era, with an overvalued foreign exchange rate, Japan was short of foreign exchange. To some extent, it saw export activities as offsetting the disadvantages exporters faced as a result of the overvalued exchange rate. References The word "processed" describes informally reproduced works that may not be commonly available through library systems. Agrawal, P., S. Gokarn, V. Mishra, K. Parikh, and K. Sen. 1992. "Learning from Tigers and Cubs." Discussion paper. Indira Gandhi Institute of Development Research. Bombay. Processed. Alam, M. S. 1989. Governments and Markets in Economic Development Strategies. New York: Praeger Publishing. Amsden, Alice H. 1989. Asia's Next Giant. New York: Oxford University Press. Aoki, Masahiko. 1988. Information, Incentives, and Bargaining in the Japanese Economy. Cambridge, U.K.: Cambridge University Press. Arnott, Richard, Bruce Greenwald, and Joseph E. Stiglitz. 1993. "Information and Eco- nomic Efficiency." NBER Working Paper 4533. National Bureau of Economic Research, Cambridge, Mass. Processed. Arnott, Richard, and Joseph E. Stiglitz. 1991. "Moral Hazard and Non-Market Institu- tions: Dysfunctional Crowding Out or Peer Monitoring." American Economic Review 81 (March): 179-90. Joseph E. Stiglitz 175 Arrow, Kenneth. 1962. "Economic Welfare and the Allocation of Resources for Inven- tion." In National Bureau of Economic Research, The Rate and Direction of Inventive Activity: Economic and Social Factors. Princeton, N.J.: Princeton University Press. Arrow, Kenneth, and Gerard Debreu. 1954. "Existence of Equilibrium for a Competitive Economy." Econometrica 22:265-90. Carroll, Chris, David N. Weil, and Lawrence H. Summers. 1993. "Savings and Growth: A Reinterpretation." Paper presented at the Carnegie-Rochester Public Policy Conference, Bradley Policy Research Center, Rochester, N.Y. April 23-24. Processed. Dasgupta, Partha, and Joseph E. Stiglitz. 1988. "Learning by Doing, Market Structure and Industrial and Trade Policies." Oxford Economic Papers 40:246-68. Greenwald, Bruce, and Joseph E. Stiglitz. 1986. "Externalities in Economies with Imper- fect Information and Incomplete Markets." Quarterly Journal of Economics 101 (May):229-64. . 1988. "Pareto Inefficiency of Market Economies: Search and Efficiency Wage Models." American Economic Review 78(2):351-55. . 1992. "Information, Finance, and Markets: The Architecture of Allocative Mecha- nisms." Journal of Industrial and Corporate Change 1(1):37-63. Itoh, Motoshige, Kazuharo Kiyono, Masahiro Okuno-Fujiwara, and Kotaro Suzumura. 1984. "Economic Analysis of Industrial Policy." Kikan Gendai Keizai (Contemporary Economy). Tokyo. Johnson, Chalmers H. 1982. MITI and the Japanese Miracle. Palo Alto, Calif.: Stanford University Press. Kim, Jung, and Lawrence J. Lau. 1993. "The Sources of Economic Growth of the East Asian Industrialized Countries." Paper presented at a Conference on the Economic De- velopment of Republic of China and the Pacific Rim in 1990 and Beyond, Stanford University, Calif. Processed. Komiya, Ryutaro, Masahiro Okuno, and Kotaro Suzumura, eds. 1988. Industrial Policy of Japan. San Diego: Academic Press. Krugman, Paul. 1994. "The Myth of Asia's Miracle." Foreign Affairs 73(November/De- cember):62-78. Kuznets, Simon. 19S5. "Economic Growth and Income Inequality." American Economic Review 45(1):1-28. Lau, Lawrence J. 1990. Models of Development, A Comparative Study of Economic Growth in South Korea and Taiwan. San Francisco: Institute for Contemporary Studies Press. Murphy, K. M., A. Shleifer, and R. W. Vishny. 1989. "Industrialization and the Big Push." Journal of Political Economy 97(5):1003-26. Nalebuff, Barry, and Joseph E. Stiglitz. 1983. "Prizes and Incentives: Towards a General Theory of Compensation and Competition." BellJournal 14(l):21-43. Okimoto, Daniel. 1989. Between MITI and the Market. Palo Alto, Calif.: Stanford Uni- versity Press. Pack, Howard, and L. E. Westphal. 1986. "Industrial Strategy and Technological Change." Journal of Development Economics 22:87-128. Rosenstein-Rodan, P. N. 1943. "Problems of Industrialization in Eastern and South East- ern Europe." Economic Journal 53(June-September):202-11. Solow, Robert M. 1957. "Technical Change and the Aggregate Production Function." Review of Economics and Statistics 39 (August):312-20. Stiglitz, Joseph E. 1974. "Theories of Discrimination and Economic Policy." In George M. von Furstenberg and others, eds., Patterns of Racial Discrimination. Lexington, Mass.: D.C. Heath. - . 1985. "Credit Markets and Control of Capital." Journal of Money, Banking and Credit 17(1):133-52. 176 The World Bank Research Observer, vol. 11, no. 2 (August 1996) - 1987. "On the Microeconomics of Technical Progress." In Jorge M. Katz, ed., Technology Generation in Latin Anmerican Manufacturing Industries. New York: Macmillan Press. . 1988. "Technological Change, Sunk Costs, and Competition." In Martin Neil Baily and Clifford Winston, eds., Brookings Papers on Economic Activity, 3 (special issue on Microeconomics):883-947. Washington, D.C.: Brookings Institution. - 1989. "Markets, Market Failures, and Development." Amnerican Economic Re- view 79(2):197-203. - 1990a. "On the Economic Role of the State." In A. Heertie, ed., The Economic Role of the State. Oxford, U.K.: Basil Blackwell and Bank Insinger de Beaufort NV. - 1990b. "Peer Monitoring and Credit Markets." World Bank Economic Review 4(3):351-66. - 1991a. "The Invisible Hand and Modern Welfare Economics." In D. Vines and A. Stevenson, eds., Information, Strategy and Public Policy. Oxford, U.K.: Basil Blackwell. . 1991b. "Social Absorption Capability and Innovation." Paper prepared for Ko- rean Development Institute 20th Anniversary Symposium, Seoul. June. Processed. . 1994a. "The Role of the State in Financial Markets." In Michael Bruno and Boris Pleskovic, eds., Proceedings of the World Bank Annual Conference on Development Economics 1993. Washington, D.C.: World Bank. -. 1994b. Whither Socialism? Cambridge, Mass: MIT Press. Stiglitz, Joseph E., and Marilou Uy. 1996. "Financial Markets, Public Policy, and the East Asian Miracle." World Bank Research Observer 11(2):249-76. Stiglitz, Joseph E., and Andrew Weiss. 1983. "Incentive Effects of Termination: Applica- tions to the Credit and Labor Markets." American Economic Review 73(December):912- 27. Vogel, Ezra F. 1991. The Four Little Dragons: The Spread of Industrialization in East Asia. Cambridge, Mass.: Harvard University Press. Wade, Robert. 1990. Governing the Market: Economic Theory and the Role of the Gov- ernment in East Asian Industrialization. Princeton, N.J.: Princeton University Press. World Bank. 1993. The East Asian Miracle: Economic Growth and Public Policy. New York: Oxford University Press. Young, Alwyn. 1993. "Lessons from the East Asian NICs: A Contrarian View." NBER Working Paper 4482. National Bureau of Economic Research, Cambridge, Mass. Pro- cessed. --. 1995. "Growth without Scale Effects." NBER Research Working Paper 4211. National Bureau of Economic Research, Cambridge, Mass. Processed. Joseph E. Stiglitz 177 IN SEARCH OF OWNERS: PRIVATIZATION AND CORPORATE GOVERNANCE IN TRANSITION ECONOMIES Cheryl W. Gray Experiments in privatizing enterprises in transition economies abound, from extensive efforts at sales to strategic owners (as in Estonia and Hungary), to programs based primarily on insider buyouts (as in Russia and Slovenia), to innovative mass privatization programs involving the creation of large and powerful new financial intermediaries (as in the Czech and Slovak republics and Poland). Each approach has inherent strengths and risks. But if the objectives are to sever the links between the state and the enterprises, to school the population in market basics, and to foster further ownership change, the initial weight of evidence seems to favor significant reliance on voucher privatization, especially given the difficulty most countries have find- ing willing cash investors. ocialism's defining characteristic was state ownership of all productive as- sets. Moving from state to private ownership and creating the conditions to improve the performance of medium-size and large enterprises are the main tasks of the transition from socialism. But privatization goes beyond a change in ownership. Programs to privatize enterprises in transition economies should be evaluated in terms of three broad dimensions: the corporate gover- nance mechanisms they create, the supporting institutions they foster, and the extent to which they create a self-sustaining economic and political reform pro- cess. Although some governments judge revenues to be an important goal of privatization, they are at best a secondary objective. The initial patterns of ownership resulting from any program of enter- prise privatization are unlikely to be optimal. They may be too dispersed, or they may be concentrated in the hands of entities unable or unwilling to use such resources efficiently. The long-run success of any privatization program therefore depends on the capacity for ownership patterns to change and evolve The World Bank Research Observer, vol. 11, no. 2 (August 1996), pp. 179-97 0 1996 The International Bank for Reconstruction and Development / THE WORLD BANK 179 to more efficient forms. Programs that spur institutional development, par- ticularly the growth of capital and asset markets, will have a distinct advan- tage in this regard. Goals of Privatization Creating "Real" Owners The primary economic goal of enterprise privatization is to create true repre- sentatives of capital-individuals or groups of individuals who clearly reap the gains from improved performance. Transferring property rights to new owners is just the first step, however, and on its own is not sufficient to guarantee changes in the behavior of managers. The new owners must also have the power, incen- tive, and ability to monitor managers and ensure that they act in the owners' best interests. Typically, in small firms the owners and managers are one and the same; in large firms, however, ownership and management are usually sepa- rate, creating the need for monitoring. Shareholder monitoring is only one of numerous constraints on managerial behavior in market economies, but it is likely to be more important in the early stages of reform in Central and Eastern European economies where markets for products, capital, and managerial talent are underdeveloped and may not exert strong competitive pressures on managers. Shareholder monitoring can be pas- sive or active. Passive shareholders simply sell their shares to discipline manag- ers, while active shareholders exert their views and vote their shares. In coun- tries where stock markets are still in their infancy, passive monitoring is unlikely to be efficient, and thus active shareholder monitoring is likely to be a critical mode of corporate governance in the near term. Furthermore, under socialism, production was so inefficient that major improvements in efficiency are likely to depend more on restructuring than on marginal changes in managerial behav- ior. Alternative patterns of corporate governance should therefore be judged on how they affect not only day-to-day decisionmaking, but also a firm's capacity for radical change and restructuring. Changes in corporate governance are not the only potential benefits that come with real owners. Privatization can also stimulate an infusion of capital, technol- ogy, ideas, and skills, complementing changes in incentives and boosting the productivity of enterprises. Whether these benefits arise depends to a large ex- tent on the technique that is used to privatize and the distribution of ownership that results. Different types of private owners-whether "insiders" or "outsid- ers," individuals or institutions, residents or foreigners-all bring different mix- tures of goals and capabilities to the firms they own. In some cases, the move from public to private may involve intermediate forms of property-neither wholly public nor wholly private-with their own rationality in the particular setting and their own distinct incentive characteristics (Stark 1996). 180 The World Bank Research Observer, vol. I1, no. 2 (August 1996) Developing Supporting Institutions Although many market economies have pursued privatization since the early 1980s, the formerly socialist economies face an especially formidable task. They must not only change ownership, but also establish the institutions of a private market economy. Socialism either crippled or reoriented these institutions to reflect the goals of central planners. Legal frameworks defining property rights, private contract regimes, fiduciary liability, dispute resolution mechanisms, and rules of entry and exit for private firms atrophied. Courts lost much, if not all, of their independence as well as their role as adjudicators of commercial disputes and enforcers of commercial laws. Banks lost their independent monitoring role over firms and became instead passive funnels for channeling state funds. "Watch- dog" institutions that provide critical information for markets to function, such as credit-rating and consumer protection services, accounting and legal profes- sions, and independent journalism, had neither reason nor permission to exist. Finally, socialism inhibited (indeed, often classified as illegal) the development of basic norms and ethics of market conduct and fiduciary responsibility on which so much behavior in advanced market economies implicitly rests. These laws, organizations, professions, and commercial norms must now be rebuilt, sometimes from scratch. Creating a Sustainable Reform Process Transforming property rights and building the institutions of a private mar- ket economy necessarily take time. These reforms must therefore be politically and economically sustainable and mutually reinforcing. Yet the often-profound tension between preexisting patterns of state enterprise control and reformers' desire to ensure a positive economic outcome can complicate privatization. On the one hand, experience shows that the design of a privatization pro- gram must take into account the interests of and distribution of power among existing stakeholders-that is, anyone who has an interest in the enterprise, whether economic or political, including managers, employees, and government bureaucrats. Incentive and efficiency problems were pervasive in all socialist economies, but the distribution of power among stakeholders varied from coun- try to country. Earlier reforms toward "market socialism" in Poland, the former Yugoslavia, and to some extent Hungary gave rank-and-file employees exten- sive powers to influence decisionmaking. In contrast, employees had very little power in East Germany or the Czech and Slovak Federal Republic (the former Czechoslovakia), where control remained firmly in the hands of management, government ministries, and the party bureaucracy until the demise of socialism, after which it shifted to the new democratic leadership. The situations in Bul- garia and the former Soviet Union were somewhere between these two extremes; some influence had devolved to workers, but bureaucrats and managers retained strong powers. Thus the Czech and Slovak Federal Republic and East Germany Cheryl W. Gray 181 could design and effectively implement top-down privatization programs; Po- land, Russia, and Slovenia had no such option. On the other hand, accommodating stakeholder interests clearly has its risks. Compromises that are made to co-opt stakeholders or overcome informational or institutional weaknesses may have negative economic or political repercus- sions down the road and may undermine long-term economic and political sta- bility. That may occur, for example, if newly privatized firms fail to restructure because new owners lack the incentives or skills for effective corporate gover- nance, if the public perceives privatization as corrupt or highly inequitable, or if privatization concentrates economic and political power in the hands of a small domestic or foreign elite rather than an expanding, independent middle class. Recent events illustrate this conflict between what is "doable" and what is optimal. In Russia, for instance, the preferences initially given to managers to garner political support for the program are proving costly. Not only is there limited evidence that managers are restructuring existing (largely insider-owned) firms, but resentment is growing over the concentrated wealth and power that has resulted from privatization. In the Czech Republic, the large state banks were encouraged to set up investment funds as a way to solidify public support for privatization (in part because they, particularly the savings banks, were among the more trusted institutions at the time). Yet this decision may prove counter- productive in the longer run because extensive economic and political power has come to rest in the hands of a few banks and funds, themselves linked to the government through both formal and informal ties. In adapting privatization strategies, certain early steps appear to increase the sustainability of reform in any setting. Countries with legacies of strong bureau- cratic control over domestic industry should move quickly to sever the old links between firms and line ministries-to cut pervasive subsidies, to weaken the ministries' control, and perhaps to abolish branch or sector ministries altogether. Old political links can be cut as well by barring former Communists from gov- ernment service for a period of time, as was done in the Czech and Slovak Fed- eral Republic. This first step should be accompanied by the quick adoption of a privatization strategy and systematic efforts to prevent the wholesale looting of the newly freed state firms before they are privatized. This is likely to require a combination of "carrots" (linking the future well-being of managers to the qual- ity of the assets they deliver to the eventual private owners) and "sticks" (im- posing strong penalties on managers who divert state assets). These actions may be easier to take in the period of "extraordinary politics" immediately following a political break with the past regime (Balcerowicz 1994). After these steps are taken, each stage of the process should ideally create the momentum and incen- tives for further progress. That occurs, for example, if the new owners lobby politicians to design and implement laws (such as corporate and securities regu- lations) that further refine and protect their new rights. Although steady progress is important for momentum, sustainability is not necessarily correlated with speed. Very rapid privatization was pushed 182 The World Bank Research Observer, vol. 11, no. 2 (August 1996) in the former Czech and Slovak Federal Republic and Russia, in large part on the theory that breaking the links with the state was the primary hurdle and that the political window of opportunity had to be seized quickly. Many aspects of these programs are impressive, and they may yet prove to be ma- jor success stories, but the initial design decisions taken to ensure speedy implementation have also produced serious problems, particularly in Russia. Their eventual economic and political impacts are far from clear. In con- trast, slower privatization programs, such as those in Poland and, to a lesser extent, Hungary, run the risk of barely getting off the ground when political receptivity is greatest and thus of stagnating before major progress can be achieved. They could nevertheless prove successful if steady progress con- tinues. The slowest movers, such as Belarus, Bulgaria, Romania, and Ukraine, run the greatest risk. Because transition governments are weak, the failure to commit to a formal privatization program leaves the door open for man- agers to strip assets or direct income flows. The economic injustice of such spontaneous privatization may eventually lead to a political backlash that will undermine further reform. Finally, reforms in property rights must be complemented by supportive reforms in other areas. Fiscal and monetary policies should be refined to foster a stable price system and impose hard budget constraints on firms. One important policy lesson is that any privatization strategy is likely to be fairer and work better where government subsidies are limited, inflation is controlled, and markets exert hard budget constraints on firms. If govern- ments continue to soften budget constraints even for private firms, as they do in many transition economies, the purported benefits of privatization (par- ticularly with regard to incentives) may disappear. On the microeconomic side, the strategy calls for reforms in product markets, more competitive labor markets, and institutional reforms to build effective financial markets. All three are needed to complement shareholder efforts to discipline managers. Methods of Privatization Privatizing large enterprises has proven more difficult than most observers originally envisioned. Not only are the goals complex and sometimes at odds with each other, but the firms are often ill-suited to the needs of a market economy. Many are overstaffed and inefficient. Reflecting socialism's efforts to make enterprises the providers of social assets as well as income, many are vast conglomerates with housing, medical services, and child care facilities. Because central planners wanted to economize on transaction costs, many are monopo- lies. Table 1 summarizes the various methods used to privatize medium and large firms and estimates the extent of privatization in seven countries under each method. What lessons have emerged to date? Cheryl W. Gray 183 Table 1. Privatization Techniques for Medium-Size and Large Enterprises, 1995 (percent) Sale to Management- Equal-access outside employee voucher Still in Country owners buyout privatization Restitution Other' state hands Czech Republic By numberb 32 0 22c 9 28 10 By valued 5 0 50 2 3 40 Estonia' By number 64 30 0 0 2 4 By value 60 12 3 10 0 15 Hungary By number 38 7 0 0 33 22 By value 40 2 0 4 12 42 Lithuania By number <1 5 70 0 0 25 By value <1 5 60 0 0 35 Mongoliaf By number 0 0 70 0 0 30 By value 0 0 55 0 0 45 Poland By number 3 14 6 0 23 54 Russiac By number 0 55 11 0 0 34 Note: Boldface numbers show the dominant method in each country. Data are as of the end of 1995. a. Includes transfers to municipalities or social insurance organizations, debt-equity swaps, and sales through insolvency proceedings. b. Percentage shares of the number of all formerly state-owned firms, including parts of firms restruc- tured before privatization. c. Includes assets sold for cash as part of the voucher privatization program through June 1994. d. Percentage shares of the value of all formerly state-owned firms. Data for Poland and Russia are unavailable. e. Does not include some infrastructure firms. All management buyouts were part of competitive, open tenders. In thirteen cases citizens could exchange vouchers for minority shares in firms sold to a core investor. f. Mongolia has used only voucher privatization to privatize medium and large enterprises. Every enterprise was sold for vouchers, first in a limited closed subscription to insiders and subsequently to outsiders via the stock exchange (Korsun and Murrell 1994). Source: Blasi (1994); Korsun and Murrell (1994); Kotrba and Svejnar (1994); World Bank data. Sales to Outside Investors Before the transition process got under way in earnest, most countries of Central and Eastern Europe that wanted to privatize state enterprises sought to sell them as going concerns. They were following the only known experience at the time, most notably Great Britain and Chile, where privatization through individual sales had been successful. Because capital markets were undeveloped in the transition economies, most countries hoped to sell the bulk of state enter- prises directly to large outside investors, generally strategic investors with special- 184 The World Bank Research Observer, vol. 11, no. 2 (August 1996) ized knowledge of the industry. Such trade sales were perceived to have three advantages: they would bring in revenue; they would result in real owners who had the knowledge and incentives to govern companies efficiently and the capi- tal to restructure them; and the conditions of the sale could theoretically be manipulated to take special needs into account. Although these advantages have indeed been evident in some cases, sales to outside investors have proven far more difficult than originally anticipated. Such sales can work when market institutions are in place, but they are problematic when such institutions are in their infancy. East Germany successfully priva- tized virtually all of its state enterprises through sales to outside investors, but only with massive amounts of political will and technical and financial assis- tance from West Germany. Among the transition countries, only Estonia and Hungary have managed to privatize the bulk of their state enterprises through direct sales. Poland and Romania pursued sales vigorously in their early efforts at privatization but with limited success. THE DISADVANTAGES-LESSONS OF EXPERIENCE. What are the disadvantages of this approach? First, the process is hampered by the limited amount of private capital (particularly domestic capital) available and by the poor quality of infor- mation about the condition of the enterprises. One option, which was followed widely in Hungary, is to sell to foreign investors who have sufficient capital and are willing to incur risks or to invest in information-gathering that might de- crease such risks. This somewhat controversial strategy may nonetheless be nec- essary if trade sales are to succeed, and even then, of course, there will be many state enterprises that overseas investors have no interest in buying. Foreign in- terests have tended to concentrate in certain sectors, such as automobiles, food processing, tobacco, and certain consumer products, whose international mar- ket structure tends to be dominated by large, oligopolistic firms (Kogut 1996). A second option is to require less capital up front, so that owners can pay in installments, perhaps out of the future earnings of the firm. Variants of this approach have been tried, for example, in Estonia, Hungary, and Poland. Enfor- cing future payments is often difficult, however, and renationalizing firms if such payments are not made is equally problematic. A second disadvantage is that both the process and the resulting distribution of ownership rights may be perceived as unfair. Not only are ordinary citizens unable to participate, but individual deals often look arbitrary, if not corrupt. This perception has been most notable in dealings with foreign investors, with whom packages of incentives and legal regulations have often been negotiated case by case. Third, the approach tends to be costly and slow, due to the sheer magnitude of the job of evaluating and negotiating deals for each company one-by-one and of providing follow-up monitoring to ensure that the buyers fulfill contract pro- visions. Sales have also been slowed by other uncertainties, such as who is re- sponsible for cleaning up past environmental damage and how to compensate Cheryl W. Gray 185 those who owned the property before the socialist government came to power (Rutledge 1995). Fourth, the process is complicated and often stymied completely by the difficulty of placing a value on firms to be offered for sale. Accounting stan- dards inherited from socialism were inadequate to determine the historical value of a firm-much less net present value, on which sales price should theoretically be based. Furthermore, price and trade reforms quickly reduce-if not elimi- nate-the relevance of previous experience. Potential buyers face profound un- certainty about what these companies will look like in the future. What prod- ucts will they produce, and for what markets? In what quantities and at what costs? What financing will be available? At what interest rate? Given these un- certainties, calculating even the rough value of an investment is virtually impos- sible in many cases, even with reforms in accounting techniques. Finally, insiders have used either explicit or implicit power to block sales to outsiders, particularly in countries such as Poland and Russia in which decisionmaking power had been decentralized. Furthermore, the strength of the insiders' incentives to block a sale is likely to be correlated with the potential profitability of the firm itself, and thus it may be harder to sell the better firms- exactly those for which there is likely to be greater demand from outside inves- tors. These disadvantages have been more debilitating than initially expected. The Treuhandanstalt in East Germany was able to privatize (or liquidate) its 8,500 state enterprises relatively quickly but at an enormous cost in terms of both skilled manpower and explicit or implicit subsidies to buyers (von Thadden 1994). Other countries, which lacked a benefactor of West Germany's economic strength, could move only slowly-or adopt radically different divestiture tech- niques. In five years (1990-94) Hungary was able to transfer only about one- third of its state-owned assets to private hands through formal sales programs (Pistor and Turkewitz 1996); large sales in infrastructure and energy firms late in 1995 pushed this total somewhat higher. With extensive assistance from former Treuhandanstalt officials, Estonia sold most of its 350 enterprises in three years (1992-95). These are the "successful" cases. None of the other countries of Central and Eastern Europe or the former Soviet Union have even come close to these achievements (in part because foreign investors have shown less interest). Overall experience in the region has led observers to conclude that sales, while a useful pillar in the privatization process, cannot be the sole or even the primary instrument in transition economies. PRIVATIZING THROUGH EQUITY OFFERINGS. Firms can also be sold to the public by floating shares on public stock exchanges, but this approach is limited by the capacity of the infant stock exchanges in transition economies. Furthermore, it tends to work only for the very best firms with good financial prospects and strong reputations. It is not an avenue for restructuring, not only because poorly performing firms are unlikely to be listed successfully, but also because the dis- 186 The World Bank Research Observer, vol. 11, no. 2 (August 1996) persed ownership structure that results is unlikely to create incentives for own- ers to make the necessary changes. Poland has perhaps had the most success with this approach but still has privatized only about two dozen firms in this manner. Initial public offerings are clearly not the answer to the need for rapid and large-scale privatization, but on the margin they can help develop capital markets. Management-Employee Buyouts A second important method of privatization is the sale or giveaway of all or part of the company to its managers and employees. Croatia, Georgia, Poland, Romania, Russia, and Slovenia have all relied primarily on management- employee buyouts, sometimes using government-issued vouchers to provide the liquidity for insiders (and a few outsiders) to purchase shares. To speed up privatization, Hungary implemented a similar small program in 1993 to supple- ment its emphasis on trade sales. Although the governments of Lithuania and Mongolia did not originally set out in this direction, many firms divested through voucher privatization programs became, in effect, management-employee buyouts (Korsun and Murrell 1994). An important advantage of this approach is its feasibility and political popu- larity. In countries in which insiders already had extensive power, those insiders have generally been able to retain their influence during the transition period and effectively maintain a veto power over privatization decisions. In some coun- tries this veto power is explicit; in Poland, for example, employee approval is required for a privatization plan to go forward. In most countries, however, such veto power is implicit. To garner political support, governments have often offered generous benefits to insiders. A second potential advantage is the one stressed by most proponents of em- ployee share-owning plans in advanced market economies. Insider ownership can be both more equitable and, under certain conditions, more efficient (Hansmann 1990; Earle and Estrin 1996; Shleifer and Vasiliev 1996). It can be more equitable because it rewards those who do the work-ironically, the argu- ment at the very heart of socialism. It can be more efficient because it aligns the incentives of owners and workers. Managers and employees may be willing to work harder, monitor each other more carefully, and push for greater produc- tivity if they reap the residual gains. THE DISADVANTAGES IN TRANSITION ENVIRONMENTS. These potential advan- tages are counterbalanced by serious disadvantages that are particularly acute in transition settings. First, the process may be inequitable, as employees, par- ticularly those in good firms, reap most of the benefits. One study, for example, calculated that the 19 percent of Russians employed in privatized firms obtained 56 percent of equity sold through June 1994, while the 81 percent of Russians who had only vouchers ended up with only 15 percent of the equity (Blasi 1996). Cheryl W. Gray 187 Second, giving preferences to insiders inhibits-and may even eliminate- competition in the privatization process. Insiders are generally unable to bring new skills and new capital to the company, and socialist managers may have few of the skills needed in a market economy. Because potential outside owners who may be more qualified are not given the chance to participate, the resulting ownership pattern is likely to be suboptimal for the economy as a whole, at least initially. Research in Central Europe confirms that firms privatized to insiders are less likely to restructure and invest than firms sold to outsiders (Earle and others 1994; Barberis and others 1995). Furthermore, changes in ownership patterns may be blocked if managers try to stop employees from selling their shares. Even if employees are free to sell, there may be few buyers. If corporate law and disclosure rules are underdeveloped and thus provide little protection for outside shareholders, as is true in virtually all transition countries (and in some advanced market economies), outsiders may be unwilling to invest in firms with significant insider ownership. Russia's voucher privatization program, completed in mid-1994, resulted in insider ownership of about two-thirds of the shares of privatized companies. Some managers have tried, albeit illegally, to prohibit workers from selling their shares to outsiders. Others have used less transparent means to block participa- tion by employees or outsiders. For example, managers may attempt to change the form of the company from joint stock to limited liability, because the latter restrains sales of shares to outsiders. Alternatively, they may try to convince employees to put their shares in a trust and assign their voting rights to the managers of the firm. Even when the rights of workers to vote their shares are not restricted, managers may-and do-convince workers that incumbent man- agement is on their side but that outsiders will fire them if they are allowed in. Finally, managers may try to get around employee ownership altogether by set- ting up new firms and using their inside information and power to transfer valu- able assets to these firms. If such a pattern is repeated on a wide scale, this form of ownership may inhibit rather than reinforce the development of a private market economy. It may also backfire politically, if the fruits of privatization become more and more concentrated in the hands of the few, unleashing grow- ing resentment among those ostensibly included at the beginning but ultimately cheated of their expected gains. ADDRESSING THE PROBLEMS WITH BUYOUTS. How can the advantages of man- agement-employee buyouts be enhanced while their disadvantages are mitigated? First, governments need to cut subsidies, liberalize prices, and institute trade reforms to force insider-owned firms to abide by the rules of the marketplace. These steps will help to ensure efficiency regardless of ownership. Second, government should encourage the entry of new firms to increase com- petition. Management-employee buyouts may work well for smaller manufac- turing and service firms in sectors that are attracting new domestic entrepre- neurs; product market competition will keep the insider-owned firms on their 188 The World Bank Research Observer, vol. 11, no. 2 (August 1996) toes. Foreign competition could potentially do the same for larger firms, but in such cases managers are more likely to turn to the state to block such competi- tion or to obtain support of one kind or another. Third, governments should adopt policies that encourage share trading and thereby develop markets for corporate control. In advanced market economies, insider (particularly worker) ownership has an inherent tendency to "degener- ate" into investor ownership over time (Earle and Estrin 1996). Whether the same tendency exists in transition environments has yet to be seen. Ownership change requires both a supply of and a demand for shares. To create a supply, shares must be immediately tradable without limitation. To create a demand, outside investors must have not only sufficient capital, but also basic informa- tion and protection against fraud and abuse by insiders. These in turn require well-functioning corporate laws, securities regulations, and accounting systems. Such shareholder protections do not arise in a vacuum but go hand-in-hand with other economic reforms (Gray and Hendley 1995). In sum, management-employee buyouts excel in their capacity to adapt to the implicit or explicit demands of existing stakeholders, but are less effective in creating corporate governance mechanisms or in artracting new capital and skills. For firms that cannot survive without restructuring, the conflicts of interest that confront insiders may prove particularly unwieldy. In such cases, insiders may look to the state for help and, given political pressures, the state may be willing to listen. This route may thus work better for viable firms that can generate internal funds for investment and may be suitable for small firms without politi- cal clout. Indeed, in the case of the latter, employees may be more willing to take painful wage cuts to preserve the company (Earle and Estrin 1996). In the case of large distressed firms with major capital needs, however, this strategy is unlikely to generate the resources, incentives, and capabilities necessary to un- dertake large-scale change. These pros and cons are particularly apt for firms in which insiders have majority interest. There are, in contrast, strong advantages and relatively few disadvantages to giving insiders minority ownership rights. One clear advantage is that privatization is more likely to be perceived as fair if workers share in any upside gains. Another is that employee-owners can monitor managers or out- side owners (such as investment funds or foreign owners with significant minor- ity stakes) who may otherwise have an incentive to loot the firms or stifle com- petition with other firms under their ownership. These advantages are important, considering the political fragility and the general weakness of watchdog institu- tions in virtually all transition environments. Equal-Access Voucher Privatization A third approach used widely in transition countries is voucher, or "mass," privatization. The government gives away or sells low-priced vouchers that can then be used to purchase shares in companies, thereby eliminating the problem Cheryl W. Gray 189 at the core of the sales approach-the shortage of domestic capital. So-called equal-access voucher programs embrace outsiders as well as insiders. This form of privatization has been-or soon will be-implemented in Armenia, Bulgaria, the Czech Republic, Kazakhstan, Lithuania, Moldova, Poland, Romania, Slovakia, and Ukraine. Well-designed voucher privatization can overcome many of the problems encountered with the various sales techniques, notably the perceived unfairness, the shortage of domestic capital, and the difficulty of placing a value on assets. Because the voucher approach can proceed rapidly, it can simultaneously stimu- late the development of market institutions, create new owners, and reorient the interests of existing ones toward further reform. Furthermore, it can speedily cut the links between enterprises and the state that both inhibit restructuring and put fiscal pressures on the state. The main concern when this method of privatization was first proposed, apart from the question of revenue, was its questionable capacity to develop real own- ers with proper incentives for effective corporate governance and with access to new capital and skills for restructuring. The concern over corporate governance arose in part from the very notion of vouchers, that is, that one did not value what one did not pay for. More fundamental, however, was the fear that the resulting distribution of ownership would be inefficient and would interfere with the development of strong ownership interests. Experience has shown, however, that a wide variety of ownership patterns can result from voucher privatization. Perhaps more important, such initiatives, if well-designed, can stimulate the development of capital markets and stock market trading, thus fostering further ownership change and speeding up the development of corporate control. It can, in effect, privatize the privatization process. COMBINING SALES AND VOUCHERS. All transition economies have chosen to follow several privatization routes simultaneously, albeit with different empha- ses. The earliest, biggest, and most successful equal-access voucher program to date has been that of the Czech Republic, which has transferred the bulk (in value) of its state enterprises through this route. Mongolia used vouchers to privatize 70 percent of its large enterprises. Romania's 1991 privatization pro- gram was much smaller; only 30 percent of the shares of eligible firms were involved. The intention, unrealized and replaced in 1995 by a second and larger mass privatization attempt, was to transfer the remaining shares to strategic owners who could effectively govern and restructure the enterprise. Poland's recent mass privatization plan is smaller still, covering only about 500 compa- nies, or fewer than one-tenth of state-owned enterprises. Larger programs have certain advantages, in that they can include both more firms and a greater diversity of firms. To ensure value to participants and thus gain more political support, while at the same time divesting firms that might not attract cash offers, some of the most profitable firms should be included in the program, along with some of the more marginal ones. Perhaps more impor- 190 The World Bank Research Observer, vol. 11, no. 2 (August 1996) tant, larger programs can achieve a greater degree of privatization in a shorter period of time. DECIDING WHETHER AND HOW A FIRM WILL PARTICIPATE. When a voucher privatization program is announced, who decides whether and in what form a particular firm will participate? As with the size of the program more generally, this decision evolves in large part from the balance of political interests and powers in the particular country. The government of the former Czech and Slo- vak Federal Republic chose which firms would participate, and the Czech gov- ernment continued to apply this principle after the country split up. In each case, the central government decided on the mix between voucher auctions and other forms of transfer (primarily sales to strategic investors and restitution to former owners) but based that decision on bids submitted by competing bidders and prepared by them with little government involvement. Thus, the process of designing privatization programs for individual firms was decentralized in a com- petitive framework, but the final decision process was controlled at the top. This approach, attractive both economically and practically, appears to have worked well in a politically centralized environment where there were no strong inside stakeholders. Poland and Romania (in its 1991 program) both attempted to follow a more centralized approach by giving the government broad powers to decide which