EDITORS Stanley Fischer and Dennis de Tray MANAGING EDITOR Shekhar Shah CONSULTING EDITOR Philippa Shepherd The World Bank Annual Conference on Development Economics is a forum for discussion and debate of important policy issues facing developing countries. The conferences emphasize the contribution that relevant policy, empirical, and basic economic research can make to understand- ing development processes and to formulating sound development policies. Conference papers are written by researchers outside the Bank and at the Bank. The conference series was started in 1989. The Proceedings of the World Bank Annual Conference on Development Economics is published as a special supplement to the World Bank's professional economics journals, The World Rank Economic Review and The World Bank Research Observer. The Editorial Boards of the Review and the Observer do not review the Proceedings, nor are the papers published here subject to the peer review process through which regular submissions to these journals must pass. Summaries are included of the floor discussions following each paper and of the roundtable discussion; they attempt to convey the sense and substance of what was discussed, interventions by participants from the floor, and responses by panelists. They have not been reviewed by the authors, the discussants, or the participants concerned. The views and interpretations expressed in this Proceedings volume do not necessarily represent the views of the World Bank or its member countries and should not be attributed to the World Bank or its affiliated organizations. All current subscribers to The World Bank Economic Review and The World Bank Research Observer receive this supplement to the journals. Subscription orders for the two journals and for individual copies of the Proceedings supplement can be placed by filling out one of the coupons at the end of this Proceedings volume and sending it with payment to World Bank Publications. O 1990 The International Bank for Reconstruction and Development I THE WORLD BANK. All rights reserved. Manufactured in the United States of America. ISSN 1014-7268 ISBN 0-8213-1461-0 Material in this journal is copyrighted. Requests for permission to reproduce articles should be addressed to the Director, Publications Department, World Bank, Washington, DC 20433, U.S.A. The World Bank encourages dissemination of its work and will normally give permission promptly and, when the intended reproduction is for noncommercial purposes, without asking a fee. Per- mission is not required to make photocopies for classroom or similar use. The World Bank Economic Review is indexed in Current Contents 1Social & Behavioral Sciences, the Social Sciences Citation Index, the Journal of Economic Literature, and the Public Affairs information Service. The World Bank Research Observer is indexed in the Journal of Economic Literature, the Standard Periodical Directory, the Public Affairs Information Service, Inc., and, online, by the Economic Literature Index and DIALOG. It is available in microform through University Microfilms, Inc., 300 North Zeeb Road, Ann Arbor, MI 48106, U.S.A. WORLD BANK ANNUAL CONFERENCE ON DEVELOPMENT ECONOMICS 1990 April 26 and 27, 1990 Washington, D.C. Introduction Stanley Fischer and Dennis de Tray Keynote Address: Development Policy Research: The Task 11 Ahead Manmohan Singh Developing Countries and the Uruguay Round of Trade Negotiations Marcelo de Paiva Abreu Comment, Andrzej Olechowski 47 Comment, Gary P. Sampson 49 Floor Discussion 55 Saving in Developing Countries: Theory and Review Angus Deaton Comment, Fumio Hayashi 97 Comment, Steven B. Webb 101 Floor Discussion 105 Social Sector Pricing Policy Revisited: A Survey of Some Recent Controversies Emmanuel Jimenez Comment, Benno j. Ndulu 139 Comment, Nicholas Stern 143 Floor Discussion 149 ivContents The Role of Institutions in Development Brian Van Arkadie Comment, John Nellis 177 Comment, Pranab Bardhan 181 Floor Discussion 187 The Noncompetitive Theory of International Trade and Trade 193 Policy Elhanan Helpman Comment, T. N. Srinivasan 217 Comment, Nancy Barry 223 Floor Discussion 227 The Policy Response of Agriculture Hans Binswanger Comment, Auishay Braverman 259 Comment, Alberto Valdks 263 Floor Discussion 269 Roundtable Discussion P R O C E E D I N G S O F T H E W O R L D B A N K A N N U A L C O N F E R E N C E O N D E V E L O P M E N T E C O N O M I C S 1 9 8 9 Introduction Stanley Fischer and Dennis de Tray The papers in this volume were presented at the first annual World Bank Con- ference on Development Economics, held April 27-28, 1989, in Washington, D.C. The conferenceseries brings researchers from the Bank's member countries together with Bank staff to stimulate interaction and exchange of ideas and information. I'articipants at the 1989 conference came from a range of research, academic, and policymaking institutions in both developing and developed coun- tries. The ultimate object of the series is, by enhancing the knowledge base, to improve both member country and Bank policymaking. Outside researchers and Bank researchers have much to offer each other. Academics and other outside researchers can and do bring different and new ways of thinking to the Bank. That is especially important for an institution as large, as busy, and as involved in policy in so many countries, as the Bank. The temptation for Bank researchers is to look inward, to talk to each other, and to assume that shared habits of thought are the right ones. Although academics too share certain modes of thought, the differences between the two groups are potentially productive. In addition, the best academics are probably closer to the cutting edge of research; they are also more free to think speculatively and to develop new theories and fresh insights. One of the purposes of the conference series is to expose Bank economists to recent developments in economics whose policy implications may not yet be clear but which are beginning to affect thinking outside the Bank and which may one day become the conventional wisdom. One example in this volume is the paper on recent developments in strategic trade theory by Elhanan Helpman. The topic is extremely important, the Bank has well-known policy views in this area, and the new theories seem to have the potential for reversing those views. The paper generated a very lively discussion, during which both Helpman and the discussants agreed that the knowledge needed for a country to benefit from the possibilities implied in the new theories is probably too detailed to make such policies empirically useful. But it is far better to reach that conclusion on the basis of an informed discussion of the theory and its empirical requirements than on a priori grounds. The Bank for its part brings to the outside research community an extremely rich and varied research agenda and an unparalleled wealth of practical knowl- edge about developing countries. The Bank's research agenda springs mainly from its operational experience and is therefore down to earth. Part of the 2 Introduction research agenda comes from an attempt to anticipate future problems, but even here the Bank tends to be practical, asking for the operational implications of each paper and research project. Of course, one can go too far in that regard, for not every useful piece of research has an immediate payoff; the right question is whether a particular piece of research will eventually have real-world impli- cations, rather than whether it has immediate operational relevance. Nonethe- less, this challenge to produce research that is operationally relevant imposes an extremely useful discipline on researchers in and out of the Bank. Development problems are far too difficult for any one institution to believe that it has the final answers to them. The Bank's approach to development is - - pragmatic and open-minded. In making policy decisions, and in thinking about development strategies, the Bank has to search for better ways of helping our member countries, to entertain new ideas, to question old ones, and to listen, talk, and learn from critics. Being open-minded, however, does not mean being softheaded. After thinking and talking, the Bank and its member countries have to act, one way or the other, in light of the relevant analytical and policy-related arguments. And because decisions have to be made, the Bank cannot agree with everyone on every topic. Nonetheless, the Bank has to keep on reexamining the way it does business and be prepared to change if necessary. This conference series is one way for the Bank to listen to and ask for help from outside re- searchers in dealing with the difficult development issues that our member coun- tries face. To help in this process, we have described the Bank's current research - program in an addendum to this introduction. (Requests for additional infor- - mation on the World Bank's research program should be directed to the Research Administrator's Office, World Bank, Washington, D.C. 20433, U.S.A.) This first World Bank Annual Conference on Development Economics did not focus on a single subject or theme. Instead, each author was asked to review and assess the implications of one of six distinct subjects. This decision to explore a set of issues rather than a single theme reflects a desire to stress policy issues and to communicate findings to a broad spectrum of development practitioners rather than a group of research specialists in a narrow field. Future conferences are expected to continue this broad coverage. The conference began with a keynote address by Manmohan Singh, secretary general of the South Commission. Singh identified some of the major areas of challenge for researchers dealing with development: the role of the state, the management of the public sector, and reform of the international trading regime. Though he did not discuss them in detail in his address, he also cited food security and human resource development as deserving priority attention. Singh sees the critical question for the modernization of the state as not so much its size as the overall quality and effectiveness of its impact. He felt that better management of the public sector and its careful delineation would be critical for private sector development as well as for the prospects for reform in the socialist economies: both issues he felt would need location-specific answers. For the trading system, Singh asked whether developed country markets would Introduction 3 still be available if developing countries which have pursued inward-looking development strategies were to switch to export-orientation based on labor- intensive manufactured goods. Singh concluded by calling for an international mechanism to impartially evaluate the development performance of developing countries undergoing structural reform. Marcelo Abreu of the Catholic University of Rio de Janeiro presented the first paper, "Developing Countries and the Uruguay Round of Trade Negotiations." With the Uruguay Round well under way, Abreu asks what negotiating stances developing countries should be taking. On what issues should they stand firm and where should they be willing to make concessions?What stands should the various subcoalitions-agricultural exporters, food importers, garment manu- facturers-take? This is a complex and important set of negotiations, one in which the stakes for the developing countries are greater than ever before. Abreu's paper highlights the importance of restoring GATT discipline to the trade in goods, even while GATT is pushing ahead with the extension of GATT-like rules to services and intellectual property rights. He also sees a considerable need to focus on the issue of unilateral liberalization, under conditionality or otherwise, and its effect on developing countries' negotiating positions at the GATT. Angus Deaton of Princeton University presented the second paper, "Saving in Developing Countries: Theory and Evidence." Deaton's assigned task was to survey the literature on determinants of saving to assess its policy implications and the existing theoretical and empirical gaps. He attacked this issue by de- veloping a new framework for analyzing saving in developing countries. Deaton's approach emphasizes that motives for saving in developing countries are likely to differ from motives in developed countries. Households in developed countries save in large part to accumulate wealth; in contrast household saving in devel- oping countries, especially in rural areas, in large part is a means of guarding against declines of consumption from an already precarious consumption base. As Deaton points out, this precautionary view of saving has very different im- plications for public policy seeking to influence saving than do conventional accumulation-based models ofsaving behavior. Deaton also cites the tremendous need for better data, and his ongoing research for the C6te d'Ivoire using this precautionary saving model. Bank staff member Emanuel Jimenez explored a controversial current policy issue in the third paper, "Social Sector Pricing Policy Revisited: A Survey of Some Recent Controversies." Declining fiscal budgets and increasing demands for health and education services have forced governments and donors to look for alternatives to general revenue financing of publicly provided goods and services. The pricing of public sector goods and services is an area that generates strongly held positions both for and against, especially when social servicessuch as health and education are on the line. Jimenez clarifies the debate by drawing the distinction between ideological differences and disputes over empirical re- lationships. He points out that, even setting ideological differences aside, con- siderably more empirical evidence is needed to demonstrate that the poor can 4 Introduction be protected from the negative consequences of pricing health and education services if the momentum of social sector price reform is to be maintained. Jimenez believes that selective user charges can yield some improvements in equity as well as in efficiency, besides achieving fiscal gains. His paper, however, argues for systematic evaluations of actual attempts at social sector price reform. Brian Van Arkadie of the University of Dar es Salaam took on one of the more difficult assignments of the conference: to assess the policy implications of the literature on the role that institutions have played in promoting-or slowing-development. He chose to define the term "institution" broadly to encompass both the "rules of the game -laws and regulations-and formal M government organizations. Van Arkadie does not believe that it would be ap- propriate to come up with a "check-list of 'good' and 'bad' institutions," and in any case, the literature is long on generalizations and short on empirical evidence. His paper throws light on the factors to be considered in appraising institutional performance, and on how institutions can be changed to enhance their performance. Van Arkadie suggests that it is best to seek answers to specific and relevant questions about the role of institutions, and not attempt to draw up a grand theory about it. Elhanan Helpman of Tel Aviv University was asked to focus on the policy lessons of the recent and rapidly growing literature on strategic trade theory. Strategic trade theory appears to have radical implications for trade policy. Whereas conventional trade theory and its proponents generally support unre- stricted trade regimes, Helpman shows that in many circumstances the new theory argues for selective interventions. In the final accounting, however, both Helpman and the audience agreed that the information requirements of the new theory made it extremely difficult to implement in real-world policy settings. In the absence of much improved information on the way economies work and on their strategic behavior, conventional wisdom-minimal intervention and uni- form tariffs-seems likely to continue to be the order of the day. In the conference's final paper Hans Binswanger, a staff member of the World Bank, looked at the implications for policy of available theoretical and empirical evidence on agricultural supply response. Many recent adjustment programs have had as one of their cornerstones the rebuilding of the country's agricultural sector. Binswanger shows that the effectiveness of such policies depends on much more than "getting prices right.'' He demonstrates that in the aggregate, short- run agricultural supply responses are inherently inelastic, though the long-run responses to agricultural policy changes can be great. Without technological change or considerable infrastructural investment, price changes alone may have much less effect on overall agricultural output than hoped for. Binswanger concludes that in formulating adjustment programs, attention to agricultural growth cannot be delayed until adjustment is completed, precisely because of the long response lags involved. The final session of the conference brought together six acknowledged de- velopment experts in a panel discussion to review the current state of devel- Introduction 5 opment policy research in their own areas of expertise. Each drew the audience's attention to the existing gaps in our understanding of development processes, and outlined problems and concerns relating to developing countries that should shape new directions for basic, applied, and policy research in the future. The organization of an international conference is even more demanding than that of a national conference, and the organization of a new type of conference more demanding yet. We owe special thanks to the World Bank staff who were responsible for so efficiently organizing the conference and overseeing the com- ings and goings of its participants: Shekhar Shah, the Bank's deputy research administrator, provided overall guidance and management for the conference and this volume; Andrea Hodson organized the conference logistics. We would also liketo thank the editorial staff who put together these proceedings, especially Philippa Shepherd and Pat McNees, and the secretarial staff who provided able , support, especially Manny Jandu. Because the World Bank engages its members in a comprehensive dialog over the entire range of economic policies, it has to undertake research and maintain expertise on a broad front. Divisional and Departmental Research Current research activities are best described by reviewing the divisions and departments that make up the research wing of the World Bank's Policy, Re- search, and External Affairs complex. The names of these units provide a useful summary of the range of research interests within the Bank. We start with Bank units which deal primarily with macroeconomic and related issues. Macroeconomic Adjustment and Growth Division. About 25 percent of the Bank's current lending supports adjustment programs being undertaken by mem- ber countries. Although member country and Bank experience in this area is growing, there is much still to be learned. Two problems receiving particular attention now are the issue of the appropriate sequencing of adjustment poli- cies-especially the interactions between external and internal adjustments, and ways of minimizing the negative social consequences of adjustment. Public Economics Division. The research and policy development program of the Public Economics Division covers five broad and interrelated areas: taxation, sector pricing, public expenditures, intergovernmental fiscal arrangements, macro-fiscal linkages. It supports and strengthens the Bank's operational and technical assistance capabilities in various areas of fiscal reform, particularly in the context of adjustment programs. Trade Policy and International Trade divisions. The importance of trade to 6 Introduction development is such that two divisions serve the Bank's needs in this area, one dealing with domestic trade policies on both the import and export sides, and the other dealing with the international trade environment. The Uruguay Round of trade negotiations and continuing trade reforms in member countries ensure lively programs in each area. Financial Policy and Systems Division. The question of how to expand and restructure ailing financial systems is receiving increasing emphasis in World Bank lending and policy advice. World Development Report 1989 dramatized the serious plight of the financial systems in many developing countries, as well as the critical role these systems play in the development process. The Bank's research efforts in this area are growing rapidly and will likely continue to grow. Public Sector Management and Private Sector Development Division. World- wide experience in the last few decades emphasizes the need to increase efficiency in public sector activities. Bank and other research is focusing on ways to increase the efficiency of the public sector while preserving services, especially to the poor. Privatization is one, but certainly not the only, avenue. Although there is general agreement on the objectives in this area, there is considerable debate about the institutional changes needed to get from here to there, and research on these issues is proceeding apace. Promoting private sector activities in general is also an area of significant emphasis at the Bank. In many countries with active adjustment programs, the private sector, especially small-scale enterprises, is at the forefront of efforts to rebuild economies and to improve economic growth. What can the Bank do to encourage private development? How can countries move from highly con- trolled, public-sector-dominated economies to more responsive, diversified econ- omies in which the private sector plays a leading role?The following units address these issues. Debt and International Finance Division. It will come as no surprise to learn that the Bank is concerned with the policy and economic implications of inter- national debt, both official and commercial. Debt is also an area in which it is relatively straightforward to see the link between academic research and the policy debate, for academic research has been highly instrumental in shaping the debate. Bank-sponsored research has contributed, and will continue to con- tribute, to this debate on several fronts, including debt measurement issues. We are also concerned to look at the forms of capital flows that may emerge in the wake of the debt crisis, and the policy measures that need to be taken to facilitate such new flows. The Bank has principal responsibility among other international agencies for the compilation and dissemination of data on the external debt of developing countries and for improvements in the quality of the data. International Commodity Markets Division. The Bank's commodity markets division is responsible for forecasts of the prices of commodities produced by Bank-funded projects. It thus makes price projections for all major commodities Introduction7 and many others over long forecasting horizons. The projections are used to evaluate proposed projects and are thus immediately operationally relevant. The division maintains an active research program to improve its understanding of the operation of particular commodity markets, and it also examines commodity production and trade policies and commodity risk management and finance for individual countries. International Economic Analysis and Prospects Division. Evaluation of Bank and member country projects and policies also requires analysis and projections of the development of the international economy. This division combines the use of global models with topical research to predict short-run developments and long-run trends, and to examine alternative scenarios for the international economy. Besides acting as secretariat to the Bank's Planning Assumptions Com- mittee, which proves assumptions for all key global aggregates, the division also takes a leading role in the preparation of relevant parts of the Bank's Annual ' Report and World Development Report. The research divisions just described focus on country and international issues. Much of the Bank's research work also focuses on sectoral issues. We turn now to describe units that work in the sectoral areas. Agriculture and Rural ~ e v e l o ~ m eDepartment. The issues here have been. n t at the core of Bank research efforts for several decades. In addition, experience, not all of it favorable, with some aspects of agricultural sector lending and with rural development projects has raised renewed interest in such issues as project evaluation under uncertainty, the role of property rights, credit markets, agri- cultural trade liberalization, and rural infrastructure in promoting agricultural development. Environment Department. Academics have more and more realized that en- vironmental concerns are critical and are likely to become more important as the years pass. These concerns are not