firms would participate and how they would participate. Although this strategy was feasible in Romania, because of the country's strong tradition of centralized power, it was contrary to Poland's diffuse power structure. Indeed, managers and employees of Polish firms have maintained effective veto power over the choice of privatization method. POOLING OWNERSHIP INTERESTS. If the ownership of shares in a particular en- terprise were as widely disbursed as the ownership of vouchers, there would be little corporate governance and probably little progress in reforming enterprises. For this reason many mass privatization programs have encouraged the creation of intermediary institutions to pool ownership interests in particular enterprises. The former Czech and Slovak Federal Republic allowed free entry of private mutual funds, and more than 420 participated in the first wave of privatization. These funds competed with each other to acquire vouchers from the public in exchange for shares in the fund. The funds then invested the acquired vouchers in shares of firms being privatized at auctions. This approach, based on free entry and competition, had the advantage of reducing the state's direct control over the process. In contrast, the Romanian (1991) and Polish plans called for the government to establish a set number of investment intermediaries, staffed by managers chosen by the supervisory boards appointed by the government. The shares of the inter- mediaries were then distributed to citizens. No auctions were held. The govern- ments hoped that the intermediaries would actively restructure the firms in their Ckeryl W. Gray 191 portfolios and then sell their interests to core investors. That objective has merit, but the danger is that the intermediaries may not be subject to direct market pressures and could end up essentially as government-protected state holding companies. In 1991 Romania's state ownership fund was allocated 70 percent of the shares of each commercial company to be privatized and was directed to divest 10 percent of its holdings a year. After four years it had divested almost nothing and had barely begun the necessary restructuring. Poland's approach did not get off the ground until 1995 (after a delay of three years), and it is too early to judge results. Although the free entry and competition among funds in the Czech Republic may be arguably preferable to the more bureaucratic approach used in Romania and Poland, creating truly private funds with market-based incentives is proving extremely difficult in any transition setting. The perennial question "who moni- tors the monitors?" looms over every experiment. This challenge is difficult enough in advanced market economies. It is even more problematic in transition environments, where norms of disclosure and fiduciary responsibility are weak, and watchdog institutions and oversight mechanisms are in their infancy. Breaking the links with the state, although desirable to stimulate entrepreneurship and risk taking, also may mean weakening the capacity to monitor the monitors. In the Czech Republic, as noted, most of the largest funds were founded by, and are still connected with, the large banks through asset management con- tracts. These banks in turn continue to be closely connected with the govern- ment, both through the sizable stake owned by the state-run National Property Fund and through the government's regulatory powers. Although some non- bank funds quickly established their independence and their potential influence over managers, the bank-affiliated funds appear to be less independent and en- trepreneurial. The latter may also face a conflict of interest because the banks lend to the same firms that their funds own. On the other hand, these larger funds may be more secure investments than the more entrepreneurial funds, which might have an incentive to loot an enterprise or take other actions at the expense of fund or enterprise shareholders. To the extent that several funds own shares in a firm, they have an incentive to monitor each other (barring collusion among the funds themselves). Although limits on ownership by individual funds tend to discourage active involvement by these owners, the potential for cross- monitoring is one advantage of such limits. The Russian privatization program favored insiders but also allowed the free entry of private investment funds. Although some 600 funds were formed, they were kept much smaller than the Czech funds, and thus they have far less power and influence. In the Russian environment, with no legal safeguards, less macro- economic discipline, and strong insider control, the goals of the funds are far from clear. Their small size may lead to complex coalitions among or between them and other actors in the economy. Some funds appear to have been estab- lished primarily for trading vouchers, while others are allied with the managers of individual firms, and still others seem to be involved in seeking subsidies from 192 The World Bank Research Observer, vol. I1, no. 2 (August 1996) government. Only a minority appear interested in owning and improving the performance of enterprises in the economy (Frydman, Pistor, and Rapaczynski 1996). Intermediary institutions bring several advantages to voucher privatization programs. At a minimum they aggregate the power of individual vouchers and thereby exercise some monitoring functions associated with ownership. In addi- tion, the free entry of independent intermediaries increases private participation and stimulates competition in the market for corporate control. Finally, observ- ers hope that the funds will become the cornerstones (together with banks or even in place of them) for developing the financial infrastructure that is essential in market economies. But achieving these goals is not easy or automatic. Gov- ernments need to consider how they might regulate funds to prevent self-dealing and encourage responsible fiduciary behavior. The involvement of foreign finan- cial experts as fund managers and advisors-one advantage of Poland's approach-can help to strengthen the expertise and norms of conduct within funds. How DO CITIZENS USE VOUCHERS? In the Czech and Slovak Federal Republic, vouchers could be invested either in specific companies or in investment funds. In the Romanian (1991) and Polish programs, in contrast, investing directly in firms was not an option. (In Mongolia and in Romania's newer program, invest- ing in funds is not an option!) In Estonia citizens could use their vouchers to acquire shares in firms or to purchase their homes or land (although relatively few shares were in the end offered for vouchers). There seem to be no obvious costs-and significant benefits-to allowing wide latitude to investors. Options create competition that can spur funds to greater effectiveness, and they force citizens to become actively involved in voucher investments. In addition, options allow investors to tailor their choices to their own personal risk preferences. Although some people prefer direct in- vestments, funds have proven to be more popular investment vehicles than first expected. When the Czech and Slovak Federal Republic program was set up, most vouchers were expected to be invested directly in firms, but 72 percent of vouchers were ultimately invested in funds. Furthermore, citizens need not be limited to investing their vouchers. Trading them is also a viable option, and such trading may encourage the emergence of strong, interested owners. If trading is not permitted, immediate rights to trade the shares that are acquired with vouchers is a close substitute. Most of the voucher schemes to date have given some latitude to citizens to sell their inter- ests, whether vouchers or acquired shares. Russia allowed trading in vouchers from the beginning. The former Czech and Slovak Federal Republic forbade secondary trading by citizens in vouchers (although the prohibition was not strictly enforced) but encouraged trading in acquired shares. Such trading has developed rapidly through the Prague and other stock exchanges in the country and through off-exchange transactions. Cheryl W. Gray 193 A somewhat surprising development has been the concentrated ownership and cross-ownership that has emerged from voucher privatization in the Czech Republic. Not only is ownership concentrated in a few funds, but individual funds often own shares of directly competing firms. Furthermore, the funds, together with affiliated banks, are locked in an intricate web of cross-ownership (or sometimes self-ownership) as a result of the privatization of the banks through vouchers (Coffee 1996). Thus banks are insulated from competitive pressures, and the government continues to influence the economy through its 40 percent (or greater) residual holdings of shares in privatized banks. ORGANIZING AUCTIONS. Most voucher schemes rely on auctions to allocate enterprise shares. Voucher auctions can be organized either simultaneously or sequentially (Boycko, Shleifer, and Vishny 1994). The Czech and Slovak Federal Republic (in the first wave) and the Czech Republic (in the second wave) fol- lowed a simultaneous approach. The Bulgarian scheme proposed for 1996 is also simultaneous. Other countries, including Georgia and Russia, have generally followed a sequential approach. From an economic perspective the Czech model is more efficient, because all options are known to all bidders at the time of the auction, and the value of a voucher (in terms of purchasing power) does not vary over time, as it does in the sequential model. The simultaneous model, however, is also more complex and costly and may be infeasible in a larger country. Shares in an auction can also be allocated two ways. One is simply to divide the shares on a pro-rata basis among bidders, based on the number of vouchers issued. The second is to match the bids against some independent measure of value and distribute the shares only when bids and offers match. The Czech approach, which was a modified version of the latter, required several rounds of bidding to equate demand and supply. The result was arguably fairer but perhaps feasible only because of the relatively small size of the country, the relatively strong central control of the government, and the relatively sophisticated level of understanding in the government and the citizenry. The sale was also facilitated by the country's more stable macroeconomic situation, which meant that infla- tion was moderate and thus the valuations of firms were more meaningful. RESIDUAL STATE OWNERSHIP. Finally, voucher privatization schemes vary in the degree of residual ownership maintained by the state. Romania, for example, privatized only a 30 percent share in each enterprise (in 1991); indeed, some observers question whether this was really privatization at all. The Czech Re- public and Poland left significant minority stakes in the hands of government property funds, with a view to using these stakes later to attract strategic inves- tors (or otherwise influence events). The Polish government also had the initial power to appoint the managers and supervisory boards of the funds. If the state is to maintain a stake in firms after privatization, its share should be small and temporary, and its stance relatively passive, although it should con- 194 The World Bank Research Ohserver, vol. 11, no. 2 (August 1996) tinue to monitor the firm to prevent fraud and asset-stripping. Extensive residual state ownership and control can lead to conflicts of interest that diminish-or even nullify-the positive effects of privatization (Pistor and Turkewitz 1996). Conclusions Experiments in privatization abound, from extensive efforts to sell to strate- gic owners to programs based primarily on insider buyouts to innovative pro- grams of mass privatization. These efforts are often complemented by extensive restitution to the former owners of the nationalized property and by smaller programs of bank-led debt-equity conversions or public offerings of shares on newly emerging stock markets. Each of these approaches has inherent advan- tages and risks, and in essence the jury is still out as to which will prove best in the longer run. At present, however, if the objectives are to sever the links be- tween the state and the enterprise, to school the population in market basics, and to foster a perception of fairness, the weight of initial evidence appears to favor equal-access voucher privatization, particularly given the difficulty most countries face in finding willing cash investors. Competing and well-monitored intermediary funds are an essential component of this approach. Experience shows that formal privatization programs are only part of the picture-and often only a small part, although they have received most of the attention. Firms are breaking apart and consolidating again from state to pri- vate ownership or from one private firm to another (Bogetic and Hillman 1995; Stark 1996). As one Hungarian observer noted, this is the period of "primitive capital accumulation" in the post-socialist world. Although formal programs may lay important ground rules, the tremendous economic, legal, political, and even moral uncertainty profoundly affect-and may even overwhelm-most formal efforts at privatization, and it is beyond our ability or insight to know what the final results will be. Both the economic outcomes of these various paths and the efforts to assess them are just beginning to yield insights, and it will be years, if not generations, before a definitive story can be told. Notes Cheryl W. Gray is principal economist in the Policy Research Department of the World Bank. The author would like to thank Roman Frydman, Alan Gelb, John Nellis, Lucia Swiatkowski Cannon, and Martha De Melo for comments on earlier drafts. References The word "processed" describes informally reproduced works that may not be commonly available through library systems. Cheryl W. Gray 195 Balcerowicz, Leszek. 1994. "Poland." In John Williamson, ed., The Political Economy of Policy Reform. Washington, D.C.: Institute of International Economics. Barberis, Nicholas, Maxim Boycko, Andrei Shleifer, and Ira Lieberman. 1995. "How Does Privatization Work? Evidence from the Russian Shops." NBER Working Paper 5136. National Bureau for Economic Research, Cambridge, Mass. May. Blasi, Joseph. 1994. "Russian Privatization: Ownership, Governance, and Restructuring." In Ira Lieberman and John Nellis, eds., Russia: Creating Private Enterprises and Effi- cient Markets. Washington, D.C.: World Bank. . 1996. "Russian Enterprises after Privatization." Paper presented at the Allied So- cial Sciences Association meeting in San Francisco, January. Processed. Bogetic, Zeljko, and Arye L. Hillman, eds. 1995. Financing Government in the Transi- tion-Bulgaria: The Political Economy of Tax Policies, Tax Bases, and Tax Evasion. World Bank Regional and Sectoral Study. Washington, D.C. Boycko, Maxim, Andrei Shleifer, and Robert W. Vishny. 1994. "Voucher Privatization." Journal of Financial Economics 35(April):249-66. Coffee, John. 1996. "Investment Privatization Funds: The Czech Experience." In Roman Frydman, Cheryl Gray, and Andrzej Rapaczynski, eds., Corporate Governance in Cen- tral Europe and Russia. London: Central European University Press. Earle, John, and Saul Estrin. 1996. "Worker Ownership in Transition." In Roman Frydman, Cheryl Gray, and Andrzej Rapaczynski, eds., Corporate Governance in Central Europe and Russia. London: Central European University Press. Earle, John, Roman Frydman, Andrzej Rapaczynski, and Joel Turkewitz. 1994. Small Privatization. London: Central European University Press. Frydman, Roman, Katharina Pistor, and Andrzej Rapaczynski. 1996. "Investing in In- sider-Dominated Firms: A Study of Russian Voucher Privatization Funds." In Roman Frydman, Cheryl Gray, and Andrzej Rapaczynski, eds., Corporate Governance in Cen- tral Europe and Russia. London: Central European University Press. Gray, Cheryl, and Kathryn Hendley. 1995. "Developing Commercial Law in Transition Economies: Examples from Hungary and Russia." World Bank, Policy Research De- partment, Washington, D.C. Processed. Hansmann, Henry. 1990. "When Does Worker Ownership Work? ESOPs, Law Firms, Codetermination, and Economic Democracy." Yale Law Journal 99(8):1751-1816. Kogut, Bruce. 1996. "Direct Investment, Experimentation, and Corporate Governance in Transition Economies." In Roman Frydman, Cheryl Gray, and Andrzej Rapaczynski, eds., Corporate Governance in Central Europe and Russia. London: Central European University Press. Korsun, Georges, and Peter Murrell. 1994. "Ownership and Governance on the Morning After: The Initial Results of Privatization in Mongolia." Working Paper 95. Depart- ment of Economics, Center for Institutional Reform and the Informal Sector, University of Maryland, College Park. Processed. Kotrba, Josef, and Jan Svejnar. 1994. "Rapid and Multifaceted Privatization: Experience of the Czech and Slovak Republics." MOCT-MOST: Economic Policy in Transitional Economies (Netherlands) 4(2):147-85. Pistor, Katharina, and Joel Turkewitz. 1996. "Coping with Hydra-State Ownership after Privatization: A Comparative Study of Hungary, Russia, and the Czech Republic." In Roman Frydman, Cheryl Gray, and Andrzej Rapaczynski, eds., Corporate Governance in Central Europe and Russia. London: Central European University Press. Rutledge, Susan. 1995. "Selling State Companies to Strategic Investors: Trade Sale Privatizations in Poland, Hungary, the Czech Republic, and the Slovak Republic." CFS Discussion Paper Series 106. World Bank, Resource Mobilization and Confinancing, Information Services Center, Washington, D.C. Processed. 196 The World Bank Research Observer, vol. 11, no. 2 (August 1996) Shleifer, Andrei, and Dmitry Vasiliev. 1996. "Management Ownership and the Russian Privatization." In Roman Frydman, Cheryl Gray, and Andrzej Rapaczynski, eds., Cor- porate Governance in Central Europe and Russia. London: Central European Univer- sity Press. Stark, David. 1996. "Recombinant Property in East European Capitalism." In Roman Frydman, Cheryl Gray, and Andrzej Rapaczynski, eds., Corporate Governance in Cen- tral Europe and Russia, London: Central European University Press. von Thadden, Ernst-Ludwig. 1994. "Centralized Decentralization: Corporate Governance in East German Economic Transition." Institut fur Volkswirtschaft, Universitat Basel, Switzerland. Cheryl W. Gray 197 HOW WELL CAN METHOD SUBSTITUTE FOR DATA? FIVE EXPERIMENTS IN POVERTY ANALYSIS Martin Ravallion No one doubts that good data are essential to sound policymaking. Alas, data are invariably faulty. Methodological solutions to data inadequacies have often been proposed and implemented, but they have been tested only rarely. Yet the methods that are used may well determine the direction of policy. For example, the particular survey method used-and the way nonsurvey data are interpreted-may be critical in assessing whether a country's strategy for reducing poverty is working. This article shows how counterfactual experiments can help test the reliability of various methods of dealing with common data problems. Well-designed methods-and they need not be very complicated-can help get around the problem, although it appears that substituting method for data is a long way from being perfect. O bjective data obtained from representative surveys of living conditions are widely used to stimulate public awareness of poverty and motivate government actions to benefit the poor. Yet analysts and policymakers routinely find that these data are deficient in one or more important respects and must find credible methods for dealing with those deficiencies. Various so- lutions have been proposed that rely on certain regularities in living conditions and use a relatively small number of more easily measured variables-such as membership in certain predefined socioeconomic groups-to infer the missing data. Hidden differences in living standards may well confound such efforts. Even though the partitions commonly used in assessing poverty-such as land owner- ship or region of residence-reveal large disparities, these disparities may be weak indicators for targeting the poor. In fact, some recent research indicates that variations between socioeconomic groups are often dwarfed by differences within such groups (see, for example, Datt and Ravallion 1993; Ravallion and The World Bank Research Observer, vol. 11, no. 2 (August 1996), pp. 199-221 CD 1996 The International Bank for Reconstructon and Development / THE WORLD BANK 199 Sen 1994; and Cowell and Jenkins 1994). So targeting policies that use standard "poverty proxies" may well result in considerable leakage to the nonpoor and highly imperfect coverage of the poor. Surprisingly little effort has gone into assessing the performance of routine tools of poverty analysis. Yet the methods that are chosen to deal with inad- equacies in the data may well influence the conclusions analysts draw, including the implications for policy. For example, in research on Indonesia, it has been found that the method used to establish the poverty lines by region can dramati- cally alter the structure of the resulting regional poverty profile and hence the regional priorities for alleviating poverty; indeed, regional rankings produced by two common methods were virtually uncorrelated (Ravallion and Bidani 1994). Here the method chosen to make up for missing data would markedly influence the precise policy prescription. This article evaluates some of these techniques by applying them in situations where the data are available and asking how well they work. Fortunately, not all countries are missing the same data, so it is possible to infer how a particular remedial method would work in a "data-poor" setting by comparing how well that same method performs in a "data-rich" setting. Knowing about the strengths and weaknesses of the many methods that have been used to generate informa- tion in data-poor settings can also help researchers make choices in collecting new data. Nationally representative socioeconomic surveys can be expensive; good substitutes, if they can be found, could produce savings. Evaluating Methodology There are three generic problems with existing poverty data. First, the spe- cific surveys used are sometimes flawed; there may be problems with the ques- tions asked or the sample selected. Second, it is often the case that too few surveys were conducted or they are incomparable over time. Third, researchers often lack complementary data that are needed to interpret the results; for ex- ample, making consistent welfare comparisons from survey data can be difficult if data on prices are not also available. Ingenious-and sometimes simple-methods of dealing with these data prob- lems can often be devised, although how well they work is unknown. One way to test how well the analyst can circumvent these data problems is to devise a counterfactual experiment. The objective is to try the method in a setting where it is not in fact needed-because the required data are already available-and see how well it performs. The following section outlines five such experiments that test how well methodology can substitute for data. Several observations about these experiments are in order. First, their aim is not to develop "quick and dirty" methods; in some cases the econometric analy- sis that is used is quite sophisticated. Second, these experiments share the ge- neric problems of any empirical research. That is, the results are to some extent 200 The World Bank Research Observer, vol. 171, no. 2 (August 1996) data-specific, and their robustness in other settings is not known. And measure- ment errors may be a problem, as always. The following experiments try to assess how well the selected methods perform against data that meet the pre- vailing standards of best practice but are unlikely to be perfect. Third, even without measurement errors, reasonable people may disagree about what should be measured. Here the measure of poverty is defined as a lack of command over market goods. Although this measurement is undeniably impor- tant, it is clearly not the only dimension of well-being; command over nonmarket goods, such as some publicly provided services, may be an important omission in conventional measures (for further discussion, see Ravallion 1993). Nor do the experiments reported here constitute a comprehensive list of data problems. For example, Chaudhuri and Ravallion (1994) show the difficulties of inferring standards of living through the use of single cross-sectional surveys; and Haddad and Kanbur (1990) study the inequality within households that is often hidden in conventional surveys. Experiment 1: A Poverty Profile with Missing Prices Policy discussions about poverty are often informed by a poverty profile that shows how poverty measures vary by location, sector of employment, or some other household characteristic. Unfortunately, deficient data and methods often misinform such discussions. One of the most common-and potentially seri- ous-problems is how to make consistent comparisons of what a given unit of money is worth to poor people in one place versus another. Ideally, studies that compare incomes or expenditures over time or in differ- ent regions should adjust for differences in prices. Almost all statistical agencies now monitor prices over time in constructing the Consumer Price Index, but monitoring differences in prices by localities at a given time is far less common. This data gap can be a serious concern in developing countries where the prices of the same goods and services differ widely within the country. Without local price data, one might be tempted to ignore these differences in calculating whether people are poor. How would this omission affect the results? Is an alternative method available that would provide the missing information? Indonesia is an interesting case study for examining these questions, given its geography and likely locational price differentials. For this experiment, research- ers set out to measure the extent of poverty by region and sector of employment, with and without adjusting for differences in the cost of basic needs. Indonesia's Central Bureau of Statistics provided the raw survey and price data. Using the method described in Ravallion and Bidani (1994), the nominal poverty line was adjusted for regional differences in the cost of a bundle of basic foods consid- ered sufficient to meet minimum food-energy (caloric) requirements for good health and normal activities. An allowance for differences in nonfood spending was added as well, consistent with the spending behavior of the households involved. The results were then compared with the costs of the same items based Martin Ravallion 201 on average national prices without adjusting for regional differences in the cost of living. The results are given in figure 1, which ranks all provinces (each one split into urban and rural areas) by the head-count index (that is, the percentage of people who are living below the poverty line) based on nominal expenditures, and then plots the results against corresponding estimates with an adjustment for differ- ences in the cost of basic consumption needs. So, if the ranking of regions does not change when adjusting for the cost of living, the lines are everywhere posi- tively sloped from left to right; if there is also agreement on the cardinal values of the poverty measures, the lines increase linearly on a 45-degree line (indi- cated by the straight line in figure 1). It can be seen from figure 1 that both the poorest urban area (point A) and the poorest rural area (point B) are correctly identified. But in several cases, adjustments for differences in the cost of living lead to a change in ranking. For example, the second poorest rural area after adjusting for cost-of-living differ- ences (point C) is only the eleventh poorest without adjustment. The rank corre- lation between the head-count index using local poverty lines (with an adjust- Figure 1. Effect of Cost-of-Living Differences on Indonesia's Poverty Profile by Province Adjusted percentage of population below poverty line 60 50 . . Urban Rural 40- 30- 20 10 f11 ° 10 20 30 40 50 60 Unadjusted percentage of population below poverty line Noteo Each point is one province (urban or rural). The rank correlation between the adjusted and unadjuLsted figures is 0.88. Point A is the poorest ur-ban area; point B is the poorest rural area; point C is the second-poorest rural area after adjustment for cost-of-living differences. Source: Author's calculations using Indonesia's National Socioeconomic Sur-ey for 1990. 202 The World Bank Research Observer, vol. 11, no. 2 (August 1996) ment for prices) and that using the national mean poverty line (that is, based on national mean prices) is 0.88. If the experiment is repeated using sector of employment instead of location (urban or rural), price adjustment has much less effect because prices do not vary by employment (figure 2). Also note that for the sectoral profile, if differ- ences in urban and rural costs of living can be incorporated correctly, the results with and without the adjustment for other regional differences show consider- able agreement. It is interesting to compare these results with another method that is com- monly used to compensate for missing price data when setting poverty lines, the "food-energy-intake method." Indonesia and many other countries use this method, which defines the poverty line in each region as the level of expenditure at which a predetermined level of food-energy intake is typically met in that region. This technique has the advantage that it does not require any data on prices. But because people in richer regions or sectors tend to have more ex- pensive tastes, and so spend more to reach any given caloric intake, the method can yield differences in poverty lines far in excess of the cost-of-living differences of the poor (Ravallion 1993). Ravallion and Bidani (1994) compared the food- energy-intake method with a price-adjusted poverty profile for Indonesia and Figure 2. Effect of Cost-of-Living Differences on Indonesia's Poverty Profile by Employment Adjusted percentage of population below poverty line 40 35- 30 - 25- 20 '5- 0 5 10 15 20 25 30 35 40 Unadjusted percentage of population below poverty line Arote: Each point is a subgroup of households defined by principal sector of employment. Source: Author's calculations using Indonesia's National Socioeconomic Survey for 1990. Martin Ravallion 203 found that the differences in poverty lines between urban and rural sectors im- plied by the food-energy-intake method were large enough to cause a reversal in all poverty measures between urban and rural areas and numerous reversals in the ranking of provinces. After adjusting for differences in the cost of basic consumption needs, Ravallion and Bidani found a greater incidence, depth, and severity of poverty in rural areas than in urban areas. The food-energy-intake method indicated the reverse, however, with more poverty in urban areas. The ranking of regions (with each province divided into urban and rural areas) by the two methods was very different. These inconsistencies arise because food-energy intake is determined by many other factors besides real incomes, such as relative prices, tastes, and activity levels, that can influence the poverty lines obtained by the food-energy-intake method in ways that have nothing to do with differences in purchasing power over basic consumption needs. Group A may be able to afford more of all basic needs than Group B, and yet, because of the differences in the way the two groups allocate their money (say, group A buys more expensive calories because they taste better or are more convenient), the food-energy-intake method can find that group A is poorer (Ravallion 1993). The method solves the problem of missing price data, but at a high cost in terms of the usefulness of the resulting poverty profile for informing policy decisions. Some methods of dealing with missing price data work far better than others. And in some cases the best practice turns out to be the simplest. In the ex- periment above, not adjusting at all for regional differences in the cost of living gave better predictions of which regions or sectors were poorest than did the more sophisticated food-energy-intake method. Even without data on prices by location, one can often make a reasonable guess as to how much more it would cost the rural poor to achieve the same standard of living in urban areas and to build this calculation into the poverty profile. This assumption can be based on a quick price survey that compares a few typical urban and rural areas. Experiment 2: Monitoring Poverty Despite Gaps and Lags in Survey Data There are often long lags between surveys, making it difficult to track a country's performance in fighting poverty. So developing a set of easily moni- tored, aggregate indicators of poverty that could serve as an alternative would be useful. An experiment in India compared the information provided by a set of aggregate indicators with survey data to see whether the findings from the two techniques were consistent. This experiment used data supplied by India's National Sample Survey Orga- nization, which has provided tabulations of the distribution of consumption per person since the mid-1950s. The surveys have been carried out at varying fre- quencies ranging from less than a year to more than five years. (For example, data were not collected between mid-1978 and 1983.) The consumption mea- 204 The World Bank Research Observer, vol. 11, no. 2 (August 1996) sures are comprehensive, and the surveys appear to follow sound and consistent practices. The National Sample Survey (NSS) has been the main instrument for monitoring poverty in India, and a large body of literature uses these data (re- viewed by Ravallion and Datt 1995). Analysts have had to rely almost exclu- sively on the survey agency's own tabulations of the raw data, which typically are not published until two to three years after the survey. Only rarely have the raw data been released, and in those cases the lags between collection and re- lease appear to have been even longer. Given these lags and the gaps between survey rounds, the question is whether poverty in India can be monitored using more readily available leading indica- tors from other sources. The answer would also help determine how frequently survey data should be collected. Some simple forecasting experiments based on NSS data can try to answer the question. The analysis focuses solely on the rural poor, whose income is derived from their own farms (which often do not produce enough for their own food needs) and from wages received from working on other farms. Thus two potential leading indicators for monitoring poverty are the agricultural wage rate and agricultural output per acre. Both are typically available well in ad- vance of tabulations from the NSS data. Poverty measures, such as the head-count index, show that poverty increased in rural India during two periods-the early 1970s and the late 1980s (Ravallion and Datt 1995). How well could data on average wages and farm output have predicted these setbacks for the poor? The forecasting model used here to an- swer that question is based on a model of rural poverty outlined in detail in Ravallion and Datt (1995). The salient features of the model for the current discussion are summarized in box 1. The model tracks the data well, but good within-sample performance need not mean good post-sample forecasts. To estimate the accuracy of this method- ology as a predictor, two types of forecasts were considered. * Type 1: This experiment entailed making forecasts one year ahead, assuming that the lag in the availability of NSS tabulations on the distribution of per capita consumption is one year longer than the lag in the availability of data on the real agricultural wage rate and agricultural output per acre. The extra lead time for these two variables would make it possible to predict poverty measures before the NSS data are released. * Type 2: This experiment comprises a sequence of dynamic forecasts, again assuming that real wage and agricultural output data are available but that the NSS data are only available at longer intervals; one must thus rely partly on past forecasts of the poverty rate. (The previous year's poverty measure is one of the variables used to forecast the next year's poverty measure; see box 1.) Table 1 gives the results of three forecasting exercises for the head-count in- dex. (The results of forecasting two other poverty indexes-the poverty gap and Martin Ravallion 205 Box 1. Tlhe Forecasting Model A separate model was estimated for each of three poverty measures; the head-count index, showing the percentage of the population below the poverty line; the poverty gap index, reflecting the average distance below the poverty line; and the squared poverty-gap index, in which those distances are squared to make the measure sensitive to changes in inequality among the poor (Foster, Greer, and Thorbecke 1984). The model was dynamic, in that the most recent past poverty measure was used as one of the predictors. Consumption-based poverty measures may not adjust instantaneously to current variables, so this feature is important. A nonlinear regression method was used to deal with the uneven spacing of the surveys over time (Ravallion and Datt 1995). The lagged poverty measure was a persistently significant predictor of the current measure. The other two predictors were the real agricultural wage rate (deflated by the Consumer Price Index for Agricultural Laborers), and average agricultural output per acre in both the current and the last survey year. A time trend was also included to pick up the effects of any time-trended omitted variables. Although other variables were tried but had no additional explanatory power, it is clear that there are omitted variables. For example, from the data available it was not possible to construct a series for employment on rural public works schemes (which have historically been important in famine relief). Furthermore, lags and other limitations of existing public finance data appear to entail at least as large a lag as India's National Sample Survey (NSS) data, thus diminishing their usefulness as leading indicators. These forecasts could probably be improved, but it would entail considerable extra effort. The model was first fitted to data for a core sample comprising all previous NSS rounds from 1958-59 until the date beyond which the forecasts were to be made. This entailed re- estimating the model for each experiment. Within-sample performance of the models was good; R2 was 0.98 or higher when using the full sample of twenty surveys, and R2 for the subsamples used for forecasting was in the range 0.90-0.98. the squared poverty gap (see box 1)-were not very different and are not re- ported.) The results for type 1, in panel A, show forecasts of the poverty measures for 1987-88 (using a model based on the data up to and including 1986-87); for 1988-89 (using the data up to 1987-88); and for 1989-90 (using the data up to 1988-89). The percentage forecasting error ranges from a 7 percent underesti- mation to a 2.5 percent overestimation. The model failed to forecast the (small) increase in the head-count index between 1986-87 and 1987-88, predicting a drop instead. Panel B gives type 2 forecasts for the same three years based on the same model as the 198 7-88 forecast in panel A (using the seventeen survey rounds up to 1986-87). From 1987-88, new forecasts are drawn from the previously fore- cast poverty measure rather than from the actual measure. (This is the correct way to assess forecasting performance when there are gaps of more than a year between survey rounds.) Naturally, in this case, past forecasting errors are built into future forecasts, so a sizable drift of the forecasts from the actual values is possible relative to year-ahead forecasts. The first year in panel B is, of course, 206 The World Bank Research Observer, vol. 11, no. 2 (August 1996) Table 1. Forecasting Poverty in Rural India One year ahead Three years ahead The mid-1970s reversal (Tyvpe 1, Panel A) (Type 2, Panel B) (Type 2, Panel C) Year Percent Percent Percent Year (Panels A and B) Actual Forecast error Actual Forecast error Actial Forecast error (Panel C) 1986-87 34.19 n.a. n.a. 34.19 n.a. n.a. 55.21 n.a. n.a. 1970-71 1987-88 34.99 33.50 -4.3 34.99 33.50 -4.3 55.52 59.52 7.2 1972-73 1988-89 34.63 32.07 -7.4 34.63 30.99 -10.5 55.72 63.13 13.3 1973-74 1989-90 29.65 30.38 2.5 29.65 27.22 -8.2 49.29 59.50 20.7 1977-78 n.a. Not applicable. Note: The table gives both actual (survey-based) and forecasted (model-based) measures of the head-count index, the percentage of people living below the poverty line. Source: Author's calculations. 0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ the same as in panel A. The second year is similar. But the third year drifts considerably farther; the underestimation of poverty in 1988-89 magnifies the downward bias in the 1989-90 forecast. We know that the living standards of the poor deteriorated in the early 1970s as a result of a drought in western India. How well does the model forecast this event? Real agricultural wages and average farm yields fell in the early to mid- 1970s. (See panel C for type 2 forecasts from the model using data up to 1970- 71.) The model predicts the increase in poverty to 1973-74, but it overestimates the effect by a wide margin. Poverty did not increase nearly as much as the aggregate data on wage rates and yields suggest. The model correctly predicts the end of the crisis, showing a decline in poverty by the late 1970s. By the next survey round, however, it has drifted off track by a wide margin. In summary, even highly aggregated data on agricultural wages and yields contain information that is relevant to predicting poverty outcomes before new household survey data are available or in lieu of them. But these are by no means perfect indicators. Although poverty measures can be forecast with toler- able precision one year ahead of the release of the survey data, quite sizable drift can arise after just a year or two. In one case a model using these indicators fails to predict an increase in poverty, while in the other it does predict it, but sug- gests far worse outcomes than the eventual survey data. Possibly the forecast could be improved by adding more variables, although the number of surveys available and the lags and gaps in the availability of other data will no doubt constrain such efforts in practice. It is likely that methods such as these are cheaper than conducting new surveys, but this saving must be weighed against the loss of accuracy in monitoring poverty more than a year ahead. Experiment 3: Rapid Appraisal Methods In a comprehensive socioeconomic survey, the bulk of the interview time is devoted to various consumption and income components. For example, a typi- cal survey for the World Bank's Living Standards Measurement Study contains many pages of questions on consumption and income sources (Ainsworth and van der Gaag 1988; also see Demery and others 1992 on a survey instrument developed for African countries). Furthermore, calculating consumption and (particularly) income from the raw data is not easy. Some analysts argue that there are far better ways to assess living standards-and a wide range of op- tions, including those described by Grootaert and Marchant (1991), Kumar (1993), Chambers (1994), and Narayan and Srinivasan (1994, ch. 20). During the last ten years or so, a technique known as "rapid-appraisal" has emerged as a popular alternative to more comprehensive surveys based on sta- tistical sampling. The idea is to use very short interviews and direct observation to obtain a set of objective and subjective indicators of welfare. Some of the rapid appraisal methods include focus group and community interviews as well as less formal individual or household-level interviews (see Kumar 1993). Here 208 The World Bank Research Observer, vol. 11, no. 2 (August 1996) I focus solely on rapid household-level surveys, including subjective questions. Some exponents are convinced that rapid appraisal can be just as accurate as a full-blown survey (Chambers 1994)-an assertion that is difficult to verify un- less both methods are applied to the same sample. Such an approach is rarely taken, however. For this experiment, some typical rapid-appraisal questions were added to a high-quality conventional survey: the 1993 Jamaica Survey of Living Condi- tions, collected by the Statistical Institute of Jamaica. The survey already in- cluded many objective questions commonly covered in rapid-appraisal inter- views; to these were added subjective questions on whether respondents considered the family's consumption of various categories of goods (food, hous- ing, clothing, transport, health care, and schooling) to be "less than adequate," "more than adequate," or "just adequate." These questions, which required less than one page of the questionnaire and roughly five minutes of extra time, pro- duced a response rate of virtually 100 percent. Forty-four percent of sampled households responded that their food consumption was "less than adequate for their family's needs," although the proportion rises to about two-thirds for the poorest quintile. A similar pattern in the differences between responses by "poor" and "rich" is evident from the answers to most of the other subjective questions (table 2). How well can these subjective indicators predict a standard objective mea- sure of consumption expenditure derived from the same survey? A regression model can be used to predict consumption per person, based on answers to subjective questions-several of which turn out to be highly significant predic- Table 2. Suimmary of Answers to tbe Suibjective Welfare Questionzs for Jamaica (percent) All Poorest Richest Answver respondentts qoiintile qznintile Food is inadequate 44.2 67.1 19.1 Housing is inadequate 48.7 71.7 26.2 Clothing is inadequate 40.6 64.1 16.3 Transportation is inadequate 51.5 65.4 34.0 Health care is inadequate 45.0 62.3 23.0 Schooling is inadequate 51.8 58.2 52.1 Food and housing are inadequate 32.5 57.2 10.4 Food, housing, and clothing are inadequate 25.7 48.9 6.8 Food, housing, clothing, and transport are inadequate 20.0 40.8 4.9 Food, housing, clothing, transport, and health care are inadequate 18.4 40.2 3.6 Food, housing, clothing, transport, health care, and schooling are inadequate 13.7 32.4 1.4 Note: Survey respondents were ranked according to household expenditure per person. Souirce: Author's calculations based on the primary survey data from the 1993 Jamaica Survey of Living Conditions, collected by the Statistical Institute of Jamaica. Martin Ravallion 209 tors of consumption. The combined explanatory power of these subjective indi- cators is still rather modest, however; they could only explain 20 percent of the variance in log consumption per person. Can the model's explanatory power be increased by adding some easily moni- tored objective indicators? For instance, the interviewer can readily observe cer- tain characteristics of the dwelling and can ask simple "yes or no" questions about it and about the respondent's possessions. From the rest of the Jamaican survey, I picked twenty-five short questions that could probably be answered in ten to fifteen minutes, including: * Information supplied from the interviewer's own observations, such as location (Kingston, other urban, rural); type of dwelling (detached, semi- detached, apartment); and material of outer wall (wood, stone, brick) * Questions, probably answerable by any adult household member, including number of people in the household; level of schooling completed by the head of the household (primary, secondary, tertiary); number of rooms in the dwelling; whether it is owned or rented; whether the household owns or shares a toilet; whether it owns or shares a kitchen; source of drinking water (inside tap, outside tap, public standpipe, well); and ownership of various consumer durables (such as telephone, television, washing machine, refrigerator). Adding this set of easily measured objective indicators to the regression model for predicting consumption considerably improves the total explanatory power; the additional indicators explain 57 percent of the variance in consumption. To test the performance of these rapid-appraisal indicators further, I reesti- mated the model on a smaller sample and used it to predict the consumption of the rest of the original sample. Such an experiment is closer to a real-world setting in which relationships revealed by a small rapid-appraisal survey of a few easily monitored indicators are used to make predictions about a far larger population. (Some actual or proposed methods of targeting transfers aimed at reducing poverty have worked this way.) For this experiment, the augmented model described above (combining subjective and short-answer objective ques- tions) was reestimated on a random subsample comprised of every fourth house- hold in the full sample. This model was then used to predict consumption per person in each household in the reserve sample, which included the remaining three-quarters of the original sample (1,270 households). Table 3 compares the actual and predicted values for households grouped by quintiles of consumption per person. (So, for example, 69 of the 254 households in the poorest quintile in terms of actual consumption were predicted by the model to be in the second poorest quintile.) The model predicted 46 percent of the households in the poorest quintile to be in other quintiles, although it pre- dicted 81 percent to be in the poorest two quintiles (54 percent correctly in the poorest quintile plus 27 percent incorrectly in the second poorest). If one splits all the quintiles into deciles (not shown here), one finds that in 60 percent of 210 The World Bank Research Observer, vol. 11, no. 2 (August 1996) Table 3. Joint Distribution of Actual and Predicted Values (number of households) Predicted consumption per person Actual Poorest Second Third Fourth Richest quintile qzeintile quintile quiintile quintile quiintile Total Poorest 137 69 36 11 1 254 (54) (27) (14) (4) (0) (100) Second 66 83 60 35 10 254 (26) (33) (24) (14) (4) (100) Third 32 61 70 66 25 254 (13) (24) (28) (26) (10) (100) Fourth 8 31 60 86 69 254 (3) (12) (24) (34) (27) (100) Richest 1 10 28 61 154 254 (0) (4) (11) (24) (61) (100) Note: There are 1,270 households in the sample; 254 in each quintile. Figures in parentheses are percentages of the total for that quintile. Source: Author's calculations based on the primary survey data from the 1993 Jamaica Survey of Living Conditions, collected by the Statistical Institute of Jamaica. cases the model incorrectly predicts the decile of those actually in the poorest decile, although only 25 percent are predicted to be in the fourth and fifth deciles. Among the middle quintiles (see, for example, the third quintile in table 3), the extent of misclassification is sizable, as indicated by the relatively large entries above and below the diagonal. The model does best for the richest quintile, predicting 61 percent of those households accurately. When the experiment was repeated without the subjective welfare questions, the method's performance deteriorated, but not by much. Most of the work is being done by the rapid-appraisal objective indicators. There are undoubtedly other subjective or objective proxies that are easily surveyed and that could add to the explanatory power of this method. Against this possibility are two reasons to believe that these experiments have probably overstated the performance of rapid-appraisal methods. First, these experiments show what the best result would be in predicting consumption from such indica- tors; the weights on the various rapid-appraisal indicators have been chosen optimally (in the sense of minimum sample error variance in predicting con- sumption). In practice, analysts typically do not have a complete integrated sur- vey on which to estimate optimal parameters. In such cases they would have to select a set of weights by some subjective means. With little else to go on, the set of weights chosen for this purpose may be far from ideal. Second, rapid-appraisal strategies often do not use rigorous sampling meth- ods; indeed, many practitioners explicitly reject such methods (see Casley's 1993 comments, however). This means that sampling biases will add to the impreci- sion of the rapid-appraisal method in practice. Martin Ravallion 21 1 Overall, these experiments suggest that a rapid-appraisal survey lasting, say, twenty minutes on a relatively small sample can yield some significant predic- tors of the more expensive objective surveys. A significant share of the variance of consumption will almost certainly be left unexplained, however, and the pre- dictions from such methods are unlikely to be highly correlated with actual stan- dards of living. As with other short-cut methods described above, these limita- tions to the accuracy of rapid-appraisal methods will have to be weighed against the potential cost savings. Experiment 4: Decomposing Aggregate Indicators without an Integrated Survey Many socioeconomic indicators are available only in a highly aggregated form, that is, by provinces or countries. Yet one would like to know how they vary between socioeconomic groups, such as the poor and the nonpoor, or between urban and rural residents. This calculation would be easy if one had a survey that collected all the relevant information for the same households. Is it possible to estimate the decomposition reasonably well without such a survey? Bidani and Ravallion (forthcoming) propose a method that can be used when the distribution of the population across the relevant subgroups (such as urban and rural) is known. An econometric model is estimated in which the observed aggregate indicator (such as the infant mortality rate) across countries or regions is modeled as a function of the population distribution by subgroups (such as the proportion of poor). Using this approach, the infant mortality rate of poor and nonpoor households can be estimated when all that is known is the overall av- erage infant mortality rate for each country and the proportion of people who are poor. Bidani and Ravallion apply the method to find out how health indica- tors differ between the poor and the nonpoor. They find that the average life expectancy of the poorest two-thirds of the population in their sample of thirty- five countries is nine years less than that of the nonpoor and that their children are 50 percent more likely to die before their first birthday. Prescott and Jamison (1985), who use a similar technique to examine mortality rates and health-ser- vice availability in China, found that rural residents had worse health and used health care less often than did urban residents. (The Prescott-Jamison method is not used here; for further discussion, see Bidani and Ravallion forthcoming.) As with the earlier experiments, the most convincing way to test for bias in this method is through a counterfactual experiment. For this experiment, the accuracy of the urban-rural breakdown of provincial socioeconomic aggregates was tested. Many statistical agencies publish socioeconomic data at the level of the local admin- istrative jurisdictions, but they rarely break down these figures into the categories one would like, such as urban and rural. Indeed, it may be administratively difficult to do so. Yet based on data in those countries that provide such information, large differences in socioeconomic outcomes, as well as in access to and use of public services, are known to exist between urban and rural areas. 212 The World Bank Research Observer, vol. 11, no. 2 (August 1996) How accurately can the urban-rural decomposition of socioeconomic aggre- gates be estimated? Five experiments were carried out for this article; four used data from provincial Indonesia, and one used data from the state level in India. In all cases, the aggregate provincial indicator, the actual urban and rural values of the indicator, and the shares of the urban and rural populations in the prov- ince or state were known. The aggregate indicators were first related by a simple regression to the share of the population in urban and rural areas (using the "random coefficients" estimators described in Bidani and Ravallion forthcoming). The estimated co- efficients represent the average urban and rural indicators across all the prov- inces. Table 4 gives the results for the actual and estimated decompositions of the means in each of the five cases. In only one case-the percentage of well- nourished children in Indonesia-are the estimates very close to the actual means. In the remaining cases, the rural mean is underestimated and the urban mean is overestimated. Clearly, one of the problems with this test is that relevant factors that influ- ence the socioeconomic indicator may have been omitted, thereby altering the model's estimates of the effect of the included variables. I therefore revised the model to account for some of these factors, as shown in table 4, with significant improvements in the es;ifnated relationships. For food expenditure in Indone- sia, the augmented regression used total consumption expenditure and average household size. For mean expenditure in Indonesia, the augmented regression used mean income. For income in Indonesia, the augmented regression used mean total expenditure. For the percent below a nutrition-based poverty line in India, the augmented regression used mean expenditure and the Gini index of expenditure. In the augmented models, I have deliberately aimed by trial and error to obtain the best estimates of the subgroup means using these explana- tory variables from available data sets. In practice-when the true means are, of course, unknown-one would be unlikely to do as well as these estimates. This experiment is an example in which the simplest methodology is likely to produce the least reliable results. The estimated urban and rural means are sub- stantially closer to the actual means under the augmented regressions than un- der the simpler unaugmented regressions (table 4). Clearly, the urban-rural decompositions can be heavily biased unless the model is expanded to include a potentially wide range of other relevant variables. If augmentation can be done, the decomposition method is capable of yielding reasonable substitutes for the missing data. Experiment 5: Assessing the Prevalence of Poverty witbout a Survey This experiment was designed to show how accurately aggregate economic and social indicators estimate the prevalence of poverty. For this study, the pro- portion of people below the poverty line in each of twenty-two countries (esti- mated from household survey data) was compared with forecasts of the same Martin Ravallion 213 Table 4. Actual and Estimated Average Indicators for Rural and Urban Areas Regressing on population Regressions augmented to Actual shares alone include other variables Item Rural Urban Rural Urban R2 Rural Urban Adjusted R2 1. Percentage of Indonesian children 46.16 59.46 46.31 57.62 0.515 n.a. n.a. n.a. under five deemed to be well-nourished in 1987 2. Per capita food expenditure in 17,690 22,735 15,770 29,277 0.285 18,086 22,184 0.902 Indonesia (in rupiah/person/month) 3. Mean expenditure in Indonesia 25,379 41,167 21,491 54,155 0.162 24,789 43,710 0.963 (in rupiah/person/month by province) 4. Income in Indonesia 18,922 31,453 16,375 39,912 0.507 18,777 31,770 0.986 (in rupiah/person/month) 5. Percentage of population below 28.42 24.93 26.25 32.67 0.002 27.96 26.42 0.973 nutrition-based poverty line in India (by state, 1983) n.a. Not applicable. Note: Random coefficient regression methods were used; see Bidani and Ravallion (forthcoming) for details on the estimation method. Source: For item 1, BPS (1989); for items 2 and 3, authors' calculations from the National Socioeconomic Survey of Indonesia, 1990; for item 4, authors' calculations from the National Socioeconomic Survey of Indonesia, 1984, with data on population proportions based on BPS (1989); for item 5, Datt and Ravallion (1993). poverty measure in each country derived only from aggregate data. The poverty line was set at $1 a person a day (converted into local currencies to assure pur- chasing power parity in 1985). Using the methodology documented in Ravallion, Datt, and van de Walle (1991), regressions were estimated on each of the twenty- two possible samples, obtained by leaving out one country at a time. These are counterfactual experiments, meaning that the poverty measure for the particu- lar country is assumed not to be known, although it is known for each of the other twenty-one countries. The predictor variables were private consumption per capita from the national accounts evaluated at purchasing power parity and at official exchange rates, the level of urbanization, the infant mortality rate, life expectancy at birth, and the proportion of women in the labor force. Some of these variables are of interest in their own right; infant mortality, for ex- ample, can be an important indicator of well-being (such as access to public health care), but here I am concerned only with the variable's usefulness in pre- dicting consumption poverty. Figure 3 shows that the predictions from the aggregate data are highly posi- tively correlated with the survey-based estimates; the simple correlation coeffi- cient is 0.87. Nonetheless, the absolute errors in predicting the prevalence of Figure 3. Actual and Predicted Poverty Measures across Countries Percentage of population consuming less than US$1 per day 70 60 5° . ....Survey estimate -o-Predicted 40- 30- 20- 10 0 2 4 6 8 10 12 14 16 18 20 22 Country (ranked by survey-based estimate) Note: The predicted measure for each observation is based on a regression that excludes that observation. Source: Author's calculations from primary data for twenty-two countries (Ravallion, Datt, and van de Walle 1991). Martin Ravallion 215 poverty are often quite large. Indeed, the average absolute error as a percentage of the original survey estimate was 49 percent. The poorest few countries were correctly identified, although there was considerable reranking. For example, the aggregate indicators suggested that the eleventh poorest country (based on the survey estimates) was almost the least poor. For the country with the largest absolute error, the survey estimate of the head-count index was 57 percent, while the aggregate economic and social indicators predicted a figure of 39 per- cent. These discrepancies arise from two factors. First, the extent of inequality varies from country to country, and this variation is hard to pick up without distributional data from a survey. Second, the relation between social indicators and consumption-based poverty measures differs from country to country; some countries where a large share of the population is poor provide effective public health care and thus have good social indicators, such as low infant mortality, while others do not. These differences can make it hard to assess the extent of poverty without a household survey. It appears that the readily available eco- nomic and social aggregates can give, at best, a rough idea of the prevalence of poverty. Conclusions Much effort goes into making up for inadequate data on poverty; the tools used range from various rapid-appraisal methods to sophisticated econometric models. But in using such methods, how do we know that the cure is any better than the disease? Or how do we decide which cure works best? Surprisingly little effort has gone into rigorous testing of these methods. New approaches- aiming to implement some new methodological paradigm or just to improve an existing survey-are routinely introduced without prior testing. Little more than blind faith guides the policymaker's interpretation of results. In many ways the experiments reported here represent a limited investiga- tion. It is not feasible to span all the methodological choices that matter, al- though these experiments may be suggestive in other applications. The results may not be robust to measurement error or applicable in other settings; there are still precious few opportunities for tests of this sort. And there are gains and losses that I have not been able to quantify. The imprecision introduced by im- perfect data and methods may appreciably lower the expected benefits from efforts at targeting the poor. Nor have I addressed the costs of data collection, which must be weighed against the benefits of greater accuracy in measure- ment. Nonetheless, the results do suggest that there are some relatively easily moni- tored proxies for poverty that can help identify and monitor poverty when sur- vey data are unavailable or inadequate. These proxies rely on certain plausible theoretical or common sense relationships that are expected to hold between 216 The World Bank Research Observer, vol. 11, no. 2 (August 1996) conventional survey-based poverty data and more aggregated data from other sources (or from simpler rapid-appraisal methods). The experiments reported here suggest that data inadequacies can be at least partially surmounted by care- ful use of such methods (table 5). Even with rather poor data, the situation facing a creative analyst is rarely hopeless. The results also point to the limitations of these methods. Omitted variables may well account for such a large share of the variation in standards of living that the method tends to be a poor substitute for better data. Even a rather narrowly defined objective indicator of welfare, such as consumption or income, is determined by a large number of variables, only a small subset of which are amenable to monitoring by these methods. In addition, the variation in welfare at any given income level is so broad that one must remain somewhat pessimis- tic about the scope for really credible monitoring of poverty in many data-poor settings. Martin Ravallion 217 Table 5. Summary of Experiments and Results Data problems Proposed solutiorns The tests Results 1. You have a good Either (1) ignore the problem and First estimate the regional poverty Method (1) is better than (2). The consumption survey, use one nominal poverty line in all profile for Indonesia using errors in ranking using (1) need not but no locational cost- regions, or (2) anchor the lines to locational data on the costs of he large, particularly if one can at of-living index. food-energy requirements so that basic consumption nceds, then see least make a reasonable assump- the local line is the expenditure how well the profile can be tion for the urban-rural cost-of- level at which requirements are predicted without these data. living difference. But method (2) met on average in each region. can be way off the mark when aiming to construct a profile of absolute poverty. 2. There are gaps and Identify proximate causes of Using time series data for rural The wage rate and farm yield are lags in survey-data poverty that can he monitored India, sce how well the model good predictors of rural poverty availability. using more readily available predicts beyond the period over measures for India. One can aggregate data. which it was estimated. Use real forecast poverty fairly well one agricultural wage rate and average year ahead of the survey. But farm yield as the predictors. sizable drift can arise after one year. 3. A full-size objective Use both subjective and objective A full-sized conventional survey for Roughly half of the variance in socioeconomic survey rapid-appraisal methods, from Jamaica also included a short consumption can be explained this is not feasible, but a short questionnaires, and small module of subjective and simple way. But the method still entails rapid appraisal is. samples. objective questions, so the results large errors; for example, in 60 can be compared. Comparc percent of cases the rapid-appraisal prcdictions for consumption. method incorrectly predicts the decile of those actually in the poorest decile. The method was better at identifying rich house- holds than poor ones. 4. You would like to Use econometric analysis to Data on the urban-rural decompo- Simple models perform badly. But know how various decompose the average indicator sition of socioeconomic aggregates provided the model is augmented indicators differ by subgroups; the population mean by region for Indonesia and India to allow for other determinants, between subgroups but indicator is regressed on the can be used to test the method. these data suggest that reasonable your survey did not ask proportions of people who are Actual values are compared with predictions are possible. for those indicators. poor and nonpoor (say), allowing the model's predicted decomposi- for the structure of the regression tion. errors. 5. You do not have any kind of survey, but you For countries in which you have For each of 22 countries with The experiment suggests that the want to at least assess the required data, construct a surveys, see how well poverty can ranking will rarely be far off, but how prevalent poverty regression model to predict the be predicted using aggregate data. that the absolute forecasting errors is in a country. poverty rate as a function of the For each country, calibrate the will be large in some cases. readily observed socioeconomic model on the other 21 countries. aggregates. Notes Martin Ravallion is with the Policy Research Department of the World Bank. He would like to thank Benu Bidani, Shaohua Chen, Gaurav Datt, Christiaan Grootaert, Margaret Grosh, Stephen Howes, Emmanuel Jimenez, Dominique van de Walle, Jacques van der Gaag, and Qing-hua Zhao. Statistical agencies in twenty-three countries have helped by collecting data used here, but special mention should be made of the Statistical Institute of Jamaica for agreeing to add to its survey the questions used in Experiment 3. References The word "processed" describes informally reproduced works that may not be commonly available through library systems. Ainsworth, Martha, and Jacques van der Gaag. 1988. Guidelines for Adapting the LSMS Living Standards Questionnaires to Local Conditions. Living Standards Measurement Study Working Paper 34. Washington, D.C.: World Bank. Bidani, Benu, and Martin Ravallion. Forthcoming. "Decomposing Social Indicators Using Distributional Data." Journal of Econometrics. BPS (Biro Pusat Statistik). 1989. Welfare Indicators. Jakarta. Casley, Dennis J. 1993. "Introduction." In Krishna Kumar, ed., Rapid Appraisal Methods. A World Bank Regional and Sectoral Study. Washington, D.C.: World Bank. Chambers, Robert. 1994. "The Origins and Practice of Participatory Rural Appraisal." World Development 22(7):953-70. Chaudhuri, Shubham, and Martin Ravallion. 1994. "How Well Do Static Welfare Indica- tors Identify the Chronically Poor?" Journal of Public Economics 53(3):367-94. Cowell, Frank, and Stephen Jenkins. 1994. "How Much Inequality Can We Explain? A Methodology and an Application to the USA." Distributional Analysis Research Programme Working Paper 7. London School of Economics, London. Processed. Datt, Gaurav, and Martin Ravallion. 1993. "Regional Disparities, Targeting and Poverty in India." In Michael Lipton and Jacques van der Gaag, eds., Including the Poor. Wash- ingtoni, D.C.: World Bank. Demery, Lionel, Jean-Luc Dubois, Christiaan Grootaert, and Tim Marchant. 1992. "An- notated Questionnaire." In Ghislaine Delaine and others, eds., The Social Dimensions of Adjustment Integrated Survey: A Survey to Measure Poverty and Understand the Effects of Policy Change on Households. Social Dimensions of Adjustment in Sub- Saharan Africa Working Paper 14. Washington, D.C.: World Bank. Foster, James, J. Greer, and Erik Thorbecke. 1984. "A Class of Decomposable Poverty Measures." Economtetrica 52:761-65. Grootaert, Christiaan, and Timothy Marchant. 1991. The Social Dimensions of Adjust- ment Priority Survey: An Instrument for the Rapid Identification and Monitoring of Policy Targeting Groups. Social Dimensions of Adjustment in Sub-Saharan Africa Work- ing Paper 12. Washington, D.C.: World Bank. Haddad, Lawrence, and Ravi Kanbur. 1990. "How Serious Is the Neglect of Intra-house- hold Inequality?" Economic Journal 100:866-81. Kumar, Krishna. 1993. "An Overviexv of Rapid Appraisal Methods in Development Set- tings." In Krishna Kumar, ed., Rapid Appraisal Methods. A World Bank Regional and Sectoral Study. Washington, D.C.: World Bank. Narayaxi, Deepa, and Lvra Srinivasan. 1994. Participatory Development Tool Kit: Train- ing Materials for Agencies and Communities. Washington, D.C.: World Bank. 220 The World Bank Research Observer, vol. 11, no. 2 (August 1996) Prescott, Nicholas, and Dean Jamison. 1985. "The Distribution and Impact of Health Resource Availability in China." International Journal of Health Planning and Man- agement 1:45-56. Ravallion, Martin. 1993. Poverty Comparisons. Fundamentals in Pure and Applied Eco- nomics, vol. 56. Chur, Switzerland: Harwood Academic Press. Ravallion, Martin, and Benu Bidani. 1994. "How Robust Is a Poverty Profile?" World Bank Economic Review 8(1):75-102. Ravallion, Martin, and Gaurav Datt. 1995. "Growth, Wages, and Poverty: Time Series Evidence for Rural India." Policy Research Department, World Bank, Washington, D.C. Processed. Ravallion, Martin, Gaurav Datt, and Dominique van de Walle. 1991. "Quantifying Abso- lute Poverty in the Developing World." Review of Income and Wealth 37(Decem- ber):345-61. Ravallion, Martin, and Binayak Sen. 1994. "Impacts on Rural Poverty of Land-Based Tar- geting: Further Results for Bangladesh." World Development 22(6):823-38. Martin Ravallion 221 DEFORESTATION AND FOREST LAND USE: THEORY, EVIDENCE, AND POLICY IMPLICATIONS William F Hyde Gregory S. Amacher William Magrath The topic of deforestation is seldom examined from the perspective of prices and responses to resource scarcity. This omission creates important errors in policy. Resource scarcity induces investments in both commercial and sub- sistence uses of the forest once prices overcome the costs of establishing prop- erty rights, forest management, and the returns from alternative agricultural uses of the land. Therefore deforestation will induce price increases and in- vestments in forestry well before deforestation attains its physical limit. These prices and costs will alter the boundaries among several important classes of forest land: sustainable private forestry, the forested commons, unsustain- able open-access forests, and unused residual forest. The greatest imnpact on the world's forests will come from refocusing the policy dialogue on the cost factors that determine these boundaries, including agricultural support poli- cies, local concentrations of nonmarket environmental resources, and policy failures that distort incentives to invest in forestry. In locations where refor- estation induces large price changes, policymakers must remain attuned to the likelihood that deforestation-induced changes in the prices of forest prod- ucts and forest policies may cause significant shifts in the activities of the poorest people. C onserving scarce forest resources is a challenge for both high- and low- income countries. Contemporary forest policy and forest management both reflect this challenge. Forest policy reflects international concern with the pressures of deforestation, including trade in tropical timber, the con- version of forests to agricultural uses, and the effects of deforestation on climate change, biodiversity, and local communities dependent on forest resources. For- est management features the transition to managed forests as deforestation causes The World Bank Research Observer, vol. 11, no. 2 (August 1996), pp. 223-48 C) 1996 The International Bank for Reconstruction and Development / THE WORLD BANK 223 a decline in the natural environment and includes questions of investments in research on natural habitats and silvicultural practices. Either deforestation or its relative, forest land use, is at the core of most contemporary forestry problems. An analysis of these problems may begin by reviewing the economic classifications of forest land use, paying particular at- tention to the thresholds that separate the classes of forest land and that define changes in the levels of deforestation and reforestation. Von Thunen's (1875) classic work on the economic geography of forests underlies this analysis. Von Thunen's insights suggest that the increasing scarcity implied by deforestation will cause the values of forest-based resources (timber, fuelwood, fruits and nuts, forage, and fodder) and environmental benefits (such as climate change, biodiversity, and erosion control) to rise until forests eventually compete well with some agricultural uses of the land. As a result, the world will never reach the physical limits of deforestation. Observations from a wide variety of economies, social settings, and ecosys- tems support the basic contention that land managers respond to the increasing scarcity of forest products by planting trees and making other private invest- ments in forestry. For example, lumber prices in the United States have risen at an annual rate of 1.8 percent for 120, and perhaps 170, years (Barnett and Morse 1963; Ruttan and Callahan 1962). This price increase has induced the technical change that allows loggers to use ever-lower quality timber growing ever farther into the interior, and it has also induced private investments in 5 million hectares of forest plantations in the South and the coastal Pacific North- west in the last thirty years. In Malawi, in a very different example, the rate of deforestation exceeds 3.5 percent annually, and fuelwood prices have risen by more than 5 percent a year for the last decade. Fuelwood may now consume 20 percent of the cash income of subsistence households in rural areas. Malawi's deforestation has occurred largely on open-access forest (land where property rights are nonexistent or unenforced), where there is no conservation incentive. Small landowners are responding by planting trees on their own lands at a rate that may offset all deforestation within ten years (Hyde and Seve 1993). Similarly, investors have responded to higher prices by planting industrial forests in Chile (Vincent and Binkley 1992), Costa Rica (Rice 1993), Kenya (Scherr 1995; World Bank 1992), and Vietnam (Byron 1993). And farmers have planted trees to meet subsistance needs in Cape Verde (Krutilla and Hyde 1995), China (Yin 1994), Madagascar (Larson 1992), Nepal (Bluffstone 1995; Gautam and others forthcoming; Amacher and others 1993b), and the Philippines (Templeton 1992; Bensel and Remedio 1993). This evidence suggests that new guidelines be adopted for formulating forest policy. Today's focus on the divergence of private and social values, the absence of secure property rights, and the imposition of fees for forest use is critical in specialized cases. But the policy dialogue could focus more usefully on the rela- tive differences in the costs of forest land; on local rather than general concen- trations of nonmarket forest-based environmental benefits, on public policy fail- 224 The World Bank Research Observer, vol. 11, no. 2 (August 1996) ures that distort market incentives to invest in forestry, and on the broader implications of the eventual use of forest land for subsistence-oriented popula- tions. The Simple Model Growing populations, increasing food consumption, and government policies that encourage agriculture all generate increasing demands for commercial agri- cultural land. They also push populations of shifting cultivators farther into the forested interior, where soils are often thinner and the shifting cultivators must either move and clear the forest more frequently or manage an increasingly degraded environment. Except for the difficulty of clearing trees, little deters agricultural expansion because the adjacent forest is unclaimed public domain, the legally unrecognized possession of the indigenous population, or the rela- tively unmanaged and unsecured responsibility of a government ministry. Meanwhile, commercial timber interests and those pursuing other forest re- sources follow a similar pattern of incentives, first harvesting at the margin between farmland and forest, and then moving ever deeper into the interior until they reach a point where harvest and access costs consume the full value of the standing resource. They repeat- or pulse-harvest this land in subsequent years whenever sufficient new natural growth justifies it. An unused residual of natu- ral forest generally lies beyond the boundary of economically viable forest; ac- cess and harvest costs for this residual exceed the value of the land for either market or subsistence production. Figure 1 provides a graphic description of this problem for a simple landscape consisting of agricultural areas and forests. The vertical axis represents the value of the land, and the horizontal axis represents the distance from the market. The most accessible land is adjacent to the local market (for commercial goods) or to the subsistence household (for nonmarket consumption goods). Highly val- ued agricultural land with secure property rights gradually gives way to har- vested forest land and eventually to unclaimed and unharvested open-access natural forest. The functions in figure 1 reflect in situ resource values. The agricultural value function reflects the accumulated net discounted value of land for agricultural production, and the natural forest function reflects the net discounted value for the standing forest resource (either timber or other wood or nonwood prod- ucts). Agriculture occupies all the land between the origin and point A, or until agricultural value declines to a level below the forest value. Thereafter, forestry becomes the preferred economic activity. Forestry remains economically viable until its in situ value declines to zero at point B, and the entire market or subsis- tence forest value is dissipated in harvest and access costs. Some land beyond the extensive agricultural margin at point A may have a positive value for long-run production of forest resources. Private individuals William F. Hyde, Gregory S. Amacher, and William Magratb 225 Figure 1. The Relationship between Land Value and Market Access Land value per hectare ricultural land value function Forest land value function Value center A A' C B or market Distance or decreasing access NVote: Point A is the point at which agricultural land value equals forest land value; point B is the point at which short-run forest land value equals zero; point C is the point where agricultuLral land value declines to zero; the area bevond B represents residtual natural forest of no private value. and communities will establish clear rights to this part of the forest if the long- run value exceeds the costs of establishing and protecting the property rights. Somewhere between points A and B, the costs of establishing and protecting permanent rights to the land exceed the in situ value of any products harvested from the forest, and it becomes an open-access resource. Thereafter, the forest value function in figure 1 reflects only short-run or extractive resource values. Generally, the costs of establishing and protecting permanent rights to the land are greater than the value of the land for long-run forest production for all land beyond point A (agricultural uses reclaim the land up to point C, and all of the forest between points C and B is an open-access resource). This is an important point. It suggests why observed rates of harvest, defores- tation, and land conversion are not as socially excessive as they sometimes seem, and also why establishing property rights is not an easy solution. Figure 1 shows that the values of standing forest resources are low relative to other land uses. Forest resources may also be dispersed. For example, fruit and nut-bearing trees 226 The World Bank Research Observer, vol. 11, no. 2 (August 1996) tend to be scattered, and some high-value tropical timber species occur only on an occasional hectare. The costs associated with establishing and maintaining property rights for these resources can easily exceed their in situ values. Fences, forest guards, and roadblocks may help protect these rights, but trespass and theft are common even in the forests of industrial countries. Inevitably, at some point between A and B in figure 1, the value of the forest will be positive but less than the cost of secure property rights. Open-access forestry is efficient in this region. Finally, the standing forest located beyond point B is also an open-access resource, but extraction from this forest is not economically attractive. This forest is usually the stewardship responsibility of the forest ministry or national park agency. Many of the forests of interior Alaska, northern Canada, Siberia, and some of the Rocky Mountains in the United States, and most remaining tropical forests, including much of the Amazon, fall into this category. This economically residual resource accounts for a significant part of most recorded measures of global forest inventories despite its uneconomic nature because the universal convention for forest inventories is a physical measure taken without regard for access costs. Therefore, access and harvest costs for industrial users and the opportunity costs of labor for subsistence forest collectors create very real limits on the extraction of resources, and the costs of ensuring tenure create a very real burden for alternative strategies for resource management. These arguments and the description shown in figure 1 accommodate many nonmarket forest values as well as the obvious commercial forest values. Fuelwood, forage, fodder, fruits, and nuts, for example, actually do exchange in local markets, and subsistence users of the forest choose the levels of their mar- ket participation according to the opportunity costs of the labor used in collect- ing the fuelwood, forage, and so on. Therefore, the costs of deforestation, land conversion, and property rights do not reflect the magnitude of the losses that would be implied by an absence of markets.' Nevertheless, some nonmarket environmental values do remain (notably watershed protection or erosion con- trol, biodiversity or genetic reserves, and carbon sequestration to protect against global climate change), and these will be discussed later. Scarcity and Increasing Prices Deforestation, climate change, and biodiversity are largely concerned with the deterioration of the residual stock of natural forest to the right of point B in figure 1. This is a dynamic problem. It is affected by policy changes, and it will change as resource extraction continues over time. It is reasonable to expect that extraction from the natural forests will continue until the prices of forest products attain a level that induces investments in trees or forests. The key issue for policy formulation should be the relative magnitudes of eventual prices for forest products, the eventual location of active tree or forest management, and William F. Hyde, Gregory S. Amacher, and William Magrath 227 the final combined inventory of planted trees and natural forests-or the final status of world deforestation. In brief, as the extraction of forest resources continues, the margin of the natural forest (point B in figure 1) must shift outward (to the right), forcing the residual natural forest to decline. If society continues to consume forest resources, then extraction must continue; access and harvest costs, including the opportu- nity costs of subsistence labor, must rise; and the market price of forest products must rise with them. As the stock of natural forest declines and access costs and market prices continue to increase, the function representing forest resource value in figure 1 rises until market prices (or the opportunity costs of subsistance house- holds) eventually justify the costs of investments in trees and of securing and protecting their property rights. If these combined costs are less than the price represented by the forest land value gradient at some point to the right of A (say A'), new plantations will appear between that point and point A to its left. Land Tenure Rigbts Agricultural land extends to point A, where the agricultural gains are equal to the costs of securing agricultural property rights. If the cost of securing the rights to forest property is comparable to the cost of securing agricultural rights- which is the general case-a discontinuity in the use of forest land can be antici- pated. Once the initial harvest of the natural forest between points A and B has been completed, agricultural use extends to point C. The extraction of forest products continues at the ever more-distant margin (point B shifts to the right), and the values of market and subsistence forest-based goods continue to in- crease. Eventually prices of forest products will overcome the costs of securing rights to forest property and the costs of local reforestation and forest manage- ment. Once these prices rise further to the level of the agricultural opportunity, trees will compete as an agricultural crop in the neighborhood of point A. This pricing sequence argues that trees will become an efficient activity somewhere closer to the market center than some extensive agricultural activities. The increasing recognition that private investment responds to declining natural forest stocks and increasing prices for forest resources is usually as- sociated with fuelwood and other subsistence forest products (Hofstad 1995; Mercer and Soussan 1992; Cline-Coal, Main, and Nichol 1990; Dewees 1989). Farmers commonly plant trees near their households or in rows along well- traveled fences, paths, and nearby embankments to provide better protec- tion for their higher-value forest resources. (Regular surveillance is more difficult and theft is easier at the more distant fringe of the household's agri- cultural lands.) Rural households follow this strategy in developing coun- tries around the world. Their private plantings may consist of only a few trees (not enough to be called forest "plantations"), but in a few cases these private plantings have been sufficient to replace all the resources removed from the open-access natural forest. 