new to the Bank, but they pose many important positive and normative research issues, at both the microeconomic and macroeconomic level. The environment and environmental economic issues have been singled out as a priority for Bank research during the next few years. Infrastructure and Urban Development Department. Current infrastructure concerns relate to transport, water supply, and sanitation. In transport, research focuses on pricing, regulatory, and financing issues, private sector participation, infrastructure management and maintenance, enterprise reform, and the role of freight logistics management in improving trade performance. Sectoral concerns on water supply and sanitation relate to institutional development and priva- tization, financing, service provision and costs, and usage and disposal patterns as they affect the environment and the possibilities of recycling. Urban development is growing in importance as the developing world's pop- ulations move to urban areas, putting increasing pressure on the urban infra- 8 Introduction structure. Research focuses on urban policies for macroeconomic adjustment and growth, with four priorities: urban infrastructure investment and its effects on productivity and growth; land and housing markets and their impact on private saving and investment; municipal finance and links to financial markets; and housing finance. New research initiatives also relate to the urban environ- ment, waste management, and private sector participation. Industry and Energy Department. The Bank's work on industrial issues has changed as emphasis has shifted toward policy-based lending and away from production in the public sector. The challenge for research is to find policies that will enable the industrial sectors of developing economies to grow and cope with international competition, rather than to search for winners and losers to prescribe to individual member countries. Energy issues have also changed with the times. Today, the main areas being researched are underpricing of energy, energy sector inefficiencies and subop- timal performance, inadequate resources for expansion programs, and weak financial condition of electric utilities. Energy staff are seeking corrective meas- ures involving private sector participation, regulation, and greater autonomy for utilities. Considerable effort is also being directed at minimizing the adverse environmental impacts of providing an adequate energy supply. Population and Human Resources Department. Human resources,hold the key to economic growth and development. The Bank's research directions in the area of human resources are best summarized by the titles of the divisions that make up the Population and Human Resources Department: Education and Employment; Women in Development; Population, Health, and Nutrition; and Welfare and Human Resources. The Bank allocates substantial resources to these cross-cutting themes, and research and policy work in these areas focus on issues of quality of service provision, cost-effectiveness, user demand, and how to implement and sustain policy change. The Central Research Program Bank staff manage all, and carry out most of, the socioeconomic research sponsored by the institution. To supplement its own divisional resources, the Bank maintains a centrally administered Research Support Budget (RSB) that provides resources to Bank units to undertake specific research projects. RSB funds are used mainly to support research collaborations with outside research- ers, and this process is one of the main avenues through which non-Bank re- searchers become involved in Bank research. There are two basic requirements for projects financed through the RSB. The first is that the project must have ultimately discernible real-world implications for developing countries. The second is that the project be rooted within the Bank, specifically that it be sponsored by a Bank unit, which will administer it and take responsibility for its successful completion. These requirements are Introduction 9 designed to ensure that Bank-sponsored research projects are both relevant to the Bank's concerns and well managed. This insistence on operational relevance distinguishes the Bank from other sources of research and research funding. Ultimately, the Bank tries to ensure that research undertaken with its resources has at least a reasonable chance of improving the development prospects of its member countries through either improved policy formulation or better project design. The Research Support Budget sets priorities for developing research in areas likely to be important in coming years. The present list of priorities is the environment, socialist economy reform, and private sector development. The development of research capacity in developing member countries is also re- ceiving special attention. The Research Administrator's office takes an active role in seeking to develop the research program in these priority areas. The priorities are reviewed and updated as research programs develop. This brief overview of the Bank's research structure underscores the Bank's broad commitment to and need for policy-oriented research. (Of course, Bank project design also draws heavily on traditional scientific and engineering re- search, and the Bank supports such research quite heavily through mechanisms such as the Consultative Group on International Agricultural Research. Given the nature of this conference, we have restricted our review of Bank research activities largely to economics.) The breadth of the Bank's research interests and concerns reflects the nature of development economics: although often seen as a subdiscipline of economics akin to labor economics or international trade, in fact it embodies all economic subdisciplines, distinguishing itself by applying these subdisciplines to a particular set out countries. Because development eco- nomics is not a separate discipline, experts in virtually any of the traditional economic and other social science subdisciplines can contribute to "develop- ment" research if they direct their expertise to the specific circumstances-the institutional and social character-of developing countries. P R O C E E D I N G S O F T H E W O R L D B A N K A N N U A L C O N F E K E N C E O N D E V E L O P M E N T E C O N O M I C S 1 9 8 9 KEYNOTE ADDRESS Development Policy Research: The Task Ahead Manmohan Singh The postwar period of development is a history of triumph-not of failure. The increase in life expectancy, the fall in infant mortality, the rate of growth, the achievements in any number of developing countries-nobody at the end of the Second World War would have expected so much. A billion people are still hungry, but it would now be two billion without the achievements that have been made [Rosenstein-Rodan 19841. Postwar progress on the development front until the closing years of the 1970s did much to shake the world's faith in the Marshallian dictum "Natura non facit salt~irn~~ (nature does not make leaps). The relation between the world of ideas and the world of action is by no means unambiguous, but it is fair to assert that the pioneers of development economics did a lot to sustain faith in the possibilities of economic progress in the developing countries. World War II to the 1980s: An Era of Optimism The basic thrust of development economics has been activist-that develop- ment can be managed. The analysis of strategic factors in development, of development strategies, and of investment criteria has thus sharpened the focus for policy intervention as well as yielding valuable insights into the growth process. The earlier postwar works singled out capital accumulation and the allied process of saving as central to the management of development. The strategic role of investment is still unquestioned. But later works, in bringing out the vital role of technical progress, entrepreneurship, and human resource development, have enlarged our understandng of growth as the product of a more complex process. The later literature has also sought to correct an earlier tendency to assign rather a passive role to the rural sector. Rising agricultural productivity is now regarded as crucial in the successful management of development. Manmohan Singh is secretary general of the South Commission, Geneva. The views expressed are the personal views of the author and should not be attributed to the South Commission. O 1990 The International Bank for Reconstruction and Development i .IHE WORLD BANK. 12 Keynote Address At the international level, analysis of the relation between the periphery and the center, and of how gains from participation are distributed in the interna- tional division of labor, spurred reformulation of the theory of international economic policy, particularly of protection. It also stimulated the search for new international arrangements for more equitable management of North-South re- lations. The establishment of several international institutions-the United Na- tions Development Programme, the International Development Association, the three regional development banks, and the United Nations Conference on Trade and Development--owes much to the inspiration of the pioneer work in devel- opment economics. The optimistic spirit of mainstream development economics--expressed in the view of some of our pioneers that self-sustained growth could take off within a single generation-provided the psychological underpinning for substantial mobilization of latent national energies, and for international support for the development cause during the first three postwar decades. The 1980s: A Change in Climate The events of the 1980s have assailed that optimism. The economies of East Asia and Southeast Asia have continued to grow rapidly, and the slower-moving countries of South Asia have also managed to improve their performance. But most countries in Africa and Latin America have retrogressed sharply. The result is a profound disillusionment, and skepticism about prevailing development strategies, policies, and programs. The crisis of development challenges the entire development community to come up with strategies to put the affected countries back on the path of sustained growth. Development research must assist this effort. The search will not be easy. Several obstacles stand in the way, which have to be surmounted or neutralized to revive growth impulses and then to sustain their momentum. First among these obstacles is a world trading environment far less expan- sionary than in the golden years from 1945 to 1973, when annual export volume growth at a steady 6 to 7 percent provided a climate hospitable to the expansion of manufactured exports from developing countries. It is highly questionable whether the emerging international environment will be as friendly. Second, the 1980s have seen a sharp deterioration in the export of capital from developed to developing countries. For several developing countries, the net transfer of resources from industrial nations is now negative. The burden of external debt is now a recognized obstacle to the revival of growth in most countries of Africa and Latin America. Yet no viable solutions have emerged from the voluminous discussions of this subject in recent years. The economic chaos and uncertainty created by the debt overhang strongly encourage capital flight at the same time as they deter the infusion of fresh capital from abroad. The climate is not conducive to large-scale private international flows, except to East and Southeast Asia. With the possible exception of Japan, no industrial country seems inclined to step up official flows-ither bilateral or multilat- eral-to the developing world. A third difficulty that developing countries have to face is the negative fallout from accelerated technical progress in developed countries-such as reduced use of raw materials per unit of final product. In addition, technological advances in areas such as microelectronics, computer numerically controlled machines, and computer-aided designs threaten the traditional division of products into labor-intensive and capital-intensive categories. As a result, cheap labor may cease to be the dominant determinant of comparative advantage, to the detriment even of developing countries making traditionally labor-intensive products, such as textiles. With a few exceptions, the technological distance between the developing and the developed countries is widening. Neoclassical international trade theory, by postulating identical ~roductionfunctions for different products in various coun- tries, assumes this problem away. That this assumption is unrealistic becomes obvious when we consider that technical knowledge is often private property, that new advances in science and technology are increasingly being privatized, and that developed countries are now making concerted attempts to coerce developing countries into conforming their domestic legislation on the protection of intellectual property to the interests of technology exporters. We must face up squarely to the consequences of unequal international access to technical knowledge. On the domestic front too the present situation has its shortcomings. First is the serious pressure on resources from population growth rates, currently 2 to 3 percent or more. Sustained development over a period of about twenty-five years was earlier expected to alter attitudes toward childbearing fundamentally and thus bring a reduction in birth rates that would offset the welcome reduction in death rates attendant on the process of development. This change in attitude has not ma- terialized in most countries outside East Asia. Fertility rates are declining, but far too slowly. While many developing countries are faced with a net outflow of capital, their domestic requirements for capital per unit of output may well rise. Often, the capital stock is technologically obsolescent. And past neglect both of proper maintenance and of timely replacement squeezes resources still more. The in- creased demand for social overhead capital associated with urbanization has a similar impact. Furthermore, capital requirements have often been kept low by neglecting environmental protection in the process of resource use. In several developing countries, land and water degradation now seriously threatens the sustainability of growth-but if compensatory measures are adopted, the cost of output is bound to increase. Investment and savings associated with a given growth rate will therefore need to be much higher than in the past. Finally, the inducement to save is itself under pressure because the influential, upper-income groups in developing countries cannot resist the temptation to adopt the life styles and consumption patterns of the societies of the post- 14 Keynote Address industrial age. These demands will, no doubt, siphon resources from essential investments, while undermining the cohesion and unity in developing societies. The widening social gap between the ruling minority and the great mass of people can erode the moral authority of these elites to command willing accep- tance of restraint on the growth of consumption. As a result, inflationary pres- sures may become endemic, reinforcing in turn the demand for authoritarian political structures, which inevitably bring further alienation and instability. The Task Ahead The task of reviving and sustaining development in the developing world in the decade that lies ahead is formidable. It will demand highly innovative meas- ures to counteract the negative influences now at work. An improvement in the external economic environment would help, but cannot be counted on. Devel- oping countries must think in terms of an efficient policy framework that will sustain the pace of development even if the international economic environment fails to improve. The need for structural reform to promote greater mobilization and more efficient use of domestic resources is inescapable. In what follows, I deal with two issues of reform prominent in current debate: the role of the state and the reform of the trade regime. (Food security and development of human resources are no less important but are not discussed here for want of time.) The discussion leads into consideration of a third bone of contention-concerned, again, with reform, but this time reform of the as- sistance mechanism itself, rather than the countries assisted: the question of how (and how much) to impose conditions on development loans. My analysis is not intended to be exhaustive. Its objective is limited. I accept the necessity of reform, but I argue that no single development model can be applied universally. The agenda must be tailored to the specific requirements of each country. I also argue that some of the now fashionable reform packages, with strong ideological overtones, distract attention from the fundamental task of modernizing the state for the successful management of development. In the final section, I discuss the case for an independent, objective international mech- anism for laying down performance norms and assessing the performance of countries that are receiving external assistance. In the postwar years, most developing-country governments have been activist in their management of development. Mainstream development economics has on the whole supported this activism on the grounds of pervasive market im- perfections, externalities, and discontinuities and deficiencies in the private sec- tor's risk-taking and entrepreneurial ability. As a result, the state's role has expanded, both in the regulation of private activity and in the actual operation of strategically important enterprises. In the 1980s, the setbacks to development have eroded confidence in past development strategies and brought this state activism under fire. Even more influential has been the reassertion of a strong pro-private sector ideology in the major developed countries, which has inevitably spilled over to the inter- national financial institutions. The developing world is now being called upon to accept a development and structural adjustment model in which the state's role is drastically reduced. A voluminous literature now argues that excessive state interference in developing countries has distorted the price structure, pro- moted inefficient resource use, and given rise to undesirable rents for influential pressure groups. The tendency now is to blame the development setback of the 1980s entirely on the cumulative process of excessive state dominance of de- velopment. This view is clearly one-sided, for it ignores the traumatic im- pact-particularly in Latin America and Africal+f international events beyond the control of developing countries (such as the virtual cessation of international bank lending, the steep rise in real rates of interest, and the collapse of inter- national commodity prices). Also overlooked is the highly interventionist role of the state in the most successful economies of East Asia (on this point, see Anglade and Fortin 1987). Even so, any objective observer must concede that developing countries often operate an overextended state apparatus, forgetting that administrative capa- bilities are one of the scarcest factors of production. The result is that scarce administrative resources are spread too thinly to achieve the basic development objectives. Furthermore, the assumption implicit in much of Anglo-Saxon eco- nomics that the state is a platonic guardian of the public interest does not hold good all the time. In this regard, political economists would do well to pay more systematic attention to the influence of organized pressure groups on government decisionmaking in developing countries. The role of the state in the economic life of the developing countries thus needs to be reappraised, but the appraisal must be marked, not by strong ideological overtones, but by pragmatic consid- eration of feasible alternatives. My own guess is that a good case can be made for substantial deregulation in many developing countries. Judging by the Indian experience, for instance, such regulatory measures as industrial licensing and import controls could be loosened with much social and economic advantage. However, there are no universal rules to determine the proper mix of regulation, promotion, and re- liance on market forces. The agenda has to be country-specific-an area offering much scope for fruitful research. 1. Professor Albert Fishlow (1987) has shown that the relative severity of the external shock has been an important reason why Latin American economies have performed less well in the post-1980 period than East Asian economies. 16 Keynote Address Modernization of State Apparatus In practice, the issue is not so much the size of government as its quality and effectiveness. Most developing countries lack an alternative to a strong, pur- poseful government providing a strategic sense of direction for the development process. An example is food policy, recent discussions of which have been dominated by a concern to get prices right. Certainly, remunerative farm prices are a linchpin of a workable agricultural strategy. Farmers do respond to price incentives. But the aggregate supply function of farm produce as a whole (as opposed to that of any single crop) is likely to be greatly influenced by the state of available agricultural technology, credit, and marketing arrangements, and by the man- agement of water resources. The efficient performance of all these activities calls for collective action and often large investments by the public sector. The mod- ernization of the state is therefore crucial to the management of development. The state's capacity for autonomous action should then be enhanced in stra- tegic areas that cannot be left to the care of market forces-a vision not incon- sistent with shedding regulatory functions that may have outlived their useful- ness. But there are several other critical considerations. A firm commitment to the rule of law and respect for human rights is one. The improvement of selection, training, and skill formation in public services is another. The maximum possible decentralization and use of appropriate consultative devices to involve local communities at the grass-roots level is yet another important dimension. A reform of the tax system designed to provide for a built-in elasticity and buoyancy in the tax structure is a high priority, as is the establishment of cost-effective delivery systems to provide basic social services, such as health and education. The final issue is the provision of institutional safeguards to ensure that the state performs its essential functions efficiently and equitably, and to curb the ma- nipulative power of organized vested interests. In this context, the role of de- mocracy and what has come to be known as "glasnost" in the Soviet Union is critical. Deregulation is often advocated as a means of depoliticizing economic proc- esses. But in real life, politics cannot be wished away in the management of development. So, we should try to get the politics right. Deregulation, when accompanied by competition, can help reduce the grip of organized vested in- terests on the machinery of the state. But competition, as Adam Smith recognized long ago, cannot be regarded as a natural outcome and requires a political framework for its enforcement and effectiveness. A reform package has to be much more comprehensive than simple deregulation. In the final analysis, only a multidisciplinary and multidimensional approach can devise credibile, coherent, and legitimate reform. There is no alternative to ~oliticalreforms designed to promote participatory and people-centered ap- proaches to development. People want to be consulted, involved, and taken into confidence. Modernizing of the state means reforming institutions to make all this possible. For by now, it is fairly obvious that reform by stealth is not going to work. Management of the Public Sector Even the most conservative advocates of the private sector have recognized that the long gestation period for social overhead capital projects, the large investments required, and the state of capital markets in developing countries make these projects less than enticing to the private sector. For this reason, public sector involvement in expanding social overhead capital has not provoked much resistance. Many developing countries, however, take the more expansive view of in- cluding in the public sector's purview a number of industries considered to be of strategic importance. State intervention has sometimes been justified on grounds similar to those applicable to social overhead capital, such as barriers to entry and lack of private initiative (Levy 1988). Another argument is that, because profits constitute the most important source of accumulation in a dy- namic economy and are an important determinant of income distribution, their socialization and the prevention of their wasteful consumption by private cap- italists could at once accelerate accumulation and help reduce inequalities of income and wealth. And another justification for state involvement is that public enterprises can set the pace for strengthening technological capabilities and accelerating technical progress.' The surge of pro-private sector ideological fervor since the late 1970s has rekindled the old antagonism toward the public sector of developing countries. With public enterprises increasingly viewed as an inefficient drag on the econ- omy, privatization now figures prominently in the programs of structural reform sponsored by the World Bank and the International Monetary Fund. Anti-public sector sentiment cannot, however, be dismissed as purely an expression of ideological bias. There have been visible gaps in the performance of the public sector. Results of public sector operation in India, for example, have fallen far short of original expectations. The enterprises have been effective in bringing some balance to the regional distribution of industry-an important consideration in managing a polity as diverse as India-and have been pioneers in introducing new technology. But, perhaps because of the monopolistic en- vironment in which they often operate, they have not kept abreast of techno- logical developments. The biggest disappointment has been their failure to gen- erate adequate surpluses to finance sustained expansion of investment. Both operational deficiencies and restrictions on their freedom to vary prices have contributed to this outcome. 2. All these arguments played an important role in the expansion of the public sector in India during the First and the Second Five-Year Plans (seeSingh 1986). 18 Keynote Address But if further expansion of the public sector is not the answer, neither is wholesale privatization of existing public enterprises. Even in countries where the private sector is quite well developed, such as India, private sector manage- ment even of unsophisticated industries, such as textiles, has been far from encouraging. The privatization experience of developed capitalist countries has little relevance for developing countries with limited entrepreneurial and risk- taking capabilities. Moreover, economic policies do not operate in a political vacuum. Often, proposals for large-scale privatization do not enjoy wide public support. Thus, in practice, there is often no alternative but to get on with the task of improving the efficiency of public enterprises. How best to manage the public sector is a key issue of contemporary devel- opment policy. Can new ground rules be devised for giving autonomy to public enterprises, so that they can function without day-to-day government interfer- ence in their operation? Is it realistic to expect government, as owner, to pass an ordinance of self-denial, of noninterference in the day-to-day operations of public enterprises? Will managers-whose appointment and removal are in the hand of government--exercisetheir powers or play safe by consulting frequently with government authorities even on day-to-daymatters?And, since many public enterprises function in a noncompetitive environment, how can effective pricing policies be evolved which neither reward inefficiency nor provide undeserved monopoly rents? The answers to these profoundly important questions are again likely to be location-specific. Here, then, is another area for fruitful multidis- ciplinary research. The outcome of efforts to improve the working of the public sector depends a great deal on the character of the government. Under a regime that regards the state as the private property of rulers, the development potential of the public sector cannot be realized. With a rotten state structure, public enterprises cannot be pacesetters of development. But then a rotten state structure is also unlikely to provide a hospitable climate for the functioning of private enterprise by the rules set down in orthodox textbooks. It should not be difficult to make a realistic assessment of the limitations of the "crony capitalism" which flourishes in a number of developing countries. Nor should we perhaps overlook the potential for reform of defective state structures as an inescapable part of the process of growing up. In the final analysis, the quality of a polity is a major determinant of the pace of development. And the quality of a polity cannot be divorced from the quality of the state. Privatization is no panacea for all the ills of the public sector. It must not distract attention from theessential nation-building task of modernizing state structures. There are no shortcuts to prosperity and progress. Citing the success of the East Asian countries' export-oriented development strategies, international financial institutions have been pushing the developing countries toward more export-oriented policies for managing their external pay- ments. Intense pressure has been brought to bear on developing countries to devalue and liberalize their import regime so as to enhance the relative profit- ability of exports. A dynamic export sector is clearly important for the efficient management of the economy, since import needs inevitably rise at least as fast as national income in the process of development. Thus, there is certainly a strong case for protection levels to be rationalized in line with perceived long-term comparative advantage and, over time, to be reduced to provide the stimulus of foreign competition in domestic markets. And the anti-export bias implicit in the incentives of tariff and exchange rate systems needs to be removed. Industrialization based on the export of labor-intensive products has contrib- uted to the growth of output and employment in the labor surplus economies of East Asia. And such countries as India have missed valuable opportunities by underestimating the growth potential of labor-intensive industrial products and by investing too little in export efforts. In many African countries, however, domestic supply conditions-including the level of development of human re- sources-do not favor export-led industrialization based on labor-intensive products. They have to rely on the export of products intensive in natural ~~~~~~~~~~ften produced with the help of transnational corporations-which cannot generate the same dynamic effects on income and employment as labor- intensive industrial products. So, what scope is there for growth of exports of labor-intensive manufactures from developing to developed countries? Once demand conditions improve, supply bottlenecks can be overcome. But will markets still be available if all the countries that have pursued inward-looking development strategies switch over to export-oriented strategies based on labor-intensive manufactures? This is a critical empirical question: a reliable answer to it can have a material influence on future development strategies. The beginnings of an answer may be indicated by the priority accorded to developing countries' interests in the Uruguay Round of multilateral trade negotiations. IV. DEVELOPMENT ASSISTANCE AND CONDITIONALITY In the 1980s both international financial institutions and bilateral donors have massively increased the conditionality attached to their development assistance. In principle, one cannot object to performance norms and conditionality designed to ensure that this assistance is used to promote genuine development. Such an assurance can only strengthen the hands of those in the developed countries who believe in a concerted global attack on world poverty. But the developing coun- tries also need an assurance that any monitoring and policy stipulations are based on objective analysis and are in no way colored by ideological preference and prejudice. The international credit mechanism is in the final analysis a highly political mechanism. Creditors-particularly those who act as lenders of last 20 Keynote Address resort-are in a strong bargaining position and are often able to impose their views on borrowers. Only an impartial international evaluation of development performance can inspire the confidence and respect of both donors and recipi- ents. The rationale for such a mechanism is well set out in the essay from which I quoted at the outset: Today we have competence, finance and no democracy in the international banks-and democracy and no finance in the United Nations. The 1954 ECLA report proclaimed the need for a separation of programming and financing. An independent body-not responsible to either creditors or debtors-should evaluate the programmes, and resources should be allo- cated according to that verdict. The World Bank has a good staff. . . but the developing countries have no confidence in the vote of its board because creditor countries have the overwhelming majority; the developed coun- tries, on the other hand, have no confidence in the United Nations. It is part of national sovereignty for each nation to limit its own rights. There will be no satisfactory solution to this problem without some sort of ar- bitration. Only an International Development Council-an International Court of Economic Justice-can solve the problem. The Committee of IX of the Alliance for Progress was an attempt to apply such an International arbitration. It failed because of sabotage on both sides, but all great ideas first fail. All progress is first proclaimed to be impossible, but is then realized [Rosenstein-Rodan 19841. The experience of the 1980s suggests that this may well be an idea whose time has come. At all events, it should figure prominently on the agenda for development policy research. Anglade, Christian, and Carlos Fortin. 1987. "The Role of the State in Latin America's Strategic Options." CEPAL Review 31 (April): 219-44. Fishlow, Albert. 1987. "Some Reflections on Comparative Latin American Economic Performance and Policy." World Institute for Development Economics Research Working Papers 22 (1):42. Levy, Brian. 1988. "The State-Owned Enterprise as an Entrepreneurial Substitute in Developing Countries: The Case of Nitrogen Fertilizer."World Development 16: 10. Rosenstein-Rodan, Paul N. 1984. "Natura Facit Saltum: Analysis of the Disequilibrium Growth Process." In Gerald M. Meier and Dudley Seers, eds., Pioneers in Develop- ment. New York: Oxford University Press. Singh, Manmohan. 1986. "The Quest for Equity in Development." R. R. Kale Memorial Lecture, Gokhale Institute of Politics and Economics. Pune, India. P R O C E E D I N G S O F T H E W O R L D B A N K A N N U A L C O N F E R E N C E O N D E V E L O P M E N T E C O N O M I C S 1 9 8 9 Developing Countries and the Uruguay Round of Trade Negotiations Marcelo de Paiva Abreu This paper, which summarizes recent developments in the Uruguay Round of mul- tilateral trade negotiations, addresses the key questions developing countries must con- sider in their negotiating:What are the trade and welfarecostsand benefits of protection for developing countries?How will developingcountries be affected by, and what might they concede in connection with, tariffs and tariff escalation, the protection of textiles and apparel, such market access issues as antidumping and countervailing duties, and such 'gray area" measures as voluntary export restraints? What concessions should they want-and be willing to make? What do they stand to gain or lose from the so- called new themes of the Uruguay Round: trade-related investment measures, intellec- tual property, and services?What do net importers of food products stand to lose or gain from reduced agricultural protection? How will dismantling the Multifibre Ar- rangement affect developing countries? How will overall trade liberalization affect them? And should trade issues be negotiated in isolation from such factors as the debt crisis and conditions imposed by international agencies? The aim of this paper is to assess the effect of protectionist policies, in both industrial and developing countries, on the interests and influence of developing countries in the current Uruguay Round of multilateral trade negotiations under the General Agreement on Tariffs and Trade (GATT). The GATT rests on three pillars. Two of them-the most favored nation prin- ciple (MFN), which automatically extends bilateral concessions to all GATT par- ticipants, and the prohibition of nontariff trade restrictions-are well in line with theoretical requirements. But the third-the principle of reciprocal conces- sions (in effect the "balancing" of reductions in import tax revenues)-hasno adequate rationale in trade theory. Indeed, this principle encourages negotiating tactics that in some respects impede liberalization. Marcelo de Paiva Abreu is a professor of economics at Catholic University, Rlo de Janeiro. He wishes to thank officials of GATT and of the Brazilian Government for their cooperation. The paper does not necessarily reflect the views of these officials. O 1990 The International Bank for Reconstruction and Development / THE WORLD BANK. 22 Developing Countries and the Uruguay Round The first ~illar-the MFN clause-has been eroded by the organization of free trade or preferential trade zones. The second-the "tariffs only" principle-was shaken right from the start by article XI:2(c)of the original rules, which allowed imposition of quantitative restrictions on imports of agricultural products (see Abreu and Fritsch 1987, pp. 24-29). This explicit breach of GATT principles reflected the Geneva negotiators' need to reconcile their drive for free trade with the widespread commitment of industrial country governments (notably the United States) to price support programs, export subsidies, and import restric- tions for temperate zone products. The original GATT rules did not do the'same for manufactures. Controls were only permitted under circumstances either of exceptional balance of payments difficulties or of '"disruptive" import growth-the so-called injury clause im- posed by the U.S. Congress as a condition for approval to negotiate the postwar trade treaties (see Gardner 1969, p. 159). Thus, once balance of payments conditions stabilized in Europe in the late 1950s, it was through the "injury" argument that most restrictive practices against manufacturing imports crept in. A landmark in the process was the U.S.-sponsored Long Term Agreement in cotton textiles in the early 1960s,a quantitative trade restrictive agreement that totally contradicted GATT principles. This was the beginning of a long history of textile and clothing- protection which culminated in a series of Multifibre - Arrangements (MFAS). More recently, such inherently illegal quantitative restric- tions negotiated outside the GATT have been generalized and thinly disguised as "voluntary export restraints" (VERS). The third pillar of the GATT-the principle of reciprocity-has remained in place, and it remains a problem. Although successive multilateral trade nego- tiations (MTNS) achieved impressive results in reforming tariffs until the mid- 1960s, it became increasingly clear that in practice the operation of the liberal and formally equitable rules of the GATT was distributing the benefits of trade liberalization unevenly. The traditional approach has been to measure the value of tariff concessions as equivalent to import volumes in a given year multiplied by the tariff rate changes granted on those products. This practice implied that in the "reciprocal bargaining" process established at MTNS, the substantive concessions favored industrial nations or trading blocs (which exchanged bilat- eral concessions that were generalized through the MFN clause), and excluded products of export interest to developing countries because they were not in- teresting as a basis for exchanging concessions between developing countries (see UNCTAD 1968, p. 94; GATT 1979, pp. 120-22). GATT came thus to be seen as a rich men's club from which developing countries derived little advantage. Although such deficiencies, identified by the Haberler Report (GATT 1958, pp. 8-12) were formally on the GATT agenda by the late 1950s, no practical changes were immediately forthcoming. The only noticeable change was the rather formal recognition, inserted under a new Part IV in 1965, of the possibility of special treatment for developing countries. The developing countries' growing disillu- sionment gradually undermined the GATT'sposition as a forum for the discussion of North-South trade relations, finding expression at the first U.N. Conference on Trade and Development (UNCTAD) in 1964 in demands that industrial coun- tries extend a Generalized System of Preferences (GSP) to all developing countries. Beginning in 1971, all industrial countries introduced GSP schemes by 1976. Developing countries, which had been marginal participants in the Kennedy Round (1964-67)were more active in the Tokyo Round (1973-79).The Tokyo Round brought tariff reductions, codes on nontariff barriers, and the "frame- work agreement." The framework agreement, on which the developing countries were especially active, provided through its "enabling clause" a standing legal basis for GSP to breach the most favored nation principle. The agreement also made it easier for developing countries to adopt trade measures to foster par- ticular industries. In return, developing countries agreed to a "graduation" prin- ciple which related the capacity to make concessions to level of development (see Winham 1986, pp. 141-46, 274-80). The codes negotiated in the Tokyo Round to counter rising nontariff protec- tion ended up by undermining the MFN clause because the principle of MFN conditionality was raised to try to limit MFN treatment to signatories of specific codes. The rights of nonsignatories to MFN were explicitly recognized by the GATT in 1979, but the trade policies of some contracting parties did not seem to reflect this decision (seeHudec 1987, pp. 81 ff., and Winham 1986, pp. 355- 60). For many years the trade policy interests of the group of developing countries could be reasonably described as convergent. But as the economic structure of many of these countries has grown more heterogeneous, so has their trade structure. Many developing countries are still basically producers of commod- ities, but quite a few are not. Their agenda for trade negotiation therefore differs. Even between primary producers, differing commodity export structure-for instance, between temperate and tropical agricultural commodities-can mean conflicting aims for trade policy. Discriminating preferential treatment results in divisive tensions between developing countries with a similar export structure. This paper tries to take these differences into account. 11. GATT NEGOTIATIONS IN THE 1980s The balance of priorities reflected in the 1982 GATT Ministerial Declara- tion-the forerunner of the Uruguay Round of MTNS-had shifted quite drasti- cally by the launching of a new round in 1986. A backlog of unresolved is- sues-nontariff barriers, agricultural subsidies, and other problems relating to trade in goods-had originally headed the agenda. By 1986 these had yielded the limelight to the "new themes -a set of issues selected by the United States M in a strategic move to adapt the rules governing direct investment and intellectual property rights to a changing environment affecting the growth opportunities and the competitive edge of U.S. firms. 24 Developing Countries and the Uruguay Round New Themes: TRIPS, TRIMS, and Trade in Services The new themes were a somewhat heterogeneous bunch of issues, some of which had been only marginally treated by the GATT in the past, covering trade- related aspects of intellectual property rights (TRIPS), trade-related investment measures (TRIMS), and trade in services (formore detailed discussion, see Abreu and Fritsch 1988). ("High technology" goods, initially included, were dropped from the list of "new themes," as it turned out to be difficult to show how they differed from other goods from the point of view of GATT rules.) With the support of Japan, and more equivocal backing from the other industrial countries, the United States pressed for the inclusion of the new themes in the agenda for the next MTNS. Some developing countries resisted all three; most opposed inclusion of trade in services. TRIPS-regulated mainly by international conventions under the jurisdiction of such agencies as the World Intellectual Property Organization (WIPO) and UNESCO-had traditionally been of little interest to GATT. Dissatisfied with the enforcement of the rules and with their allegedly increasing infringement, es- pecially in semi-industrialized countries, the industrial countries included TRIPS in the 1982 Ministerial Declaration with a view to bringing them under the aegis of GATT rules and enforcement capabilities. This early initiative met with strong developing-country resistance in a clash of views that has continued and is unlikely to be soon resolved. TRIMS. GATT discussions on TRIMS centered on the legality of national regu- lations that require foreign firms to export a given amount of their output, or to purchase a given amount of their inputs or equipment from domestic suppliers. A GATT panel established in 1984 concluded that the export performance reg- ulations are not inconsistent with ATT rules, but that the import content ob- G ligations were inconsistent with Article III:4. Developing countries have reserved their position. Trade in services. In the 1982 ministerial session, the United States pressed hard for discussion of enlarging GATT to cover trade in services to be included in the work program. Opposition from many developing and even some indus- trial countries, on grounds of insufficient information, deferred discussion of the issue until the 1984 session, to give time for national studies and stimulate the exchange of information. Placing trade in services within the GATT framework was initially the most divisive of the new themes. Developing countries felt that the unresolved tra- ditional issues that originally headed the agenda in the 1982 Ministerial Dec- laration should not have been displaced by the debate on services. And they feared that the issue was likely to strengthen the hand of the industrial countries in the new round of negotiations. Developing-country misgivings were shared by some members of the European Community (EC), whose support for putting services into the GATT framework was less than wholehearted because of the complex legal and technical problems involved, and because the theme covers a large number of sector-specific issues that encroach on the territory of other international organizations. The Negotiations: 1986-1989 When the ministers arrived in Punta del Este in September 1986, they had before them two formal agenda proposals. One, tabled by Colombia and Switz- erland-with overwhelming support from developed countries and substantial support from developing countries-included all the new themes in a single track. The other, tabled by the G-10 coalition-a group of developing countries formed by Argentina, Brazil, Cuba, the Arab Republic of Egypt, India, Nica- ragua, Nigeria, Peru, Tanzania, and Yugoslavia-included none of them. The Colombian-Swiss proposal had in fact foundered before Punta del Este, when the EC withdrew its support to the wording on agriculture. The eventual compromise reached at Punta del Este in the Uruguay Round Declaration (GATT 1986) distinguished trade in services from the other subjects formally encom- passed in the negotiations, including the other new themes. With GATTSecretariat support, the ministers established a special Group on Negotiations on Services to carry out negotiations in this area and make recommendations to the Trade Negotiations Committee. This arrangement, however, has little hope of heading off developing countries from exchanging concessions in services for concessions in the trade of goods. Other negotiating groups were to deal with: tariffs; nontariff measures; prod- ucts based on natural resources; textiles and clothing; agriculture; tropical prod- ucts; GATT articles; MTN agreements and arrangements; safeguards, subsidies, and countervailing measures; trade-related aspects of intellectual property rights, including trade in counterfeit goods; trade-related investment measures; dispute settlement; and functioning of the GATT system. The results of two years of negotiations were presented to the Montreal Mid- Term Ministerial meeting of December 1988 (see GATT 1988a, 1988b). The meeting ended deadlocked on four issues: agriculture, intellectual property, tex- tiles and clothing, and reform of the safeguards system. On agriculture, the U.S. position that all trade-distorting subsidies affecting agricultural products should be eliminated within a specified time frame was unacceptable to the EC (predictably, in view of earlier French-inspired intran- sigence on export subsidies, and continuing EC insistence on maintaining a dual price