228 The World Bank Research Observer, vol. 11, no. 2 (August 1996) Commercial forest plantations in industrial countries also tend to follow this model. Although they do not compete with the highest agricultural values, they often produce a more valuable crop than agricultural uses such as grazing. The land-use sequence (see figure 1) ranges from high-value agriculture to commer- cial forest plantations to livestock grazing to timber harvests (from the remain- ing natural forests) to an unharvested residual forest. Sedjo and Lyon (1990), for example, project that this pattern will prevail for the foreseeable future, with at least half the world's harvest of industrial wood in 2050 originating from natural forests. Therefore, periodic harvests from natural forests will play a large role in industrial timber supply beyond 2050. Classification of Forest Land The investment response to the increasing scarcity of forest resources creates four or possibly five categories of forest land. The first is an economically re- sidual standing natural forest to the right of point B in figure 1. The second is a region of mature natural forest from which timber or other forest products are currently harvested (the neighborhood of point B itself). The land area and for- est inventory in this second category are highly responsive to short-run changes in prices, harvest and access costs, or forest sector policies. Higher prices for forest products, technical changes that reduce extraction costs, new roads into previously inaccessible regions, and the reduction of public policy constraints on harvesting each expand the area and volume of current resource extraction. Declining markets and tighter policy constraints reduce the size of this category of forest land and decrease or delay current extraction. These two categories of land use are the common focus of most forest policy discussions. The third category of forest land covers the region of previous resource ex- traction to the left of point B. It supports repeat or pulse harvests as the depleted natural forest grows to some minimum economic size. Access to the resources in this region remains open because the cost of obtaining and protecting property rights exceeds the potential return to forest investment. A fourth category of forest land may appear in the neighborhood of the ex- tensive agriculture (to the right of point A) if the expected in situ return on forest investments exceeds the costs of protecting rights to forest property. This land-use category is more likely in subsistence communities where common ownership offers a lower cost and a more secure regime of property rights than do individual private rights. The initial cost of converting forests to agriculture identifies the boundary between this fourth category of forest land and exten- sive agricultural uses of land. This category cannot exist when the cost of secure property rights exceeds the expected return on investment in forest resources. The prices and costs of forest products continue rising until increasing demand eventually shifts some forest production for both subsistence and industrial uses to a fifth category of higher-valued (previously agricultural) land. The in situ forest product prices for William F. Hyde, Gregory S. Amacher, and William Magrath 229 this land must justify the cost of trees managed as an agricultural crop, includ- ing the important cost of secure property rights. This land-use category is a recent phenomenon in forest history that was unnecessary before deforestation- induced scarcity.2 These observations suggest that three categories of forestry are sustainable: commercial plantations, household tree plantings, and the uneconomic residual natural forest. The regions of current resource extraction and pulse harvests will always be subject to open-access extraction because the cost of securing tenure in these regions exceeds the value of the resource. These regions can become sustainable only with effective, and perhaps expensive, public regulation or for- estry ministry management. That expense may be unacceptable in many devel- oping countries, and the degradation of forests in these regions may be inevi- table and economically efficient. The Empirical Evidence Data on these five categories of land use are exceedingly difficult to obtain. Table 1 gives a rough impression of the importance of each category in a cross- section of five countries. These data were collected in a variety of formats. None were collected with the intention of identifying the economic distinctions of land use, and they seldom provide the means to estimate the areas of pulse harvests. These areas of degraded forests no longer appear in the forest statistics of many countries. Nevertheless, a few points are clear. Plantations and trees planted for household uses account for significant areas, even in some develop- ing countries, and the areas of uneconomic residual forest are extensive in all five sample countries. The evidence on nontimber products is limited; the evi- dence in table 1 refers to harvests of commercial timber. The margin of natural forests is an important source for commercial harvests in the United States. It is Table 1. Classification of Forested Regions (thousands of hectares) Commercial Pulse Current Uneconomic and harvests of harvests from residual household forest natural natural Country plantations products forests forests Belize 3 - 3 2,000 Philippines 290 - 316 7,830 Chile 1,400 - Negligible 16,200 United States 2,000 - 1,250 238,400 Finland 12,650 7,000 - 4,000 - Not available. Source: Derived from Chomitz and Gray (1994); Republic of the Philippines (1987); Hyde and Cham- berlain (1995); U.S. Department of Agriculture (1992); Aarne (1995); Bailey (1995); World Resources Institute (1994). 230 The World Bank Research Observer, vol. 11, no. 2 (August 1996) the site of agricultural land conversion, if not for commercial timber, in Belize and the Philippines. The evidence in table 1 is static. The real support for this argument depends on changes in land use and the increasing scarcity of forest resources over time. Scattered evidence on rapidly changing patterns in land use for both industrial and developing countries supports the contention that scarcity does generate investments in forestry. Industrial Forestry: The United States There is no reliable long-term record of stumpage prices (the value of standing timber) in the United States, but lumber prices have risen at an inflation-corrected annual rate of 1.8 percent since 1870 and perhaps longer. This is the only long-term real price increase in U.S. history for any primary natural resource (Barnett and Morse 1963; Ruttan and Callahan 1962). Of course, prices cannot continue to climb indefinitely; Barnett and Morse (1963) and Berck (1979) predicted that the rate of increase would decline, and more recently Sedjo and Lyon (1990) contended that the upward movement came to an end during the past twenty years. As the stock of natural forests was drawn down during this long period, its deple- tion drew the attention of the American Association for the Advancement of Sci- ence in 1873 and provided the justification for the Forest Reserve Act of 1897, which became the basis for the National Forest System. The U.S. Forest Service, in approximately decennial projections from 1909 to the present, continues to antici- pate timber shortfalls, but Clawson (1979) pointed out that Forest Service projec- tions have never been correct, largely because they always underestimate future production and price-induced improvements in wood use and harvest technologies. Libecap and Johnson (1979) observed that settlers of the forested American frontier harvested the forests and claimed the land as soon as the land had value- to the point at which all value was dissipated in the costs of acquiring the prop- erty rights. Furthermore, in separate analyses, Johnson and Libecap (1980) and Berck (1979) observed that fears of excessive harvesting from the natural for- ests in the Great Lakes region from 1880 to 1920 and from the Pacific North- west since 1950 were unwarranted. Indeed, for these two great timber- producing regions, real stumpage prices rose at less than the rate of interest, and harvests of the natural stock occurred at rates considerably below expected re- turns on private capital. After a long period of apparently rational private ex- traction, the accompanying price increases have slackened, forest investments have become profitable, and the modern U.S. forest industry supports more than 5 million hectares of private plantations.3 Industrial Forestry: Other Countries Chile is another industrial wood basket of the late twentieth century, and it provides a similar story-although without the supporting econometric litera- William F. Hyde, Gregory S. Amacher, and William Magrath 231 ture. In the nineteenth century, the demand for mine timbers led to extensive depletion of the country's natural forests. The large mining interests responded by planting experimental forests to supply their future needs. Agricultural culti- vation kept the natural forest in abeyance, and pulse harvesting of the natural forest growth largely satisfied the remaining domestic demand until the 1950s. Since then, expanding demand from Europe, the development of Chile's ports, and (arguably) a government program of forestry assistance, have provided a supporting environment for more than 1.4 million hectares of forest planta- tions. These plantations are generally located at the margin of good agricultural land; but within much easier acccess to the ports and processing facilities than the natural forests (Vincent and Binkley 1992; Amacher and others 1996). Several developing countries have begun the transition to plantation produc- tion of industrial wood. In Costa Rica the conversion of forests to agricultural land and the growth of industrial and subsistence forest consumption have con- tributed to deforestation. Little of the original forest remains, and the remaining natural forest is insufficient to ensure sustained operation of the local wood- processing industries. One firm, recognizing the potential long-run gains to con- tinued operation of its mills, has invested in natural forest management on 10,000 hectares in an area in which it has secure property rights (Rice 1993). In Kenya 90 percent of the industrial wood came from the natural forest in the 1950s. Today 80 percent comes from 180,000 hectares of pine and cypress plantations, and less than 10 percent is harvested from natural forests (World Bank 1992). Although it is clear that land conversion for agriculture is the pri- mary source of deforestation, it is reasonable to assume that forestry only re- cently began to offer the satisfying financial return necessary to induce private investment. Scherr's (1995) historical evidence for western Kenya paints a clear picture of scarcity-induced household investments in trees following permanent settlement in that region early in this century. In Malawi, tobacco farmers are the largest consumers of industrial wood, both for posts and for fuel in flue curing. They relied on natural forests on customary lands (lands largely open to first claimants) until Malawi's 3.5 per- cent annual rate of deforestation and more than 5 percent increase in the price of fuelwood induced both a new wood-saving technology for curing tobacco and self-sufficiency in wood production. Some of the new forest plantations are in large blocks in less populous regions where forest trespass is not a serious problem. Other tree plantings, in more populated regions where trespass is more likely and the protection of forest property rights is more difficult, tend to be in fence rows, farm yards, and along paths and roads near farm operations (Hyde and Seve 1993). In the Mekong delta of Vietnam, small farmers have responded to the high export price for wood chips (for pulping) by introducing a new tree species, eucalyptus camalduensis, for sale to the new export market. They cultivate this new species in private stands and along paths and paddy edges, but always close to the household where private investments are most secure (Byron 1993). 232 The World Bank Research Observer, vol. 11, no. 2 (August 1996) Forest Production for Household Consumption Farmers in many developing countries have responded to higher prices by planting trees to satisfy household consumption. Several examples from Nepal serve as nice illustrations, in part because Eckholm (1976) made the world so aware of deforestation in that country. Most of the evidence is cross-sectional and is restricted to fuelwood, although many other forest resources (fruit, nuts, forage, and fodder) consumed by subsistence households also exchange in local markets. The patterns of forest extraction and investment for these resources are probably similar to those observed for fuelwood. Shifting cultivators receive much of the blame for Nepal's deforestation, but careful observations show that they respond to increasing scarcity by settling, establishing common and then private property rights and ultimately introduc- ing more land-intensive and longer-term agricultural technologies (Eggertsson 1990 reviews this literature). The Gurung population of the Annapurna Sanctu- ary followed precisely this pattern during the course of this century (Stevens 1988). Early in the century this region was isolated, even for Nepal, and the Gurungs cut the forest, grew subsistence crops for a few years, and moved on to cut more forest. Over time population pressures, deforestation, declining agri- cultural yields, and growing familiarity with new agricultural technologies in- duced major social changes. The indigenous population is now settled. Private property rights are well defined, and the Gurungs use agricultural technologies such as terracing that are widespread throughout Nepal. Some Gurung land- owners plant trees, and the remaining forest has stabilized. Gautam and others (forthcoming) surveyed small landowners in five commu- nities in Nepal's hills. Many farmers own trees and use them for fuel and con- struction, as a source of fruit and fodder, and for other purposes. Their private plantings of trees have increased in the past ten years as the forest inventory on common lands has declined. More often, it was the wealthier and better edu- cated landowners who planted trees and adopted other conservation techniques. Amacher, Hyde, and Joshec (1993) examined household production of fuelwood in two hill districts of Nepal. Fuelwood prices, distance from common forest lands, and the standing forest inventory differed significantly in the two districts. Not all households bought or sold fuelwood, but many did, and all had the opportunity to participate in local markets. Households in the district char- acterized by lower prices, briefer travel time, and a larger forest inventory (all indicating that scarcity is less serious) relied on fuelwood from the forested com- mons. Households in the district where fuelwood was scarcer produced more fuelwood on their own lands. Scarcity was also more likely to induce households in the latter district to substitute combustible agricultural residues for fuelwood and to adopt improved cook stoves. A broader survey of households in twenty- nine (of fifty-nine) districts in Nepal's more populated hill and tarai regions produced econometric production and consumption results consistent with the first analysis and with the hypothesis that subsistence households respond to William F. Hyde, Gregory S. Amacher, and William Magrath 233 increasing resource scarcity by increasing their plantings of trees on private land (Amacher, Hyde, and Kanel forthcoming). In another study on Nepal's tarai, Kanel (1993) reported annual price in- creases of 16 percent for fuelwood and more than 20 percent for construction wood. He also reported a 1.3 percent rate of annual deforestation regionwide and total deforestation of the region's public forestlands. Private plantations totaled 10,000 hectares in 1981 but expanded to 150,000 hectares by 1991. Barnes and others (1993) made similar observations for the more populated periurban areas of Java, and Krutilla, Hyde, and Barnes (1995) drew similar conclusions for an even broader collection of thirty-three periurban areas in Africa, Asia, the Middle East, the Caribbean, and Latin America. The counterargument to scarcity-induced investment is that some species and associated forest resources grow only in natural environments. These species are unresponsive to scientific intervention and cannot be domesticated. Brazil nuts are the common example, but Viana and others (1996) have shown that deforestation and higher prices induce successful plantations even in this ex- treme case. Industrial examples of rubber, oil palm, and coffee trees, and subsis- tence economy examples of many fuelwood, forage, and fodder species are only better-known cases of the same effect. Nonmarket Values and Resource Degradation The argument that market-induced responses to increasing scarcity will cor- rect many forest allocation problems is appropriate for commercial timber as well as for many nontimber forest resources. It leaves unaddressed three issues of great importance: erosion control, biodiversity or genetic reserves, and the control of global climate change. Controlling Erosion Erosion is a local problem caused by livestock, shifting cultivation, and logging activities on forested uplands. It affects the adjacent lowlands and watercourses within a watershed. Its solutions must also be local. One solu- tion arises with the shift from forest extraction at the economic fringe to plantation management in higher-value lowland areas. This shifts the erod- ing activity away from more fragile upland environments and also provides tree cover for more heavily used lowlands. Both steps lessen erosion, but the problem will remain in many areas. Two empirical analyses demonstrate the potential for regional economic gain. Anderson (1987) estimated the expected effects of reforestation on wind control and increased agricultural produc- tivity in northern Nigeria. Yin and Hyde (1995) calculated an 8 percent gain in agricultural productivity attributable to increased tree cover in northern China since 1974. 234 The World Bank Research Observer, vol. 11, no. 2 (August 1996) The evidence from industrial countries (for example, McConnell 1983 and Miranowski 1984) argues that individual landowners correct for the problems of erosion when they own both the sources and the effects of soil loss. This finding suggests that any system that improves private rights for forested prop- erty and enables communal arrangements for addressing transboundary prob- lems would also decrease erosion. Where improved property rights and commu- nity-imposed landowner rules are insufficient, reforestation and broader restrictions on human access to the forested uplands may be necessary. Enforce- able property rights are a prerequisite, however, for effective broader controls and successful public interventions. Their absence may justify custodial respon- sibility by the forest ministry, but effective administration of custodial responsi- bilities is an uneasy challenge. Few forest ministries have shown that they are up to it. Preserving Biodiversity Biodiversity and genetic reserves are broad international values conferred by highly specialized and generally local forest resources. There is some evidence of a rising international market for the protection of genetic resources (Simpson, Sedjo, and Reid, forthcoming) and some evidence that these values are not large in any aggregate sense (Sedjo 1992; Mendelsohn and Balick 1995). In any event, the market is thin, and these resource values often have to be protected with specialized management criteria or in specialized local forest reserves. The prob- lem is to arrange secure rights for specialized habitats. Protecting biological diversity is all the more difficult because neither flora nor fauna respect property boundaries. Moreover, protecting a few select indi- viduals or even an ecosystem does not guarantee either diversity within a species or the species' long-term survival. Protecting biodiversity requires land-use ar- rangements that are more complex than fences and permanent restrictions on forest management in national parks and wildlife preserves. Furthermore, the protective land-use arrangements must change depending on the type of forest. In tropical forests, local endemism and species specializa- tion are greater. The key issue is ensuring undisturbed regeneration cycles in individual, site-specific populations. In boreal forests the key issue is a shifting forest mosaic with respect to species and age structures. Species turnover or progression is more rapid in boreal forests. Therefore, tropical forests may con- tain more species diversity, but boreal forests may require larger areas to pro- tect a single species (Vehkamaki and Simula 1995). As figure 1 shows, the critical forested regions for both erosion and ge- netic reserves are the open-access forest between points C and B and the region of extraction from mature forests in the neighborhood of point B. The residual natural forest beyond point B is undisturbed by forestry activities; its soil and genetic reserves also remain undisturbed. Forest plantations, in the region of point A, protect the soil, but this region long ago lost most of William F. Hyde, Gregory S. Amacher, and William Magrath 235 its contributions to natural genetic diversity. It is in extensive agriculture (A- C) or the open-access forest (C-B) and the exploitable mature natural forests (the neighborhood of B) where local enclaves of nonmarket environmental values such as watershed protection (erosion) and genetic diversity justify public policy interventions. Global Climate Change The importance of global climate change is an unresolved issue, but, if it is important, there can be no doubt that trees provide an important carbon reserve and protection against further global change. All trees store carbon. The storage of carbon increases as trees are planted, grown, cut, stored, and new trees planted. Incorporating a value for protection against global climate change enhances the value of standing forests. In terms of figure 1, it raises the forest value functions and shifts point A to the left, which means that it would have been socially optimal for commercial forest to compete successfully with agriculture at lower forest product prices and at an earlier time. It expands the socially optimal area of land in commercial forests. The social valuation of carbon sequestration, however, is difficult to calculate and even more difficult to impose on land man- agement because this value is held by many consumers who have little personal contact with the forest itself. Furthermore, policies addressing global climate change would also have significant administrative costs and high costs per tree or hectare protected. Therefore, one might hypothesize that correcting those policies in other sectors that have large deleterious spillover effects to forestry and improving the security of property rights to the forest may cause greater shifts in land use at points A and B and save more trees for carbon sequestration than would altering prices and policies to reflect optimal adjustments for global change. Policy Implications The first conclusion of a model that shows responses to rising prices is that deforestation is not the ultimate issue often portrayed in popular discussions. Rather, the market sets a limit on the extent of deforestation once prices rise to a level equal to the "backstop" costs of plantation forestry (that is, the cost of planting and managing trees or forests plus the cost of ensuring the rights to this investment). Important delays may precede the initial investment in tree planting. The first delay occurs while the natural forest is drawn down to the point where its har- vest and access costs equal the plantation backstop price. Investor uncertainty due to the long production cycles typical for forests may cause an additional delay. These delays could be arguments for policy intervention, but there is no empirical evidence of their economic importance. One problem is the difficulty 236 The World Bank Research Observer, vol. 11, no. 2 (August 1996) of measuring the aggregate scarcity response of many small landowners. Byron (1984), for example, reported that three-quarters of all fuelwood consumed in Bangladesh comes from forest stands that are too small to be included in the national forest inventory. Sedjo and Lyon (1990) found that the economic limit on deforestation has been attained for industrial timber demands. Nevertheless, they project that pulse harvests of the world's residual forest stock will provide the majority of industrial timber for the foreseeable future. There is no similar aggregate anal- ysis for worldwide consumption by subsistence communities. In two self- contained local markets where deforestation has been extreme, the situation appears to have moderated. Bluffstone (1995) found that deforestation is stable in Nepal, and Hyde and Seve (1993) projected that plantation investments may offset forest removals in Malawi within ten years. Therefore, continued world- wide deforestation is not as critical an issue as it sometimes seems, either for market-based industrial timber values or for subsistence household values for a wide variety of forest resources. Spillover Policy Effects Policies designed for other sectors of the economy often spill over to increase deforestation. Macroeconomic policies that discourage either long-term invest- ments and or a preference for holding real assets, policies that encourage agri- cultural production, and policies that affect the establishment and transfer of land tenure are common examples. Forestry's low value means that policies with small economic impacts on other sectors tend to have relatively large unintended spillover effects on forest land use. In terms of figure 1, the shallow slopes of the land-use value functions mean that small increases or decreases in the levels of any of these functions cause larger shifts to the left or right in the location of points A and B and therefore larger shifts in forest land use. Thus, the effects of policy spillovers may be more notable in terms of hectares of forest than in terms of net forest value. A comparison of Chile and Argentina provides a good example of macroeco- nomic policy impacts. Large parts of Chile and Argentina have similar forest- growing conditions. Argentina has the advantage of proximity to their mutual European markets. (Chilean exports must pay the duty for passage through the Panama Canal.) Nevertheless, Chile is a major wood exporter, while Argentina's wood exports are almost nonexistent. Argentina's high inflation and real inter- est rates and its unstable macroeconomic policies discourage long-term invest- ments in all sectors, including forestry. Chile's stable macroeconomic environ- ment permits long-term planning with confidence, and one result is extensive investment in plantation forests. Chile's macroeconomic policies raise the value of forest plantations, provide an earlier incentive for forest plantations, and expand the region of successful plantations, marked in figure 1 by the neighbor- hood around point B.4 William F. Hyde, Gregory S. Amacher, and William Magrath 237 For many countries, subsidies on agricultural inputs and price supports on agricultural outputs ensure that the market value of agricultural land is greater than its social optimum. This shifts point A in figure 1 to the right, increasing the land area devoted to agriculture, extending the amount of time before forest values are competitive with agriculture, delaying investments in trees, and de- creasing the number of hectares of tree crops. There are few empirical analyses of the effects of agricultural policies on the forestry sector, but extensive world- wide intervention in agriculture supports an argument that agricultural policies can be important sources of deforestation. Rules for establishing land tenure are the third example of policies de- signed for other purposes that can have an unusual effect on forestry. Any government's formal rules of tenure are designed to be general to any land use or claimant. The costs they impose are relatively constant per hectare for any land use; thus their relative impacts are often greatest on low-value land uses such as forestry. There is no problem where a country is still develop- ing, the rules are informal, and local customs for recognizing land claims arise on their own. This has been the experience in the agricultural conver- sion of forest land in West Africa, and the resulting land-use patterns are efficient (Migot-Adhola and others 1991). The legal codes of industrial countries and developing countries in Asian with long histories of formal government structure spell out the rules for establishing tenure. These codes often constrain the initial allocation of claims on the forest frontier, but the final allocation is efficient wherever the code permits initial claimants to transfer the land to higher-value uses. Libecap and Johnson (1979) showed that this was the experience for western U.S. expansion in the nine- teenth century. Government policy required the initial claimants to be small farmers, but these claimants had full rights to exchange their new property after some minimal period, and many farmers (and speculators who called them- selves farmers) quickly sold their rights to other economic enterprises. Often, however, the formal rules of tenure establish preference among initial claimants and restrict land transfers by the claimants. For example, many coun- tries permit claims on the forest frontier only with evidence of capital improve- ments such as fences and forest removal. This rule prevents claims for natural forest management. Binswanger (1989), Mahar (1989), and Schneider (1993), for example, showed that preferential treatment for the livestock industry (and the implicit restriction on comparable claims for forest management) played an important role in Amazonia's deforestation in the 1980s. Restrictions on land transfers can be an important problem in countries with established formal institutions but without the means to enforce the formal claims on the forest. For example, Feder and others (1988) showed that the inability of Thailand's Royal Forestry Department to enforce government claims on the forest, together with its reluctance to transfer land to trespassing (and probably higher-valued) users, created inefficiencies in land use. In general, government restrictions on transfers to high-valued agricultural uses also remove the incen- 238 The World Bank Research Observer, vol. 11, no. 2 (August 1996) tives for conservation practices and long-term agricultural management by set- tlers at the forest fringe in many Asian countries. Because the forest ministries are ineffective in restricting illegal access to the forest, squatters and other tres- passers engage in short-term agricultural and forestry practices that increase deforestation and increase erosion (Cruz, Francisco, and Conway 1988; Amacher and others 1995). Two hypotheses emerge from this discussion. First, restoring macroeconomic stability and correcting policy failures that inadvertently spill over to affect for- estry may have a greater positive effect on the forestry sector and on the amount of residual land that remains in forest than do all preferential forestry sector policies. Second, permitting transfers of land use rights to the highest-valued use and permitting future land exchanges as values eventually instruct would im- prove long-term land use, limit erosion, and improve the conditions of indig- enous peoples. Neither hypothesis suggests a total solution; both suggest that current policy distortions are worse than an unfettered market. Forest Rent The magnitude and allocation of forest charges (rent) are subjects of some contention. Gillis (1988) and Vincent (1990) argue that forest rent is large, that it belongs to the government, and that its assignment makes a difference in efficient levels of output. Paris and Ruzicka (1991) and Hyde and Sedjo (1992) argue that its magnitude varies from case to case and that its best allocation (to the government or to forest concessionaries) is a distributive matter of undetermined merit, but one that does not alter efficient land use. These arguments are valid for the publicly owned share of the forest. The rent in question is generally associated with mature natural forests and com- mercial harvests (in the neighborhood of point B in figure 1). Net resource values in this region are too low to justify long-term management. The fact that most public forest lands went unclaimed by earlier commercial interests is evidence that the rents accruing to public forest management are not al- ways large and economically important. Where the forest ministry claims the property rights and charges royalties for resource extraction, it is often the case that either the minimum fee is too high and there are no bidders or that the ministry must subsidize harvests. Both con- ditions suggest a low value for standing forest resources and point to the finan- cial efficiency of open-access management. The U.S. Forest Service, for example, recently noted that 62 of its 156 national forests suffer net financial losses on overall timber sales. Many more national forests suffer net losses on individual timber sales. In other examples, the Philippine Bureau of Forestry Development operates at an overall net loss on its timber sales (Boado 1988), and the indus- trial forests of northeastern China suffer a similar shortfall. Maintaining mill employment is the key objective in northeastern China, and the government William F. Hyde, Gregory S. Amacher, and William Magrath 239 subsidizes industrial forests to ensure harvest levels sufficient to protect mill employment. If this argument is correct, it also begs the question of why the payments for some new timber concessions are so high. Where large rents are apparent for lands at the margin of previous forest resource extraction, they must be created from some recent relaxation of a restriction on forestry activity. In this case, the rent is truly a return to relaxing the restriction. (Deacon 1994 and 1995 more broadly suggests that these rents, and deforestation in general, are largely deter- mined by the uncertainty of widely variable political conditions.) If the forest had value without the restriction, it would have been harvested earlier when its net value was newly positive and still small. Surely all private operators would have had sufficient incentive to harvest each year all the way to the geographic point where their access and harvest costs depleted the entire value of the stand- ing resource. Rents in subsequent years can appear only where previous extrac- tion was incomplete, and incomplete extraction would occur only where harvest policies were restrictive. What kinds of restrictions create rents? Limited access to the forest and pub- lic policies that make harvest uneconomical are two examples. As for the former, much of the forest in northeastern Thailand, Liberia, the Amazon, and the inte- rior of Canada and the United States would not be open for harvest had roads not been built, and forest resource values alone often cannot justify these roads. The "allowable cut" policy that restricts timber harvests to biological timber rotations is an example of the latter.5 The allowable cut policy is taught at Dehra Dun, India, and Los Banos, Philippines, in two of the oldest and largest forestry schools in the world. It is common to western European and North American forest ministries and probably to most of the forest ministry officials and schools around the world. This perspective should change the debate on forest rent. In sum, forest rent is seldom large, and it is often smaller than expected because analysts overlook the costs of administering timber sales. In those important but specialized cases where the rent is large, it begs questions about the forest management activity or policy change that created the rent. Does the same policy exist elsewhere? What are its advantages? Could more rents be captured by applying the same management activity or the same change in forest policy to other timbersheds and other forests in the same region? If so, should that activity or policy be expanded? Forestry Research and the Effect of New Technologies The theme of forestry as a low-value and low-cost resource continues as an important explanatory factor, this time for its effect on locating successful op- portunities for research and technical change. Successful research is defined by technological breakthroughs and the adoption of new technologies that lower 240 The World Bank Research Observer, vol. 11, no. 2 (August 1996) production costs. Yet production costs are already very low throughout most forestlands. Therefore, the opportunities for research and technological change are small. Indeed, too often, the experimental successes of forestry research are uneconomic in broad field applications because the costs of new, higher biologi- cal yields are greater than the costs of harvesting natural forest resources. Con- sider the potential impacts of research and the likely adoption of new technolo- gies for the five categories of forest land use. * Residual forest land. Shifting cultivation and parks for recreational uses are the only productive activities. Only research on subsistence agriculture or research that extends nonconsumptive recreational and aesthetic opportunities from a given preserved area can affect them; research on production forestry cannot. * Degraded open-access forests and mature natural forests. Production costs are zero in these regions. The only costs are those relating to access and to the harvest operation itself. Therefore research impacts in this region are restricted to technologies such as improved saws or cheaper transportation systems that reduce harvest or access costs, or to institutional changes that decrease the costs of establishing and protecting secure rights to the land. The technologies serve to extend the region of current harvests farther into the residual forest (or farther to the right in figure 1). The institutional changes enable land conversion from the region of open-access pulse harvests to sustainable forest plantations. * Private tree plantings and commercial plantations. Only in these regions can growers take advantage of the full range of new cost-saving forest production technologies. In addition, wood-saving technologies such as improved stoves (in subsistence economies) and improved wood utilization practices in the mill (for industrial forestry) are candidates for adoption in regions where the price of wood is high enough to justify investments in plantations. (See Scherr 1995 and Patel, Pinckney, and Jaeger 1995 for findings on household plantations in Kenya; Amacher, Hyde, and Rafiq 1993 on seedling distribution in Pakistan; Amacher, Hyde, and Joshee 1992 on stoves in Nepal; and Barnes and others 1993 on the adoption of improved stoves in general.) The policy instructions indicate that biological research must focus on the species and forest types that are appropriate for industrial and smallholder plantings-and only in locations where such planting will soon or already occur. Fortunately, the increasing practice of farm forestry indicates that the opportu- nities for biological forestry research are increasing. Biological research should refrain from activities designed to improve species and forest products charac- teristic of the areas of mature and open-access natural forests. New biological technologies for these regions will not be adopted. Social science research and institutional change, of course, may still offer possibilities in the regions of open-access harvests that might become pri- William F. Hyde, Gregory S. Amacher, and William Magrath 241 vate property or well-managed commons. The research objective should be to find cheaper ways to establish and protect the full range of long-term property rights to these forests, including the right to legal transfer. Success- ful applications of this research would probably be concentrated in countries where established and inflexible formal institutions hamper private land claims and transfers, or in any country with serious erosion problems or important sources of biodiversity. Conclusions and Final Observations Market responses to commercial forest values and subsistence household re- sponses to available forest resources create limits to potential deforestation. Both economic intuition and empirical evidence from a broad array of industrial and developing countries, temperate and tropical, support this argument. Further- more, policy interventions to correct problems associated with forest land ten- ure, deforestation, and forest management do not necessarily improve on market-based solutions because forest resources and forest land generally have low values and are widely dispersed. Finally, forest policy interventions are of- ten ineffective because they require extensive monitoring and administrative costs-which the low-value resources at risk cannot support. This is a modest argument, and the basic points should be simple and clear. They are important only because today's policy environment overlooks them all too often. 'I'his encourages public forest management and broad-based forest policy interventions, despite general recognition of the historic inefficiencies of forest ministries and previous forestry regulations. Perhaps it is time to recog- nize that some of the merits of deregulation and structural adjustment com- monly acknowledged by policymakers in other sectors also pertain to forestry. How then do we identify the targets for further inquiry and useful interven- tion in forest policy? Such interventions typically occur at the margins of the various land-use categories. The first target is the collection of existing macro- economic and agricultural policies that spill over to discourage forest invest- ment. The second is the assortment of existing policies for establishing secure rights to forest land and for transferring these rights. Many formal rights to forest resources are static and final. They are unresponsive to economic change and to the developing economic validity of private (or well-managed common) ownership of natural forests. These policies also delay socially optimal forestry investments and conservation practices. Finally, both erosion of forested up- lands and protection of genetic reserves may require more aggressive policy proscriptions. They call for interventions that constrain market solutions, but only in specific locations where erosion extends beyond the boundaries of any single ownership or where preserving a natural forest ecosystem is desirable. It is a myth that all forests everywhere support multiple and significant nonmarket values. 242 The World Bank Research Observer, vol. 11, no. 2 (August 1996) Self-correcting adjustments to scarcity may create two new problems, par- ticularly for the poorest households in the lowest-income countries. Once local subsistence farmers begin planting trees, they must forgo other production on their scarce household land. The poorest landowners may be forced into trading trees for nutrition when they give up agricultural production for trees. Landless households may suffer even more because they do not have the option of plant- ing trees on their own lands. Expanding markets for contract labor may solve the latter problem (Bluffstone 1995), but where the markets for contract labor are not expanding, the only options for landless households may be to trespass to obtain forest products or to incur the opportunitv costs of collecting from the ever-more-distant open-access forests. The alternative scenario is that some land- less households will supply forest labor, and the value of their labor will increase as the value of deforestation and forest products increase. In either scenario, the status of the landless, and not deforestation, may be the most serious result of increasing forest scarcity-and it is relatively unexamined as yet.6 Notes William F. Hyde is professor and Gregory S. Amacher is assistant professor in the For- estry Department at Virginia Polytechnic Institute and State University; William Magrath is senior economist in the Agriculture and Forestry Production Systems Division of the World Bank. Matti Pallo, Marian S. de los Angeles, Luis Constantino, and Robert Ander- son shared critical insights during the preparation of this article. Earlier versions were presented at the annual meeting of the International Association for the Study of Common Property in Manila, June 15-19, 1993, and the European Association of Environmental and Resource Economists in Dublin, June 22-24, 1994. 1. Several recent assessments confirm this model for a relationship between access and forest cover in Belize (Chomitz and Gray 1994), Bolivia (Robbins, Kenney, and Hyde 1995), the Philippines (Liu, Iverson, and Brown 1993), and Tanzania (Hofstad 1995). 2. Nevertheless, von Thunen (1875) also places forest plantations closer than agriculture to the town and the local value center. For isolated German states in 1875, as for some subsistence agricultural communities in the 1990s, land used for these plantations had a higher net value for fuelwood production than for agriculture. 3. Harvests from the public forests are a more contentious issue. The general view is that harvests have been too slow in some regions, too rapid in others, and seldom at rates justified by either private or social criteria. See Berck (1979), Hyde (1980), and Repetto (1988). 4. The analyses of general macroeconomic and trade policy effects on the forestry sector are scant. Binswanger (1989) identified a variety of fiscal and monetary policies that en- couraged deforestation in the Brazilian Amazon. Browder (1987) observed that Brazil's general export promotion policy encouraged a predatory system of advance timber pur- chasing that, for one company alone, extended harvests on marginal forest land in Rondonia by 300,000 hectares. Boyd and Hyde (1989) concluded that in the United States a favor- able capital gains tax policy had more effect than direct forest sector policies. 5. "Allowable cut" has many meanings. It variously can mean long- or short-run planned timber sales, successful bids for a harvest over a period of time, the model for determining a "planned" sale, or reported harvests for a preceding year. The term must be used with care. The general model taught in forestry schools and applied by forest ministries around the world begins with a biological definition of productive timberland that generally over- states the economic forest land base. It estimates harvests from these lands as some variant William F. Hyde, Gregory S. Amacher, and William Magrath 243 of the volume or area of mature forest divided by the harvest age at biological maturity, plus average annual growth for the full area. Biological rotations depend on the species, but they generally exceed economic rotations by 35 percent or more. For land that is economic, this model sharply constrains harvests. Because all land in the analysis is not economic, the net impact on harvest levels is uncertain and varies from case to case. The net economic effect is always negative. See Davis and Johnson (1987) or Hyde (1980) for statements of the allowable cut model and Hirshleifer (1974), Samuelson (1976), and Hyde (1980) for detailed critiques. 6. The exceptions in the economic literature are Kumer and Hotchkiss (1988), Barnes and Qian (1994), Larson and others (1996), and the work of N. S. Jodha, summarized in Jodha (forthcoming). References The word "processed" describes informally reproduced works that may not be commonly available through library systems. Aarne, Martti, ed. 1995. Yearbook of Forest Statistics. Helsinki: Finnish Forest Research Institute. Amacher, Gregory S., Luis Constantino, William F. Hyde, and Monica Rios. 1996. "Mac- roeconomic Effects of Natural Resource Enhancement in Small Open Economies: Chile's Forest Incentive Program." Department of Forestry, Virginia Polytechnic Institute and State University, Blacksburg, Va. Processed. Amacher, Gregory S., M. Conception Cruz, Wilfredo Cruz, and William F. Hyde. 1995. "Environmental Motivations for Migration: Population Pressure, Poverty, and Defor- estation in the Philippines." Department of Forestry, Virginia Polytechnic Institute and State University, Blacksburg, Va. Processed Amacher, Gregory S., William F. Hyde, and Bharat R. Joshee. 1992. "The Adoption of Consumption Technologies under Uncertainty: The Case of Improved Stoves in Nepal." Journal of Econiomnic Development 17(2):93-105. . 1993. "Joint Production and Consumption in Traditional Households: Fuelwood and Agricultural Residues in Two Districts of Nepal." Journal of Development Studies 30(1):206-25. Amacher, Gregory S., William F. Hyde, and Keshav R. Kanel. Forthcoming. "Household Fuelwood Demand and Supply in Nepal: Choice between Cash Outlays and Labor Opportunity." World Development. Amacher, Gregory S., William F. Hyde, and Mohammed Rafiq. 1993. 'Local Adoption of New Forestry Technologies: With an Example from Pakistan's Northwest Frontier Prov- ince." World Developmetnt 21(3):445-54. Anderson, Dennis. 1987. The Economics of Deforestation. Baltimore: Johns Hopkins Uni- versity Press. Bailey, Ronald, ed. 1995. The True State of the Planet. New York: Free Press. Barnes, Douglas F., Keith Openshaw, Kirk R. Smith, and Robert van der Plas. 1993. "The Design and Diffusion of Improved Cooking Stoves." World Bank Research Observer 8(2):119-41. Barnes, Douglas F., and Liu Qian. 1994. "Social Impacts of Energy Policies." In Douglas Barnes, J. Dowd, Liu Qian, K. Krutilla, and William Hyde, eds., "Urban Energy Tran- sitions, Poverty, and the Environment." World Bank, Industry and Energy Department, Washington, D.C. Processed. Barnett, Harold J., and Chandler Morse. 1963. Scarcity and Growth: The Economics of Natural Resource Availability. Baltimore: Johns Hopkins University Press for Resources for the Future. 244 The World Bank Research Observer, vol. 11, no. 2 (August 1996) Bensel, Terrence G., and Elizabeth M. Remedio. 1993. "Patterns of Commercial Woodfuel Supply, Distribution, and Use in the City and Province of Cebu." Regional Wood En- ergy Development Programme Working Paper. Food and Agricultural Organization, Bangkok. Processed. Bcrck, Peter. 1979. "The Economics of Timber: A Rcncwablc Rcsourcc in the Long Run." Bell Journal of Economics 10(2):447-62. Binswanger, Hans. 1989. "Brazilian Policies That Encourage Deforestation in the Ama- zon." Working Paper 16. World Bank, Environment Department, Washington, D.C. Processed. Bluffstone, Randall. 1995. "The Effect of Labor Markets on Deforestation in Developing Countries under Open Access: An Example from Rural Nepal." Journal of Environ- mental Economics and Management 29(1):42-63. Boado, Eufresina L. 1988. "Incentive Policies and Forest Use in the Philippines." In Robert Repetto and Malcolm Gillis, eds., Public Policy and the Misuse of Forest Resources. Cambridge, U.K.: Cambridge University Press. Boyd, Roy G., and William F. Hyde. 1989. Forestry Sector Intervention: The Impacts of Public Regulation on Social Welfare. Ames, Iowa: Iowa State University Press. Browder, John 0. 1987. "Brazil's Export Promotion Policy (1980-1984): Impacts on the Amazon's Industrial Wood Sector." Journal of Developing Areas 21:285-304. Byron, Neil. 1984. "Peoples Forestry: A Novel Perspective of Forestry in Bangladesh." ADAB News (Association of Development Agencies in Bangladesh) 11(2):28-35. -1993. Personal records. Chomitz, Kenneth M., and David A. Gray. 1994. "Roads, Land, Markets, and Deforesta- tion: A Spatial Model of Land Use in Belize." Policy Research Working Paper 1444. World Bank, Policy Research Department,Washington, D.C. Processed. Clawson, Marion. 1979. "Forests in the Long Sweep of American History." Science 204:1168-74. Cline-Coal, R. A., H. A. C. Main, and J. E. Nichol. 1990. "On Fuelwood Consumption, Population Dynamics and Deforestation in Africa." World Development 18(4):513-27. Cruz, Wilfredo D., H. A. Francisco, and Z. T. Conway. 1988. '"The Onsite and Down- stream Costs of Soil Erosion in the Magat and Pantabangan Watersheds." Journal of Philippine Development 15(1):48-85. Davis, Lawrence S., and K. Norman Johnson. 1987. Forest Management. New York: McGraw-Hill. Deacon, Robert T. 1994. "Deforestation and the Rule of Law in a Cross Section of Coun- tries." Land Economics 70(4):414-3Q. - 1995 "Assessing the Relationship between Government Policy and Deforestation." Journal of Environmental Economics and Management 28(l):1-18. Dewees, Peter A. 1989. "The Woodfuel Crisis Reconsidered: Observations on the Dynam- ics of Abundance and Scarcity." World Development 17(8):1159-72. Eckholm, Eric P. 1976. Losing Ground: Environmental Stress and World Food Prospects. New York: W. W. Norton. Eggertsson, Thrainn. 1990. Economic Behavior and Institutions. New York: Cambridge University Press. Feder, Gershon, Tonraj Onchan, Y. Chalamwong, and C. Hongladarom. 1988. Land Ownership Security, Farm Productivity, and Land Policies in Thailand. Baltimore: Johns Hopkins University Press. Gautam, K. H., B. R. Joshee, R. L. Shrestha, V. K. Silwal, M. P. Suvedi, and L. P. Uprety. Forthcoming. "Some General Features of Household Forestry in Nepal." In William F. Hyde and Gregory S. Amacher, eds., Social Forestry: An Empirical Introduction from South and Southeast Asia. New Delhi: Oxford and IBH. William F. Hyde, Gregory S. Amacher, and William Magrath 245 Gillis, Malcolm. 1988. "Indonesia: Public Policies, Resource Management and the Tropi- cal Forest." In Robert Repetto and Malcolm Gillis, eds., Public Policy and the Misuse of Forest Resources. Cambridge, U.K.: Cambridge University Press. Hirshleifer, Jack. 1974. "Sustained Yield Versus Capital Theory." In Barney Dowdle, ed., Proceedings of a Conference on the Economics of Sustained Yield Forestry. Seattle: University of Washington, College of Forestry. Hofstad, Ole. 1995. "Woodland Deforestation by Charcoal Supply to Dar es Salaam." Agricultural University of Norway, Forestry Department, As. Processed. Hyde, William F. 1980. Timber Supply, Land Allocation, and Economic Efficiency. Balti- more: Johns Hopkins University Press. Hyde, William F., and James L. Chamberlain. 1995. "Who Gains from Privatizing the U.S. Forest Service?" Journal of Forestry 94(8):22-26. Hyde, William F., and Roger A. Sedjo. 1992. "Managing Tropical Forests: Reflections on the Rent Distribution Discussion." Land Economics 68(3):343-50. Hyde, William F., and Juan Seve. 1993. "The Economic Role of Wood Products in Tropi- cal Deforestation: The Severe Example of Malawi." Forest Ecology and Management 57(2):283-300. Jodha, N. S. Forthcoming. "Common Property Resources and the Dynamics of Rural Pov- erty: Field Evidence from the Dry Regions of India." In William F. Hyde and Gregory S. Amacher, eds., Social Forestry: An Empirical Introduction from South and Southeast Asia. New Delhi: Oxford and IBH. Johnson, Ronald N., and Gary D. Libecap. 1980. "Efficient Markets and Great Lakes Timber: A Conservation Issue Reexamined." Explorations of Economic History 17:372- 85. Kanel, Keshav R. 1993. "Market Opportunities in Nepal's Tarai." Report prepared for Finnish International Development Authority, Kathmandu. Processed. Krutilla, Kerry, William F. Hyde, and Douglas F. Barnes. 1995. "Urban Energy Consump- tion and Periurban Deforestation." Forest Ecology and Management 74(2):181-95. Kumer, Shubh K., and David Hotchkiss. 1988. "Consequences of Deforestation for Women's Time Allocation, Agricultural Production, and Nutrition in Hill Areas of Nepal." Re- search Report 69. International Food Policy Research Institute, Washington, D.C. Pro- cessed. Larson, Bruce A. 1992. "Changing the Economics of Environmental Degradation: Lessons from the National Environmental Action Plan Process." Winrock International Work- ing Paper. Washington, D.C. Processed. Larson, Bruce A., Glenn Rogers, Joy Green Larson, and Grenville Barnes. 1996. "Access to Forests and Child Health in Zaire." Forest Science 42(1):3-9. Libecap, Gary D., and Ronald N. Johnson. 1979. "Property Rights, 19th Century Federal Timber Policy, and the Conservation Movement." Journal of Economic History 39(1):129-42. Liu, Dawning S., Louis R. Iverson, and Sandra Brown. 1993. "Rates and Patterns of Deforestation in the Philippines." Forest Ecology and Management 57(1):1-16. Mahar, Dennis J. 1989. "Government Policies and Deforestation in Brazil's Amazon Re- gion." Working Paper 7. World Bank, Environment Department, Washington, D.C. Processed. McConnell, Kenneth E. 1983. "An Economic Model of Soil Conservation." American Journal of Agricultural Economics 64(1):83-89. Mendelsohn, Robert, and Matthew Balick. 1995. "The Value of Undiscovered Pharmaceu- ticals in Tropical Forests" Economic Botany 4(2):223-28. Mercer, Evan, and John Soussan. 1992. "Fuelwood Problems and Solutions." In N. P. Sharma, ed., Managing the World's Forests. Dubuque, Iowa: Kendall/Hill. 246 The World Bank Research Observer, vol. 11, no. 2 (August 1996) Migot-Adholla, Sham, Peter Hazell, Benoit Barel, and Frank Place. 1991. "Indigenous Land Rights Systems in Sub-Saharan Africa: A Constraint on Productivity?" World Bank Economic Review 5(1):155-75. Miranowski, John A. 1984. "Impact of Productivity Losses on Crop Production and Man- agement in a Dynamic Economic Model." American Journal of Agricultural Economics 66(1):61-71. Paris, Remy, and Ivan Ruzicka. 1991. "Barking Up the Wrong Tree: The Role of Rent Appropriation in Sustainable Tropical Forest Management." Environment Office Oc- casional Paper 1. Asian Development Bank, Manila. Processed. Patel, H. Sandeep, Thomas C. Pinckney, and William K. Jaeger. 1995. "Smallholder Wood Production and Population Pressure in East Africa: Evidence of an Environmental Kuznets Curve?" Land Economics 71(4):516-30. Repetto, Robert. 1988. "Subsidized Timber Sales from National Forest Lands in the United States." In Robert Repetto and Malcolm Gillis, eds., Public Policy and the Misuse of Forest Resources. Cambridge, U.K.: Cambridge University Press. Republic of the Philippines. 1987. Philippine Statistical Yearbook. Manila: National Eco- nomic Development Authority. Rice, Richard. 1993. Personal communication. Robbins, Eugene S., Robert N. Kenny, and William F. Hyde. 1995. "Forest Concessions and Industrial Policy in Bolivia." Report to the Undersecretariat of Natural Resources/ Bolfor, La Paz. Processed. Ruttan, Vernon W., and J. C. Callahan. 1962. "Resource Inputs and Output Growth: Comparisons between Agriculture and Forestry." Forest Science 8(1):68-82. Samuelson, Paul A. 1976. "Economics of Forestry in an Evolving Society." Economic Inquiry 14:466-92. Scherr, Sara J. 1995. "Economic Factors in Farmer Adoption of Agroforestry: Patterns Observed in Western Kenya." World Development 23(5):787-804. Schneider, Robert. 1993. "Land Abandonment, Property Rights, and Agricultural Stability in the Amazon." LATEN Dissemination Note 3. World Bank, Latin America and the Caribbean Technical Department, Environment Division, Washington, D.C. Processed. Sedjo, Roger A. 1992. "Property Rights, Genetic Resources, and Biotechnological Change." Journal of Law and Economics 35:199-213. Sedjo, Roger A., and Kenneth S. Lyon. 1990. The Long-Term Adequacy of World Timber Supply. Washington, D.C.: Resources for the Future. Simpson, David, Roger A. Sedjo, and John Reid. Forthcoming. "Valuing Biodiversity for Use in Pharmaceutical Research." Journal of Political Economy 104(1). Stevens, Stan. 1988. "Sacred and Profaned Himalayas." Natural History 97(l):26-35. Templeton, Scott. 1992. "Major Determinants of Farmer Use of Semi-Permanent Soil Ero- sion Controls in Upland Areas of the Philippines." Ph.D. dissertation. University of California, Berkeley, Agricultural Economics Department. Processed. U.S. Department of Agriculture. 1992. Agricultural Statistics. Washington, D.C.: Govern- ment Printing Office. Vehkamaki, Seppo, and Markku Simula. 1995. "Economic Effects of Biodiversity in For- ests." Paper presented at the International Union of Forest Research Organizations World Congress. Helsinki University and Finnish Forest Research Institute, Tampere, Finland. August. Processed. Viana, Virgilio M., Ricardo A. Mello, Luiz M. deMorais, and Nilson T. Mendes. 1996. "Ecology and Management of Brazil Nut Populations in Extractive Reserves in Xapuri, Acre." Departmento Ciencas Florestais, Universidade de Sao Paulo, Brazil. Processed. Vincent, Jeffrey. 1990. "Rent Capture and the Feasibility of Tropical Forest Management." Land Economics 66(2):212-23. Willianm F. Hyde, Gregory S. Anmacher. and Willian Magratb 247 Vincent, Jeffrey R., and Clark S. Binkley. 1992. "Forest-Based Industrialization: A Dy- namic Perspective." In N. P. Sharma, ed., Managing the World's Forests. Dubuque, Iowa: Kendall/Hill. von Thunen, Johann H. 1875. The Isolated State in Relation to Land Use and National Economy Berlin: Schmaucher Zarchlin. World Bank. 1992. World Development Report 1992: Development and the Environ- ment. New York: Oxford University Press. World Resources Institute. 1994. World Resources, 1994-95: A Guide to the Global Envi- ronment. New York: Oxford University Press. Yin, Runsheng. 1994. "The Farmer Tree Planting Response to Political Uncertainty in China." Ph.D. dissertation. University of Georgia, College of Forest Resources, Athens. Processed. Yin, Runsheng, and William F. Hyde. 1995. "Trees as an Agriculture Sustaining Resource: Evidence from Northern China." University of Georgia, College of Forest Resources, Athens. Processed. 248 The World Bank Research Observer, vol. 11, no. 2 (August 1996) FINANCIAL MARKETS, PUBLIC POLICY, AND THE EAST ASIAN MIRACLE Joseph E. Stiglitz Marilou Uy Many factors contributed to the rapid growth of the economies of East Asia in the past quarter century. This article examines one important aspect of that growth-commonly referred to as the "East Asian miracle"-public policies affecting the financial markets. East Asian governments intervened extensively in financial markets at all stages of their development. What sets their actions apart from those of other developing countries that have not fared as well? We do not bave the information to answer conclusively what effect particular actions had (that requires a counterfactual test of what growth would have been wvithout the particular intervention). But we can identify the market failures the East Asian governments addressed, assess some of the theoretical reasons why each policy might be growth enhancing, and provide some data attesting to the impacts of the policy. Several characteris- tics of fintancial sector interventions in East Asia stand out: they incorpo- rated design features that improved the chances of success and reduced op- portunities for abuse; interventions that did not wvork out wvere dropped unhesitatingly; and policies wvere adapted to reflect changing economic conditions. T he economies of East Asia, beginning with Japan and followed by the Four Tigers-the Republic of Korea, Hong Kong, Singapore, and Taiwan (China)-and then by Malaysia, Indonesia, and Thailand, have grown so rapidly during the past quarter century that their growth has been called the "East Asian miracle." What lessons for other countries can be derived from this experi- ence, assuming that it was not really a miracle, but rather a consequence of some well-designed programs and policies? In particular, what can be learned from the policies affecting financial markets? East Asian governments have intervened inten- sively in the operations of their financial systems. They have helped create financial markets and institutions, regulated them heavily, and directed credit to some indus- The World Bank Research Observer, vol. 11, no. 2 (August 1996), pp. 249-76 © 1996 The International Bank for Reconstruction and Development / THE WORLD BANK 249 tries and away from others. These actions have been intended to mobilize savings and to affect the allocation of investment, but their effects have extended well be- yond the capital market: the "prize" of scarce credit, awarded for good export performance, has provided strong investment incentives. Careful examination of the interventions provides insights into the impor- tance of specific design features that increased the likelihood that the interven- tions would work and decreased the likelihood of abuses.1 Government flexibil- ity was also important. When programs failed, the interventions behind them were abandoned, and as economies changed, so did the role of governments. These general precepts, more than the particular interventions adopted, may prove to be the most important insights that other developing countries can borrow from the East Asian miracle. Government interventions were directed at two broad objectives. The first is straightforward: making financial markets and institutions work better. With- out the intermediation provided by capital markets, firms would have to rely solely on retained earnings for their investments, and firms' expected marginal private returns from investment would diverge markedly, especially in the short run. When capital markets work well, marginal returns are equated in all sec- tors and firms. Moreover, by spreading and pooling risks more broadly, capital markets lower risk premiums, so firms can undertake investments with greater risks and higher expected returns. But even if the marginal private returns from investment are equated in all sectors and firms, capital may not be allocated efficiently if there are systematic deviations between private and social returns. Thus the second objective of gov- ernment intervention in financial markets is to correct any resulting misalloca- tion of resources. Government financial policies are of three types: creating markets and finan- cial institutions; regulating them; and providing rewards (subsidies or access to credit or foreign exchange, often on preferential terms) to firms, groups, or in- dustries that undertake priority activities or perform in an exemplary manner. The East Asian economies had high national saving rates, achieved largely by voluntary actions, and they were able to invest their savings in ways that yielded high returns. Government interventions in the financial market that promoted savings and the efficient allocation of capital were central to these successes. Five of the more important interventions are examined here: promoting savings, regu- lating banks to fortify their solvency, creating financial institutions and markets, enforcing financial restraint, and intervening directly in the allocation of credit. Promoting Savings East Asian governments promoted national saving in several ways, from cre- ating financial institutions and regulating them to running small fiscal deficits or even surpluses. 250 The World Bank Research Observer, vol. 11, no. 2 (August 1996) Creating Postal Savings and Provident Funds The postal saving systems in Japan, Malaysia, Singapore, and Taiwan (China) were the most important of the institutions governments created to promote savings. These systems attracted multitudes of small savers by giving them secu- rity and convenient access. When Japan's postal saving system was created in 1875, other financial institutions had commonly excluded small savers or dis- couraged them by paying low interest rates or requiring minimum deposits too high for small savers (Mukai 1963). The postal saving system also provided convenient access through an extensive branch network of post offices, espe- cially in rural areas (Yoshino 1992, Shea 1993, and Chiu 1992). Japan further encouraged postal savings by exempting the interest paid on postal deposits from income tax (until 1988); Taiwan (China) since 1965 and Korea do the same below a certain threshold. The postal saving banks mobilized huge amounts of saving-up to 25 percent of national saving in Japan since the 1950s, 20 percent in Taiwan (China), and 12 percent in Singapore. In Japan postal saving was particularly important dur- ing periods of financial distress, when confidence in commercial banks waned. During the 1920s, for example, when the banks became unstable, households shifted their savings from banks to postal savings. In 1920 demand deposits in banks were six times the amount held in postal savings; by 1930 they had dwindled to just twice that amount. Although the positive role of postal saving is generally acknowledged, the compulsory pension plans of Malaysia and Singapore are more controversial. Two rationales are given for such plans. First, most developing countries lack private annuity markets, and the few that exist are barely functional. Indexed annuities, for example, are rarely available, and nonindexed annuities are often priced unattractively because of large transactions costs and problems of adverse selection. In Malaysia and Singapore not only did the government provide the annuities, but it also required individuals to participate in the plans. This leads to the second rationale: even when annuities are available, people may not save enough for their old age, and governments end up assisting elderly people who are unable to support themselves. To avoid this free-rider problem, government may require citizens to have at least a minimum level of savings for retirement. Unlike the social security systems of Europe and the United States, East Asia's programs were fully funded rather than pay-as-you-go. The distinction is im- portant because fully funded systems are more likely to increase the national saving rate. Also setting them apart from U.S. and European systems were the extremely high required contribution rates (up to 50 percent of salaries in Singapore and 28 percent in Malaysia) and the fact that the savings covered not just retirement but other purposes as well, such as providing the down payment for housing. The impact of provident funds on aggregate saving depends on how much volun- tary saving they displace. In East Asia displacement does not appear to have been a Joseph E. Stiglitz and Mariloui Uy 251 problem: private pension schemes grew alongside public ones as individuals pur- sued ways to support increased consumption during retirement. At worst the pen- sion funds' negative effects were so small that aggregate saving still increased. Dekle (1990) finds little empirical support of a negative effect of pension funds on private saving in Japan, while Noguchi (1985) finds a small negative effect. Singapore's Central Provident Fund had a positive effect on aggregate saving: had provident fund contributions remained at 10 percent, their required level in 1966, instead of rising to 35 percent during the 1970s and to 50 percent during the 1980s, average saving rates over the past two decades would have been about 4 percentage points lower (Monetary Authority of Singapore 1991). Regulating to Encourage Saving Governments used three types of regulations to influence national saving rates. Some regulations were designed to discourage consumption, some to enhance the safety and soundness of private financial intermediaries, and some to trans- fer resources from households to corporations. RESTRICTIONS ON CONSUMER CREDIT. Most East Asian governments discour- aged consumption by deliberately preventing mortgage markets and other in- struments of consumer credit from developing. With little or no consumer credit available to purchase housing, consumer durables, and other goods, households were forced to save the full amount if they wanted to buy a house or make other large purchases. Although demand for consumer durables increases with house- hold incomes, consumers could not borrow to buy these goods, so savings as a share of income rose rapidly. There is evidence that constraints on liquidity and on the development of consumer credit markets have significant effects on sav- ing and may explain some of the differences in saving rates across countries (Jappelli and Pagano 1994). Once households acquire the necessary consumer durables, their saving rates stabilize and even drop slightly. PRUDENTIAL AND OTHER BANK REGULATIONS. Prudential and other bank regu- lations significantly lowered the likelihood of bank failure. This increased secu- rity led to larger deposits. FINANCIAL RESTRAINT. The conventional wisdom is that financial repression depresses saving, particularly deposits in financial institutions. But moderate repression-what this article calls financial restraint-may actually increase sav- ing. Lowering interest rates transfers incomes from households to corporations, and because the corporate sector has a higher propensity to save, aggregate saving increases. This positive redistribution effect is partly offset by the nega- tive interest rate effect: as long as saving responds to changes in interest rates (the interest elasticity of saving is positive), lowering interest rates reduces house- hold saving. But the empirical evidence suggests that this effect is small (see 252 The World Bank Research Observer, vol. 11, no. 2 (August 1996) Ishikawa 1987 on Japan, Nam 1991 on Korea, and Sun and Liang 1982 on Taiwan, China) and that it is overwhelmed by the redistribution effect (Balassa 1989; Giovannini 1983; Gupta 1984). Although East Asian economies kept interest rates below equilibrium levels, financial repression was moderate compared with that practiced in many other developing countries with interest rate controls. Average interest rates in East Asia have been moderately negative at worst; many economies in other regions have had highly negative rates (figure 1). The policies associated with financial restraint also increased saving because governments such as Korea's pitted firms against each other to see which could achieve the highest rate of exports and investment. Winners were rewarded with access to cheap credit and the rents associated with other artificially created scarcities. Achieving high rates of investment required high rates of corporate saving. Financial restraint also helped promote household saving by making the financial system more stable. Financial repression (accompanied by entry re- strictions) enhanced the profitability-and thus the stability-of financial insti- Figure 1. Average Real Deposit Rates in Selected Countries, 1978-91 Hong Kong Indonesia Japan Korea, Rep. of M alaysia i_i____ i_i_i _ i Singapore I Taiwan (China) i j Average rate Thailand 6i i Standard deviation France Germany Sweden Switzerland . United Kingdom ii United States i Argentina i __i_i Bolivia Chile Ghana - Mexico Nigeria i i _ j j j a -60 -40 -20 0 20 40 60 80 100 Percent Note: Data for Hong Kong are for 1973-91. l)ata for Thailand are for 1978-90. Data for Argentina exclude the extreme inflation years of 1984, 1989, and 1990. Data for Bolivia exclude the extreme inflation years of 1985 and 1986. Deposit rates for Ghana are unavailable for 1989 and 1990 and unavailable for Switzerland for 1978-80. Source: IMF (1995). Joseph E. Stiglitz and Marilou Uy 253 tutions and, by boosting the franchise value of banks, provided strong incentives for banks to undertake prudent investments. Keeping the Economy Stable and Deficits Low Macroeconomic policies in East Asia are far more stable than those in most other developing countries. Macroeconomic stability has a positive effect on saving for a variety of reasons. Because most countries do not have fully indexed accounts, macroeconomic stability, particularly low rates of inflation, reduces the variability of return on saving; a more secure return may increase saving. High and extremely volatile inflation, which often generates large negative real interest rates, is particularly likely to discourage saving. The East Asian econo- mies have generally maintained positive and stable real interest rates on depos- its, paralleling interest rate trends in the United States and other industrial coun- tries (see figure 1). Governments' tendency to run small deficits or large surpluses contributed to macroeconomic stability, and the budgetary surpluses contributed directly to high levels of national saving. All East Asian countries have maintained consis- tently high public saving as well as growing private saving (figure 2). Did Interventions Increase Saving Rates? That East Asian governments undertook a variety of actions to increase na- tional saving is clear; what is less clear is whether saving rates were higher than they otherwise would have been. Certainly, the region's remarkably high saving rates suggest that government policies had an effect. Studies examining demo- graphic patterns indicate that only part of this difference can be explained using standard life-cycle models. Several authors have emphasized cultural factors in explaining cross-country variations in saving (Horioka 1990). Although these factors may play a role, they do not explain why saving rates were so much higher during the past three decades than in, say, the period before World War II. Surely, entire cultures did not suddenly change. Rapid growth is one of the distinguishing characteristics of the high- performing East Asian economies. Is it possible that high growth caused high saving, rather than the other way around? Studies have shown a strong correla- tion between saving and growth (Carroll, Weil, and Summers 1993) and be- tween saving and investment and investment and growth (World Bank 1989). But these correlations do not reveal the direction of causation (nor do they rule out the possibility that other factors are the cause of the observed changes).2 We performed standard tests (called Granger causality tests) to ascertain whether income predicts saving or whether saving more accurately predicts income. The results show income growth to be a better predictor of saving in Indonesia, Japan, Korea, Taiwan (China), and Thailand, but the results were ambiguous for Hong Kong and Malaysia. (In Singapore income growth was not a predictor 254 The World Bank Research Observer, vol. I1, no. 2 (August 1996) Figure 2. Public and Private Savings in Selected Developing Couintries Indonesia 1981-88 Japan 1955-80 i 1981-8 ., Malaysia 1961-80 1981-90 . . Singapore 1974-80 1981-90 _ Thailand 1980-85 1986-87 Argentina 1980-85 r 1986-87 .3 Public saving Brazil 1980-85 rivate saving 1986-87 Chile 1980-84 1985-87 - Ghana 1980-86 . 1985-87 Mexico 1980-87 ' -5 0 5 10 15 20 25 Percentage of GDP Source- Corbo and Schmidt-Hebbel (1991); Singapore Department of Statistics (various years); Lin (1991); World Bank data; Japanese Economic Planning Agency data. of the rapid increase in saving rates, a result consistent with a 1991 study by the Monetary Authority of Singapore showing that demographic factors and the policies of the Central Provident Fund accounted for the increase.) Finding empirical support for a progression from income growth to saving growth is important because it suggests a virtuous cycle in which high growth leads to high saving and high saving leads to high growth. Thus government policies may be able to transform an economy with low income growth and saving to one with high income growth and saving. Moreover, government poli- cies that enhance productive investments and raise income may have a com- pound effect on growth by increasing saving rates. Advocates of less government interference in the market claim that such in- terventions are both ineffective and undesirable. Why interfere with individuals' preferences? There are two broad answers. First, many government interven- tions represent attempts to remedy market failures, to make markets work bet- ter, or to create institutions to fill in gaps left by markets. In particular, the social returns to saving and investment may exceed the private returns, as is the case when there is learning from new investments and the benefits of the learn- ing are not fully appropriable by the investor. Second, saving rates affect the Joseph F. Stiglitz and Marilou Uv 255 intergenerational distribution of income. Markets may yield (Pareto) efficient resource allocations under ideal circumstances, but the distribution of income- including the intergenerational distribution of welfare-yielded by markets is in no way ideal. Regulating Banks to Enhance Their Solvency East Asian governments have imposed several regulations to enhance the sol- vency of financial institutions, thereby improving both the saving rate and the efficiency of resource allocations. As noted earlier, security is as important to saving as is the rate of interest. Individuals may save even if they have no confi- dence in any financial institution, but the returns are likely to be lower than could be obtained through a well-functioning financial system. Moreover, well- functioning financial institutions are essential to efficient financial intermedia- tion, channeling funds to their most efficient use in an economy. There is ample evidence that financial crises occur with remarkable frequency in the absence of government intervention. Private monitoring apparently does not suffice to prevent a financial crisis. Moreover, no single financial institution will exercise sufficient care on its own to avoid financial distress. Information failures (individuals may not be able to sort out good banks from bad, for lack of complete information) and credit links among banks mean that the effects of financial failures may spread far beyond an individual bank in distress-there are large negative externalities. These difficulties are exacerbated by the moral hazard problems that arise with undercapitalized financial institutions, because such institutions are more likely to take large risks-they have less to lose if a loan goes bad than better-capitalized institutions. (The market failure rationale for government intervention is treated at length in Stiglitz 1994.) Prudential Regulations to Keep Financial Institutions Sound Japan, Hong Kong, and Singapore began strengthening prudential regula- tions (sometimes referred to as regulations for safety and soundness) during the 1970s; Malaysia, Taiwan (China), and Thailand followed suit in the 1980s and Indonesia in the 1990s. The stringent regulations common to these economies were adopted for three sets of reasons. Some countries, such as Singapore, rec- ognized the importance of sound prudential regulations early on, not only for their internal capital markets but also for international commerce. Singapore has greatly benefited from its stringent regulations-financial services account for about 17 percent of its gross domestic product. The financial system's size is largely attributable to the confidence of the foreign financial and business com- munity. Even countries that do not aspire to being a regional financial center have recognized that closer links with international markets require that domes- tic banks meet international standards. 