system with different prices for exports and home consumption). The deadlock in agriculture galvanized Argentina and the other Latin American members of the Cairns group of agricultural free traders (see section 111) into action. Their activities eventually achieved agreement to shelve the results so far obtained by eleven negotiating groups, pending the results of further consulta- tions and negotiations to be held in early April 1989. In April the deadlock on agriculture was broken by U.S. acceptance of the EC refusal to commit themselves 26 Developing Countries and the Uruguay Round to ending subsidies, and more flexibility from the EC on the freezing of protection in the short term. On intellectual property, as on agriculture, the gap between the extreme po- sitions after Montreal was wide. The industrial countries continued to urge that - GATT'srules and disciplines in this area be enlarged, and enforcement as well as dispute settlement improved, while the large developing countries-Brazil and India-insisted that WIPO was the proper forum to deal with the matter. The industrial countries' views prevailed in April 1989: it was decided that negoti- ations should proceed in the GATT, and that discussion of which international organization would be in charge would be postponed to the end of the Round. As the safeguard issue was disposed of through an agreement on the nego- tiating group's program of work, pressure mounted on the developing countries to reach agreement on textiles and clothing. The outcome of negotiations on textiles and clothing was disappointing: it was agreed that within the time frame of the Uruguay Round a decision will be reached on modalities of integration of this sector into the GATT. This is to include the MFA (see News of the Uruguay Round 1989, pp. 8, 21). Ironically, divergences among industrial countries at Montreal over agriculture troubled the negotiations more than the differing stands of developed and de- veloping countries at Montreal on the new themes. On the prime bone of con- tention-services-negotiations advanced steadily, to the visible delight of the director general of GATT (interview, MOCI 1989). The inclusion of the principle of national treatment of foreign suppliers in the agreed Mid-Term text is a major breakthrough and an important concession by the developing-country G-10 coalition on services. The developed countries for their part have toned down their insistence on a multilateral framework for trade in services by accepting the proviso that before such a framework is accepted "concepts, principles and rules will have to be examined with regard to their applicability to individual sectors and types of transactions to be covered by the multilateral framework" (see News of the Uruguay Round 1988, pp. 40-43). Results in Montreal in other groups under the Group of Negotiations on Goods were mixed. Some of the "successful" groups owed their achievement more to the elaborate ambiguity of agreed drafts than to any substantive advance in negotiations. Main results of interest to developing countries seem to be taking shape in relation to tropical products involving $25 billion (all dollars are U.S. dollars; billion = 1,000 million) in trade, tariff reduction on the order of 30 percent, and the transformation of nontariff into tariff barriers. Some advance is to be expected on more institutional GATT issues, such as the improvement of dispute settlement machinery and the functioning of the GATT system. The latter will involve efforts to improve the GATT trade policy review mechanism and to strengthen its links with other multilateral organizations such as the World Bank and the International Monetary Fund and will entail greater ministerial involvement in the GATT (see News of the Uruguay Round 1988, pp. 26-39). The most relevant costs of protection for developing countries as a whole relate to the value of forgone exports, displaced by protection in industrial countries, and the deadweight losses entailed by protection of their own domestic markets, which distorts production and consumption decisions. (See Bhagwati 1987a on why the computation of deadweight losses is likely to underestimate the costs of protection.) But specific issues raise differing concerns for different developing countries. The discussion is thus organized thematically in subsec- tions covering the main issues that affect the interests of developing countries, and the main areas where concessions might be exchanged. Tariffs, Tariff Escalation, and Preferences Tariffs have become less important in industrial countries owing to agreed reductions in previous MTNS-they now average around 5 percent. But this decline has been at least partly offset by the rise in nontariff barriers, and the average nominal tariff hides important variations that in general tend to hurt the trade interests of deve~o~ingcountriesmost. The effect of tariff peaks, high internal taxes, and tariff escalation on processed tropical products is well known. If a 10 percent ceiling were set for tariffs, imports of developed countries would rise by 1.5 percent, as against 4.9 percent if all tariffs were eliminated (see Erzan and Karsenty 1987). Internal taxes on tropical products in developed countries amounted to $5 billion in 1983, excluding $22 billion on tobacco (see Commonwealth Secretariat 1987, p. 14). The processing of tropical products in developing countries is heavily penalized by the escalation of tariffs (and non- tariff measures) in developed countries. The result is increased protection of value added, which twists the worldwide distribution of value added along processing chains in favor of the industrial countries. This has prompted com- pensating export taxation by developing-country exporters in a cumulative trend that restricts the market for tropical goods (see Cable 1987b, tables 22-1 and 22-2; Yeats 1987). Trade and welfare gains related to some processed tropical products such as roasted coffee are likely to be significant (see ValdCs and Zietz 1980, p. 34). Many exporters of tropical products enjoy preferential entry in developed markets. Tariff reduction, which erodes these advantages, may thus be opposed by participants in preferential arrangements, though concessions such as those on internal taxes in the EC may avoid such difficulties. Despite its institutional drawbacks (forexample, its limited inclusion of textiles and agricultural products and its restrictive safeguards and rules of origin) the General System of Preferences (GSP)is important to the expansion of developing- country exports-more because it creates trade than because it diverts it (see Karsenty and Laird 1986). It mainly benefits the larger developing economies such as Brazil, Hong Kong, Korea, and Taiwan. The major donor countries have 28 Developing Countries and the Uruguay Round instituted a policy of graduating country-product pairs as a direct consequence of the enabling clause of the framework agreement of 1979. Their argument is that the distribution of GSP should be equitable and that, as some developing countries become competitive, their preferential treatment should be withdrawn in favor of the least developed countries. The argument is contradicted by the evidence that trade in graduated products tends to be diverted either to developed countries or to the more advanced developing countries (see MacPhee 1986, pp. 10-12). Experience has repeatedly (and not surprisingly) shown that the devel- oping countries who enjoy best access to developed markets are those relatively less able to supply the products, and vice versa. As developed countries increas- ingly emphasize full reciprocity, some of the large developing countries seem to be reconsidering their interest in GSP, as they feel the balance of benefits and costs shifting against them. The proliferation of preferential trade agreements among the major trading nations is increasingly undermining the GATT. Such arrangements are traditional EC policy, and have spread over former colonies, the Mediterranean Basin, and countries of the European Free Trade Association (EFTA). U.S. preferential agree- ments have been with Caribbean countries, and the more recently negotiated free trade area with Canada. The possibility looms of preferential arrangements between the United States, Japan, and the Asian newly industrialized economies as an alternative to a stalemate in the GATT. These would probably provoke defensive preferential arrangements by the EC and other major trading blocs, and they could lead to the disintegration of the multilateral system (see Fritsch 1989; Luyten 1988). Levels of tariff protection in developing countries are generally very high (see Laird and Yeats 1987, table 13-2),but consideration of these costs is deferred to the section on quantitative restrictions and administrative controls below, because these play a much more important role than tariffs. Agricultural Protectionism In the current negotiations, the discussion of agricultural protectionism centers on industrial-country policies that disrupt trade in temperate agricultural goods. Developing-country exports of agricultural raw materials and tropical agricul- tural commodities which do not compete with the output of developed countries are relatively unaffected by such ieasures and this fall outside the area of agriculture in the GATT. The highly protectionist agricultural policies of the developed countries strangle efficient agriculture not only by providing closed markets for inefficient producers-mainly through variable levies and quanti- tative restrictions-but also through export subsidies required to dispose of surplus production. These policies depress world prices of agricultural products significantly. Their costs in the large countries belonging to the Organisation for Economic Co-operation and Development (OECD)in 1984-86 averaged $216 billion yearly; the United States and the EC spent about $80 billion each and Japan $50 billion. Consumers mostly bear these costs in the EC and Japan; in the United States it is mainly the taxpayer who pays the bill (see OECD estimates quoted in Kelly and others 1988, p. 140). Nominal rates of protection in developed countries are high, especially in Europe and Japan. Weighted averages for consumer prices yield nominal pro- tection coefficients of 1.56 for the EC, 1.81 for other European countries, 2.08 for Japan, and 1.17 for the United States (see World Bank 1986, pp. 112-113). Recent estimates of producer subsidy equivalents (PSES) (which try to encompass a wide spectrum of distorting measures to calculate the subsidy required to maintain constant farmers' incomes) were 14.5 percent for Australia, 68.9 per- cent for Canada, 40.1 percent for the EC,and 28.3 percent for the United States (see Kelly and others 1988, p. 141; on PSES and variants see Josling and Tan- germann 1988). Inefficient agriculture is endemic in the EC, Japan, and many small European economies, but the United States is also far from blameless in its protection of inefficient production of rice, sugar, wool, cotton, certain processed meats, and dairy products. And the United States has a long-standing bad record of market disrupting activities, recently worsened by the introduction of the Export En- hancement Program in answer to competitive pressures from EC agricultural exports.' By contrast, economic policy in developing countries tends to have an anti- agricultural bias. The distortions come from a variety of policies: artificially low prices paid by marketing boards, taxation of exports, inefficient domestic pro- duction of inputs, and overvalued exchange rates that reduce the cost of com- petitive imports. This bias is the rule in many small developing economies, and is reflected in the PSES of Argentina (50.1 percent) and Nigeria (44 percent) (World Bank 1986). Some of the more advanced developing countries, for in- stance Brazil (PSE 4.2 percent), have adopted more balanced policies, or even policies biased, like those of developed countries, in the opposite direction (for instance, Indonesia, Korea, and Mexico with PSES of 38.3,58.5, and 39.5 percent respectively). Agricultural protectionism in developed countries has serious consequences for prices, trade volume, and welfare, as does the very different intervention of the developing countries. (Estimates of these impacts are known to be very sensitive to model specifications, but the general picture of what liberalization would bring is nevertheless clear. See Valdis 1987, p. 575). The impact on prices and trade volumes of a hypothetical end to intervention in certain commodities (table 1) illustrates the point. The figures on how the trade benefits and losses from agricultural liberali- zation in the developed countries will be distributed among developing countries 1. Warley's remark (Warley 1976, p.322) remains valid: "America's enthusiasm for a Liberal trade regime for farm products is not only a late conversion but is also highly selective. It focuses on those commodities in which the United States is an exporter." 30 Developing Countries and the Uruguay Round Table 1. The Effects of Liberalization of Selected Commodities on International Prices and Trade Volume in Specific Countries or Groups of Countries, 1985 (percent) Coarse Beef and Impact Wheat grains Rice veal Dairy Sugar Price change EC 1 3 1 10 12 3 Japan o o 4 4 12 3 United States 1 3 0 0 5 1 OECD 2 1 5 16 27 5 Developing countries 7 3 12 0 36 3 All 9 4 8 16 67 8 Trade-volume change EC 0 4 0 107 34 5 Japan 0 3 30 57 28 1 United States 0 14 2 14 50 3 OECD 1 19 32 195 95 2 Developing countries 7 12 75 68 330 60 All 6 30 97 235 190 60 -- Note: This includes the effectof an end of intervention in agricultural markets and not only trade intervention. Source: World Bank (1986, p. 129). are fragile and not necessarily compatible with the best aggregate estimates. If developed countries liberalized all trade measures affecting agriculture, agricul- tural exports by developing countries of beef, wheat, sugar, and maize would increase by 533 percent, 146 percent, 103 percent, and 52 percent, respectively (Zietzand ValdQ 1986, p. 43).Estimates for a 50 percent reduction by developed countries of trade barriers on temperate agricultural products other than those mentioned above suggest that the impact is not very significant, except for wine (see ValdCs and Zietz 1980, p. 34). Table 2. Efficiency Gains of Different Economic Blocs from Different Agricultural Liberalizations of Selected Commodities, 1985 (billions o f dollars) Liberalization in: Industrial Developing All Efficiencynains in countries countries countries Developing countries 11.8 28.2 18.3 Industrial countries 48.5 10.2 45.9 East European nonmarket economies 11.1 13.1 23.1 All 25.6 4.9 41.1 Source:World Bank (1986,p. 131). But the gains are not evenly distributed. A crucial finding of recent research is that liberalization of trade in agriculture, if restricted to developed countries, would hurt the developing countries as a whole (table 2). The winners when liberalization is restricted to developed countries are a few of the large developing countries, such as Argentina and Brazil; the main losers are Korea, Sub-Saharan Africa, and some countries of the Middle East. It is liberalization in developing countries themselves that improves their welfare as a group. - - How a country-developed or developing-stands on the issue of protection for temperate products will vary according to its efficiency, income per capita, whether it is a net importer or a net exporter, and the importance of such goods in its total exports. Australia and New Zealand-fficient developed agricultural - producers-are hurt by the rise of protection. The United States is a mixed case, inefficient in some agricultural activities and competitive in others. Developed economies that are inefficient producers can be classified into two types: those that protect domestic output and disrupt world agricultural markets by heavily subsidizing exports (for instance, the EC and, for rice, Japan) and those that do not export their inefficient output (typically, EFTA members). Developing countries too can be roughly divided into two groups: net ex- porters (of varying degrees of efficiency and dependence on agricultural exports, ranging from such efficient and dependent countries as Argentina to such less efficient and lessdependent ones as Brazil)and net importers, which will continue to enjoy low import prices if agricultural protection remains unassailed. The Cairns group of "free trading"developing and developed countries formed in August 1986 is one manifestation of this fragmentation of interests. The members-Argentina,Australia, Brazil, Canada, Chile, Colombia, Fiji, Hungary, Indonesia, Malaysia, New Zealand, Philippines, Thailand, and Uruguay-are mainly net exporters of temperate agricultural products that have consistently pressed to dismantle agricultural protectionism. Textiles and Clothing The present Multifibre Arrangement (MFA IV), to run until 1991, has a long history. Since 1959, successivearrangements in the GATT have de facto legitimized textile and clothing protection, allowing industrial countries to impose quan- titative restrictions on an increasing range of such exports from developing countries. The justification was that these exports were damaging output and employment in the developed countries. In fact, capital deepening, made possible by economic rents generated by import restraints and investment subsidies, is more to blame for contraction of these industries (seeSilberston 1984, chap. 7). - . Even inhibited by the MFA, the share of exports of textiles and clothing- from developing countries in the relevant world markets has increased. The share of textile exports has increased less than that of clothing because the competitive position of developed countries in the more capital-intensive textile industry is much stronger. Developing countries (including China) in 1985 supplied $29.5 billion in textiles and clothing to the developed countries, that is, roughly 40 32 Developing Countries and the Uruguay Round percent of the latter's imports (grossof intra-ECand intra-EFTAtrade). MFA rules have been circumvented by upgrading exports or "quota hopping '-investing 7 in nonquota developing countries. Developing countries imported about $30 billion in 1985, of which 40 percent was from developed countries. The high tariffs or quantitative restrictions generally imposed on these imports are based on claims of balance of payments difficulties. (See Cable 1987a, especially p. 620; 631-32 on imports into developed countries; and Kelly and others 1988, pp. 74-75 for protection in developing countries.) The protectionist lobby in the importing countries is helped by the fact that some developing countries are often lukewarm about a return to competition. In countries where quotas are distributed according to past performance, or where export licenses are auctioned, exporters or governments reap the rents generated by artificial scarcity. This freezing of potential comparative advantage means that some developing countries actually oppose a return to competition, fearing reductions in market shares that have been sustained by the inertial rules of the MFA. Considerable empirical work on the impact of the MFA, mainly based on partial equilibrium analysis, suggests substantial benefits from liberalizing trade in tex- tiles and clothing. Kirmani, Molajoni, and Mayer (1984) estimated that, if all trade barriers were removed, developing-country exports would expand by 82 percent (textiles) and 93 percent (clothing)-resultsroughly in line with those obtained by UNCTAD. About half the trade expansion generated by removing restrictions in developed countries would be in textile and clothing products (Laird and Yeats 1986, p. 29).Consumer costs of protection in the United States alone, including losses in consumer surpluses and higher prices for imports as well as domestic output, were estimated at nearly $20 billion, and net welfare costs at about $8.1 billion in 1985 (see Cline 1987, p. 191; for other estimates, see World Bank 1987, p. 151). Traditionally, analysts have suggested, on the basis of comparative production costs, that dismantling MFA restrictions would hurt relatively high-wage, middle- income countries without well-developed textile and clothing industries. Pro- duction would become concentrated in countries with low labor costs, locational advantages (the Mediterranean and Eastern Europe for the EC;the Caribbean and Central America for the United States), or with high technology comple- mented by relatively low labor costs and flexibility in fashion updating (Hong Kong) (see Cable 1986, pp. 29-30). But the empirical basis for these arguments is tenuous: experiments with a free market for textile and clothing products have been few. For example, Norway did not participate in MFA I1 and MFA 111, and textile and clothing imports were regulated only by global (not country- specific) arrangements: it is interesting that Hong Kong maintained its market share in Norway between 1978 and 1982 (around 7-8 percent) and that the only other developing country to have a market share above 1 percent during this period was China. More recent "general equilibrium" work suggests that such views should be Abreu 33 dramatically revised: developing economies as a group would gain from an abolition of MFA (see Trela and Whalley 1988, especially tables 5 and 6). For several economies, welfare gains would exceed $1 billion: Brazil ($1.03 billion), China ($2.34 billion), Korea ($2.09 billion), and Taiwan ($1.4 billion). India's and Sri Lanka's gains would be surprisingly insignificant, and if liberalization were restricted to MFA quotas, Hong Kong would suffer substantial welfare losses, and Macao, Pakistan, Singapore, and Thailand very minor ones. The implication is that countries that fear absolute export contraction and displacement from the dismantling of the MFA would in fact experience welfare gains. So textile and clothing protectionism tends to depend for survival on domestic rent reapers in developed countries and not on the fears of high-cost textile and clothing exporters among the developing countries. Speedy dismantling of the MFA is politically unrealistic. But liberalization could begin with globalization of quotas across countries or products or (through different methods) a gradual (but scheduled) liberalization of srnall suppliers first (see Raffaelli 1989). GATT Article XVIII: Balance of Payments Difficulties and Infant Industry Protection Article XVIII(b) of the GATT allows developing countries to impose quanti- tative import restrictions if they face balance of payments difficulties. The use of this provision to block imports has been facilitated by rather perfunctory GATT surveillance of whether such measures were indeed warranted by balance of payments difficulties or were disguising a virtually permanent absolute pro- tection of inefficient sectors (see Anjaria 1987, sections I and 11).So easy has it been for developing countries to use article XVIII(b) that in recent years they have rarely had to resort to actions under article XVIII(c) (protecting infant industries). Under this umbrella, protectionism in developing countries has flourished. Nontariff measures affect 40 percent of tariff lines in developing countries (see Kelly and others 1988),in comparison with 22.6 percent for nonoil imports in developed countries. The literature tends to present the costs of protection in developing countries in terms of forgone economic growth. Such evaluations are flawed by the difficulty of disentangling the costs of protection from the costs of other economic policies, and of comparing different national experiences in different historical moments. Another method of evaluating costs is the meas- urement of effective protection rates which underline the distortions in existence in developing countries (see World Bank 1987, pp. 88-89). The results from both methods of estimation show clearly that many developing countries pro- tected their domestic industry well beyond the time needed to make infant industries competitive. The well-established GATT principle of special and differential treatment (s&D) has been much criticized recently. (For a guarded condemnation, see the Leu- twiler Report, GATT 1985, pp. 44-45. For more radical critical views, see Wolf 34 Developing Countries and the Uruguay Round 1987.) The critics argue that the S&D principle does developing countries a disservice first in allowing them to avoid making reciprocal concessions to de- veloped countries and thus not participate in the GATT system; and second in enabling them to block imports on balance of payments or infant industry grounds. The argument is based on the contention that liberalization, even if unilateral, is better than protection. Its proponents suggest that the advantages for devel- oping countries of removing the S&Dprinciple would not come from reciprocal concessions by industrial countries as their "influence ... is transparently neg- ligible": they should strive for "a fuller and more equal participation" in the GATT, but the main advantage would be to make it easier to liberalize at home (see Wolf 1987, pp. 661-65). "Modernization" of article XVIII has been urged, to take into account the "new role"of fluctuating exchange rates since the early 1970s. If this reasoning were accepted with no qualification, it would mean erasing article XVIIl(b),and if it were applied to other GATT articles in order to make them compatible with economic theory very little of the present charter would be left standing. Radical reform of the charter was not on the Punta del Este agenda, and it is not in the cards in the foreseeable future. Market Access The access of exports of developing-country manufactures to industrial coun- tries' markets is hindered by many barriers: antidumping duties (ADS), subsidy countervailing duties (CVDS), safeguards, and indeterminate measures such as voluntary export restraints (VERS). The increasing use of ADS and CVDS against developing-country exports since the early 1980s is well documented (see Finger and NoguCs 1987; Nam 1987; Laird and NoguCs 1988). There is wide agreement that ADS and cvs are used in place of safeguard measures, that their harassment content is important, and that at least in the United States preliminary determinations may be biased toward affirmative findings. The concept of constructed price is open to criticism (see Kelly and others 1988, pp. 10-11). Filing an unfair trade petition is com- monly a first step in a process which leads to a U.S. demand that a VER be imposed. The economics of the legislation is faulty in concentrating on injury to domestic producers rather than the advantage to consumers of cheaper im- ports. In practice, it favors pricing policies based on full cost, and thus it fosters rather than prevents predatory pricing policies. And the argument that the leg- islation prevents pricing policies that might in the future exploit consumers in industrial countries is unconvincing in a world with a plurality of prospective suppliers (see Finger 1987, p. 156ff.) Article XIX of the GATT states the rules for emergency action on imports of products that are injuring or threatening to injure domestic producers. Restric- tions that apply under the most favored nation principle (MFN) are allowed, but affected suppliers should be compensated. These legal safeguards have rarely been applied. Instead, arrangements such as voluntary export restrictions are used, which formally preserve GATT legality by apparently being voluntary rather than being initiated by the "injured party." There are a great many VERS: 95 in September 1986, of which 30 affected developing-country exports. The percentage of imports of developed countries from all sources affected by VERS increased between 1981 and 1986 from 6.6 percent to 45.2 percent for iron and steel products, and it remained more or less unchanged at about 9 percent for other manufactures. The cost of VERS to consumers is well documented. Less is known about the net costs to exporters because contraction of import volume is offset by rents. For clothing alone, the rents of VERS were as high as 1.4 percent of Hong Kong's gross domestic product (GDP) in 1981-83 (see Sampson 1988b, pp. 13940; World Bank 1987, pp. 149-50). ADS, CVDS, and VERS affect the more industrialized developing economies: Brazil, Hong Kong, Korea, Mexico, and Taiwan. The least developed countries are much less interested in such issues, because their main constraint is supply response rather than market availability. These arrangements create vested interests-protected inefficient domestic producers obviously, but also export quota holders who prefer a stable market unrelated to dynamic comparative advantage and enjoy the economic rents generated by the restrictions. In fact, the developing countries may be better off with VERS than without them: restricting countries may, for instance, be willing to pay enough for a VER to make exporters better off than in the pre-VERsituation, because the alternative article XIX safeguard would have to be applied to all suppliers (see Hindley 1987, pp. 698-99). The industrial countries have consistently made their return to GATT discipline away from "gray area" protective devices conditional on "selectivity." Selectiv- ity-authorization to apply safeguards to specific suppliers-would blatantly undermine the GATT ; the only alternative suggested (loosening the disciplines of article XIX, including abolishing compensation) would entail the loss of rents by exporters and lessen market access (see Hindley 1987). A revival of the Uruguay-Brazil Plan of the 1960s may be an effective re- placement for the cumbersome retaliation provision of article XIX if VERS are to be discontinued. The proposal established the principle of financial payments by developed to developing countries for violations of the General Agreement. While the number and incidence of trade restrictions will stay much the same whether the loss is paid for by the party causing it or the party suffering it, exporters will benefit from the former approach (see Dam 1970, pp. 268-70). T h e N e w Themes: TRIPS, TRIMS, and Trade in Services Of the three new themes, TRIPS (trade-related intellectual property rights) have vied with services for first place as the principal bone of contention between industrial and developing countries. Developing-country resistance to discussion of new themes has mostly been a reaction to U.S. pressure to bring trade and 36 Developing Countries and the Uruguay Round foreign investment rules applicable to services into the GATT, but since the stale- mate on services was broken, TRIPS have been leading the field by a narrow margin. TRIMS (trade-related investment measures) now seem the least likely candidate for a sustained confrontation, partly because the political and technical costs of enlarging the GATT to deal with them seem prohibitive and partly because some of the TRIM issues would in any case be covered by negotiations on trade in services. This discussion of the new themes will therefore be confined to TRIPS and trade in services.' TRIPS. Intellectual property rights are crucial to developed countries for the strategic reasons already mentioned. The United States has increasingly used the issue to justify unilateral pressure, especially on the more advanced developing countries, to obtain (preferential)changes in their property rights legislation. (A recent example is the imposition by the United States of trade-restrictive measures on Brazilian products following an investigation under Section 301 of the U.S. Trade Act concerning alleged infringement of U.S. pharmaceutical patents.) Developed countries are increasingly dissatisfied with the shortcomings of the regime for regulating intellectual property, both in its coverage and enforcement in general, and in its inadequate protection of patent and copyright, particularly in new fields like biotechnology, semiconductor chips, and software (see Benko 1988, p. 221ff.) At stake is whether regulation of intellectual property rights will be transferred to the GATT from such organizations as the World Intellectual Property Orga- nization (WIPO) or UNESCO. Developed countries see the Punta del Este nego- tiations mandate as justifying the transfer; developing countries have stressed that the trade-related aspects are limited. Beyond their political opposition to a transfer they consider against their interests, the developing countries feel that property rights legislation overprotects monopoly rights at the expense of issues vital to themselves, such as access to technology and limitation of exports. The interests of the several groups of developing countries differ on the TRIPS issue. Some, like Hong Kong, favor policies similar to those advocated by de- veloped countries. Among those more disposed to negotiate some, such as Ar- gentina and Colombia, are mainly interested in the issue as a pawn in other negotiations; for others (Korea and Mexico) the issue is important in itself. Brazil, Egypt, and India are particularly reluctant to see the GATT setting and enforcing rules in this field. As with services, discussion of the issue is hampered by the lack of reliable estimates of the economic impacts: the principal U.S. document on the issue (U.S. ITC 1988) reports total "losses" of $23.8 billion-an unchecked figure reported by U.S. firms using unknown procedures. Lack of credible evidence tends to concentrate the negotiations on principles and frustrates any progress towards consensus. 2. For a detailed treatment of the TRIMS issues at stake in the Uruguay Round, see Commonwealth Secretariat (1988). It has been suggested that developing countries, rather than resisting the pro- posed intellectual property agenda and trying to maintain the present position as one of equilibrium, should take into account that, if there is no GAIT agreement on TRIPS, the present position will deteriorate, because developed countries will become much more aggressive in their rule-enforcing bilateral efforts. A closely related argument is that a new arrangement will be reached irrespective of developing-country resistance, if need be on the fringes of GATTand based on conditional MFN rules. Such initiatives must menace progress in other negotiating groups and, more generally, further threaten the major GATT principles. Break- through in this difficult area probably depends on cross-issue negotiations, since developing countries stand to gain little otherwise. Services. The U.S. emphasis on liberalizing trade in services arises from sig- nificant structural changes taking place in industrial countries in the producer services-telecommunications; engineering; financial and legal consultancy; in- surance, banking, and other financial services; advertising; distribution; and data processing. Advances in communication and information technology have had a profound impact on the competitiveness and foreign expansion of firms that provide such services. U.S. firms want to expand and compete abroad, and they can only do so by being near the customer (U.S. Congress 1986, p. 43). But most countries restrict the foreign provision of services. The misgivings of the developing countries-particularly those G-10 countries such as Brazil, Egypt, India, and Yugoslavia that have a substantive and im- mediate interest in the issue, both as importers and exporters-are rooted in two distinct sets of arguments. The first set concerns the backlog of unfinished business on trade in goods. Developing countries argue that this bacltlog should be tackled before proceeding to the services negotiation, so as to avoid cross bargains which are bound to weaken their bargaining position on the traditional themes. And the G-10 countries point out that discussion of services (apart from the strictly legal point that services are clearly outside the scope of the General Agreement), is bound to raise questions about right of establishment, national treatment, and other complex and politically sensitive issues. In fact, the intro- duction of the discussion on services in the GATT was seen as a blatantly one- sided approach to issues relating to foreign direct investment crucial to devel- oping countries, such as right of access to technology in the developed countries and a code on restrictive business practices by transnational corporations. The contradictory U.S. stance on these themes in the United Nations, where the United States has effectively blocked discussion of a code of conduct for trans- national corporations, has also been noted (see Maciel 1986, p. 90). Last but not least, the agenda initially proposed by developed countries concentrated unduly on capital-related services and excluded labor-intensive services that are of much more interest for developing countries. The second set of arguments put forward by G-10 countries (seeBatista 1987, p. 1) is that too little is known about transactions in international services to predict the implications of trade liberalization. Trade and industrial policies 38 Developing Countries and the Uruguay Round toward the rapidly changing producer-services sectors are clearly crucial for economic development. First, as intermediate inputs, the provision of these serv- ices at internationally competitive prices is important to maintain efficiency and export growth. Second, these new activities have important backward linkages with the production of hardware and technological capability in the domestic industrial sector. Assessing the benefits a particular country might gain from liberalizing trans- actions in services is hampered by conceptual problems and the paucity of data on the structure of protection and the prevalence of nonprice restraints. The developing countries' stand against trade liberalization in this area has been built on the assumption that static gains will be unevenly distributed, since compar- ative advantage is concentrated in a few developed countries and developing countries would be thwarted from realizing their comparative advantage (see Nayyar 1986). Understanding of what is at stake has advanced in the last few years: the stand of countries such as Brazil and India is no longer seen as mere filibuster. The opposing views have acted as a powerful stimulant to clearer thinking on how to advance negotiations. But empirical work on the advantages of liber- alization has not kept pace with these advances. Estimates of the costs of pro- tection are almost as fragmentary and incomplete as they were when the United States started to press for inclusion of the issue in the agenda of the new round. Sectoral lobbies in the developed countries, especially the United States, re- sponding positively to the initiative of a handful of more active developing countries, have begun to lay the groundwork for the developed countries to develop a more balanced proposal, in line with the Punta del Este decision that the multilateral rules on trade in services should promote economic growth for all and contribute to the growth of developing countries (see Richardson 1988, p. 9). Signs of receptiveness in developed countries to proposals freeing the flow of labor services, and proposals mentioning the need to assure an adequate flow of technologies, suggest that there are grounds for developing countries to begin to believe that there is something to negotiate. Developing countries may, as many have noted, pay a high price for abstaining from negotiating (see Bhagwati 1987b, p. 565ff.). As merely obstructive nego- tiating tactics began to lose momentum, through repeated use or flagging support in the capitals, and the uncompromising stand of developed countries began to thaw, the idea that developing countries should assume a position of demandeurs gained strength (see Sampson 1988a, p. 108). The demands in question relate to specific sectors, such as the possibility of technological absorption through joint ventures as well as improved market access. The advantages of improved availability of services for competitiveness in the supply of goods could be another basis for negotiation. There is scope for cooperation in establishing new rules that would fulfill the Punta del Este mandate in its entirety. (See, for instance, for proposed principles of behavior by producers, appropriate regu- lation, and development compatibility, Richardson 1988, pp. 8-10.) Abreu 39 IV. COALITIONS: OLD AND NEW Together, developing countries constitute a more important market than the United States: if united, they would obviously be a force to reckon with in the GATT negotiations. But though coalition formation by developing countries has a long history in other multilateral agencies, coalitions have been less common in the GATT , where informal consensus rather than United Nations-style divided vote is the usual procedure for reaching a decision. And a coalition encompassing all developing countries would be harder to achieve than in the past, when the interests of developing countries were much more homogeneous than they are today. In fact, the only defined coalition of exclusively developing countries to emerge in the 1980s has been the G-10 Group, whose objective was to block the inclusion of services in the new Round's agenda. An active coalition since Punta del Este, as noted above, has been the Cairns Group of countries against agricultural protectionism, an issue-based group of both industrial and developing c~untries.~But hopes that other issue-based co- alitions would follow this example have proved unfounded. The so-called Hotel de la Paix group, whose membership roughly coincides with the group sup- porting the Swiss-Colombian draft in 1986, is by no means based on issues, and while joint proposals have been presented in certain GATT negotiating groups (those on safeguards and natural resources), these initiatives do not seem to presage more formal coalition formation. (SeeHamilton and Whalley 1988, pp. 36-37.) The prospects for developing-country coalitions based on concrete economic aims can be gauged by examining their convergent interests (for a previous attempt, see Kahler and Ode11 1988). Developing countries are demandeurs in four major GATT fields: textiles, tropical products, agriculture, and market access; they may become demandeurs in services but are unlikely to do so for the other new themes. As demandeurs in services, the core G-10 countries probably have reasons to revive their coalition. Textile liberalization is the only issue that, according to new evidence, would interest all developing countries (though not with the same intensity). Major economies to benefit include Brazil, China, Indonesia, Korea, Singapore, and Taiwan, but surprisingly not India. Unfortunately, substantive discussion of this vital issue has been delayed by the renewal of MFA to 1991. Developing countries are divided on both tropical products and agriculture. Countries that are members of preferential trading areas are less interested in liberalizing tropical products than nonmembers, since liberalization would erode their preferences. Interests diverge even more over agriculture. Food importers such as Korea, Sub-Saharan Africa, and some countries of the Middle East would lose from liberalization. Their trade losses are not very significant if compared 3. Differences of views within the Cairns Group should not be underestimated, especially in connection with S&D. 40 Developing Countries and the Uruguay Round with gains by major suppliers, but net welfare losses in connection with grains are substantial. Liberalizing trade in agriculture would benefit a few large de- veloping countries such as Argentina and Brazil. Improving market access in developed countries for imports of manufactures from developing countries interests mainly the Asian newly industrialized econ- omies and a few Latin American countries (for instance, Brazil and Mexico). But even here interests are not necessarily entirely convergent, since the products affected tend to differ, the Asian exports being concentrated in more techno- logically sophisticated goods. A GATT-related issue of interest for a large group of developing countries is the foreign debt constraint. Trade-debt links in the current negotiations are now restricted to the awkward issue of article XVIII(b)and tangentially to the mon- itoring of commercial policies being discussed in relation to the functioning of the GATT system. Highly indebted countries, especially in Latin America and Africa, would like to see their foreign debt servicing eased by debtor-country concessions over market access, but such developments are unlikely. Such fragmentation of interests makes a strong coalition of developing coun- tries unlikely, unless the more advanced developing countries decide that the advantages of such a coalition are worth the costs of making some concessions. The more diversified the interests of a country, the more active it is likely to be in searching for such a coalition. Developing-country commitment to trade policies that enhance market effi- ciency is growing, partly as a result of conviction, partly in response to conditions imposed by multilateral agencies. Developed countries emphasize liberalization of obstacles to the flow of services and foreign investment, rather than to trade in goods. Only for agricultural goods is there a major trading country-the United States-with a special interest in liberalizing trade, and even here the U.S. initiative is likely to be impeded by the protectionist interests of the EC, Japan, and the smaller European economies. Besides showing a patent disinclination to tackle the backlog of unresolved GATT issues, developed countries, especially the United States, have been shifting their policy in a direction that short-circuits the multilateral trade system through a net of bilaterally negotiated preferential arrangements. Conversely, the idea of a "level playing field" for all GATT members raises the specter of full reci- procity-as opposed to what has been called first-difference reciprocity (see Bhagwati 1987b, p. 564)-with S&D as a main target, and it threatens devel- oping countries' claims that, since so much of the protection backlog consists of de facto disrespect of GATT law by developed countries, it should be rolled back at no cost in terms of new concessions by developing countries. The present multilateral system is a direct consequence of U.S. trade policies since 1934, and U.S. emphasis on the most favored nation clause. The system is far from perfect, but has on balance permitted considerable reduction of trade barriers and fast growth of trade. A host of illegal, barely legal, and legal exceptions to the rules have been allowed from the start. At present, the mul- tilateral system is under serious threat from the U.S. Omnibus Trade Bill, with its mercantilist emphasis on the need to redress unbalanced trade through bi- lateral trade instruments that do not conform with GATT rules. Unless the U.S. government modifies this legislation, the multilateral system is in grave danger. The interest of developing countries is best served by strengthening the GATT, not undermining it. Developing-country commitment to GATT'slegal framework is not, as is sometimes claimed, lip service. It is in line with their fragile bargaining position with their major industrialized trade partners. Uneven distribution of gains and losses among different developing countries creates vested interests against negotiating liberalizing policies in the GATT. Such difficulties can only be surmounted if all parties gain something in the process. Developing countries are demandeurs in tropical products, and likely to obtain concessions. Substantial advance in reducing agricultural protectionism is es- sential for advance in negotiations as a whole. The losses suffered by the small developing countries will have to be considered and compensated either directly or indirectly. Textiles and clothing are too important to be left out of the negotiations. The developed countries are in a very weak position to ask de- veloping countries to liberalize tariffs if they are not prepared to reciprocate with a long-term commitment to discontinue the MFA and reduce the relevant tariffs. In a constructive negotiation, developed countries would need to concede something on article XVIII over the market access issue--certainly on disciplines concerning parts (b) and (c) and possibly a time restriction on the use of quan- titative restrictions and a legalization of the use of nondiscriminatory tariff surcharges. Article XIX is perhaps the most intriguing pending issue in the GATT, since the avoidance of safeguards has consolidated a low-level equilibrium and no party feels strongly enough to press for the reform of the rules. It is not altogether clear how to evaluate changes in GATT institutional matters. In principle, improvement in enforcement and dispute settlement should assure balanced application of such new provisions and consequently the support of - - those contracting parties more interested in strengthening the GATT. In TRIPS and TRIMS, the developed countries are demandeurs, and it is difficult to see how developing countries could be lured from their defensive position since they do not stand to gain from rule setting and enforcement. ~ L c will h depend on how much developed countries offerconcessions in other negotiating groups. Services, however, seem to leave scope for an exchange of concessions involving detailed negotiations on a sector by sector basis. A trade liberalization in developed countries in 1983 would have increased their imports by about 12 percent (roughly $30 billion) (Laird and Yeats 1986, p. 29). It is easy to imagine fluctuations of exchange rates, interest rates, and the level of economic activity in the developed countries having a similar impact on the exports of developing countries. The drawbacks of restricting negotiations 42 Developing Countries and the Uruguay Round to trade topics, to the exclusion of related issues (such as the debt problem) that are crucial to many developing countries, need to be considered. The matter of linking trade and debt questions is certainly vexing, In 1985 the authoritative Leutwiler Report (GATT 1985, p. 49) stated that "the health and even the maintenance of the trading system ... are linked to a satisfactory resolution of the world debt problem . ." In the early days of the debt crisis it . was naively thought that the debtors' leverage in obtaining access to the creditor countries' markets would increase. In the event, commercial banks not only refrained from lobbying to improve market access for debtor countries' exports, they even turned initial ideas about the trade-debt link upside down by backing U.S. insistence on obtaining rights to establish service industries in developing countries. Another trade-debt complication is the apparent contradiction be- tween GATT'straditional reciprocal basis of negotiation, and unilateral liberal- ization arising from conditions imposed on borrowers by multilateral lending agencies. These tariff reductions even if not bound are unlikely to be taken into account as concessions in the future. Export performance in some of these indebted economies since the beginning of the decade has been at least as good as those of the Asian newly industrializing economies, but their GDP per capita stagnated. Views on liberalization tend to differ over timing and sectoral distribution rather than its inherent validity. Trade liberalization by highly indebted devel- oping countries without a corresponding liberalization by developed countries requires bigger devaluations than a concerted move by both. (Sachs 1987 has cogently advanced the prior claims of fiscal equilibrium and price stability over trade liberalization, stressing the impact of devaluation on the public deficit and on the level of inflation.) Or the reduction in trade surpluses could be compen- sated by much larger transitory financial support for liberalization reform than is envisaged at present. (Anjaria 1987 finds trade liberalization a worthy justi- fication for conceding fresh foreign finance, but he believes that this role is already played by the IMF.) The debt question is paramount for many GATT members. For them to participate meaningfully in the Round, the present un- stable debt position needs to be settled in such a way as to segregate old and new debt and start the process of restoring normalcy to world financial markets. Fragmentation of the GATT and the multilateral trading system based on the MFN principle would not be in the interest of developing countries. The weakest have most to fear from the abandonment of rule. To strengthen the GATT, developing countries need to launch more positive negotiating programs, and more often adopt the position of demandeurs. The need to liberalize and restruc- ture is by no means restricted to developing countries. There isscope for mutually beneficial negotiation. Abreu, Marcelo de Paiva, and W. Fritsch. 