256 The World Bank Research Observer, vol. 11, no. 2 (August 1996) Other countries imposed such regulations as part of the development process. During early stages governments owned or directly controlled banks and other financial institutions. With development governments gave up direct control. Increased deregulation (of interest rates and entry, for example) reduced gov- ernments' leverage on bank behavior, requiring stronger indirect prudential regu- lation in its place. Some economies introduced prudential regulations only after experiencing financial difficulties. For example, Hong Kong strengthened its prudential regu- lations after the financial crises (brought on by real estate speculation) of 1965 and 1985. Prudential regulation takes a variety of forms, each requiring a different de- gree of supervision by the regulator. Capital, net worth, and collateral require- ments are easiest to monitor, although even they involve some degree of discretion (such as in valuing collateral). At the other extreme is judging the riskiness of particular transactions, which requires the active involvement of bank examiners. CAPITAL ADEQUACY REQUIREMENTS. Capital adequacy standards are probably the most important tool governments can use to ensure the solvency of financial institutions. These standards make it less likely that liabilities will exceed assets and provide incentives for banks to undertake appropriate risks. The savings and loan debacle in the United States can be partly attributed to the high risks assumed by banks with low or negative net worth. All East Asian economies have adopted the capital adequacy requirements set by the Bank for Interna- tional Settlements. Most economies had already imposed comparable standards on their own. The one exception is Indonesia, which implemented the new guide- lines only recently. COLLATERAL REQUIREMENTS. Regulators in East Asia have also encouraged banks to impose sizable collateral requirements to reduce the risks arising from defaults. This practice has earned East Asia's banks the reputation of being "pawnshop" banks. Although collateral requirements will not keep banks sol- vent, regulators have adopted this conservative system to limit banks' ability to take risks. One unintended impact of this practice has been to tie banks to the fortunes of the real estate market because most collateral has been in the form of real estate. Thus banks have overextended their lending during periods of high asset inflation, exposing themselves to greater portfolio risks during peri- ods of declining asset value. In Japan, for example, the dramatic fall in real estate prices in the first half of the 1990s created portfolio problems for banks that had engaged in speculative real estate lending during the 1980s, when asset prices were high. LENDING RESTRICTIONS. One of the most important objectives of East Asian policymakers was to discourage speculative lending, which has become the main Joseph E. Stiglitz and Marilou Uy 257 source of financial disruptions in Hong Kong, Malaysia, Thailand, and, recently, Japan (table 1). The adverse impact of speculative lending has been aggravated by bank lending to related parties, as bank owners sought to capture the gains from their speculation. In response regulatory authorities have increasingly re- stricted lending for real estate and to related parties-as well as lending concen- trated on a few borrowers. Restrictions on related lending have been difficult to implement, however, because disclosure rules are generally poor, and in Indone- sia, Japan, and Thailand, banks and firms have interlinked ownership, and com- panies are closely held. DIRECT SUPERVISION. Except in Indonesia, central banks (often working with ministries of finance) in East Asia have done a good job of supervising commer- cial banks' loan portfolios, resulting in a lower proportion of nonperforming loans than in many other developing countries. Central banks have also super- vised bank management, restraining the entry of potentially fraudulent or incompetent lenders. Singapore takes pride in the fact that its regulatory au- thorities detected problems in the Bank of Credit and Commerce International- problems that escaped detection by allegedly more sophisticated regulators, in- cluding those in the United States and United Kingdom-and refused to allow it entry. There are major exceptions to this generally favorable picture of bank supervision, however: the financial difficulties of banks in Hong Kong, Indone- sia, Korea, and Malaysia during the 1980s were brought about in part by weak supervision. Indonesia recently introduced more rigid bank supervision when it liberalized entry for private banks and as the problem of rising nonperforming loans became apparent. The East Asian style of regulation has, at least until recently, been based more on regulatory discretion and constant interaction between regulators and banks than on the more structured rules that characterize supervision in indus- trial countries. This approach enables regulators to provide banks with feed- back on the riskiness of their portfolios. Contemporary supervision practices in most East Asian countries (Japan, Malaysia, and Thailand are good examples) seem to combine modern prudential rules with traditional interactive monitor- ing, in which supervisors elicit cooperation by using the government's leverage over branch licensing, rediscounts, and other regulations. Although allowing such discretion creates the potential for abuse, abuses do not seem to be preva- lent. In other countries, however, this style of regulation could create problems. PRUDENTIAL BEHAVIOR OF GOVERNMENT BANKS. Unlike government-owned banks in many other developing economies, those in Korea (until their privatization in 1983), Singapore, and Taiwan (China) seem to have behaved prudently. In other countries political considerations often distort the lending decisions of government-owned banks. Many such banks have ended up lending to make up the losses of inefficient public enterprises. And because governments have deep pockets and tend to recapitalize public banks when they run into 258 The World Bank Research Observer, vol. 11, no. 2 (August 1996) financial trouble, the regulators and managers of publicly owned banks are of- ten less concerned with solvency than private banks are. East Asian govern- ments took several steps to minimize these problems. Taiwan (China) avoided these risks by imposing strict collateral requirements and giving the employees of public banks incentives to act prudently, going so far as to penalize employ- ees whose loans did not perform. Korea imposed strict performance criteria to guide banks' lending decisions. In Malaysia public officials are prohibited from serving on the boards of public banks. But publicly owned banks in Indonesia and a few in Malaysia have not been properly monitored and consequently have experienced high levels of nonperforming loans. REGULATING NONBANK FINANCIAL INSTITUTIONS. Regulators in East Asia and other developing countries have also been concerned with the appropriate regu- lation of nonbank financial institutions such as merchant banks, leasing compa- nies, and cooperatives. Since the early 1980s these institutions have grown in number, but they have been less closely regulated than commercial banks, lead- ing to several major insolvencies (see table 1). For instance, during the first half of the 1980s, Malaysia's deposit-taking cooperatives and Thailand's numerous finance companies became insolvent. These failures threatened the solvency of commercial banks, which often used nonbank financial subsidiaries as a way to avoid close scrutiny by regulators. In response to these insolvencies regulators have increased their supervision of nonbank financial institutions. Protecting Banks from Competition Nearly all governments regulate the entry and operations of financial institu- tions to ensure that new entrants and incumbents are safe and solvent. But most East Asian governments (with the exceptions of Hong Kong, Singapore, and recently Indonesia) have gone further, restricting entry by new domestic com- petitors and by foreign banks. As a result few new banks have been established, so the financial sector has expanded largely by licensing new branches of exist- ing banks. Japan's banking system has expanded enormously since the 1950s even though no new banks have been allowed in Japan since then. In Korea and Taiwan (China) only a few new private banks have been licensed in the past few years; competition comes primarily from nonbank financial institutions and the curb market. Not surprisingly, banking in many East Asian economies-Indo- nesia, Korea, Taiwan (China), and Thailand-has been highly concentrated. Governments had several reasons for imposing entry restrictions. The most widely cited reasons relate to prudential concerns. Entry restric- tions develop from the view that governments must ensure that only trustwor- thy bankers handle depositors' money. This view is reinforced by the not en- tirely valid belief that financial systems with a small number of large banks are less risky than systems with a large number of small banks (Vittas 1991). Many policymakers fear that excessively competitive systems-with low profit rates- Joseph E. Stiglitz and Marilou Uy 259 Table 1. Nature, Causes, and Resolution of Bank Crises in East Asia Fconomny Nature of finiancial distress Causes Nature of bailout or rescue I-Hong Kong Nineteen deposit-taking * Large exposure to real * The government revamped regulatory and 1982-83 companies failed. estate lending, fraud and auditing system and liquidated troubled mismanagement, and weak deposit-taking companies. prudential regulation. Hong Kong Four banks, including a IHigh international interest * The government took over the larger 1983-86 major international bank, rates. banks and introduced new management, became insolvent. including top executives seconded from *Large exposure to real the largest commercial bank. Those banks estate lending and spillover not taken over by the government effects from 1983 crisis received credit from other commercial because these banks owned banks. deposit-taking companies as subsidiaries. Malaysia The failure of one deposit- * Fraud and speculation in * The government rescued 24 insolvent 1985-88 taking cooperative in 1986 real estate and stocks. cooperatives and consolidated and caused runs on 32 (of 35) merged weak finance companies. The others. In addition, 4 (of 38) * Deterioration in terms of central bank injected fresh equity capital banks and 4 (of 47) finance trade. and replaced management of some banks. companies were also in financial distress. Overall, 10.4 percent of banking system deposits were affected. Thailand Government's cost to hail * Fraud and speculation on The government liquidated 24 finance 1983-87 out 50 finance companies real estate and exchange companies and merged another 9, and the was estimated at US$190 rate transactions. central bank took over the other 17 and million, or 0.48 percent of sold them to new investors (including GNP. other banks). Five commercial banks * High concentration of * The government bought some shares of accounting for 24 percent of unsecured insider loans. troubled banks. commercial bank assets were in financial difficulties in * High international interest * To provide emergency loans to troubled 1986-87. rates. banks, the government created a "lifeboat fund" financed by contributions from commercial banks. Taiwan (China) Four trust companies and 11 * Cooperatives arbitrating * Healthier banks took over management 1983-84 cooperatives failed. from an artificially steep or bought the shares of failed banks. yield curve. Singapore Domestic commercial bank's * Macroeconomic reasons. * The government worked out a two-year 1982 nonperforming loans rose to write-off period (using tax breaks). about US$200 million, or 0.63 percent of GDP. Japan A central bank report * Excessive exposure to real * The government encouraged mergers of 1991 estimated the size of problem cstate lending (90 percent weaker banks with healthier ones. loans of the top 21 banks to of bad loans), and a steep be between 3.5 percent and decline in real estate * Groups of banks provided emergency 4.8 percent of banking prices. loans to weaker banks. system assets. Informal estimates of the amount to be * Inadequate prudential * Nonperforming loans were to be trans- written off are as much as supervision. Banks were ferred to a separate financial institution, 1.5 percent of banking able to increase their and the cost of the write-offs was to be system assets. exposure through loans to shared among commercial banks. their nonbank affiliates. Source: World Bank (1 993a). ON are also excessively fragile.3 Restricting competition increases profits, and higher profits strengthen the banking system (provided banks do not simply distribute profits to shareholders). Higher profits also increase a bank's franchise value and its incentive to maintain its reputation, thus encouraging more prudent be- havior. Japan's limits on entry appear to have been motivated largely by pru- dential concerns. From the Meiji era (starting in 1868) until the 1920s, the Japa- nese banking system was highly unstable-banks were numerous, bank failures (especially of small banks) occurred periodically, and consumer confidence in banks declined. The government responded to a spate of bankruptcies in the 1920s by encouraging mergers and by regulating banks more closely. Between 1937 and 1940 the number of banks was halved, then halved again through mergers (Dekle 1992). Banks also formed cartels to set interest rates, further boosting their profitability. The second set of reasons relates to efficiency concerns. Some bank regula- tors argue that larger (and fewer) banks are more efficient in intermediation. This argument is justified by economies of scale in information gathering and monitoring, but it does not provide a convincing basis for restricting entry. If economies of scale exist, government intervention should not be required to realize them. Besides, protection encourages monopolistic and inefficient prac- tices, undermining the efficiency-enhancing arguments for restricting entry. A third argument, this one pertaining to entry of foreign banks, is that do- mestic banks need to be protected until they can compete on an equal basis with established foreign banks. That requires more than just learning banking tech- nology. Savers may have more confidence in foreign banks even if domestic banks are equally efficient in providing services and equally capitalized, putting domestic banks at a competitive disadvantage. Two factors amplify concerns about foreign entry. The first is that lending patterns may differ as a result of different knowledge. Foreign banks may be more familiar with foreign-owned firms, for example, and may thus exclude domestic firms or charge them higher interest rates. The opposite may be true for domestic banks. Accordingly, the government may want to encourage de- posits into domestic banks. The second concern about foreign entry is that do- mestic banks may be more subject to "window guidance"-that is, they may be more responsive to the monetary authorities' efforts to control money supply through discount rates or reserve requirements. Entry restrictions are popular because they give governments a powerful dis- cretionary tool for influencing the behavior of banks. Banks have an incentive to respond to government requests lest they forgo the additional profits from asset growth. Because the potential penalties and powers that governments can exer- cise over domestic banks are greater than those they can exercise over foreign banks, governments have an incentive to restrict the entry of foreign-owned banks. This differential power is reflected in the assertion by some East Asian bankers that they would not mind the entry of foreign banks if the banks com- peted on a level playing field. In their view the demands government puts on 262 The World Bank Research Observer, vol. 11, no. 2 (August 1996) them-which it cannot put on foreign banks-puts them at a disadvantage. East Asian governments have exerted influence over domestic banks not only in the level of lending (as is common in industrial countries) but also in the direction of lending-toward investments where social returns are viewed as being particu- larly high. These governments have attempted to balance the advantages of competition (greater efficiency) with the perceived disadvantages (loss of discretionary power and lower profits, perhaps leading to less stable financial institutions). In some cases governments have been able to strike a reasonable balance. The enhanced stability of the financial system as a result of entry restrictions led to financial deepening, while the cost of the reduced competition was reflected more in re- duced innovation than in higher lending margins or spreads.4 And there is con- siderable evidence that governments were able to guide the allocation of credit, partly as a result of their discretionary powers (for example, with respect to branch banking). Creating Development Banks and Financial Markets to Fill Credit Gaps During the past few decades East Asian governments have helped develop financial markets by creating financial institutions to fill gaps in the types of credit private entities provide. Most countries have established long-term credit banks and specialized institutions providing credit for agriculture, small firms, and housing. Some East Asian governments have also established commercial banks that cater to a specific group of borrowers (such as Malays or the Islamic community). Here we focus on development banks. Banks offering long-term credit have been among the most common government-created financial institutions. The Japanese government created the Industrial Bank of Japan in 1902 in part because of the absence of alter- native sources of long-term credit for business investment (such as bond and equity markets). The government also recognized that commercial banks were poorly suited to extending long-term credit. Why did the government choose to create long-term credit banks instead of trying to create securities markets? Some observers have emphasized the key advantages that banks have over markets, advantages that are particularly im- portant at early stages of development. Most important, banks have the institu- tional capacity and incentives to monitor business borrowers closely; such moni- toring becomes critical when there is no well-developed industry of financial analysts. Close supervision is needed to assess when to extend and when to withdraw credit and to distinguish between situations where profits are low because of bad luck or an economic downturn and those where they are low because of bad management. Without such close supervision, the banks might be reluctant to take on risk and might focus on short-term profits instead. Joseph E. Stiglitz and Marilou Uy 263 Governments set up long-term credit institutions rather than lending or in- vesting directly in firms largely because they believed that a certain amount of independence would enhance the performance of banks and firms. Long-term investments require selection and monitoring. Government agencies are not de- signed to screen and monitor commercial projects and might be subject to po- litical influence. The creation of the Industrial Bank of Japan reflected a recog- nition that long-term credit banks had served as effective monitors of firms in other countries, especially among firms not affiliated with major conglomerates (Packer 1992). Relationship of Development Banks with Government Governments did more than create the development banks. They also pro- vided assistance, particularly in developing sources of funds during the banks' early years. For example, the Japanese government initially bought a substantial share of the bonds issued by private long-term credit banks and was a catalyst in ensuring that other private banks and financial institutions subscribed to these bonds. It allowed development banks to issue long-term bonds or debentures, whose market the government helped create. This privilege helped redress the mismatch between the maturity structure of the banks' assets and their liabili- ties, a problem that had plagued commercial banks. Limiting competition en- abled the long-term credit banks to obtain funds more cheaply than they other- wise could have. The government went even further, encouraging government units and commercial banks to purchase the long-term bonds, which allowed the development banks to obtain funds at an even lower rate. The Thai govern- ment provided similar privileges to its private long-term credit banks. Why do governments form both public and private development banks? The main advantage of private development banks is that they are further removed from government, although the government can still exercise considerable influ- ence. The Industrial Bank of Japan, for example, could choose projects accord- ing to its own commercial criteria, but it had to select firms from within priority industries identified by the government. There are tradeoffs: the closer the link between banks and the government, the easier it is for the government to exer- cise influence and the less likely it is that commercial criteria will be employed. Private development banks may be more credible in setting commercial criteria for investment projects, but private ownership places an additional monitoring burden on government to guard against misappropriation of government assis- tance. Japan's private and public development banks illustrate how lending ac- tivities are divided. The privately owned Industrial Bank is a major lender of long-term funds to industries that are not necessarily being promoted by gov- ernment, and it has acquired a substantial reputation within rhese industries. The government-owned Japan Development Bank, on the other hand, focuses its lending on industrial activities that have received the highest subsidies, such as sea transport, mining, electric power, and transport machinery. 264 The World Bank Research Observer, vol. 11, no. 2 (August 1996) The Influence of Development Banks Development banks lent substantial funds on a long-term basis in Indonesia, Japan, Korea, and Taiwan (China), but not in Hong Kong, which has no devel- opment banks, or Thailand, where industrial development banks hold only 1 percent of the assets of the financial system. The Korean Development Bank accounted for about a third of all loans and guarantees in the 1970s. Taiwan's Bank of Communications accounts for about half the assets of the banking sys- tem. In Japan the development banks together accounted for about two-thirds of outstanding loans for equipment investment in the 1950s and about half in the early 1960s. The Japan Development Bank alone accounted for 45 percent of equipment lending in the early 1950s. Since the 1950s, however, the Japan Development Bank's lending has accounted for only 2 percent of newly lent funds (Kawaura 1991; Horiuchi and Sui 1993). In recent years the share of development lending has been small even in new growth industries. DEVELOPMENT LENDING AS A COMPLEMENT TO COMMERCIAL LENDING. For sev- eral reasons, however, development banks have been more influential than their small share in lending might suggest. Because they had close ties to the govern- ment, their lending provided information to entrepreneurs and other banks on the areas that the government was promoting. In addition, other financial insti- tutions valued information on the development banks' choice of clients (as dis- tinct from sectors). This signaling effect works, of course, only if development banks have sound institutional reputations, which they did in Japan, Singapore, Taiwan (China), and, by and large, Korea. Development banks did more than provide information. They also initiated loan syndication and other forms of cooperative lending with private banks. Development bank lending often created the perception of "risk sharing"-gov- ernment support made it more likely that government assistance would be forth- coming in the event of a problem (Yasuda 1992). The perception of implicit government insurance of priority activities was fortified by several government bailouts within priority sectors. At the government's urging, the Industrial Bank of Japan led syndications for firms in distress and forced their restructuring to avoid liquidation or external takeovers. In Indonesia the government bought 35 percent of the equity of a large cement plant, a priority activity, when the plant became financially troubled (Kunio 1988). To protect firms and banks, Korea extended cheap loans during the 1980s to firms that were unable to meet their debt obligations. Thus development lending complemented private sector lending rather than displaced it. To see how the Japan Development Bank's lending to medium-size and large firms affected corporate investment, Horiuchi and Sui (1993) ana- lyzed a sample of 477 medium-size firms listed in the Tokyo Exchange in 1965. They found that an increase in a Japan Development Bank loan to a firm in- duced more than an equal increase in the firm's investment spending. Moreover, Joseph E. Stiglitz and Marilou Uy 265 private banks extended credit more actively to firms that borrowed from the development bank than to those that did not. Japan Development Bank loans preceded the increase in a firm's borrowing from private banks, further evi- dence of the signaling effect. Japan Development Bank borrowers also enjoyed more favorable borrowing conditions from other banks and were less sensitive to changes in the cost of capital than were firms that did not borrow from Japan Development Bank. Even private long-term credit banks seemed to follow the lead of the public development banks. Of 161 listed companies that had a private long-term credit bank as its primary lender in 1967, nearly half also obtained loans from the Japan Development Bank, and 20 percent obtained loans from the Export- Import Bank of Japan. Establishing a causal link between development bank and commercial bank lending is difficult, but several pieces of evidence support the existence of such a link, beyond the evidence on timing noted earlier. Incentives for cooperative financing between Japanese development banks and commercial banks were strong, for example. Commercial banks that cofinanced projects initiated by the long-term credit banks received preferential treatment in purchasing the finan- cial debentures issued by the long-term credit banks. That development banks influenced lending is important, of course, only if their lending patterns differed from those that commercial banks would have chosen on their own. There is evidence that this is the case. As noted earlier, the discrepancies between private and social returns may be large. Even if such discrepancies did not exist, however, governments believed that they did and expected substantial gains from channeling resources to prior- ity areas where social returns exceeded private returns. Development banks were one of their main tools for channeling such resources. Priority sectors varied across countries and over time. Most countries gave some priority to exports; the Korean Development Bank was a major financial intermediary for loans to heavy and chemical industries in the 1970s, and both the Bank of Communications in Taiwan (China) and Singapore's Development Bank have been active in financing high technology. SPECIALIZED DEVELOPMENT BANKS. Most East Asian governments have also created specialized banks in areas where private lending has been viewed as inadequate, most notably in agriculture and small-scale enterprises. Thailand's agricultural development bank, for example, caters to small farmers who do not have access to commercial bank lending. The bank has reached 80 percent of potential agricultural borrowers (including most small farmers), even though its lending to agriculture is much smaller than total commercial bank lending. Loan rates are slightly lower than commercial bank rates and substantially lower than informal market rates. The bank operates on a cost plus basis (over subsidized funds) and has prudently reduced lending in response to nonrepayment. The bank has been financially sound despite high operating costs and the expected 266 The World Bank Research Observer, vol. 11, no. 2 (August 1996) poor repayment record. To increase farmers' access to formal credit, the Thai government has complemented financial reforms with legal reforms that enable small farmers to use their land as collateral for loans. Because markets for mortgages-especially for low-income housing-are underdeveloped, East Asian governments have also created financial institutions for housing finance. Japan's postwar government created the Housing Corpora- tion, whose lending has accounted for a growing share of the government's fis- cal investment and lending program. Singapore is probably the most prominent example of intervention in housing finance. In 1960 the government established the Housing Development Board to build and provide subsidized mass housing, which policymakers viewed as essential to maintaining social stability. The gov- ernment subsequently allowed would-be buyers to use part of their provident fund contributions to purchase the subsidized housing. The housing subsidies may have resulted in lower wages (particularly in a country such as Singapore, where wages are determined through national bargaining), but the lower wages boosted profits and, consequently, retained earnings. Moreover, the housing program increased social stability by raising living standards and increasing the net worth of households, giving them a greater financial stake in their society. Why Have Development Banks Succeeded in East Asia? Many developing countries have been unsuccessful in promoting develop- ment banks. Eighteen development banks examined in a World Bank (1989) study had, on average, half of their loans in arrears. Even in East Asia, where the experience with development banks has been mostly positive, development banks have failed. Japan's insolvent Reconstruction Finance Bureau was closed in 1952 and Thailand's Industrial Bank in 1959. Some development banks in Indonesia and Malaysia are reporting a rising volume of arrears. The most common causes of failure are political pressure to finance bad projects and poor incentives for financial institutions to screen and monitor projects. Among the key ingredients of the success of many other East Asian develop- ment banks was an insistence on commercial standards within the priority sec- tors. Successful development banks transformed themselves from government agencies financing development pro]ects into more market-oriented financial enterprises. The largest development banks of Japan, Korea, Singapore, and Taiwan (China) have consistently demonstrated such a pattern. Although government monitoring seems to have had a salutary effect in East Asia (Cho and Hellman 1993), government intervention has had a negative im- pact in other countries. This differentiated outcome is likely the result of gov- ernment efforts to shield development banks from political interference. Japan, Korea, and Taiwan (China) appointed established officials (from ministries of finance) as chairmen so that they could better withstand pressure from other parts of the government. That does not explain why these officials did not sub- vert the development banks for their own purposes, however. And the fact that Joseph E. Stiglitz and Marilou Uy 267 these officials were competent and honest does not mean that in other countries and at other times oversight by a ministry of finance will provide an adequate check against abuse. Some East Asian countries have also controlled types of lending: Thailand simply barred its development bank from lending to state enterprises. Successful development banks also instilled a high level of profes- sionalism and institutional identification in their staff, making government in- tervention-other than in establishing priority sectors-difficult. By making lend- ing to a nonperformer a criminal offense, Taiwan (China) made sure that loan officers did not give in to political pressures or abuse their discretion. Moreover, private banks often cofinanced development bank projects, thereby serving as a check on development banks' lending criteria. Thus the information flow be- tween development banks and commercial banks was reciprocal. Creating Financial Markets Only a small portion of long-term investments in East Asia have been fi- nanced by corporate bonds. Except for Thailand and Korea since 1980, bonds accounted for much less than 10 percent of the net financing of nonfinancial corporations among five high-performing East Asian economies for which data exist (World Bank 1993a). One obstacle to the emergence of bond markets in these economies is the absence of a market for government securities-because the governments do not run deficits, they do not need to borrow. With no mar- ket for government securities, there is no benchmark risk-free rate, and so mar- kets must determine both the risk-free rate and the risk premium associated with a specific corporate bond. Hong Kong's government has responded to this limitation by auctioning government bonds-even though it does not need the financing-to provide a benchmark risk-free rate and eventually help create a market for corporate bonds. Malaysia and Singapore are considering doing the same. Other East Asian economies have also taken steps to foster the growth of bond markets. Malaysia, for example, established a rating agency for bond is- sues in 1991. Hong Kong, Taiwan (China), and Thailand have strengthened their legal infrastructure for securities (bonds and equity) issues. And in Ko- rea-which has the region's most rapidly expanding bond market-the govern- ment has been issuing guarantees. Like bond markets, equity markets provide a small fraction of the net financ- ing of nonfinancial corporations in East Asia, although the relative importance of equities rose slightly in Korea and Thailand during the 1980s. In recent years East Asian governments have increased their efforts to promote stock markets. Hong Kong has strengthened disclosure requirements and implemented laws against fraud in response to financial disruptions experienced in the stock ex- change. Korea, Taiwan (China), and Thailand have provided preferential cor- porate tax measures to encourage companies to list on the stock exchange. Ko- rea introduced these incentives in the 1970s and since the 1980s has indirectly 268 The World Bank Research Observer, vol. 11, no. 2 (August 1996) promoted equity issues by encouraging firms to lower their debt-equity ratios. All of these East Asian economies have expanded their stock exchanges; Indone- sia, Korea, and Singapore have introduced over-the-counter markets. Enhancing Growth through Financial Restraint and Credit Allocation Earlier we showed how, contrary to standard arguments, financial restraint may have increased national savings. A standard argument against financial restraint has been that it impedes efficient resource allocation by preventing a free market auction from occurring (World Bank 1989; Fry 1988). As recent work on credit markets has emphasized, however, credit is not allocated by auction even in perfectly competitive markets. In a world of asymmetric infor- mation, banks do not allocate loans to the highest bidders, but rather to those borrowers they deem most likely to repay. Even when the adverse selection and incentive effects associated with higher interest rates do not induce credit ra- tioning (Stiglitz and Weiss 1981), these effects do mean that moderate financial restraint on lending rates reduces default rates and increases the social returns to lending. Financial restraint has further allocative benefits. To the extent that lower deposit rates are reflected in lower lending rates, financial restraint enhances the ability of firms to increase their equity, and hence their level of investment and their ability and willingness to take prudent risks (Stiglitz 1994). To the extent that lower deposit rates are not passed on in lower lending rates, financial re- straint enhances bank equity, and hence banks' ability and willingness to make loans. And greater bank equity enhances the stability of the financial system. One of the benefits of stable financial systems is the organization-specific na- ture of information: bank failures destroy information that is valuable to ensur- ing the efficient allocation of capital. Financial restraint also has incentive effects. Higher bank profits increase the franchise value of banks, providing strong incentives for prudential behavior. Appropriately chosen bases for allocating scarce credit can also provide strong performance incentives. In Japan, Korea, and Taiwan (China), competition for access to credit generated high marginal returns to greater effort, as measured, for instance, by exports (Stiglitz 1994). Because the shadow value of access to capital was high, this prize was valuable. The empirical evidence usually cited against financial restraint, based on cross- country regressions showing a positive relation between real interest rates and output growth (Gelb 1989), is as faulty as the theoretical arguments. This evi- dence suggests that high real interest rates are associated with increased finan- cial depth, a modest increase in savings and investment, and more productive investments (World Bank 1989). That savings and income have continued to grow in East Asian countries despite financial restraint raises some questions Joseph E. Stiglitz and Marilou Uy 269 about the findings of these studies: does anyone really believe growth would have been even faster in East Asia in the absence of financial restraint? There are three problems with these studies (see Stiglitz 1994 and Murdock and Stiglitz 1993 for greater detail). First, they fail to distinguish between large and moderate degrees of interest rate constraints. Highly negative real rates have a severely negative impact on economic performance, which the regres- sions capture, but moderate interest rate restraints may yield positive effects, which the regressions do not reflect. Indeed, when the sample is split between developing countries with high and those with moderate degrees of financial repression, the positive correlation between real interest rates and growth dis- appears for the countries with moderate financial restraint. Second, countries with severely negative real interest rates have had both severe financial repression and bad macroeconomic policies (as reflected in high rates of inflation). These policies, rather than the financial restraint, may ac- count for their poor economic performance. That suggests that the gains to growth come less from rationalizing interest rates and more from decreasing the distortionary effects of high inflation. The third problem is one of identification: real interest rates may be low not because of financial restraint, but because there are no good investment oppor- tunities. This problem can be addressed in two ways: using a simultaneous equa- tions model or trying to measure financial restraint directly. Such assessments are possible for Korea and Taiwan (China), using data on curb market rates. For these cases the degree of financial restraint does not appear to explain economic growth, although other variables, such as inflation, do. Priorities Reflected through Directed Credit All East Asian countries have directed credit in varying degrees to support industrial policies or social objectives. Countries direct credit for several rea- sons, ranging from a perceived contrast between social and private economic rates of return to more immediate concerns, such as providing national security. Like other economies, high-performing East Asian economies use two broad types of intervention. First, the government directs credit to priority firms, groups, industries, and activities (such as exports or high-technology projects). Second, the government directs credit for social reasons, often to small farmers, small and medium-scale enterprises, or a specific ethnic group. In both cases the gov- ernment directs credit by investing in public enterprises, using its development banks to lend to priority areas (and to signal to other financial institutions what these areas are), and compelling commercial banks to lend to designated activi- ties. Although the rationale for and categories of directed credit do not differ between the East Asian economies and other economies that have used directed credit, the high performers in East Asia have implemented their programs with more moderate credit subsidies and with institutions that enabled better selec- 270 The World Bank Research Observer, vol. 11, no. 2 (August 1996) tion (through performance criteria) and monitoring of promoted projects, re- sulting in higher repayment rates for subsidized loans. Among East Asian economies Japan and Korea most pervasively directed credit to promote specific firms and industries-Japan during its postwar recon- struction and Korea during its promotion of chemical and heavy industries in the 1970s. During the 1950s the Japanese government's financing amounted to nearly a third of new equipment lending to industry; most went to shipbuilding, electrical power, coal, sea transport, and machinery. The results of the program in Japan are controversial: many successful growth industries were not heavily supported by the government's credit programs; among those provided with credit subsidies, some (such as shipbuilding) increased their exports, while oth- ers (such as coal mining) continued to decline. The results of Koreas's policy loans also have been mixed (Stern and others 1992). Some of Korea's heavy industries, such as steel, electronics, and passenger cars, became leading export- ers during the 1980s, while others became financially distressed. The chemical and heavy industries policies also increased the concentration of wealth among conglomerates and contributed to high firm leverage (Cho and Kim 1995). Indonesia and Malaysia had few successful experiences with selective credit intervention and abandoned the schemes once the negative effects of the policy became apparent. Thailand has avoided credit programs directed at specific firms and industries. Assessing the success or failure of a directed credit program is difficult for three reasons. First, as mentioned earlier, there is usually no way of knowing whether growth would have been higher or lower in the absence of the directed credit program. Second, a good program requires risk-taking, which means that failures are inevitable. A program with nothing but successes would necessarily have been too conservative. Third, many of the returns may be long term, so current profitability may not provide an adequate measure of success. Thus the measure of Korea's chemical and heavy industry program should not be how those industries fared in the early 1980s, but what the structure of the economy looks like in the late 1990s. By the same token, low profits may reflect cyclical conditions rather than long-run prospects. Thus, although it is difficult to determine whether directed credit pro- grams were successful, the evidence shows that government lending did not simply displace private lending-it affected the allocation of resources. More- over, credit was directed to areas with high social returns. During the 1950s in Japan, for instance, the bulk of directed credit went to basic industries that supplied essential inputs for growth in other parts of the economy. Once these basic industries were developed, Japan promoted other industries (such as machine tools) whose expected spillover effects on the economy were large (JDB/JERI 1993). Many other developing countries have failed in their industrial credit pro- grams. At least six factors set the successful East Asian economies apart from others. First was their ability to change credit policies rapidly when they real- Joseph E. Stiglitz and Marilou Uy 271 ized the policies were not functioning properly. Second, unlike many developing countries that funneled a large portion of directed credit to public enterprises, the high-performing East Asian economies directed credit mainly to private en- terprises. Even in East Asian countries that made loans to state-owned firms- Indonesia, Malaysia, and Singapore-the proportions of total credit were not persistently high, the state firms tended to perform better financially than state firms in other countries, and the interest rate subsidies were not substantial (except in Indonesia). Third, all East Asian economies directed credit to indus- try based on broad functional criteria (such as whether the firm produced ex- ports), typically using objective performance measures. Fourth, credit was usu- ally more common than outright subsidies, which were limited. Fifth, directed credit was more limited than elsewhere. Although directed credit amounts to as much as 75 percent of the loans of financial institutions in some countries, even in Korea (which used directed credit most aggressively) directed credit amounted to only about 40 percent of total credit, and in Japan it never exceeded 15 per- cent. Sixth, monitoring was more effective, so default rates were lower. Can the East Asian Experience Be Replicated? Most East Asian governments' financial sector interventions were meant to remedy market failures. Such failures occur even in industrial countries, and governments impose extensive regulation to deal with them. Market failures are usually more significant in developing countries, and governments' ability to correct them is more circumscribed. What is remarkable is that East Asian gov- ernments undertook actions (such as prudential regulation) similar to those taken by more industrial countries, and that they did so at an earlier stage of develop- ment. Moreover, these regulatory initiatives succeeded without the abuses that often accompany them elsewhere. East Asian governments sought not to re- place markets and market forces, but to use and direct them. Government lend- ing programs complemented private lending; they did not replace or displace it. Although governments established priorities for lending-and discouraged lending for real estate and consumer goods-they still employed commercial standards. How replicable are these interventions in other developing countries? That so many East Asian economies were successful suggests that success was not just fortuitous, the result of, say, unusually good civil servants. The cultural diver- sity of the region makes explanations based on unique cultural factors unpersuasive (Stiglitz 1994). Many of the specific institutions, programs, and practices that contributed to East Asia's success can easily be replicated, including the region's large invest- ment in education. The resultant high level of educational attainment undoubt- edly contributed to governments' ability to execute their programs. Several of the institutions that contributed to high savings-such as the postal saving sys- tem and provident funds-could easily be introduced elsewhere. Prudential regu- 272 The World Bank Research Observer, vol. 11, no. 2 (August 1996) lations, particularly capital adequacy requirements and controls on real estate lending, are essential and replicable. The adaptability of government policies- the ability to abandon policies when they fail and to change policies with chang- ing circumstances-is clearly a lesson of general applicability, although it is hard to design institutions that capture that lesson. The main concern with activist policies such as those pursued in East Asia is abuse of political power: activist policies generally entail giving governments discretionary powers that can easily be abused. In other countries such abuse explains or at least contributes to the failure of activist programs. But many of the ways in which East Asian institutions were designed reduced their vulner- ability to political abuse-and these institutional arrangements can be repli- cated. The use of performance-based criteria for allocating credit, for example, limited discretion. Similarly, requiring that commercial criteria be satisfied to receive credit and requiring borrowers to raise part of their funds on their own and to put up their own equity are replicable practices that enhance the likeli- hood that funds will be allocated to good ventures and reduce the likelihood of political abuse. Furthermore, moderate subsidies and financial restraint reduced rents (rela- tive to those found in many developing countries), further attenuating incen- tives for abuse. So did the extensive use of competition (Stiglitz 1994). Japan used several tools, including interest rate regulation and competition from non- bank financial institutions, to curtail potentially high rents from entry restric- tions. At bottom, however, is a more fundamental issue: why were the governments of East Asia able to implement policies that lessened the potential for abuse? The answer lies in political economy, which is beyond the scope of this article. A few studies have touched on this issue (Campos 1993), but much remains to be done. Notes Joseph E. Stiglitz is chairman of President Clinton's Council of Economic Advisers, on leave from Stanford University, where he is professor of economics, and Marilou Uy is with the South Asia Country Department of the World Bank. This article was written as part of the World Bank's project on the East Asian miracle. The views expressed in it are solely those of the authors and in no way rcprcsent those of thc organizations with which they are affiliated. The authors gratefully acknowledge the research assistance of Kevin Murdock, as well as the comments of participants at the many seminars at which portions of this article were presented. The authors also acknowledge the assistance of Brian Casabianca in preparing the figures. 