1987. "Brazil, Latin America and the Car- ibbean." In J. Whalley, ed., Dealing with the North: Developing Countries and the Global Trading System. London, Canada: University of Western Ontario. 1988. "New Themes and Agriculture in the New Round: A View from the South." Texto para Discussao 188. Rio de Janeiro, Brazil: Department of Economics, Catholic University. Anjaria, S. J. 1987. "Balance of Payments and Related Issues in the Uruguay Round of Trade Negotiations." World Bank Economic Review 1, no. 4 (September): 669-88. Batista, P. N. 1987. "Trade in Services: Brazilian View of the Negotiating Process." Statement Made at the General Debate in the Group of Negotiations on Services by Head of the Brazilian Delegation, Geneva, GATT. Processed. Benko, R. P. 1988. "Intellectual Property Rights and the Uruguay Round." World Econ- omy 11, no. 2: 217-32. Bhagwati, J. N. 1987a. "Economic Costs of Trade Restrictions." In J. Michael Finger and Andrzej Olechowski, eds., The Uruguay Round: A Handbook for the Multilateral Trade Negotiations. Washington, D.C.: World Bank. . 1987b. "Trade in Services and the Multilateral Trade Negotiations." World Bank Economic Review 1, no. 4 (September):549-69. Cable, V. 1986. "Textiles and Clothing in a New Trade Round." London: Common- wealth Secretariat. Processed. . 1987a."Textilesand Clothing in a New Round of Trade Negotiations." World Bank Economic Review 1, no. 4 (September): 619-46. . 1987b. "Tropical Products."InJ. MichaelFingerand AndrzejOlechowski,eds., The Uruguay Round: A Handbook for the Multilateral Trade Negotiations. Wash- ington, D.C.: World Bank. Cline, W. R. 1987. The Future of World Trade in Textiles and Apparel. Washington, D.C.: Institute of International Economics. Commonwealth Secretariat. 1987. "Tropical Products in the Uruguay Round: An Over- view." London. Processed. .1988. "Trade-Related InvestmentMeasures in the UruguayRound: Some Issues for Consideration." London. Processed. Dam, K. W. 1970. The GATT: Law and the International Economic Organization. Chi- cago: University of Chicago Press. Erzan, R., and G. Karsenty. 1987. Products Facing High Tariffs in Major Developed Market Economy Countries: An Area of Priority for Developing Countries in the Uruguay Round? UNCTAD Discussion Paper 22. Geneva. Finger,J. Michael.1987. "Antidumping and AntisubsidyMeasures."InJ. Michael Finger and Andrzej Olechowski, eds., The Uruguay Round: A Handbook for the Multilateral Trade Negotiations. Washington, D.C.: World Bank. Finger, J. Michael, and Julio NoguCs. 1987. "International Control of Subsidies and Countervailing Duties." World Bank Economic Review 1, no. 4 (September): 707- 25. Fritsch, W. 1989. "The New Minilateralism and Developing Countries." In J. J. Scott, ed., Free Trade Areas and U.S. Trade Policy. Washington, D.C.: Institute for Inter- national Economics. Gardner, R. N. 1979. Sterling-Dollar Diplomacy. New York: Oxford University Press. GATT (General Agreement on Tariffs and Trade). 1958. Trends in International Trade. A Report by a Panel of Experts. Geneva. 44 Developing Countries and the Uruguay Round .1979. The Tokyo Round of Multilateral Trade Negotiations. Geneva: GATT. .1985. Trade Policies for a Better Future [The Leutwiler Report]. Proposals for Action. Geneva. .1986. Press Release 1396, September 25. . 1988a. MultilateralTrade Negotiations. The Uruguay Round. Group of Ne- gotiations on Goods, Report to the Trade Negotiations Committee Meeting at Min- isterial Level, Montreal, December 1988. M T N . G N G Geneva. / ~ ~ . . 1988b. MultilateralTrade Negotiations. The Uruguay Round. Group of Ne- gotiations in Services, Report on the Trade Negotiations Meeting at Ministerial Level, Montreal, December 1988. MTN.GNG/~~.Geneva. Hamilton, C., and J. Whalley. 1988. Coalitions in the Uruguay Round: The Extent, Pros and Cons of Developing Country Participation. National Bureau of Economic Research Working Paper 2751. Cambridge, Mass. Hindley, B. 1987. "GATT Safeguards and Voluntary Export Restraints: What Are the Interests of Developing Countries?"World Bank Economic Review 1, no. 4 (Septem- ber): 689-705. Hudec, R. 1987. Developing Countries in the GATT Legal System. London: Gower for the Trade Policy Research Centre. Josling, T., and S. Tangermann. 1988. "Measuring Levels of Protection in Agriculture." Twentieth International Conference of Agricultural Economists, Invited Papers, vol- ume 2. August 24-31, Buenos Aires. Processed. Kahler, M., and J. Odell. 1988. "Developing-Country Coalition Building and Interna- tional Trade Negotiations." In J. Whalley, ed., Rules, Power and Credibility. London, Canada: University of Western Ontario. Karsenty, G., and S. Laird. 1986. The Generalized System of Preferences:A Quantitative Assessment of Direct Trade Effects and Policy Options. UNCTAD Discussion Papers 18. Geneva. Kelly, M., N. Kirmany, M. Xafa, C. Boonekamp, and P. Winglee. 1988. Issues and Developments in InternationalTrade Policy. International Monetary Fund Occasional Paper 63. Washington, D.C. Kirmani, N., P. Molajoni, and T. Mayer. 1984. "Effects of Increased Market Access on Selected Developing Countries' Export Earnings: An Illustrative Exercise." Interna- tional Monetary Fund M 84/54. Washington, D.C. Processed. D Laird, S., and Julio NoguCs. 1988. "Trade Policies and the Debt Crisis." World Bank International Economics Department. Washington, D.C. Processed. Laird, S., and A. Yeats. 1986. The UNCTAD Trade Policy Simulation Model: A Note on the Methodology, Data and Uses. UNCTAD Discussion Papers. Geneva. .1987. "Tariff Cutting Formulas-andComplications." InJ. Michael Finger and Andrzej Olechowski, eds., The Uruguay Round: A Handbook for the Multilateral Trade Negotiations. Washington, D.C.: World Bank. Luyten, P. 1988. "Multilateralism versus Preferential Bilateralism: A European View." In J. J. Scott, ed., Free Trade Areas and U.S. Trade Policy. Washington D.C. Institute for International Economics. Maciel, G. A. 1989. "0 Brasil e o GATT." Contexto International 3 (January/July). MacPhee, C. R. 1986. "Effect of Competitive Need Exclusions and Redesignation under the U.S. Scheme of Generalized Preferences." UNCTAD/ST/MD/~~.Geneva. Processed. MOCI (Switzerland).1989. No. 849, January 2. Nam, G. 1987. "Export-Promoting Subsidies, Countervailing Threats, and the General Agreementon Trade and Tariffs."World Bank Economic Review 1, no. 4 (September): 727-43. Nayyar, D. 1986. "International Trade in Services: Implications for Developing Coun- tries." Exim Bank Commencement Day Annual Lecture. Bombay. News of the Uruguay Round. 1988. "Text adopted by Ministers in Montreal." MTN.TNC/ ~ ( M I N )December 9. Reproduced in , News of the Uruguay Round 23 (December 14). News of the Uruguay Round. 1989. "Text adopted by Ministers in Montreal." ~ ~ ~ 1 1 1 , April 21. Reproduced in News of the Uruguay Round 27 (April 24). Raffaelli,M. 1989. "SomeConsiderations on the Multi-Fibre Arrangement: Past, Present and Future." Geneva: GATT. Processed. Richardson, J. B. 1988. "Some Thoughts on Dealing with Development in an Agreement on Trade in Services." Paper presented to the Roundtable on the Role of the Service Sector in the Development Process, Schloss Fuschl, Austria, July 1-3. Sachs, J. D. 1987. "Trade and Exchange Rate Policies in Growth-Oriented Adjustment Programs." ~ ~ F - W o rBank Symposium on Growth-Oriented Adjustment Programs. l d Washington D.C., February. Processed. Sampson, G. 1988a."Developing Countries and the Liberalization of Trade in Services." In J. Whalley, ed., Rules, Power and Credibility. London, Canada: University of Western Ontario. . 1988b. "Nontariff Barriers FacingDeveloping Country Exports." InJ. Whalley, ed., Rules, Power and Credibility. London, Canada: University of Western Ontario. Silberston, A. 1984. The MFA and the UK Economy. London: Her Majesty's Stationery Office. Trela, I., and J. Whalley. 1988. "Do Developing Countries Lose from the MFA?" De- partment of Economics, University of Western Ontario. Working Paper 8804C. Lon- don, Canada. Processed. U. S. Congress Office of Technological Assessment. 1986. Trade in Services, Exports and Foreign Revenues. Washington, D.C. U. S. International Trade Commission. 1988. Foreign Protection of Intellectual Property Rights and the Effect on U.S. Industry and Trade. Washington, D.C. UNCTAD (U.N. Conference on Trade and Development) Secretariat. 1968. The Kennedy Round Estimated Effects on Tariff Barriers. Doc. ~ ~ / 6 / r e v .New York. l . Valdts, Alberto. 1987. "Agriculture in the Uruguay Round: Interests of Developing Countries." World Bank Economic Review 1, no. 4 (September):571-93. Valdts, Alberto, and J. Zietz. 1980. Agricultural Protection in OECD Countries: Its Cost to Less-Developed Countries. International Food Policy Research Institute, Research Report 21. Washington, D.C. Warley, T. K. 1976. "Western Trade in Agricultural Problems." In A. Shonfield, ed., International Economic Relations of the Western World 1959-1971. Volume 1. Pol- itics and Trade. London: Royal Institute of International Affairsand Oxford University Press. Winham, G. R. 1986. International Trade and the Tokyo Round Negotiations. Princeton, N.J.: Princeton University Press. Wolf, Martin. 1987. "Differential and More Favorable Treatment of Developing Coun- tries and the International Trading System." World Bank Economic Review 1, no. 4 (September):647-68. 46 Developing Countries and the Uruguay Round World Bank. 1986. World Development Report 1986. New York: Oxford University Press. .1987. World Development Report 1987. New York: Oxford University Press. Yeats, A. 1987. "The Escalation of Trade Barriers." In J. Michael Finger and Andrzej Olechowski, eds., The Uruguay Round: A Handbook for the Multilateral Trade Ne- gotiations. Washington, D.C: World Bank. Zietz, J., and Alberto ValdCs.1986. The Costs of Protectionism to Developing Countries: An Analysisfor SelectedAgricultural Products. World Bank Staff Working Paper 769. Washington, D.C. P R O C E E D I N G S O F T H E W O R L D B A N K A N N U A L C O N F E R E N C E O N D E V E L O P M E N T E C O N O M I C S 1 9 8 9 Andrzej Olechowski I always find it very exciting to discuss the GATT and the Uruguay Round, and I do so whenever I am in Geneva or Washington. I rarely do it in Warsaw. The irony is that I am heading a department which in the Polish administration is responsible for GATT issues. This observation sets the tenor of my intervention. I would like in these brief comments to look at the issues cogently discussed by Professor Abreu specifically from the point of view of Poland-that is, from the point of view of a country that is medium-size, developing, heavily indebted, and undergoing a major po- litical and economic reform aimed at internal and external liberalization. Given the above characteristics, the GATTshould be very important for Poland. First, it should secure free accessfor Polish products to the major export markets. Second, through article XIX it should protect our exporters from unrestrained protective actions in the importing countries-a feature particularly important for a middle-income country, which, to a large extent, exports products and services considered "sensitive" by the importing countries (such as steel, ship- building, petrochemicals). Third, it should provide guidelines for domestic pol- icies and regulations, and impose discipline on the ways trade policy is carried out. Thus, in Poland we see the GATT the same way as its founding fathers-as a strong commitment by each participating country to keep its markets open to imports. The safeguard clause provides a way to maintain this general com- mitment in the face of unusual trade developments affecting isolated industries. In practice, the GATT is not effective in fulfilling its role. Owing to certain features of Poland's protocol of accession, the most favored nation (MEN) treat- ment in some countries-notablythe United States and the European Communitjr countries, or major trading partners-is viewed as a unilateral concession and therefore open to political maneuvering. Secondly, the GATT safeguard proce- dures do not shield our exports from import-restrictive actions in the form of "voluntary"export restraints (VERS). Poland, afterJapan, issubject to thelargest number of VERS, which cover many agricultural, textile, steel, and other industrial Andrzej Olechowski is Director, Ministry of Foreign Economic Relations, Government of Poland. The views expressed are the author's and not those of the Government of Poland. O 1990 The International Bank for Reconstruction and Development 1THE WORLD BANK. 48 Comment products. Finally, because of these restrictions and the general lack of interna- tional discipline, "GATTconsistency" is not a persuasive argument when decisions on domestic policies and regulations are made. It is often adhered to only su- perficially, while substance and practice remain in conflict with GATT principles. How, in this context, do the politicians and the general public in Poland view the Uruguay Round? I believe that the prime minister has only a vague idea of what the Round is about, not to mention the president or the public. But two issues could attract considerable attention and make a significant impact on the Polish economy: agriculture and services. Both sectors are facing radical reform in Poland in the shape of a thorough demonopolization and extensive privatization. The reform is meeting strong resistance from pressure groups who cite short-term decline in production and uncertainty about external conditions as the main grounds for their opposition. Their resistance would be much easier to overcome if there were (evententative) Uruguay Round agreements as to the principal futureconditions for international trade in these sectors. The agreements would need to be accompanied by strengthened commitment to the safeguard rules. Otherwise, guided by past experience, politicians would hesitate to risk opening domestic markets for ag- ricultural products and services to external competition, with no guarantee that other countries would do the same. These considerations are also germane to some other developing countries. Many of them, faced by the collapse of the central planning concept and attracted by the successes of the market economies, are rethinking their economic systems and development strategies. In many respects, many developing countries are now at the same stage that industrial countries were when the GATT was for- mulated. Unfortunately, the developed countries have moved on to a stage where (often very narrow) reciprocity has become the dominant issue. P R O C E E D I N G S O F T H E W O R L D B A N K A N N U A L C O N F E R E N C E O N D E V E L O P M E N T E C O N O M I C S 1 9 8 9 Gary P. Sampson As mentioned in Professor Abreu's useful and comprehensive review, the December 1988 Mid-Term Ministerial Review in Montreal did not reach con- sensus in four of the fifteen Uruguay Round negotiating groups: agriculture, textiles, intellectual property, and safeguards. All these areas are important to developing countries for different reasons. Because the Uruguay Round is a political undertaking, the process was put "on hold." Agreement in the out- standing areas was finally reached at a meeting of the Trade Negotiations Com- mittee in April 1989. Since Professor Abreu completed his paper before the April meeting, my comments supplement his paper and report on the current state of play. With respect to agriculture, at Montreal, the United States and the Cairns Group (four developed and ten developing countries) proposed the long-term elimination of restrictions on market access and other trade-distorting policies, such as subsidies. The Cairns Group also proposed that in 1989 and 1990, short- term measures should be adopted to freeze and gradually reduce farm support measures. The European Community, however, emphasized the need for short- term measures based on existing policies to reduce support for agriculture. For the United States, agreement on long-term measures to eliminate farm support was a prerequisite for any discussion of such short-term measures. As for the long term, the European Community proposed to stabilize world markets by reducing the negative effects of agricultural support measures and rebalancing external protection policies. This fell short of the U.S. proposal for long-term elimination of farm support. Not surprisingly, the agreement reached in April represents a compromise: in the long term, the objective is a "substantial progressive reduction in agricultural support." Commitments are to be negotiated for import access, subsidies and export competition, and export prohibitions and restrictions. In the short term, farm support is to be frozen at current levels of domestic and export support and protection. Gary Sampson is Director, Group of Negotiations on Services Division, ATT. The views expressed G are those of the author and not necessarily those of the organization for which he works. O 1990 The International Bank for Reconstruction and Development / THE WORLD BANK. 50 Comment Governments are to come forward with negotiating proposals by the end of 1989 on a list of topics that represents a formidable research agenda for policy- oriented agricultural economists. Topics include ways to adapt existing farm support (for instance, moving to tariffs and decoupling income support from production levels), how to take account of the possible harm the reform process might do to developing countries that are net importers of food, and the form of and use to which measures of aggregate farm support will be put. The task is daunting. The complexity of interest groups involved within and across coun- tries is staggering, and most known intervention measures are cyrently being employed in this sector. It seems fair to say that there were four breakthroughs in the negotiations on agriculture. First, for the first time in the negotiating history of the General Agreement on Tariffs and Trade (GATT) all forms of agricultural protection are now on the table. Second, there is definitely scope for the special status of agriculture to disappear over time. Third, there is agreement to freeze existing levels of protection and reduce them in the future. Fourth, and in some respects most important, governments are now engaged in a permanent state of nego- tiation. As for textiles, negotiators at Montreal faced the issue of how far governments were prepared to commit themselves to dismantling the Multifibre Arrangement (MFA) and in what period of time. In the April agreement there is, for the first time, a clear commitment to negotiate an end to the MFA and to start phasing out the network of bilateral restraint arrangements in 1991. I would tend to take issue with Abreu's assessment that textiles are unlikely to play a prominent role in the Uruguay Round because the present Multi-Fibre Arrangement is to end only in 1991. It could be argued that the expiration date of the present MFA allows countries to make a negotiated removal of the MFA part of the total negotiating package. In fact this was a consideration when the last expiration date was negotiated. The challenge is to find a mechanism that would permit the gradual undoing of the damage from three decades of bilateral restraint arrangements and the return of textile trade to an open, liberal trading system in which decisions to produce and consume respond to relative prices rather than bilaterally negotiated limits on quantitative restraints. This may seem a good topic for an undergrad- uate term paper, but the issues are complex. Many participating countries see the existing arrangement as representing some balance of perceived interests in the importing and exporting countries. To be acceptable, the proposed mech- anism should maintain this balance during the phaseout period. As Abreu notes, a principal issue in negotiations about trade-related aspects of intellectual property rights (including trade in counterfeit goods) is the insti- tutional question of whether a new set of rules involving enforcement and set- tlement of disputes will be negotiated in the GATT in spite of the previous role of organizations such as the World Intellectual Property Organization. The fundamental point is that different countries see their interests as better repre- Sarnpson 51 sented in one institution or the other. The compromise struck in April was to continue the GATT negotiations but to decide at the end of the Uruguay Round which institutions should implement the rules. I agree with Professor Abreu that in many ways negotiating a new safeguards clause (Article XIX) is "the most intriguing pending issue in GATT." Indeed, for any self-respecting economist the goal should be to establish under what cir- cumstances sudden surges of imports may be legitimately restrained for a short period without damaging the long-term interests of the economy. The procedures should avoid insulating producers from market forces that herald the need for healthy changes in patterns of production and consumption. The management of world trade in textile products provides clear evidence that such insulation only creates vested interest groups and exacerbates long-term adjustment prob- lems. As Abreu notes, issuessuch as selectivity (selectiveapplication of safeguard protection to specific countries) prevail, and are a major concern for some developing countries that see themselves as potential candidates for such selective treatment. The compromise in the April Mid-Term text is largely procedural; it was agreed that a draft text was to be prepared by the chairman of the Nego- tiating Group on Safeguards in conjunction with