1. Many of the theoretical and technical details of the analysis presented in this article, especially on the market failure framework, are discussed in Stiglitz (1985, 1989, 1994, and 1996) and Greenwald and Stiglitz (1986). 2. Standard economic theories have identified several reasons why savings rates might depend in part on income growth rates (Carroll, Weil, and Summers 1993), although the sign of the relationship is ambiguous. First, as growth rates increase, households become less confident that such growth will be sustained; it becomes more probable that growth Joseph E. Stiglitz and Marilou Uy 273 will fall, and so individuals are induced to put away more for a "rainy day." Second, as growth rates increase, the degree of uncertainty increases, which may induce households to save more. Third, at higher growth rates, the return to capital may be higher, and higher returns to capital may elicit higher savings rates. And fourth, lags in adjustment (habit formation) may result in consumption not keeping up with income. That is perhaps the most persuasive reason for the high savings rate. Other arguments suggest that faster growth rates would be associated with lower savings rates. With higher growth rates, individuals need to save less for the future. 3. To be sure, in a world with perfect information, depositors would recognize the risks associated with undercapitalized and low-profit banks and would deposit their money with such institutions only if the higher rate of interest paid on deposits compensated for the risk. These higher required deposit rates would serve as a barrier to entry, ensuring that even without government intervention there would not be "excessive" entry; evidence for this exists for some countries. Still, in many of the countries of East Asia, it appears that-at least at their stage of development-there would have been far more entry in the absence of government intervention. 4. Even with entry restrictions, competition remained keen in most East Asian countries (for Japan, see Sakakibara, Feldman, and Harada 1982). Japanese banks have lower mar- gins than banks in the United States (Ueda 1992) or Germany. Gross interest markups as well as net operating costs (as a percentage of total assets) of banks in Korea, Malaysia, and Thailand were in line with those in industrial countries in the early 1980s and lower than in many other developing countries, notably Turkey and selected countries in Latin America (Hanson and Rocha 1986). On the other hand, World Bank studies (World Bank 1993b) of the banking systems in Indonesia and Korea find that protection has led to less efficient and innovative practices. Moreover, even in Japan there has been less innovation than in the United States, for example, in introducing derivative securities. References The word "processed" describes informally reproduced works that may not be commonly available through library systems. Balassa, Bela. 1989. "The Effects of Interest Rates on Savings in Developing Countries." Policy Research Working Paper 56. World Bank, Policy Research Department, Wash- ington, D.C. Processed. Campos, Jose Edgardo. 1993. "The Institutional Foundations of High-Speed Growth in the High-Performing Asian Economies: Insulation Mechanisms and Public Sector- Private Sector Relations." Background paper for The East Asian Miracle. World Bank, Policy Research Department, Washington, D.C. Processed. Carroll, Chris, David N. Weil, and Lawrence H. Summers. 1993. "Savings and Growth." Paper presented at the Bradley Policy Research Center's Carnegie-Rochester Public Policy Conference, Rochester, N.Y. Processed. Chiu, Paul C. H. 1992. "Money and Financial Markets: The Domestic Perspective." In Gustav Ranis, ed., Taiwan from Developing to Mature Economy. Boulder, Colo.: Westview Press. Cho, Yoon Jee, and T. Hellman. 1993. "Government Intervention in Credit Markets: An Alternative Interpretation of Japanese and Korean Experiences from the New Institu- tional Economics Perspective." World Bank, Financial Sector Department, Washing- ton, D.C. Processed. Cho, Yoon Jec, and Joon-Kyung Kim. 1995. Credit Policies and the Industrialization of Korea. World Bank Discussion Paper 286. Washington, D.C. Corho, Vittorio, and Klaus Schmidt-Hebbel. 1991. "Public Policy and Savings in Develop- ing Countries." Journal of Development Economics 36(1):89-115. 274 The World Bank Research Observer, vol. 11, no. 2 (August 1996) Dekle, Robert. 1990. "Do the Japanese Elderly Reduce Their Total Wealth?" Journal of the Japanese and International Economies 4(September):309-17. 1992. "Raising Saving Rates: Lessons from Japanese Experience." Background paper for The East Asian Miracle. World Bank, Policy Research Department, Washing- ton, D.C. Processed. Fry, Maxwell J. 1988. Money, Interest, and Banking in Economic Development. Balti- more, Md.: Johns Hopkins University Press. Gelb, Alan. 1989. "Financial Policies, Growth, and Efficiency." Policy Research Working Paper 202. World Bank, Policy Research Department, Washington, D.C. Processed. Giovannini, Alberto. 1983. "The Interest Elasticity of Savings in Developing Countries: The Existing Evidence." World Development 11(7):601-7. Greenwald, Bruce, and Joseph E. Stiglitz. 1986. "Externalities in Economies with Imper- fect Information and Incomplete Markets." QuarterlyJournal of Economics 101(2):229- 64. Gupta, Kanhaya L. 1984. Finance and Economic Growth in Developing Countries. Lon- don: Croom Helm. Hanson, James, and Roberto Rocha. 1986. "High Interest Rates, Spreads, and the Costs of Intermediation: Two Studies." Industry and Finance Series 8. World Bank, Washing- ton, D.C. Processed. Horioka, Charles Y. 1990. 'Why Is Japan's Household Saving Rate So High? A Literature Survey." Journal of the Japanese and International Economies 4(March):49-92. Horiuchi, Akiyoshi, and Qing-yuan Sui, 1993. "The Influence of the Japan Development Bank Loans on Corporate Investment Behavior." Journal of the Japanese and Interna- tional Economies 7(December):441-65. IMF (International Monetary Fund). 1995. International Financial Statistics. Washington, D.C. Ishikawa, T. 1987. "Savings: Structural Determinants of Household Saving and the Finan- cial and Tax Systems." In K. Hamada, M. Kuroda, and A. Horiuchi, eds., Macroeco- nomic Analysis of the Japanese Firm. Amsterdam: North-Holland. JDB/JERI (Japan Development Bank and Japan Economic Research Institute). 1993. "Policy- Based Finance. The Experience of Postwar Japan, Final Report for the World Bank." Tokyo. Processed. Jappelli, Tullio, and Marco Pagano. 1994. "Saving, Growth, and Liquidity Constraints." Quarterly Journal of Economics 109(February):83-109. Kawaura, Akihiko. 1991. "Identification of the Size and Direction of the Post-War Credit Allocation in Japan." World Bank, Financial Sector Development Department, Wash- ington, D.C. Processed. Kunio, Yoshihara. 1988. The Rise of Ersatz Capitalism in South East Asia. New York: Oxford University Press. Lin, See Yan. 1991. "Savings Investment Gap, Financing Needs, and Capital Market De- velopment." Bank Negara Malaysia, Kuala Lumpur. Processed. Monetary Authority of Singapore. 1991. "Savings-Investments Balances in Singapore: Determinants and Medium-Term Outlook." Economics Department, Singapore. Pro- cessed. Mukai, Yurio. 1963. "Development of Postal Savings and Its Factors" (in Japanese). Kinyu Keizai 83(December). Murdock, Kevin, and Joseph E. Stiglitz. 1993. "The Effect of Financial Repression in Econo- mies with Positive Real Rates: Theory and Evidence." Background paper for The East Asian Miracle. World Bank, Policy Research Department, Washington, D.C. Processed. Nam, Sang-Woo. 1991. "Korea's Financial Policy and Its Consequences." Paper presented at the East-West Center's Workshop on Government, Financial Systems, and Economic Joseph E. Stiglitz and Marilou Uy 275 Development: A Comparative Study of Selected Asian and Latin American Countries, Honolulu. October. Processed. Noguchi, Y. 1985. "Pensions, the Tax System and the Saving Behavior of the People." In Kenkyu Hopoku Gaiyo [Summary of Research Findings]. Research Project on the Fu- ture Prospects of Government-Managed Voluntary Life Insurance, Tokyo. (In Japa- nese.) Packer, Frank. 1992. "The Role of Long-Term Credit Banks within the Japanese Main Bank System." Columbia University, New York. Processed. Sakakibara, E., R. Feldman, and Y. Harada. 1982. The Japanese Financial System in Com- parative Perspective. Washington, D.C.: U.S. Government Printing Office. Shea, J. D. 1993. "Development and Structural Change of the Financial System." In Patrick Hugh and Yung Chul Park, eds., Financial Development-Japan, Korea and Taiwan. New York: Oxford University Press. Singapore Department of Statistics. Various years. Yearbook of Statistics. Singapore. Stern, Joseph, Ji-Hong Kim, Dwight Perkins, and Jung-Ho Yoo. 1992. "Industrialization and the State: The Korean Heavy and Chemical Industry Drive." Harvard Institute for International Development, Cambridge, Mass., and Korea Development Institute, Seoul. Processcd. Stiglitz, Joseph E. 1985. "Credit Markets and the Control of Capital." Journal of Money, Credit, and Banking 17:133-52. . 1989. "The Economic Role of the State." In Arnold Heertje, ed., The Economic Role of the State. Oxford, U.K.: Basil Blackwell. - 1994. "The Role of the State in Financial Markets." In Michael Bruno and Boris Pleskovic, eds., Proceedings of the World Bank Annual Conference on Development Economics 1993. Washington, D.C.: World Bank. . 1996. "Some Lessons from the East Asian Miracle." World Bank Research Ob- server 11(2):151-77. Stiglitz, Joseph E., and A. Weiss. 1981. "Credit Rationing in Markets with Imperfect In- formation." American Economic Review 71(3):393-410. Sun, C., and M. Liang. 1982. "Savings in Taiwan, 1953-1980." In Kwoh-Ting Li and Tzong-Shian Yu, eds., Experiences and Lessons of Economic Development in Taiwan. Taipei: Institute of Economics, Academia Sinica. Ueda, Kazuo. 1992. Institutional and Regulatory Frameworks for the Main Bank System. Tokyo: University of Tokyo. Vittas, Dimitri. 1991. "Measuring Commercial Bank Efficiency: Use and Misuse of Bank Operating Ratios." Policy Research Working Paper 806. World Bank, Country Eco- nomics Department, Washington, D.C. World Bank. 1989. World Development Report 1989: Financial Systems and Develop- ment. New York: Oxford University Press. . 1993a. The East Asian Miracle: Economic Growth and Public Policy. New York: Oxford University Press. . 1993b. "Korea Financial Sector Study." East Asia and Pacific Regional Office, Country Department I, Industry and Energy Operations Division, Washington, D.C. Processed. Yasuda, Ayako. 1992. "The Performance and Roles of Japanese Development Banks." Stanford University, Directed Readings for Professor Joseph E. Stiglitz, Stanford, Calif. Processed. Yoshino, N. 1992. "History and the Role of Post Office Savings and the Role of Govern- ment Financing in Japan." Keio University, Department of Economics, Tokyo. Pro- cessed. 276 The World Bank Research Observer, vol. 11, no. 2 (August 1996) CREDIT POLICIES: LESSONS FROM JAPAN AND KOREA Dimitri Vittas Yoon Je Cho The success of policy-based credit programs in Japan and the Republic of Korea suggests that credit policy can be an effective instrument for economic development. Why, then, have credit policies failed in so many countries, and what factors explain their relative success in Japan and Korea? Both economic and institutional factors appear to be important in the success or failure of credit policies. Essential economic factors include a reliance on the private sector, a bias toward industrialization, an orienta- tion toward export production, the encouragement of domestic competi- tion, and a commitment to price stability. Crucial institutional factors include extensive and frequent consultation between government and the private sector, effective monitoring systems, and, most important, a clear and credible plan for economic development. Although several countries have included one or more of these factors in their programs, the experi- ence of Japan and Korea suggests that a comprehensive network combin- ing all or most of these factors may be necessary for the successful imple- mentation of credit policies. irected credit programs that give loans on preferential terms and condi- tions to priority sectors were a leading tool of development policy in the 1960s and 1970s. The realization that most of these programs had cre- ated distorted economic incentives among both lenders and borrowers led to a reconsideration of their rationale and effectiveness during the 1980s. Countries around the world found that the programs had stimulated projects that were capital intensive, that preferential funds were sometimes used for nonpriority purposes, and that the programs had increased the cost of funds to nonpreferential borrowers. In addition, policy-based credit programs had provoked a decline in financial discipline that led to low repayment rates and a swelling of budget deficits. Once introduced, moreover, policy-based credit programs proved diffi- cult to eliminate. The World Bank Research Observer, vol. 11, no. 2 (August 1996), pp. 277-98 © 1996 The International Bank for Reconstruction and Development / THE WORLD BANK 277 This negative assessment conflicts, however, with the experience of directed credit programs in Japan and the Republic of Korea. Government officials in these countries argue that subsidized interest rates and government involvement in directing credit are warranted under several circumstances: when the economy shows a significant discrepancy between private and social benefits; when the investment risk of particular projects is too high to attract private lenders; when the lack of reliable economic information discourages private lending to small and medium-size firms; and when infant industries face large social set-up costs (OECF 1991). They suggest that the main constraint facing new or expanding enterprises is limited access to credit at reasonable terms and conditions and that policy-based lending and other forms of industrial assistance such as grants and tax reduction can overcome this constraint. Theoretical Underpinnings of Policy-Based Lending In an ideal world in which economic information is complete and readily available, the financial system is passive. Investors fund the projects that yield the highest returns, and neither governments nor financial institutions need to improve the allocation of credit. In the real world, however, information is highly imperfect and costly to acquire, and the allocation of credit suffers from the unequal distribution of information, the costs of monitoring and verification, and the costs of default or contract enforcement. Under these conditions, credit is not necessarily allocated to its best use. Informational asymmetries give rise to the possibility that credit may be given to unviable candidates (adverse selection), that it may be awarded to irrespon- sible parties (moral hazard), that some players will attempt to receive, without cost, the benefits of credit allocation (free riding), and that incentives arising from the credit program itself may conflict with one another or with program goals. These problems may be further compounded by uncertainty about project returns and by dynamic externalities, which occur when external, or social, ben- efits increase faster than do private benefits. The potential for difficulties of this kind justifies intervention by governments and financial institutions in the allo- cation of credit-even though their ability to allocate credit efficiently will also be constrained by the lack of information. Economic theory has made considerable progress in recent years in under- standing these phenomena and has bridged the wide gap that once existed between theory and practice (Calomiris and others 1992; Calomiris and Himmelberg 1994, 1995; Cho and Hellmann 1994). Economic theory now stresses, for example, the role of market imperfections in explaining why firms rely on internally generated funds (retained earnings) and other forms of "inside" finance (finance provided by owners, managers, and banks that have access to information not available to the public). Reliance on inside finance is especially pronounced for young growing firms and for new indus- 278 The World Bank Research Observer, vol. 11, no. 2 (August 1996) trial sectors in developing countries, and it is clearly a factor constraining industrial growth. Lending by Banks In the absence of full information, banks tend to allocate credit to firms with reliable track records or available internal funds, even if other firms present better investment opportunities. Financial intermediaries, especially (but not only) commercial banks, can acquire information that is superior to that of outsiders by developing and maintaining close long-term relationships with their custom- ers. They can thus play an important role in screening projects, monitoring be- havior and outcomes, and managing corporate distress. These potential advantages of bank lending, however, depend on the behav- ior of the bankers involved and the incentives and regulations that govern their operations. In many countries, commercial banks favor lending for low-risk ac- tivities, such as self-liquidating, short-term working capital and trade finance, or for high-risk, but more speculative, projects with short payback periods, such as real estate development. That is especially true for countries where informa- tion on corporate performance is difficult to obtain, but it is also true for the United Kingdom, the United States, and the Scandinavian countries, which do not suffer from such problems. Commercial banks are generally less willing to finance high-risk projects with long payback periods, even if these projects may yield higher overall returns. They are generally also reluctant to finance small firms that lack adequate collateral, even though such firms may be more inno- vative and promising than others. Lending by Governments A government role in allocating credit can be justified on two grounds. First, directed credit programs can be a preferred or superior industrial policy instru- ment for reaping positive externalities (that is, for increasing benefits across the economy). If firms lack access to credit, other industrial policy tools, such as tariffs and subsidies, that may rely on cost and profit incentives to increase production could prove ineffective. Second, the government has a comparative advantage in directing the alloca- tion of credit. Government agencies (often in direct collaboration with private industrial associations and research institutes) may have better information on sectoral prospects than do individual private firms. They may therefore have an advantage in screening projects, as well as in monitoring behavior and outcomes- although that will depend partly on the relative efficiency of the government agencies and the financial intermediaries. In addition, government costs of en- forcing contracts, through taxation and police powers, are likely to be lower than costs for private intermediaries. The power to tax, moreover, may make it possible for the government to internalize benefits from specific lending policies Dimitri Vittas and Yoon Je Cho 279 that private intermediaries cannot capture. This advantage increases if there are technological spillover effects that neither the firm nor its intermediary can cap- ture but that the government can claim through future taxes. Ultimately, however, the advantages depend on the motivation and the effi- ciency of the government involved. Governments do not always "do the right thing." Government involvement in credit allocation can, and often does, result in rent seeking by borrowers, corruption by bankers and government officials, and crowding out of other worthwhile projects. An important issue in the study of policy-based lending is how governments can prevent rent-seeking behavior by borrowers from undermining the growth objectives of government policies. Experience with directed credit policies varies widely. In Japan and Korea government intervention in credit markets is deemed to have been effective and beneficial for economic growth and development. In the vast majority of devel- oping countries, however, credit policies have failed to promote growth and have given rise, instead, to severe market distortions. Size and Scope of Programs Most studies of directed credit policies focus on the size of the programs, the level of interest rates, and, especially, the level of subsidies. But focusing too narrowly on these aspects of credit policy may be misleading. In many cases, the government's influence on credit allocation may be much stronger than these levels indicate. Japan, for example, avoided highly negative real interest rates (rates adjusted for inflation) during its early period of economic development and had relatively small (explicitly labeled) directed credit programs. The government's role in credit allocation was nonetheless significant; in this early period Japan combined policy-based credit with extensive financial regulation to create a rigidly segmented financial system that favored lending to industry but discouraged lending for speculative purposes, real estate development, or consumption (Vittas and Kawaura 1995). Similarly, one cannot assess the degree of government intervention in a country over time by simply looking at the level of real interest rates. Korea, for example, doubled the level of interest rates in 1965, yielding highly posi- tive real rates. Its action was interpreted by many as financial liberalization, but the interest-rate rise actually strengthened the government's role in allo- cating credit by shifting funds from the unregulated curb (informal) market to the banks, which were more tightly controlled by the government (Cho and Kim 1995). Program Management Just as good or bad management affects the performance of firms, good or bad governance affects the success of credit policies. Although economists have recognized some merit to government intervention in specific markets, they have 280 The World Bank Research Observer, vol. 11, no. 2 (August 1996) often equated failures in these markets with a simplistic notion of government failure. Precisely what characteristics contribute to good economic management and good policy implementation? Good management, even for an economy, does not always mean the least management or least intervention. It does, however, require effective incentive schemes to motivate behavior beneficial to the economy; it also requires a sup- portive institutional environment, an arrangement for close consultation and coordination between the government and business sector, and effective mecha- nisms for monitoring and enforcing performance. Government control of credit may also be understood in light of governance structures. Indeed, credit alloca- tion was used as a powerful instrument for governance of industrial firms in Korea and, to a lesser extent, in Japan (Cho and Hellmann 1994; Vittas and Kawaura 1995). Government's role in credit policy should also be understood in a dynamic context. In the early stages of economic development, when many markets are missing, when those that exist are highly imperfect, and when private institutions are poorly developed, government intervention can help stimu- late healthy economic growth. In the later stages of development, when the private industrial sector has become more sophisticated and markets are better developed and more robust, the merits of government intervention diminish significantly. Credit Supports or Grants? Assuming that the government is able to identify and confront the prob- lems arising from market imperfections, why should it use credit supports (preferential access to credit or subsidized interest rates) rather than grants (direct outlays through the government budget) to address these problems? The answer relates to both efficiency and flexibility. In disbursing grants, the government must select the individual firms that deserve support. In dis- bursing credit subsidies, the government need only signal the sectors that deserve support and leave the selection of individual firms to the banks. In addition, directed credit as an instrument of industrial policy depends partly on the timing of the subsidy, not simply the amount. Subsidies may be allo- cated more flexibly than grants, and they may take into account the perfor- mance of subsidized firms or industries. Good performance can be rewarded by rolling over debt or extending new debt; bad performance (including the diversion of funds to nonpriority uses) can be punished by reducing or termi- nating support. In addition, it is not clear, despite claims to the contrary, that budgetary outlays are more open to scrutiny than are credit subsidies. Both kinds of support risk politicization. In a country where individual credit allocations are based on economic criteria, operating through credit subsi- dies may be both less politicized and more flexible than operating through the budget. Dimitri Vittas and YoonJe Cho 281 Credit Policy in Japan Japanese credit and industrial policy has evolved in response to the changing needs and structure of Japan's economy (Vittas and Wang 1991;JDB/JERI 1994; Vittas and Kawaura 1995). Three phases are usually identified: the reconstruc- tion period, from 1945 to 1955, when industrial policy and the direct govern- ment allocation of funds were most significant; the high-growth period, from 1955 to 1973, when government policy operated less directly, although the fi- nancial system was rigidly segmented and subject to wide-ranging controls; and the liberalization period, from the mid-1970s to the present, when policy be- came less interventionist, and a slow but steady process of financial liberaliza- tion began. Japan's early emphasis on individual industries has given way to a recent concentration on programs that cross industrial boundaries. The main focus of industrial policy in the 1950s was the expansion of capacity, followed in the 1960s by modernization and technological upgrading, in the 1970s by restruc- turing and adjustment at both the company and industry levels, and in the 1980s by diversification of the industrial structure. Throughout, Japanese credit (and industrial) policy seems to have had four specific industrial objectives: to pick and support "winning" industries, especially in markets in which Japan could enjoy a dynamic comparative advantage; to phase out industries in which Japan was no longer internationally competitive; to support small firms; and to pro- vide the industrial infrastructure necessary for growth. The extensive financial support given some traditional and relatively ineffi- cient industries, such as agriculture, can be seen more as a social policy objec- tive, or a response to political pressures, than as a component of an active indus- trial policy. Government regulation and intervention in the financial sector did not focus exclusively on securing cheap finance for the most dynamic sectors, but instead seemed to aim for a balance among the claims of different sectors. Basic Features During the high-growth era, several features of the Japanese financial system, although not unique in themselves, combined to produce a system quite distinct from Anglo-American or European financial systems. These features included the preponderant use of indirect finance; the reliance of the large city banks on credits from the Bank of Japan to fund their loans to industrial corporations; the overborrowing, or high leverage, of industrial companies; and the artificially low level of interest rates (Suzuki 1980; but see Kuroda and Oritani 1980, Horiuchi 1984, and Ikeo 1987, who challenge the particularity of these charac- teristics). Other distinctive features of the Japanese financial system include the role played by the main bank system (in which certain banks were not only main lenders to specific companies but also the main shareholders and providers of other financial services), the close relations between banks and industry, the 282 The World Bank Research Observer, vol. II, no. 2 (August 1996) different roles played by debt and equity in the Japanese financial system, and the important financial intermediary role played by large conglomerate groups, especially the general trading companies, in channeling funds to small firms at the periphery of different industrial groups. Because the financial sector in Ja- pan was highly segmented during this period, the government had considerable control over the allocation of financial resources. The Japanese authorities did not, however, impose strict directed credit programs on private financial insti- tutions. The most distinctive element of Japan's industrial policy has been the coop- erative relationship between government and industry, each of which has recog- nized the necessity of striving toward common goals. This cooperation has been reflected most clearly in the emergence of the "deliberative council system," by which councils made up of industry representatives, former bureaucrats, aca- demics, and others have provided the public and private sectors with a forum for coordinating and developing policy directives. These councils were well estab- lished by the 1960s, when they decided almost all important industrial policies. Another important aspect of Japan's directed credit programs has been the high quality of loan appraisal and project oversight. Loan approval is based on detailed reviews of the projects to be financed and evaluations of the history and character of the firms involved. Once projects are approved, no payments are made without adequate documentation. Close cooperation between develop- ment and commercial banks then ensures continuous monitoring of the perfor- mance of borrowers and allows development banks to take early action if loan repayments are in arrears. The general economic success of Japan has also meant that most borrowers have made substantial profits and have had little difficulty in repaying their loans. Three general characteristics of Japanese credit policy seem particularly sig- nificant in accounting for its success. The first is the Japanese government's respect for the market economy. A precondition for a successful policy-based financial system is the existence of private business and financial structures that can be supplemented with policy-based funding. Japan's prewar experience as a market economy and its determination to establish postwar economic reform created an environment that encouraged entrepreneurship within the existing private sector. The second characteristic (particularly evident as Japan's high-growth period unfolded) is the close relation between policy-based finance and the government's economic plans. The preconditions for this relationship are a public savings sys- tem (in Japan, the postal savings system) and a vehicle for the efficient alloca- tion of funds (in Japan, the Fiscal Investment and Loan Program). The third characteristic has been the Japanese government's respect for mana- gerial independence, assuming the existence of a sound financial institution and management. Although the Japan Development Bank has had inherent limita- tions as a government-related financial institution, it has the autonomy to make funding decisions on a neutral and fair appraisal basis. Dimitri Vittas and Yoon Je Cho 283 The Reconstruction Finance Bank and the Dodge Plan The Japanese government established several agencies to assist in the distri- bution of policy-based funds. The first of these was the Reconstruction Finance Bank, which was created immediately after the war to provide finance to prior- ity industries. It accounted for 84 percent of the total funding for capital invest- ment during the postwar period and for 16 percent of the working capital needs of major industries such as coal, electric power, fertilizers, iron and steel, ocean shipping, and textiles. This financing facilitated the recovery of production of these high-priority industries and paved the way for Japan's economic recovery. The Reconstruction Finance Bank was funded, however, by bonds underwrit- ten by the Bank of Japan, a strategy that fueled inflation, which in turn made the fixed interest rates on priority loans highly negative. The loss of control over inflation, combined with a high incidence of delinquent loans (as well as finan- cial scandal at the reconstruction bank), led the Japanese authorities to suspend the bank's operations in 1949, at which time the Dodge Plan for economic sta- bilization was introduced. The Dodge Plan was intended to achieve a central government surplus and a unified and stable exchange rate. The plan initially caused a deep economic recession, with reductions in exports and investment, company closures, and production and personnel cuts. Successful implementation of the Dodge Plan, however, imposed fiscal balance, removed price controls, and contained mon- etary expansion. Although the policy brought about deflation, it created the stable macroeconomic environment necessary for the subsequent implementa- tion of policy-based lending and other industrial policies leading to the rational- ization and modernization of Japanese industry. Other Directed Credit Institutions In the early 1950s the Japanese authorities established several policy-based financial institutions to provide funding for industrial investment, housing de- velopment, and other purposes. THE FISCAL INVESTMENT AND LOAN PROGRAM. To avoid the inflationary impli- cations of financing these activities through monetary creation, these institu- tions were funded with resources collected through the extensive postal savings and annuities network and channeled through the Trust Fund Bureau as part of the Fiscal Investment and Loan Program. The funds available to the program amounted to 4 percent of gross national product in the 1950s and increased to more than 8 percent in the early 1990s. This increase reflects the growing im- portance of the postal savings funds. Policy-based finance through the trust fund accounted for 13 percent of total lending in the mid-1950s, fell to 10 percent in the 1960s, rose to 15 percent in the 1970s and 1980s, and declined to 12 percent in the early 1990s. 284 The World Bank Research Observer, vol. 11, no. 2 (August 1996) The funds have increasingly been used over the years for social (or, at least, nonindustrial) purposes, especially for financing housing development. The trust fund's share of the total supply of funds for new industrial equipment fell from about 30 percent in the mid-1950s, to 20 percent for most of the following two decades, to about 12 percent in recent years. The most favorable interest rate offered by policy-based financial institutions at the beginning of the 1950s was 3.5 percentage points lower than the private sector long-term prime rate; the least favorable was the same as the prime rate. In addition, policy-based loans had much longer maturities (up to twelve years) and did not require the compensating balances that often substantially increased the cost of private finance, especially for smaller firms. Although policy-based loans continued at fixed rates of interest after the 1950s, success in maintaining macroeconomic and price stability avoided the recurrence of the highly negative interest rates that had bedeviled the reconstruction bank's early operations. THE JAPAN DEVELOPMENT BANK. The successor, in a sense, to the Reconstruc- tion Finance Bank, the Japan Development Bank was given a managerial inde- pendence the first bank had never had. The government assured the first gover- nor of the Japan Development Bank that he would not have to bend to political pressure to fund nonviable projects and that loan decisions would be left to the professional judgment of bank staff and officials. The government was respon- sible for annually establishing the basic policy for the operation of funds and for conducting regular annual audits. The bank allocated project funds, monitored their spending, and assessed their impact. This system of multiple checks pre- vented the misuse of policy-based funds and enabled the development bank to keep its loan losses low. The Japan Development Bank made relatively few mistakes in selecting loan projects. Despite its specialization in long-term industrial finance, the bank's loan losses during the high-growth era were much lower than those of the com- mercial and trust banks, which concentrated on short-term loans and more di- versified loan portfolios. The bank experienced write-offs of 0.09 percent of average loans outstanding from 1951 to 1955 and of only 0.01 percent from 1956 to 1965. One caveat should be noted, however. Some of the loan losses incurred in declining industrial sectors, such as coal mining and, later, ship- building, were transferred to the Japanese government and absorbed by the gen- eral budget. This may explain the unusually low level of loan losses for the period, not only by the development bank, but also by most commercial banks. When the development bank was established, the equity capital contribution from the government accounted for a substantial portion of its total funds. Be- cause statutory reserves increased in proportion to increases in its loan balance, the bank's financial composition continued to be favorable. It could therefore offer a preferential interest rate in line with policy demands, without being sub- sidized by public finance. This strong financial position further guaranteed its managerial autonomy. Dimitri Vittas and Yoon je Cho 285 Empirical Evaluation Two recent empirical studies lend support to the argument that policy-based finance in Japan was effective in stimulating initial growth and encouraging private investment in growing firms in priority industries. Horiuchi and Sui (1993) compared the investment behavior of rnedium-size firms receiving development bank assistance for the 1964-88 period with firms of similar size not receiving such funding. They found that the year of initial lending was associated with increased private investment and that within three years, firms began to move away from dependence on development bank lending to rely more on private banks. Horiuchi and Sui also found that directed credit was most effective for firms that did not have main bank affiliations. Calomiris and Himmelberg (1995) examined the effect of policy-based fi- nance for 1963-91 for the machine tool industry, an industry selected for its high potential for spillover effects through technological innovation and learn- ing. The authors based their study on data for firms, collected with the support of the Japan Development Bank. Because the data set excludes firms that closed during the period, identifying positive effects from policy-based finance is some- what difficult, particularly for the 1960s and 1970s, when machine tool produc- ers underwent considerable consolidation and less-productive firms shut down or were merged with other firms. The level of credit sought by machine tool producers declined over the con- solidation period. It averaged 27 percent of capital from 1965 to 1974 but fell to 10 percent from 1975 to 1991. Similarly, total long-term debt for these produc- ers fell in relation to capital, from 41 percent before 1975 to 26 percent after- ward. Directed credit, which accounted for only a small part of total long-term credit for this period, fell from an average of 3 percent of capital before 1978 to 1 percent after the mid-1980s, or from more than 7 percent of all long-term credit to less than 4 percent. A comparison of lending to general machinery producers by the Japan Devel- opment Bank, by private long-term lenders, and by the government-affiliated Export Import Bank of Japan provides interesting insights. Development bank lending declined, from between 3.7 and 5.3 percent of capital in the late 1960s to between 0.8 and 2.6 percent in the 1980s. This clear drop in support is consis- tent with the premise that government credit relaxed borrowing constraints and helped firms to become seasoned credit risks. Credit from the private Industrial Bank of Japan ranged from between 5.8 and 8.9 percent of capital in the late 1960s to between 4.0 and 6.6 percent in the 1980s. Thus, despite the growing recourse of Japanese firms to the Euromarkets in the 1980s, their reliance on private Japanese funding did not experience the same decline as their use of credit from the development bank. In contrast to borrowing from the development bank, borrowing from the export-import bank registered a large increase, from between 0.4 and 0.9 per- cent of capital in the 1960s to between 3.8 and 19.9 percent in the 1980s. The 286 The World Bank Research Observer, vol. 11, no. 2 (August 1996) growing reliance on these loans must be associated with the increasing maturity of the industry and the greater part played by exports, and perhaps overseas operations, in the 1980s. Support from government sources thus showed con- siderable flexibility in responding to changes in the structure and orientation of the industries and firms receiving credit support. Similar patterns are observed for other types of machine tool producers. Calomiris and Himmelberg also found that government funds had not been captured at either the industry level or firm level. Directed credit was usually provided to a firm only once and for a brief period (80 percent of firms received one-time credit for an average period of less than eight years). They also found that government credit was provided to large, growing, capital-intensive firms with high investment rates. Directed credit appears to have bolstered the posi- tive characteristics of these firms, reinforcing the process of consolidation, in- vestment, and technological change desired by the government. The authors also found that government credit was positively correlated not only with pri- vate credit, but also with reinvestment by the firms themselves. These results are weaker than those Calomiris and Himmelberg reported in 1994 for a shorter period (1982 to 1991). The authors' use of a different methodology in conduct- ing the empirical tests seems to be the main reason for the weaker results re- ported in 1995. Another reason may be a selection bias in the sample used for the 1963-91 period, which did not include firms that closed in the 1960s and 1970s. If those firms were low-investment, poor-performance firms and those that survived were the ones more likely to have received government credit, the effect of government credit would be underestimated. In addition, accounting data for the earlier years are probably less reliable. For the nearly thirty-year sample period studied by Calomiris and Himmelberg, a government loan of 100 yen produced an average investment of 60 yen. The effect of credit from the Japan Development Bank was even stronger: a 100 yen loan produced a 150 yen investment as well as 44 yen in long-term loans from private sources. During the 1960s credit from several government agencies shared equal responsibility for the overall effect of government credit; during the 1970s and 1980s, credit from the development bank had the largest and most signifi- cant