the GATT Secretariat and pre- sented to the Negotiating Group by June 1989. It is hard to think of new theoretical or empirical research on safeguards which would be useful. Generations of economists have argued that market disturb- ances reflecting changing patterns of comparative advantage are all part of the normal workings of the market, while market disturbances related to dumping and subsidies can be dealt with through other procedures. If there is to be government intervention on import surges, it should be designed to facilitate rather than retard the process of market-led structural change. I would like to make a few comments on the area of my own responsibility in GATT: negotiations to create a multilateral framework for trade in services in order to progressively liberalize this trade and promote the economic growth of all trading partners and the growth of developing countries. I will concentrate on three issues of importance to developing countries that emerged in the Mont- real discussions. First, there is no clear definition of what constitutes trade in services (variously described as trade in invisibles, intangibles, and so on), so the nature of trans- actions to which the multilateral framework will apply is something to be ne- gotiated. One way of defining trade in services is to draw a parallel with trade in goods-that is, something (presumably the service itself) must cross the border for trade to take place (as when some telecommunications services are traded). A definition at the other end of the spectrum would embrace those service transactions that require a foreign presence and a cross-border movement of factors of production (labor and capital) to be marketed internationally (for example, retail banking services). Some developing countries seem to have been of two minds. Opting for a narrow definition would meet some concerns. For example, it would minimize the impact of foreign firms on nascent infrastructural 52 Comment industries that for a variety of reasons some countries wish to maintain even if they are internationally uncompetitive. At the same time, only a broad definition would ensure the flow of resources and technology necessary to promote de- velopment and therefore fulfill the negotiating objectives of the Punta del Este Declaration. Not surprisingly, lines tended to be drawn according to whether individual countries preferred outward or inward development strategies. The question was not settled at Montreal, but the door was certainly opened to a broad definition of trade in services. The text provides that future work in the negotiating group will proceed on the basis of trade in services involving cross- border movement of services, of consumers (as in tourism), and of factors of production, where this is essential to sell the service abroad. A second issue important to developing countries is sectoral coverage of the multilateral framework in services. Some developing countries have been under the impression that sectors in which they possess a comparative advantage will not be covered either because of sensitivities (for example, about labor mobility in construction services) or because little can be done in such an arrangement to help expand their trade (such as tourism). In Montreal, the question of coverage was dealt with by agreeing not only that no sector would be excluded from the arrangement but also that sectors of export interest to developing countries should be specifically included in the arrangement. Third, with respect to a point raised in the paper, some developing countries have been reluctant to engage in negotiations to liberalize trade in services. I agree that "opposing views concerning trade in services negotiations acted as a powerful stimulant to clearer thinking concerning ways to advance the negoti- ations." Abreu points out some reasons for this resistance, but basically the varying degrees of enthusiasm probably reflect the fact that some countries consider the link between liberalization and development to be tenuous at best. In the view of these countries, a framework supportive of development would need provisions that take account of the difficulties developing countries face in immediately implementing full obligations under the arrangement, and that introduce concepts that would strengthen the link between liberalization and development. The Montreal text opens the way for introducing concepts that will minimize the damage of rapid liberalization in developing countries. It recognizes, for example, the need for rules and procedures for developing countries to extend market access progressively in line with their development situation. And the door is also open for provisions to increase the developing countries' partici- pation in world trade in services and expand their own exports of services by strengthening their domestic capacity in services and making the sector more efficient and competitive. The challenge facing negotiators from developing countries today is clear. What provisions can be written into the multilateral framework to ensure that developing countries become more efficient and competitive in providing services via the negotiated progressive liberalization of service activities in their countries? Sampson 53 Given the dearth of information on which to base conclusions, there is certainly scope for imaginative thinking. Finally, the diversity of developing countries' interests, which Professor Abreu stresses in a more general context, is also apparent in the services negotiations. But all developing countries seem to agree on one important point: any devel- opment provisions should be an integral part of the agreement itself and not an addendum (as with Part IV of the GATT) or a list of exemptions from obligations (such as special and differential provisions). P R O C E E D I N G S O F T H E W O R L D B A N K A N N U A L C O N F E R E N C E O N D E V E L O P M E N T E C O N O M I C S 1 9 8 9 One participant suggested that making debt an essential element of the GATT negotiations would overload already complex negotiations. Presumably the debt crisis will be resolved by another mechanism (guided by the World Bank and the International Monetary Fund) in the next two or three years, he said. The point of the Uruguay Round negotiations, on the other hand, is to establish rules governing trade in the last years of this century and the first years of the next. Abreu responded that steps taken to coordinate efforts in the Bank and the IMF were inadequate in an explosive situation. Abreu's paper maintained that tariff reductions as part of conditionalities imposed on countries borrowing from the World Bank are unlikely to be con- sidered as concessions later. To the participant who suggested that this is not true if the tariffs are bound, Abreu pointed out that in many countries tariffs cannot be bound because the governments committing themselves to liberali- zation are not credible. Another participant suggested that if nontariff barriers aren't addressed, it doesn't matter if tariffs are bound. Abreu agreed that it is useless to bind tariffs if GATT Article 18-B provides developing countries an easy way out. To discussprotectionism in the developingcountries, he said, something must be done about Article18-B. As for the thorny question of a tradeoff between Articles 18 and 19, Abreu said it is unclear what Article 19 offers or how that deadlock will be broken, because of vested interests in maintaining voluntary export restraints. Sampson (discussant)said that solutions proposed to deal with the problem of nontariff barriers include (1)retariffication to replace voluntary export re- straints and (2)temporary tariff quotas with a quantitative restriction (imports could be brought in if the penalty tariff were paid, but the penalty tariff would eventually be phased out and replaced by a tariff-based system that would eventually be bound). On the points raised by Sampson in his comments on Abreu9spaper, Abreu agreed that something is being done about textiles in the GATT-but too little and too late. He reemphasized that new evidence suggests that all developing countries would gain from a termination of the Multifibre Arrangement (MFA). A member of the audience commented that it is a big jump from the observation that everybody has something to gain from abolishing the MFA to the conclusion that there is room for a coalition. This session was chaired by Herminio Blanco, undersecretary of trade, government of Mexico. 01990 The InternationalBank for Reconstruction and Development /THE WOIRLDBANK. 56 Floor Discussion of Abreu Paper Abreu said that the evidence about textiles breaks new ground but is still controversial, and even he wonders whether Brazil would benefit from a phase- out of the MFA. He noted that this represented an opportunity for somebody to do some careful empirical work that would suggest a more traditional division of interests on textiles-for example, with China and India for, and Hong Kong, the Republic of Korea, and Taiwan against dismantling the MFA. On the point about the prospects for a timed phaseout of the MFA, one participant stated that only the threat of other kinds of measures-typically safeguard actions-propel developing countries into these voluntarily negotiated agreements. To make predictions about something concrete happening in tex- tiles, he said, we must look at what else is happening in the trading system in the safeguards area-so observing what happens to safeguards such as steel antidumping measures becomes important. Phasing the MFA out in such a way that countries feel they are getting a fair deal is crucial to the phaseout's success, said Sampson. For example, some countries extract rent in the trading transaction. Giving that up means trading it off against something else, such as expanded market access elsewhere. Those devising a plan to phase out the MFA are trying to find an objective way to estimate the value of nontariff measures as a basis for negotiating change. The ultimate objective-a bound, nondiscriminatory, most-favored-nation tariff- would be considerably less restrictive than the voluntary export restraint ar- rangements now in place. One participant commented on how much more active developing countries had been in this Round compared with the Kennedy and Tokyo Rounds, and how much they had been able to influence the agenda. Sampson agreed, saying that five developing countries-Argentina, Brazil, Chile, Colombia, and Peru- were able to put the negotiations on hold in Montreal. In the services-negotiating group Sampson thought this an important development, because a number of developed countries think that over time the developing countries will be unable to live up to the obligations that come out of the agreed text and will drop out. Sampson did not think that was the way the negotiations would develop. If they do, it would be a bitter fight, because the developing countries were wedded to the Montreal Text-and the far-reaching Montreal Text on services was not put on hold. Several members of the audience were less optimistic than Abreu and (espe- cially) Sampson about the results of the Uruguay Round. One asked what global mechanisms were being put in place to ensure that all of this was not an exercise in futility-observing that little seemed to have changed since the Tokyo and Kennedy Rounds. Another asked what would happen if the real action took lace among the United States, Japan, and the European Community (EC). In the same vein, a third suggested that actions speak louder than words and that the actions of the three major actors in the game suggested that regional trading arrangements are going to be more important than a multilateral trading ar- rangement within GATT. Floor Discussion of Abreu Paper 57 Sampson admitted that discipline of the multilateral trading system could be eroded by actions of the Super-301 type, Europe in 1992, the U.S.-Canada bilateral trade agreement, regional liberalization, and an equivalent of the Or- ganisation for Economic Co-operation and Development for the Pacific Rim countries. But he felt that many people believed these possibilities were one way to apply pressure to revive the multilateral trading system. Only time would tell. According to Sampson, in the negotiating group on services the real fight would probably not be about North-South issues but about such issues as whether liberalization in financial services will be on the basis of reciprocity or national treatment between the United States and the EC; or whether the major telecom- munication country suppliers will be able to maintain their state monopolies; or, in civil aviation, whether the extensive network of bilateral restraint ar- rangements will remain or will be replaced by some sort of multilateral agree- ment. We have signs, Sampson said, that the fight will be fought at a high political level. But he also thought that everyone, at least in the services talks, hoped to make existing agreements conform with whatever emerges successfully from the Uruguay Round. Olechowski (discussant) said that he was not particularly optimistic about the outcome of the negotiations. Furthermore, he had difficulty in reconciling what one participant said about the Uruguay Round framing rules for the next century and Sampson's comment that once the rules are agreed upon, the current policies will be made to conform to them. Responding to participants' comments about misplaced optimism, Abreu said that his paper implies that the behavior of the U.S. trade negotiators is slightly schizophrenic and that their bilateral and multilateral policies are contradictory. He exhorted the United States to lead the way toward multilateral trade liber- alization, because it is too much to ask a high-inflation indebted country to adjust to unilateral liberalization. One participant asked Abreu how all of this would help us decide the things we have to decide tomorrow, or next week, or by the end of the century? He asked Abreu to provide an analytical framework to help people at the World Bank decide which approaches on specific issues would strengthen and reform- or weaken and destroy-the GATT system. Abreu responded that it is not easy to translate the paper into immediate policy actions, but that the World Bank could help deepen understanding of many of the issues being negotiated, particularly the so-called new themes. The basic U.S. document on intellectual property is fundamentally weak, because it is merely a summary of what the industry claims about losses incurred in coun- terfeiting and the like. Abreu felt that establishing an intellectual property system on the fringes of GATT might weaken GATT, but that he was walking on thin ice with this view. Similarly, Abreu felt that a lot more had to be learned about services. In 1982, the United States had difficulty convincing people that services should have an important position on the agenda. Abreu thought GATT should include seririces, 58 Floor Discussion of Abreu Paper because no international organization was doing the job globally, although UNCTAD had been involved in some aspects of services. He believed there would probably be a big political wrangle about intellectual property rights, because the World Intellectual Property Organization would probably fight. And al- though the developed countries felt that GATT was the ideal venue for intellectual property, the developing countries did not. It is not easy to know whether extending GATT discipline to services and intellectual property will advance GATT rules. The first obligation, Abreu felt, was to restore GATT legality over trade in goods. The contracting parties had, of course, decided that inclusion of new themes should not be made conditional on solution of the backlog; but Abreu nonetheless felt that the backlog must be dealt with. Abreu suggested that the Bank should analyze the long-term effects of liber- alization in a country such as Brazil, taking into account the possibility that liberalization could backfire and strengthen protectionist lobbies. This line was not being followed in the establishment of conditionalities. One participant saw negotiations and actual liberalization in services as two very different issues. He argued that it would be extremely difficult to find meaningful criteria to selectively choose services to liberalize. If to do business globally you have to have factor mobility, national treatment, and the right to establish enterprises, and therefore services should be liberalized, then you are really changing the rules of the game and discussing something far beyond GATT. At one extreme, it could imply that the nation-state is no longer relevant in resource allocation and decisionmaking. The issue of whether nation-states should give up their sovereignty to allow national treatment and full factor mobility across all borders is much more important than GATT. What is the use of firms in developed countries having access and national treatment on services, he continued, if currencies aren't convertible, if they can't repatriate their profits, and so on. To make this workable we will have to change the conditions for doing business. Some countries are not prepared to accept the pure application of GATT rules and principles to services, Sampson responded. For example, it makes sense, and there is a very specific reason, to apply national treatment to goods (under Article 3, a concession granted at the border, such as tariff reduction, should not be negated by a government restriction once inside the border); but services don't pass through customs houses, so offering national treatment to providers of services amounts effectively to free trade-with very different implications. Sampson agreed that the problem of selective liberalization of services must be dealt with. The Montreal Text indicates that specificity of purpose, discrete- ness of transaction, and limited duration are to be considered in determining, sector by sector, what will or will not be a service transaction. Overriding all these considerations is the understanding that national policy objectives will be respected; but there is also the sense that countries should not unilaterally erect Floor Discussion of Abreu Paper 59 obstacles to trade that they consider appropriate for their own national objec- tives. Blanco (chair)concluded that more theoretical work is needed to design ways for those countries undertaking unilateral liberalization to get automatic rec- ognition and credit from GATT-S that a country coming to the negotiating O table with a 100-percent tariff would not have more negotiating power than a country which has independently reduced its tariff to 20 percent. Trade gains would be speeded up if the reaction to unilateral liberalization efforts were automatic liberalization from other countries. Political pressures from domestic interest groups could be neutralized if a country knew it would receive automatic credit and recognition for liberalization. Blanco recommended that the World Bank support theoretical work on the following: a The new issues-services and intellectual property rights. a The implications of the most-favored-nations principle for block formation, country size, and negotiating possibilities for small and large countries. a How much have developing countries gained by special and differential treat- ment? What operating methods are available to make this clause work for each negotiating group? Finally, the chair thought that GATT should be encouraged to do more theo- retical work on how to cease creating acronyms such as OGS TRIMS, and TRIPS. F , P R O C E E D I N G S O F T H E W O R L D B A N K A N N U A L C O N F E R E N C E O N D E V E L O P M E N T E C O N O M I C S 1 9 8 9 Saving in Developing Countries: Theory and Review Angus Deaton In the literature on economic development, much of the interest in saving has been focused on the relation between saving and growth. But saving is not only about accumulation. It is about smoothing consumption in the face of volatile and unpre- dictable income, and helping to ensure the living standards of poor people whose lives are difficult and uncertain. This paper develops a model of households which cannot borrow but which accumulate assets as a buffer stock to protect consumption when incomes are low. Such households dissave as often as they save, do not accumulate assets over the long term, and have on average very small asset holdings. But their consumption is markedly smoother than their income. Much of the evidence is as consistentwith this view of saving as it is inconsistentwith standard views of smoothing over the life cycle, and with explanations of the link between saving and growth in terms of life-cycle saving behavior. Consumption smoothing is also a useful way of thinking about government policy, where volatility in the world prices of taxed com- modities can generate sharp fluctutationsin government revenues as well as realloca- tions of revenue between the private and public sectors. Many important policy issues in developing countries hinge on issues of consumption and smoothing, and research on these issues is currently likely to be more productive than work on the relation between saving and growth, at leastuntil we haveamore satisfactorytheory of economic growth. I can think of four good reasons for studying saving in developing countries separately from saving behavior in developed economies. At the microeconomic level, developing-country households tend to be large and poor; they have a different demographicstructure; more of them are likely to be engaged in agriculture; and their income prospects are much more uncertain. The problem of allocating income over time thus looks rather different in the two contexts, and the same basic models have different im- plications for behavior and policy. Angus Deaton is a professor of economics and international affairs at the Woodrow Wilson School of Public and International Affairs and the Department of Economics, Princeton University. He is grateful to John Campbell, Avinash Dixit, Mark Gersovitz, Fumio Hayashi, Christina Paxson, Nicholas Stern, T. N. Srinivasan, and Steven Webb for helpful discussions and comments, and to Susan Collins, Christina Paxson, and Dwayne Benjamin for providingdata. He also thanks the Bradley and McDonnell Foundation for funding some of the research reported in this paper. 