impact. A possible explanation for this difference may be that other gov- ernment lenders moved away from producers with high growth potential in the 1970s and 1980s. Credit Policy in Korea Credit policy in Korea has also experienced three distinct phases. In the 1950s and 1960s it was directed toward particular activities, mainly exports and in- dustrial investment. In the 1970s it increasingly promoted specific industries, in particular, the heavy and chemical industries. The successes and failures of this phase induced a change in strategy, and in the 1980s the government became Dimitri Vittas and Yoon Je Cho 287 involved in the industrial and financial restructuring of sectors and companies in distress. As in Japan, Korean credit (and industrial) policy began to focus not only on picking winners, but also on phasing out losers. At the same time, policy was reoriented toward producing a more balanced industrial sector that would not be dominated by a few business conglomerates. Lending to small and me- dium-sized firms received greater attention than it had in the earlier stages, and credit policy was redirected toward functional activities such as research and development and investment in equipment. Basic Features Credit policy in Korea differed from that in Japan in several important re- spects. It involved significant government subsidization of the cost of borrow- ing, and the scope of directed credit programs was much broader. The Korean program made extensive use of commercial as well as development banks to channel loans to priority sectors; the government owned both types of banks. Government intervention in Korean credit policy was also somewhat coercive. The government used a strong package of tax and financial incentives to en- courage firms with minimum equity funds to enter priority industries, and it used its control of the banking system to exert strong leverage on the behavior of firms (World Bank 1987; Cho 1989; Vittas and Wang 1991; Cho and Kim 1995). In the 1950s credit policies were often implemented without clear industrial policy goals. In the 1960s, however, they were structured specifically to support exports and were linked more closely with other policy measures. The Korean government undertook a series of measures in the early 1960s to strengthen state control over finance. It nationalized the commercial banks and amended the central bank act to subordinate the Bank of Korea to the govern- ment. In 1965 it implemented interest rate reform, doubling the level of bank interest rates. This reform not only prompted the rapid growth of deposits in the government-controlled banks, but also expanded the scope of government con- trol over the allocation of financial resources, as funds shifted from the infor- mal, unregulated market to the regulated sector. In addition, the government initiated close consultations with the business sector during the 1960s and careful monitoring of the performance of sup- ported firms. The Monthly Export Promotion Meetings and Monthly Brief- ings on Economic Trends, which were chaired by the Korean president and included senior government officials, managers of industrial firms, bankers, and representatives of industry associations, constituted a forum both among ministries and between the government and the private sector. In these meet- ings, progress toward achieving policy goals was closely reviewed and a con- sensus sought on ways to address emerging problems. The economic man- agement of Korean industry thus came to resemble that of a major corporation. The banks were used as the treasury unit; the industrial sector, 288 The World Bank Research Observer, vol. 11, no. 2 (August 1996) as the production and marketing units; and the government, as the central planning and control unit. During the 1970s the government relied heavily on its control of the credit system, particularly policy-based loans to provide the heavy and chemical in- dustries with preferential access to credit at substantially subsidized rates. The authorities had intended to reduce policy interventions in the credit market in the 1980s, but overexpansion of the heavy and chemical industries in the late 1970s, coupled with the collapse of foreign markets in construction, shipping, and shipbuilding in the early 1980s, forced the government to assist in restruc- turing industrial firms that faced financial difficulties. As Korean politics be- came more democratized in the late 1980s, the emphasis of directed credit pro- grams was shifted to social programs and income redistribution. SOURCES OF FUNDS. A significant difference in the selective credit policies of Korea and Japan has been in the source of funds used for policy loans. Korea has depended heavily on central bank credit and on deposits mobilized by commer- cial banks, and much less than Japan has on fiscal funds or funds mobilized through the government, such as postal savings. In Korea only 7 to 8 percent of the total value of policy loans extended by commercial banks were financed by fiscal funds; about 35 percent were financed through central bank credit at a discounted rate. Korea's heavy reliance on money creation to finance policy- based loans may in large part explain why prices in Korea have been less stable than in Japan. Unlike Japan, Korea used foreign capital as a major source of policy-based finance. Foreign capital had a particularly strong effect on Korean economic development, moreover, because domestic savings were far below desired in- vestment levels. Many observers overlook the role of foreign capital in shaping Korean cconomic policies (including financial sector policies) and the course of development in Korea. Between 1962 and 1982, the average annual economic growth rate was 8.2 percent. A rough estimate suggests that if investmcnt had depended entirely on domestic savings during this period, the average growth rate would have becn only 4.9 percent. Without ready access to forcign capital, Korea could not have continued the repressive financial policies that appear to have accelerated economic growth-although these policies also limited the mobilization of financial resources. The Korcan government controlled foreign loan allocation as tightly as it regulated domestic credit. All foreign loans required government authorization, and their allocation was determined by industrial policy goals. In 1965 the gov- ernment revised the Foreign Capital Inducement Act to allow government-con- trolled banks to guarantee the repayment of foreign borrowing by firms. These guarantees encouraged inflows of foreign capital and technology, but they also perpetuated government intervention in the banks. To avoid default on foreign loans and possible profound disruptions in development projects, the banks of- ten had to absorb, through the rescheduling of domestic bank loans, the exter- Dimitri Vittas and Yoonje Cho 289 nal shocks that prevented domestic firms from meeting their foreign debt ser- vice. The cost of government intervention in domestic banks had, in turn, to be shared by the banks' depositors. THE SHARE AND STRUCTURE OF POLICY LOANS. Policy loans mobilized by the Korean financial system were substantial, accounting for about 50 percent of the total credit extended by domestic financial institutions in the 1970s. That pro- portion gradually decreased to about 30 percent as private nonbank financial institutions, which were not required to make policy-based loans, expanded their share in the 1980s. Policy loans accounted for about 60 percent of total lending made by government-controlled deposit money banks throughout the period. In the 1960s and 1970s, policy loans were extended mainly to the manufac- turing sector; its share of total bank credit was more than twice its share of gross domestic product (GDP), whereas the service sector's share was only about 60 percent of its share in GDP.' Within the manufacturing sector, export industries and the heavy and chemical industries received more credit in relation to their share in GDP than did light industries or industries producing for domestic consumption. Policy Effectiveness Available data suggest that Korean credit policies were effective in reducing the cost of funds to priority sectors and in enhancing their access to funds. Ex- port-oriented firms enjoyed greater access to credit and lower borrowing costs than did domestic-oriented firms. Heavy and chemical industries, despite their high risk of failure, had greater access to credit and significantly lower borrow- ing costs than did light manufacturing industries. The availability and low cost of funds aided in the rapid expansion of the export and heavy and chemical industry sectors, especially in the initial, take-off stage. This evidence does not necessarily imply, however, that selective credit sup- ports were essential for rapid economic growth. The opportunity cost of such supports, that is, the cost of benefits lost by choices not made, are very difficult to estimate. A general equilibrium analysis might provide some answers, but in Korea such an analysis would face severe limitations because substantial parts of input and output prices were controlled in the early stages of development. It is too early, in any case, to provide a full answer to this question. Korean eco- nomic development is still in progress, and the cost of past financial policies may not have been fully realized. It seems clear that export growth drove economic growth in Korea in the 1960s and 1970s and that credit support was indispensable for export growth. Thus credit support must have contributed to rapid economic growth (although a large subsidy may not have been necessary to trigger export growth). The effect of credit supports on growth of the heavy and chemical industries remains controversial, however. Although credit supports contributed to the rapid de- 290 The World Bank Research Observer, vol. 11, no. 2 (August 1996) velopment of the two industries, credit might have been used more efficiently, given the labor endowment in the 1970s, if its allocation had been better bal- anced between the heavy and chemical industries and light industry. It can also be argued, however, that had the push to develop the heavy and chemical indus- tries not taken place in the 1970s, and the two sectors not become Korea's leading export industries in the mid-1980s, Korea might not have been able to take full advantage of the appreciation of the Japanese yen and the world eco- nomic boom in the second half of the 1980s. No solid answer can be provided to this question. The impact of credit policies on industrialization is not limited to their effect on the cost of and access to credit. In an economy like Korea's, in which initial capital accumulation was low, and rapid investment growth was financed mainly by bank credits and foreign loans, firms were highly leveraged financially. As a result, any significant economic downturn would precipitate a financial crisis unless schemes existed for sharing risk between creditors and borrowers. By controlling finance, the Korean government became an effective risk-sharing partner with the industrialists. It could thus encourage their venturesomeness and induce them to have long-term horizons for their companies. The govern- ment was thus a partner, with industry and the banks, in an implicit coinsurance scheme. Without this partnership, Korea might not have been able to establish its large, internationally competitive industrial firms so quickly. This indirect effect of the government's credit policies may be more important than the sub- sidies themselves in explaining Korea's rapid industrialization. The Cost and Legacy of Credit Policies Government control of finance in Korea was not without cost. Several com- mentators have argued that the drive to develop the heavy and chemical indus- tries was overambitious and resulted in a serious misallocation of resources. These critics maintain that the priority sectors expanded production capacity too rapidly, giving firms too little time to accumulate experience and digest new technologies. The result, critics say, was that most firms experienced excess capacity, high production costs, and products so low in quality that they could not be exported. At the same time, nonpriority sectors were forced to borrow at very high rates from the informal sector. This dual nature of the credit system created a major imbalance in the industrial structure of the country (Koo 1984; Kwack 1984). Because Korea relied too heavily and for too long on credit interventions as an instrument of industrial policy, the costs were borne primarily by banks and their depositors. Commercial banks in Korea were involved so deeply in di- rected credit programs that they functioned as development banks. Managerial efficiency and quality of services were compromised, and the banks had large volumes of nonperforming loans. Nonbank financial institutions, which oper- ated more freely, expanded rapidly and superseded the banks' share in the fi- Dimitri Vittas and YoonJe Cho 291 nancial intermediation market. The expansion of the nonbanks helped to im- prove financial market operations by keeping competitive forces alive in the financial system. The problem of carelessness and irresponsibility at the commercial banks was no less serious. As long as the government was willing to rescue firms, banks had little incentive to screen projects carefully or to monitor firms' performance. Indeed, the firms supported by the government became too large and dominant to be allowed to go bankrupt. In the mid-1980s when export markets collapsed after the second oil shock and the worldwide recession, considerable financial support was given to ailing firms and industrial sectors. Although detailed data for this period are not readily available, these setbacks apparently made it in- creasingly difficult for the government to break out of a pattern of financial repression. When the subsequent expansion of the Korean economy and the increasing sophistication of its industrial structure called for a more innovative and market-oriented financial sector, this past legacy became a constraint on liberalization. Because Korea's industrial policy emphasized the economies of large-scale pro- duction to maintain international competitiveness, it led to overwhelming economic concentration within huge business conglomerates, known as chaebols. During the 1970s, it was not uncommon for these conglomerates to triple the number of their affiliates through new acquisitions. In the 1980s, in the face of growing public dis- content with excessive economic concentration, the government was forced to redi- rect policies toward redistributing income; this change in focus often involved in- creased regulation of the business activities of the large firms. Factors of Success in Japan and Korea Theory and practical experience in Japan and Korea suggest that credit policy can be an effective policy instrument for economic development. Why, then, did credit policies fail in so many countries, and what factors explain their relative success in Japan and Korea? The presence of specific economic and institutional factors seem to be essen- tial to success. Economic factors include the maintenance of price stability, an orientation toward export production, the encouragement of domestic competi- tion, a reliance on the private sector, and a bias toward industrialization. Insti- tutional factors include the use of extensive public-private consultative arrange- ments, the creation of effective monitoring systems, and, above all, the development and propagation of credible visions. Price Stability Japan and Korea have, with occasional exceptions, both been able to main- tain price stability. Price stability in itself, however, does not seem sufficient to 292 The World Bank Research Observer, vol. 11, no. 2 (August 1996) ensure the success of credit policies. Several countries in the Middle East, North Africa, South Asia, and southern Europe have also avoided high inflation rates. India, for example, has maintained better price stability than has Korea, but neither India's credit policies nor its economic performance have been as suc- cessful as Japan's or Korea's. Price stability seems clearly to be important for encouraging the growth of financial savings. India and other countries with moderate inflation have experienced a substantial deepening of their financial sectors compared with countries with similar levels of per capita income but with higher inflation. Japan, with very stable prices, has experienced steady and rapid growth of its financial system. Korea, however, with relatively high infla- tion, experienced poor growth of its financial sector in the 1970s, although that trend was reversed in the 1980s. The recent rapid growth of financial saving in China, where the macroeconomic environment has been relatively stable com- pared with other developing countries, also supports the importance of price stability. Competition, Exports, and Private Ownership The credit policies of Japan and Korea have been more successful than other countries primarily because those two countries were able to combine price sta- bility with intense domestic competition, strong export orientation, and a sig- nificant reliance on the private sector. A strong orientation toward exports has forced domestic firms in Japan and Korea to attain high levels of efficiency in order to compete internationally. It has also provided objective criteria for moni- toring the performance of individual firms and for assessing the effectiveness of credit supports. Good performance in export markets, for example, implies con- tinuing access to policy-based finance. Although both Japan and Korea have had import protection, the large indus- trial firms have engaged in vigorous competition in their respective domestic markets. Even in the economically stable countries of the Middle East, South Asia, and southern Europe, industrial production has often been sheltered from both domestic and foreign competition, has been directed toward domestic con- sumption, and has been controlled by state-owned enterprises. Industrialization and Export Promotion Credit policies in Japan and Korea have been narrowly focused and well co- ordinated with other policies. Policy measures, such as those dealing with for- eign exchange, tax, and fiscal arrangements, have been directed toward sup- porting industrialization and export promotion, the main objective of credit policies. The allocation of foreign exchange in Japan and the approval of foreign loans in Korea have been coordinated with domestic credit policies to advance industrial strategy. Credit policies in India, by contrast, have focused on redis- tributing wealth and income toward small farmers and firms and have thus off- Dimitri Vittas and YoonJe Cho 293 set other programs supporting manufacturing and exports. These conflicting policies have often resulted in an implicit taxation of large industries to support farmers and small business. Consultation with the Private Sector Government participation and leadership in organizations such as the delib- erative councils in Japan and the industrial associations in Korea have been critical in helping these countries avoid the pitfalls of credit policy. The coop- erative networks they have established have helped to ensure that scarce re- sources are allocated to activities likely to yield long-term benefits across the economy, and they have provided an effective mechanism for collective risk- sharing by industry, government, and their lenders. Monthly meetings have of- fered opportunities to collect and exchange information, to reassure individual firms and lenders of the government's commitment to particular enterprises, and to promote consensus in pursuit of industrial policy goals. Organizations such as these, however, may be subject to manipulation, corruption, and ineffi- ciency. Their use in allocating credit is an improvement over market solutions only if there are strong safeguards against abuse. Monitoring Reference has already been made to the effectiveness of monitoring in Japan and Korea. Loan approval in Japan is preceded by careful review and indepen- dent appraisal, and the use of funds is strictly monitored. Fund disbursement is based on adequate documentation, and continued access to policy-based loans depends on success in reaching objective targets, mostly in internationally com- petitive export markets. The importance of effective monitoring cannot be overemphasized. What matters for economic development is not really the mobilization and allocation of financial resources but the efficient use of those resources. Any country can mobilize resources by printing money, which it can then allocate to priority sectors. Such money creation will, of course, increase inflationary pressures, but if the resources are used well and lead to higher and more efficient levels of production, unit costs will be lowered and will offset the inflationary impact of the initial monetary financing. Vision Although economists have generally paid little attention to the broad aspects of strategy, the importance of having a credible and consistent vision for eco- nomic development cannot be overstated. In both Japan and Korea, industrial- ization and economic growth have taken precedence over the development of an efficient and modern financial sector. Although the authorities have been com- 294 The World Bank Research Observer, vol. 11, no. 2 (August 1996) mitted to ensuring the safety of deposits and the solvency of financial intermedi- aries, they have been less concerned with allowing banks and other financial intermediaries to innovate and develop new services intended to reduce the cost of financial intermediation. Banks in both countries have been subject to exten- sive controls-on bank spreads and interest rates, branching and bank mergers, and bond issues-as well as to administrative guidance that initially discour- aged lending for consumer credit and housing finance and encouraged the cre- ation of large industrial-financial groupings. In addition, both countries have encouraged lending to fast-growing, high-yield industries, such as automobiles, electronics, petrochemicals, and steel, in which income elasticity of demand has been high, technological progress has been rapid, and labor productivity has risen quickly. Their strategies have emphasized dynamic comparative advan- tage rather than static cost considerations (Ojimi 1972; Johnson 1982; Yotopoulos 1991). It is ironic that the current sorry state of Japanese banks derives from bad loans made after credit policies were relaxed in the 1980s and real estate lending was expanded exponentially. The lesson to draw from the current experience of Japanese banks, however, is not that liberalization was ill advised. Japan's in- dustrial success made inevitable a reorientation of its financial system toward real estate lending, housing finance, and consumer credit. What hurt the banks was the failure to recognize accumulating problem loans and to take early mea- sures to address the impact of that buildup on the banking system. Several other aspects of broad strategy have contributed to the success of credit policies in Japan and Korea. Both countries, for example, have empha- sized complementarities in production for exports and domestic sales. This tac- tic has allowed resources to shift to exports when balance of payments problems forced the government to curtail domestic demand and to shift back to domestic sales when the problems of paying for imported raw materials have eased. The rationale is that if factories can operate through all phases of the business cycle, they can achieve higher scales of production and lower operating costs (Johnson 1982; Yotopoulos 1991). Japan has benefited from an industrial strategy that has been adjusted over time to meet the needs of general development goals. This flexibility of focus has allowed Japan to emphasize, first, the recovery of priority production, then the modernization of equipment in heavy industries, and, later, the development of new industrial sectors, such as machine tools, with a high potential for pro- ducing positive economic side effects. The strategy has also accommodated the smooth adjustment of declining industries, the restructuring of companies fac- ing difficulties, and the rationalization of entire sectors of industry suffering from overcapacity. Credit policies in Japan and Korea have also emphasized the use of additional instruments of industrial policy to complement policy-based finance. Both coun- tries have made extensive use of incentives such as accelerated depreciation al- lowances and tax-free special reserves. These incentives have permitted success- Dimitri Vittas and Yoon Je Cho 295 ful firms in targeted sectors to retain and reinvest a larger part of their profits than firms in nontargeted sectors have been allowed to retain. Particularly im- portant have been the special reserves that are tied to past export performance, because these are linked with the overall strategy of export promotion. These incentives have reinforced the impact of credit policies and have accentuated the credibility of directed credit programs. The existence of coherent, credible visions for development in Japan and Korea has also lent substance to the consultative process. Many countries have tried to promote close consultation between government and the private sector, but with- out a clear vision for development, such exchanges have been little more than ineffective talk shops or forums for special pleading. An important contribution to this collaboration in Japan and Korea was the governments' collection and dissemination of information about long-term sectoral prospects, an activity not readily undertaken by the private industrial sector or the private securities markets. A clear vision of economic goals supported by the analysis of broadly reliable data about the prospects of particular industrial sectors reinforced the economic signaling effect of policy-based finance in Japan and Korea. Lessons from Japan and Korea Ten general lessons can be drawn from the experiences of Japan and Korea. The first six relate to "good vision," the last four to "good management." First, credit programs must be small in size, have a narrow focus, and be of limited duration with clear phase-out provisions. Second, programs must involve a low level of subsidy (if any), to minimize incentives to profit from price fluctuations and also to reduce the tax on financial intermediation that all credit programs necessarily entail. Third, programs must be financed by long-term funds to avoid inflation. Recourse to central bank credit, in particular, should be avoided ex- cept in the very early stage of development, when selective credit programs sup- ported by central bank credit might help jump-start economic growth and de- velopment. Even at this stage, however, care must be taken to prevent high rates of inflation. Fourth, programs should aim at achieving positive economic or social side effects (or at avoiding negative effects). They should therefore focus on overcoming the external finance constraint facing small or rapidly expanding firms as well as helping firms in declining industries to phase out in an orderly and timely fashion. Fifth, programs should promote industrialization and ex- port orientation, and they should be based on a competitive private sector with internationally competitive operations. Sixth, programs should form part of a broader credible vision of economic development, promoting growth with eq- uity and including a long-term strategy to develop a sound financial system operating on economic criteria. Seventh, policy-based loans should be channeled through well-capitalized, administratively capable, and autonomous financial institutions. Professional 296 The World Bank Research Observer, vol. 11, no, 2 (August 1996) management and managerial autonomy are essential. Eighth, loans should be based on clear, objective criteria that are easily monitored. Detailed project ap- praisals, close supervision of disbursements, and monitoring of performance and repayment records are essential to the success of credit programs. Ninth, policy- based programs should aim for a good repayment record and low loan losses. Tenth, programs should be supported by effective mechanisms for consultation between the public and private sectors, including the collection and dissemina- tion of basic market information. Although these lessons are important, replicating the Japanese and Korean experience may be difficult in today's financial environment. High technology coupled with the globalization of financial markets has substantially reduced the effectiveness of foreign exchange controls on capital movements and has limited the ability of government authorities to set interest rates substantially below market levels. The new World Trade Organization agreement, moreover, limits the use of credit policies. Despite these potential obstacles, however, the greatest challenge facing most developing countries in the use of policy-based finance remains the absence of the essential factors of good vision and good manage- ment that explain the success of directed credit policies in Japan and Korea. Notes Dimitri Vittas is Adviser with the Financial Sector Development Department of the World Bank. Yoon Je Cho is Vice President of the Korea Tax Institute. This article draws on the findings of a World Bank research project on the "Effectiveness of Credit Policies in East Asian Countries." These findings were published either as World Bank Discussion Papers or as Policy Research Working Papers and are cited in the list of references. We are grate- ful to our many collaborators in this project and to several staff members of the World Bank for useful insights and comments. 1. The manufacturing sector received 46 percent of total bank loans given in 1970 and S4 percent of the total given in 1980. In contrast, the service sector received only 29 percent in 1970 and 24 percent in 1980. References The word "processed" describes informally reproduced works that may not be commonly available through library systems. Calomiris, Charles W., and Charles P. Himmelberg. 1994. "Directed Credit Programs for Agriculture and Industry: Arguments from Theory and Fact." Proceedings of the World Bank Annual Conference on Development Economics 1993. Washington, D.C.: World Bank. . 1995. "Government Credit Policy and Industrial Performance: Japanese Machine Tool Producers, 1963-1991." Policy Research Working Paper 1434. World Bank, Fi- nancial Sector Development Department, Washington, D.C. Calomiris, Charles W., Charles P. Himmelberg, Charles M. Kahn, and Dimitri Vittas. 1992. "Evaluating Industrial Credit Programs in Japan: A Research Proposal." Working Paper, University of Illinois at Urbana-Champaign, Department of Finance. Processed. Dimitri Vittas and Yoon Je Cho 297 Cho, Yoon Je. 1989. "Finance and Development-The Korean Approach." Oxford Re- view of Economic Policy 5(4, Winter):88-102. Cho, Yoon Je, and Thomas Hellmann. 1994. "The Government's Role in Japanese and Korean Credit Markets-A New Institutional Economics Perspective." Seoul Journal of Economics 7(4, Winter):383-416. Cho, Yoon Je, and Joon-Kyung Kim. 1995. Credit Policies and the Industrialization of Korea. World Bank Discussion Paper 286. Washington, D.C. Horiuchi, Akiyoshi. 1984. "The 'Low Interest Rate Policy' and Economic Growth in Post War Japan." The Developing Economies 22(4, December):349-71. Horiuchi, Akiyoshi, and Qing-Yuan Sui. 1993. "Influence of the Japan Development Bank Loans on Corporate Investment Behavior." Discussion Paper Series 93-F-4. University of Tokyo, Faculty of Economics. Processed. Ikeo, Kazuhito. 1987. "Japan's Financial System: A Micro Approach." Japanese Economic Studies (Fall):60-77. JDB/JERI (Japan Development Bank/Japan Economic Research Institute). 1994. Policy- Based Finance: The Experience of Postwar Japan. World Bank Discussion Paper 221. Washington, D.C. Johnson, Chalmers. 1982. MITI and the Japanese Miracle. Stanford, Calif.: Stanford Uni- versity Press. Koo, Bohn-Young. 1984. "Role of the Government in Korea's Industrial Development." Working Paper 84-07. Korea Development Institute, Seoul. October. Kuroda, Iwao, and Yoshiharu Oritani. 1980. "A Reexamination of the Unique Features of Japan's Corporate Financial Structure-A Comparison of Corporate Balance Sheets in Japan and the United States." Japanese Economic Studies 7(4):82-117. Kwack, Taewon. 1984. "Industrial Restructuring Experiences and Policies in Korea in the 1970s." Working Paper No. 84-08. Korea Development Institute, Seoul. OECF (Overseas Economic Cooperation Fund). 1991. Issues Related to the World Bank's Approach to Structural Adjustment-Proposal from a Major Partner. Occasional Paper 1. Tokyo. Ojimi, Yoshihisa. 1972. "Basic Philosophy and Objectives of Japanese Industrial Policy." In Organization for Economic Cooperation and Development, The Industrial Policy of Japan. Paris. Suzuki, Yoshio. 1980. Money and Banking in Contemporary Japan. New Haven, Conn.: Yale University Press. Vittas, Dimitri, and Akihiko Kawaura. 1995. "Policy-Based Finance, Financial Regulation and Financial Sector Development in Japan." Policy Research Working Paper 1443. World Bank, Financial Sector Development Department, Washington, D.C. Vittas, Dimitri, and Bo Wang. 1991. "Credit Policies in Japan and Korea-A Review of the Literature." Working Paper Series 747. World Bank, Country Economics Depart- ment, Washington, D.C. World Bank. 1987. Korea: Managing the Industrial Transition. Washington, D.C. Yotopoulos, Pan A. 1991. "Exchange Rates and State-Led Capitalism: What Can the NICs Learn from Japan?" Asian Economic Journal 5(1):76-99. 298 The World Bank Research Observer, vol. 11, no. 2 (August 1996) NOW A VAILABLE World Development Report 1996: From Plan to Market Wr orld Development Report 1996: From Plan to Market steps back from the extraordinary array of recent events and policy changes in 28 for- mer centrally planned economies-those in Central and Eastern Europe and the newly independent states of the former Soviet Union, along with China, Mongolia, and Vietnam-to ask what we have learned about the ingredients of a successful transi- tion to a market economy and how they should be pursued. The Report addresses the initial challenges of transition and how different countries have responded. It also analyzes the longer-term agenda of consolidating the early reforms by developing the institutions and policies that will help new systems to develop and prosper. Published for the World Bank by Oxford University Press: English (Hardback) Stock no. 61108ER (ISBN 0-19-521108-1) $45.95 English (Paperback) Stock no. 61107ER (ISBN 0-19-521107-3) $22.95 Available in paperback: $22.95 eacb: Arabic: Stock no. 13269RO (ISBN 0-8213-3269-4) September 1996 Chinese: Stock no. 13268RO (ISBN 0-8213-3268-6) September 1996 French: Stock no. 13264RO (ISBN 0-8213-3264-3) August 1996 German: Stock no. 13266RO (ISBN 0-8213-3266-X) September 1996 Japanese: Stock no. 13267RO (ISBN 0-8213-3267-8) September 1996 Portuguese: Stock no. 13271RO (ISBN 0-8213-3271-6) September 1996 Russian: Stock no. 13270RO (ISBN 0-8213-3270-8) October 1996 Spanish: Stock no. 13265RO (ISBN 0-8213-3265-1) September 1996 World Bank Publications For payment in local currency, contact your World Bank Distributor In the U.S. contact. TheWorld Bank, PO. Box 7247-8619, Phi adelphia, PA 19170-8619. Phone: (202) 473-1155, Fax: (202) 522-2627 Shipping and handling: US$5.00. For airmail delivery outside the U.S., add US$.00 for one item plus US$6.00 for each additional item Paynent by US$ check drawn on a U S. bank payable to the World Bank or by VISA, MasterCard, or American Express. 1225 Read the Executive Summary Online! http://www. worldbank. org Click on Publications Distributors of World BRAZIL DENMARK Bank Publications PublicacoesTecnicas SarifundsLitteratur Prces and credit terms vary from lntemacionais Ltda. Rosenoems Al II 1 country to country Consult your Rua Peixoto Gormde, 209 DK- 1970 Frederksberg C local distributor before ptacing an 01 409(SaoPaulo,SP Te: (31)-351942 Fax: (11)258-699C ALBANIA EGYPT, ARAB REPUBLIC OF Adrion Ltd. CANADA Al Ahram Perat Rexhepi Str. Renouf Publishing Co. Ltd. Al Galaa Street Pall. 9, Shk, lAp. 4 1294 Algoma Road Cairo Tirana Ottawa, Ontario KI B 3W8 Tel: (2) 578-6083 Tel: (42)27419;221 72 Tel:613-741-4333 Fax: (2) 578-6833 Fax: (42)27419 Fax:613-741-439 The Middle EastObserver ARGENllNA CHINA 41, Shenf Street Oticinadel Libro Internacional China Financial & Economic Cairo Av. Cordoba 1877 Publishing House Tel: (2) 393-9732 1120BuenosAires 8, DaFoSiDongJie Fax: (2) 393-9732 Tel: (1)815-8156 Beijing FINLAND Fax: (1) 815-8354 Tel: (1) 333-8257 Akateeminen Kirakauppa Fax: (1) 401-7365 P.O. Box 23 AUSTRALIA, FIJI, PAPUA NEW FIN-00371 Helsinki GUINEA, SOLOMON ISLANDS, COLOMBIA Tel: (0)12141 VANUATU,ANDWESTERN lnfoenlaceLtda. Fax: (0) 121-4441 SAMOA Apartado Aereo 34270 URL: httpl/booknet.cultnet.iVaka/ D.A. Information Services Bogota D.E. 648 Whitehorse Road Tel: (1) 285-2798 FRANCE Mitcham 3132 Fax: (1) 285-2798 Word Bank Publications Victoria 66, avenue d'lena Tel: (61)392107777 COTE D'IVOIRE 75116 Pads Fax: (61)392107788 Centre d'Edition etde Diffusion Tel: (1)40-69-30-56/57 URL: httpYlwww.dadirect.com.au Africaines (CEDA) Fax: (1)40-69-30-68 04 B.P. 541 AUSTRIA Abidjan 04 Plateau GERMANY Gerold and Co. Tel: 225-24-6510 UNO-Verlag Graben3l Fax: 225-25-0567 PoppelsdorferAllee55 A-1011 Wen 53115Bonn Tel: (1) 533-50-14-0 CYPRUS Tel: (228)212940 Fax: (1) 512-47-31-29 Centerof Applied Research Fax: (228)217492 Cyprus College BANGLADESH 6, Diogenes Street, Engomi GREECE Micro Industres Development P.O. Box 2006 Papasotiriou S.A. Assistance Society (MIDAS) Nicosia 35, Stournara Str. House 5, Road 16 Tel: 244-1730 10682 Athens Dhanmondi RlArea Fax: 246-2051 Tel: (1) 364-1826 Dhaka 1209 Fax: (1) 364-8254 Tel: (2) 326427 CZECH REPUBUC HONG KONG, MACAO Fax: (2) 811188 National Intormation Center Asia 2000 Ltd.~ prodejna, Konviktska g Sales & Circulaton Department BELGIUM CS - 11 357 Prague I Seabird House, unit 1101 -02 Jean De Lannoy Tel: (2) 2422-9433 22-28 Wyndham Street, Central Av. du Roi 202 Fax: (2) 2422-1484 HongKong 1060 Brussels URL: http://www.nis.cz/ Tel: 8522530-1409 Tel: (2) 538-5169 Fax: 852 2526-1107 Fax: (2) 538-0841 URL: http:// www.sales@asia2000.com.hk HUNGARY R.O.Y International MALAYSIA Foundation for Market Economy P.O. Box 13056 University of Malaya Cooperative Dombovar Ut 17-19 Tel Aviv 61130 Bookshop, Limited H-1 117 Budapest Tel: (3)5461423 P.O. Box 1127 Tel: 361 204 2951 or Fax: (3)5461442 Jalan Pantai Barn 361 2042948 59700 Kuala Lumpur Fax: 36 1 204 2953 Palestinian Authority/Middle East Tel: (3) 756-5000 Index Information Services Fax: (3) 755-4424 INDIA P.OB. 19502Jerusalem Allied Publishers Ltd. Tel: (2) 271219 MEXICO 751 Mount Road INFOTEC Madras -600002 ITALY Apartado Postal 22-860 Tel: (44) 852-3938 Licosa Commissionaria Sansoni 14060Tlalpan, Fax: (44)852-0649 SPA Mexico D.F. Via Duca Di Calabria, 1/1 Tel: (5) 606-0011 INDONESIA Casella Postale 552 Fax: (5) 624-2822 Pt. Indira Limited 50125 Firenze Jalan Borobudur20 Tel: (55) 645-415 NETHERLANDS P.O. Box 181 Fax: (55) 641-257 De Undeboom/lnOr-Publikaties Jakarta10320 P.O. Box 202 Tel: (21) 390-4290 JAMAICA 7480AE Haaksbergen Fax: (21)421-4289 lan Randle Publishers Ltd. Tel: (53) 574-0004 206 Old Hope Road Fax: (53) 572-9296 IRAN Kingston 6 Kowkab Publishers Tel: 809-927-2085 NEW ZEALAND P.O. Box 19575-511 Fax: 809-977-0243 EBSCO NZ Ltd. Tehran Private Mail Bag 99914 Tel: (21) 258-3723 JAPAN New Market Fax: 98 (21) 258-3723 Eastern Book Service Auckland Hongo3-Chome, Bunkyo-ku 113 Tel: (9) 524-8119 Ketab Sara Co. Publishers Tokyo Fax: (9) 524-8067 Khaled Eslamboli Ave., Tel: (03)3818-0861 6th Street Fax: (03)3818-0864 NIGERIA Kusheh Delafrooz No. 8 URL: http://www.bekkoame.or.jp/ University Press Limited Tehran -svt-ebs Three Crowns Building Jericho Tel: 8717819 or 8716104 Private Mail Bag 5095 Fax: 8862479 KENYA Ibadan E-mail: ketab-sara@neda.net.ir Africa Book Service (E.A.) Ltd. Tel: (22) 41-1356 Quaran House, Mfangano Street Fax: (22) 41-2056 IRELAND P.O. Box 45245 Govemment Supplies Agency Nairobi NORWAY Oifig an tSolathair Tel: (2) 23641 Narvesen Information Center 4-5 Harcourt Road Fax: (2) 330272 Book Department Dublin 2 P.O. Box 6125 Etterstad Tel: (1) 461-3111 KOREA, REPUBLIC OF N-0602 Oslo 6 Fax: (1)475-2670 DaejonTrading Co. Ltd. Tel: (22) 57-3300 P.O. Box 34 Fax: (22) 68-1901 ISRAEL Yeoeida Yozmot Literature Ltd. Seoul PAKISTAN P.O. Box 56055 Tel: (2) 785-1631/4 Mirza Book Agency TelAviv 61560 Fax: (2) 784-0315 65, Shahrah-e-Quaid-e-Azam Tel: (3) 5285-397 P.O. Box No. 729 Fax: (3) 5285-397 Lahore 54000 Tel: (42) 7353601 Fax: (42)7585283 Oxford University Press SAUDIARABIA, QATAR SRI LANKA, THE MALDIVES 5 BangaloreTown Jarir Book Store Lake House Bookshop Sharae Faisal P.O. Box3196 P.O. Box 244 P.O. Box 13033 Riyadh 11471 100, Sir Chittampalam A. Karachi-75350 Tel: (1)477-3140 GardinerMawatha Tel: (21) 446307 Fax: (1) 477-2940 Colombo 2 Fax: (21) 454-7640 Tel: (1) 32105 E-mail: SINGAPORE, TAIWAN, Fax: (1)432104 oup@oup.khi.erum.com.pk MYANMAR, BRUNEI Asahgate Publishing Asia SWEDEN PERU Pacific Pte. Ltd. Fritzes Customer Service Editoral Desarrollo SA 41 Kallang Pudding Road #04-03 Regeringsgaton 12 Apartado3824 Golden Wheel Building S-1 06 47 Stockholm Lima 1 Singapore349316 Tel: (8) 690 90 90 Tel: (14)285380 Tel: (65) 741-5166 Fax: (8)21 4777 Fax: (14)286628 Fax: (65)742-9356 e-mail: Wennergren-Williams AB PHILIPPINES ashgate@asianconnect.com P. O. Box 1305 International Booksource Center lnc. S-171 25 Solna Suite 720, Cityland 10 SLOVAK REPUBLIC Tel: (8)705-97-50 CondominiumTower2 Slovart G.T.G. Ltd. Fax: (8) 27-00-71 H.V dela Costa, corner Krupinska 4 Valero St. PO. Box 152 SWITZERLAND Makati, Metro Manila 852 99 Bratislava 5 Libraire Payot Tel: (2)817-9676 Tel: (7)839472 Service Institutionnel Fax: (2) 817-1741 Fax: (7)839485 C6tes-de-Montbenon 30 1002 Lausanne POLAND SOUTH AFRICA, BOTSWANA Tel: (021)-341-3229 Intemational Publishing Service For single titles: Fax: (021)-341-3235 Ul. Piekna 31/37 Oxford University Press 00-577Warzawa Southern Africa Van Diermen EditionsTechniques Tel: (2) 628-6089 P.O. Box 1141 Ch. de Lacuez 41 Fax: (2)621-7255 CapeTown8000 CH1807Blonay Tel: (21)45-7266 Tel: (021) 9432673 PORTUGAL Fax: (21)45-7265 Fax: (021)9433605 Livraria Portugal Rua Do Carmo 70-74 ForIsubscription orders: TANZANIA 1200 Lisbon Internatonal Subscription Service Oxford University Press Tel: (1)347-4982 POrgox41095 MaktabaStreet Fax: (1 ) 347-0264 Johannesburg 2024 PO. Box 5299 Tel: (11)880-1448 DaresSalaam ROMANIA Fax: (11)880-6248 Tel: (51)29209 Compani De Librari Bucuresti Fax (51) 46822 S.A. SPAIN Str. Lipscani no. 26, sector 3 Mundi-Prensa Libros, S.A. THAILAND Bucharest Castello 37 Central Books Distribution Tel: (1) 613 9645 28001 Madrd 306 Silom Road Fax:(1)3124000 Tel: (1)431-3399 Bangkok Fax: (1) 575-3998 ~ Tel: (2) 235-5400 RUSSIAN FEDERATION http://www.tsai.es/mprensa Fax: (2) 237-8321 Isdatelstvo 9a, Kolpachniy Pereulok Mundi-Prensa Barcelona Moscow 101 831 Consell de Cent, 391 Tel: (95) 917 87 49 08009 Barcelona Fax: (95)9179259 Tel: (3) 488-3009 Fax: (3)487-7659 TRINIDAD &TOBAGO, UNITED KINGDOM ZIMBABWE JAMAICA Microinfo Ltd. Longman Zimbabwe (Pte.)Ltd. Systematics Studies Unit P.O. Box3 Tourle Road,Ardbennie #9 Watts Street Alton, Hampshire GU34 2PG P.O. Box ST125 Curepe England Southerlon Trinidad, West Indies Tel: (1420)86848 Harare Tel: 809-662-5654 Fax: (1420)89889 Tel: (4)6216617 Fax: 809-662-5654 Fax: (4) 621670 ZAMBIA UGANDA University Bookshop Gustro Ltd. Great East Road Campus Madhvani Building P.O. Box 32379 P.O. Box 9997 Lusaka Plot 16/4 Jinja Rd. Tel: (1)213221 Ext. 482 Kampala Tel/Fax: (41) 254763 The World Bank Headquarters 1818 H Street, N.W. Washington, D.C. 20433, U.S.A. Telephone: (202) 477-1234 Facsimile: (202) 477-6391 Telex: MCI 64145 WORLDBANK MCI 248423 WORLDBANK Cable Address: INTBAFRAD WASHINGTONDC World Wide Web: http://www.worldbank.org E-mail: books@worldbank.org European Office 66, avenue d'Iena 75116 Paris, France Telephone: (1) 40.69.30.00 Facsimile: (1) 40.69.30.66 Telex: 640651 Tokyo Office Kokusai Building 1-1 Marunouchi 3-chome Chiyoda-ku, Tokyo 100, Japan Telephone: (3) 3214-5001 Facsimile: (3) 3214-3657 Telex: 26838 127 9 '78082 1332795' ISBN 0-8213-3279-1 Jorgensen. Steen Lau EXC u 10-063 ( 1 )