01990 The International Bank for Reconstruction and Development 1THE WORLD BANK. 62 Saving in Developing Countries e At the macroeconomic level, both developing and developed countries are concerned with saving and growth, with the possible distortion of aggregate saving, and with saving as a measure of economic performance. But few developing countries possess the sort of fiscal system that permits deliberate manipulation of personal disposable income to help stabilize output and em- ployment. e Much of the postwar literature expresses the belief that saving is too low, and that development and growth are impeded by the shortfall. Sometimes the problem is blamed on the lack of government policy, sometimes on mis- guided policy. Saving is even more difficult to measure in developing than in advanced economies, whether at the household level or as a macroeconomic aggregate. The resulting data inadequacies are pervasive and have seriously hampered progress in answering basic questions. The discussion is organized around these four topics. I focus on areas that are either not covered in the recent excellent survey by Gersovitz (1988), or where I wish to develop a different perspective. I make no attempt to be com- prehensive where I would only be repeating Gersovitz's review. Section I develops the microeconomic framework for household saving on which the rest of the discussion is based. The analysis is within the framework of the standard life-cycle permanent income model, but it emphasizes different features of behavior to generate results far from the standard "consumption equals permanent income" story that has so far dominated empirical work in developing countries. Section I1 turns to macroeconomic saving, beginning with the relations be- tween income growth, population growth, and saving rates predicted by some versions of the life-cycle hypothesis. The section next addresses macroeconomic stabilization. In many developing countries, protecting consumption against fluc- tuations in income is a public as well as a private problem. Many governments tax primary commodity exports, so that both government and external balances are sensitive to fluctuations in output levels and international commodity prices. Although the issues are perhaps well understood, there is considerable evidence that governments in many developing countries have not been able to design policies that make the appropriate adjustments. The discussion in section 111of whether saving is too low reviews the arguments based on externalities, and inquires whether government policy, by regulating interest rates, distorts patterns of saving and financial intermediations. Much of the argument here hinges on the interest elasticity of saving. The rest of this introductory section deals with the fourth topic-the data, and how their shortcomings affect the subsequent analysis. Perhaps the worst problem that besets data on saving is that saving is not measured directly but is the residual between two large magnitudes, each itself measured with error. The consequences show up at all levels of discussion. Household survey data often show an implausibly large fraction of households Deaton 63 dissaving, though the implausibility may be associated with faulty theory as well as faulty data. The household survey data are often quite inconsistent with and typically less than the national income estimates-themselves possibly biased downward (see Visaria and Pal 1980; Paxson 1989) and certainly subject to substantial uncertainty. The standard household survey may well understate saving. The concept of income is itself extraordinarily complex, and most people in developing countries have little reason to distinguish between business and personal cash transactions. A farmer who buys seeds and food in the same market at the same time may not appreciate that, when computing income, he should only deduct the expen- diture on seeds from his receipts. Nor is a seller of street food likely to distinguish accurately between what is eaten by his customers and what by his family. A subsistence farmer, whose outgoings approximately equal his incomings, is quite likely to report that his income is zero. (Even in developed countries the meas- urement of self-employment income is notoriously inaccurate.) The problems are not entirely solved even by the detailed questioning of more sophisticated surveys, in which the surveyor, not the respondent, calculates income. And the national accounts data for household saving are not themselves reliable enough to provide a good cross-check that will show what sort of surveys do best or how they should be redesigned to do better. In the national accounts, household saving is typically a residual among re- siduals. Fry (1988) notes that in many countries illegal capital outflows take place because imports are "overinvoiced" and exports are "underinvoiced," so that if saving is calculated as the sum of domestic investment, the government surplus, and the trade surplus, saving will be underestimated by the extent of the illegal capital outflow. Because household saving is not measured directly (either there are no survey data or the statisticians mistrust them), it is derived from national saving by deducting corporate and government saving, so that errors in measuring corporate saving or investment will be absorbed into this ultimate residual. Given these deficiencies, even trends over time may be incor- rectly observed, while real changes, especially those that involve foreign trans- actions (such as increased remittances to India and Pakistan) may affect the measurement as well as the reality of saving. In India around 1980, observers were puzzled by an apparent dramatic increase in saving and investment un- accompanied by a corresponding increase in economic growth. Rakshit (1982, 1983) and a report by the Reserve Bank of India (1982)investigating the meas- urement issues concluded that the increases, though real, were greatly overstated by the accounting practices. Data problems also preclude testing even the most basic hypotheses, for ex- ample, the classical (Lewis) model in which saving is done almost exclusively by capitalists. If capitalists belong exclusively to the corporate sector, the model is almost certainly false, but in most developing countries we have no way of sorting out the "capitalistic" pockets in the household sector, where a great deal of accumulation may take place within small household-owned business. Finally, statistical practices for measuring saving are far from uniform across 64 Saving in Developing Countries countries-not even between the United-States and Japan (see Hayashi 1986), let alone across developing countries. Conventions for calculating depreciation are good examples of practices that are largely arbitrary and vary widely. The (very extensive) literature that makes comparisons between countries thus rests on peculiarly shaky ground. (Good reviews of these and other data problems are contained in Berry 1985, Fry 1988, and Gersovitz 1988.) A Simple Theoretical Model The basic framework of the simple stylized model of household saving de- veloped here-intertemporal utility maximization-isstandard in the literature. But it does not deliver the standard result, that consumption should be pro- portional to permanent income. The special assumptions under which the stan- dard result is derived seem to me to be unusually inappropriate for most house- holds in developing countries. My model therefore diverges from the textbook in four important respects. First, households in developing countries tend to be larger than households in the United States or Europe, and there is a much greater tendency for several generations to live together. At the extreme, a household might have a stationary demographic structure: old people, as they die, are replaced by those a little younger, and everyone "shifts up"-down to the youngest children, who are replaced by new births. Such a household has no need for "hump" or retirement saving, either as a vehicle for transferring income from high-productivity to low- productivity phases of the life cycle or as a means of transferring wealth between generations. Resources are shared between workers and dependents, and own- ership is passed from parents to children. As emphasized by Kotlikoff and Spivak (1981), this kind of household can internalize many of the insurance activities that would otherwise require saving. Transfers within the household can insure individuals against health risk and old age by providing what are effectively annuities, and the close relationships between the individuals concerned may mean not only that moral hazard issues are less severe than in a more individ- ualistic society but also that the quality of the protection is very high. Note also that this kind of household lives much longer than any individual and thus gives some substance to the idea of a "dynasty" of consumers. Second, income derived from agriculture is inherently uncertain, an uncer- tainty that spreads from agriculture to related occupations and affects most of the population in predominantly agricultural economies. Uncertainty at low income poses a real threat to consumption levels, a threat that is likely to exert a powerful influence on the way in which income is saved and spent. The poorer consumers are, the more risk averse they are generally supposed to be. Declining (absolute) risk aversion has important implications for the shape of the con- sumption function (see Leland 1968, Zeldes 1989b, and most recently Kimball 1988; forthcoming). The standard model in which consumption equals per- Deaton 65 manent income cannot be derived from utility maximization in such a context. Note also that the household insurance arrangements discussed in the previous paragraph are unlikely to be able to deal with income uncertainty, particularly in an agricultural setting. Even multiple earners will not provide much protection if all are dependent on local agriculture. The third divergence from the standard model is the assumption that bor- rowing is not permitted. This is an extreme simplifying assumption, but more appropriate than its opposite, that households are free to borrow and lend at a fixed real rate of interest. For the present, the fact of borrowing constraints is more important than the reason behind them, although it is not difficult to think of plausible scenarios. In a world of financial repression, there may be no credit available to nonfavored borrowers. Or borrowing rates may be so much greater than lending rates that credit is only a last resort in dire emergency. Even where there are financial intermediaries, they may be unwilling to lend for consumptioil purposes to individuals who have no collateral or to lend across agricultural seasons rather than within them. The analysis of borrowing constraints under uncertainty is the main theoretical innovation of the section, even though much of the analytical apparatus can be borrowed from elsewhere. The fourth distinction between household saving in developed and developing countries is a consequence of the previous three. In the model developed here, saving provides a buffer between uncertain and unpredictable income and an already low level of consumption. Saving here is "high-frequency,"intertemporal smoothing saving, not life-cycle-hump or intergenerational saving. The analysis is different, and so are the welfare issues, which are focused on the protection of consumption, particularly among those whose consumption levels may not be far above subsistence. The buffer analogy is also technically useful, since the model here is formally identical to the model of optimal commodity stockpiling (see Gustafson 1958 for an early treatment; also see Samuelson 1971, Newbery and Stiglitz 1981, 1982, and Deaton and Laroque 1989 for modern versions). As far as the consumption literature is concerned, the model developed here is essentially the same as that developed by Schechtman and Escudero (1977). I begin from the standard specification without borrowing restrictions whereby the household maximizes an intertemporal utility function of the form where 6 > 0 is the rate of time preference, c is total household consumption, and u(c,)is the instantaneous utility associated with consumption c,. In keeping with the dynastic view of the household, the time horizon is taken to be infinite. The expectation operator, E, is taken with respect to information at time t. The budget constraint evolves in the usual way, so that for real nonhuman wealth A,, real income y,, and fixed real interest rate r, (2) A,+, = (1+ r)(A, + y, - c,) 66 Saving in Developing Countries The fixed real interest rate is assumed to make later computation more tractable; the assumption in any case may be reasonable in many of the contexts with which I am concerned. Dynamic programming arguments applied to equations 1 and 2 yield the Euler equation, where h(ct) = u'(ct) is the instantaneous marginal utility of consumption in period t. Often, h(ct) may be taken as a price or value: when h(ct) is high, consumption is low, and goods are scarce and valuable. The concavity of the instantaneous utility function implies that h(c) 1s mon- ' otonically decreasing, or in terms of the price interpretation, that price rises with scarcity. I shall also be concerned with the case where h(c,) is (strictly) convex, so that the scarcer the commodity, the steeper the price rise, flattening out as consumption increases. At a technical level, the convexity of h(c,) is guaranteed by the usual assumption of declining absolute risk aversion. More intuitively, the convexity of the marginal utility function guarantees a precautionary motive for saving. In bad times, when consumption is low, the consequences are much worse than they are better in the good times, when consumption is high. The marginal disutility of losses in consumption near subsistence is greater than the marginal utility of gains in times of relative abundance. Individuals will therefore give up high consumption when it is possible so as to prepare for possible disasters, even if those disasters are few and far between. As can be seen from equation 3, an increase in the riskiness of future consumption will increase the right-hand side of the Euler equation, so that to restore equilibrium, the left- hand side must increase, and current consumption fall (see Sibley 1975). A precautionary motive for saving seems a necessary ingredient in modeling poor households which have the additional misfortune of facing considerable uncer- tainty in their incomes. Perhaps the most popular solution to the Euler equation is one which assumes away the precautionary motive by working with quadratic preferences. In this case h(c,) is a declining linear function but is not strictly convex. Given this linearity, and the additional assumption that the rate of interest r is equal to the rate of time preference 6, equation 3 is satisfied by the permanent income consumption function whereby consumption is the annuity value of the sum of assets and expected future income. In the form consumption equals permanent income, it has been widely applied to household behavior in developing coun- tries, (Bhalla 1979, 1980; Musgrove 1979, 1980; Wolpin 1982; Muellbauer 1982). For present purposes, however, the assumption of certainty equivalence (linear marginal utility) is unattractive because it rules out the sort of precau- tionary behavior described in the previous paragraph. Unfortunately, apart from some special cases, the Euler equation is a good deal harder to solve in cases when utility is not quadratic. For econometric work this need not be a problem; Deaton 67 Hansen and Singleton (1983) and a host of subsequent studies have used gen- eralized methods of moments techniques to estimate equation 3 directly, con- ditional on some specification of the instantaneous utility function. However, I need to develop the theory more before I can turn to empirical evidence. In important papers, Skinner (1988) and Zeldes (1989b) have shown that with a finite horizon, precautionary saving can cause major deviations from the certainty equivalent formulation in which consumption equals permanent in- come. Since borrowings must be repaid in the finite future, possibly at a time when consumption is already low and its marginal utility high, individuals with low nonhuman wealth will be hesitant to borrow at all. This consideration influences even relatively well-off households who own land, who fear having to sell and so face permanently lower incomes thereafter. In many ways, this is an attractive formilation for the analysis of low-income househdld behavior. Here I follow another tack and assume that households cannot borrow. I choose this approach partly because the resulting model is easier to handle (largely because it generates a stationary stochastic equilibrium in which con- sumption is not constant) but also because, at least for some households, bor- rowing restrictions are real and necessary to explain what we observe. House- holds in developing countries are often quite long-lived, and many face income processes that are stationary and well understood. But they do not have constant consumption-I suspect because they do not have access to infinite credit and do not therefore make plans that rely upon its availability, particularly when they most need it. Another motive for adopting a new approach is the "excess sensitivity" lit- erature on the United States (particularly Flavin 1981, Hall and Mishkin 1982, and Hayashi 1987) which has exposed a closer link between actual income and consumption than can be explained by permanent income theory. Suppose then that the basic model is extended to incorporate borrowing restrictions by adding to the utility function (equation 1) and the budget con- straint (equation 2) the inequality constraint, that for all periods t, real wealth1 cannot be negative: For most households, I also assume that 6 > r (the rate of time preference is greater than the rate of interest). Otherwise, as shown by Schechtman (1976) and by Bewley (1977), the borrowing constraints ultimately have no effect; households eventually accumulate infinite amounts of nonhuman wealth, at which point they never need to borrow, and consumption once again settles down to a constant fixed number. The assumption that 6 is greater than r has substantive economic content; it takes people to be impatient, unwilling to accumulate, even "feckless," though not of course irrational. With such pref- erences, assets are a lost opportunity for consumption and would ideally be run down quickly. But there is a benefit to holding assets: they are the insurance against having to reduce consumption to unacceptably low levels when times 68 Saving in Developing Countries are bad. Even so, the Schechtman and Bewley results indicate that in this model it is the high level of impatience that keeps people poor. Even without the ability to borrow, patient consumers will eventually become rich, at least as long as r is not driven below 6. I leave such people aside for the moment but will return to them at the end of the next section. In contrast to the simple Euler equation (3),there are now two possible cases in each time period. In case 1, the consumer would like to borrow but cannot; even if all wealth and current income are consumed, the marginal utility of an additional unit of current consumption is greater than the expected marginal utility to be derived by saving that rupee until tomorrow. In this case, con- sumption is the sum of assets and income, saving is zero or negative, no assets are carried forward, and marginal utility is not equated across periods. Formally, we have In case 2, the consumer does not want to borrow, consumption is less than total cash on hand, and the original Euler equation (3) is satisfied. Of course, this does not mean that the borrowing constraints have no effect. The expected marginal utility of future consumption is different from the unconstrained case because the future contains the possibility of being unable to borrow when it would be desirable to do so. The two cases of the Euler equation can be combined into a single expression: In the appendix, I derive the consumption function characterized by the modified Euler equation (6) in the simplest case where income is always positive and is independently and identically distributed over time. This is reasonable enough for a poor agricultural household in a stagnant economy, but it would require substantial modification for a worker facing an increasing nonstationary income stream (see Deaton 1989 for details). The solution takes the form (7) Ct = f b t ) where x, = A, + y, is the amount of cash on hand that is available either for spending today or for keeping for tomorrow. The function f(x)as characterized by equation 7 is determined by the process determining income, for example by its mean and its variance, and by the parameters of the utility function. If any of these change, so will the shape of the function. Although it is not possible to derive an explicit functional form for this consumption function, it can be com- puted for a range of values of x given a utility function and a stochastic process for income. The appendix explains the calculations; the next subsection uses the calculations tc explore the implications of the model. Deaton 69 Implications and Relation to the Evidence In the consumption functions calculated from equation A-5 in the appendix, I have assumed that income is drawn from a (truncated) normal distribution with mean p and variance a', truncated at (p - So, p + 50). The truncation has little or no practical effect, but it ensures that the marginal utility cannot become infinite. The instantaneous utility function reflects the assumption of constant relative risk aversion, that is, it is of the power form, so that the marginal utility of money function h(c)takes the form c - ~For p > 0, this marginal utility . function is convex, thus guaranteeing a precautionary motive for saving. Figure 1shows the relation between consumptionand cash on hand (assetsplusincome) for p = 2 and p = 3 given that p = 100, o = 10, r = 0.05, and 6 = 0.10. When cash is scarce, all of it is spent. At some critical point, which depends on the parameters of the problem but is rarely far from mean income, the slope of the consumption function has a discontinuity, and for higher values of cash on hand, the marginal propensity to consume is lower and continues to fall as the wealth position improves. For consumers with large assets, the liquidity constraints cease to matter, and we are back in the standard case. However, it can be shown that the slope of the consumption function will always remain greater than rl(1 + r);with 6 > r, there is always an incentive to consume assets rather than to hold them. Bear in mind that these multiperiod models do not imply that liquidity constraints mean that poor consumers simply consume their Figure 1. Liquidity-Constrained Consumption Functions Consumption 140 Cash on hand (assets plus income) - ----- Key: coefficient of relative risk aversion, p = 2; p = 3. Note: Assumes mean Income = 100, standard deviation = 10, interest rate, r = 0.05, rate of time preference, 8 = 0.10. Source: Calculated from equation A-5 in appendix. 70 Saving in Developing Countries incomes. Rather they consume all their available resources, which typically will include some assets accumulated in the past. Saving can be negative, and often is. And even when the liquidity constraints are not binding, so that the consumer does not wish to borrow, the consumption function is quite different from what it would be without the possibility of future liquidity constraints. Consumers in this model take precautions against bad years in which they know they will not be able to borrow. The resultingsaving behavior is determined by the interaction of the liquidity constraints and the degree of precautionary saving, as controlled by the parameter p. These points have been well emphasized by Zeldes (1989a, 1989b). Figure 1 shows how the consumption function shifts downward when the precautionary motive is stronger. Similar downward shifts occur as the riskiness of income increases, although increasing the standard deviation (coefficient of variation) from 10 to 15 (10-15 percent) has only a very small effect on the position of the curve. Higher interest rates also shift the function downward; in this model, assets are a buffer stock, and the cost of holding the buffer in terms of utility forgone depends on the excess of 6 over r. Higher interest rates make it cheaper to hold (this sort of) buffer stock, and so there will be more saving and higher asset levels associated with higher interest rates. Perhaps the most important feature of figure 1 is the prediction that saving will increase with "permanent income" as conventionally defined, so that the elasticity of consumption with respect to measured permanent income will be less than unity. The literature on household saving in developing countries has almost uniformly found this result (seein particular Bhalla 1979, 1980 for India; Musgrove 1979 for Latin America; Muellbauer 1982 for Sri Lanka; Betancourt 1971 for Chile; and Paxson 1989 for Thailand).The exception is Wolpin (1982) who, as Gersovitz (1988) points out, has a rather odd measure of permanent income-he assumes, in the Indian context, that permanent income is positively spatially correlated with permanent differences in rainfall, which may not be true in the presence of migration. According to the theory behind figure 1, consumption is not a function of any simple concept of wealth such as permanent income, but a nonlinear function of cash on hand, the function itself depending on characteristics of incomes and preferences. But clearly the propensity for any given household to consume out of assets will be much lower when assets are high than when assets are low, which is one interpretation of the findings. To the (limited) extent that a cross section can be interpreted as identical households with different income draws from the same distribution, saving will rise with assets, and thus with conventionally measured permanent income. Of course, different individuals in any cross section will have different income processes, and different versions of the consumption functions shown in figure 1. In the simplest case, where the distribution of income"rescales"as the mean changes, all the consumption functions will be scaled versions of the original, and saving rates will be independent of scale. But since the income and asset processes will not be perfectly correlated across individuals, saving rates will still be positively correlated with assets across all individuals. More complicated stories can be told if there is a systematic relation between the level of income and its variability. Note also that many of the richest households are likely to be those for whom 6 5 r, who have (eventually) accumulated assets and broken out of the liquidity constraints. For members of this group, consumption is growing over time, but so are assets, at least for the group as a whole. Unlike the households in figure 1,these households are responsible for net accumulation over time, and thus for household saving in the aggregate, even in a stationary income environment. The presence of some of these households in the cross section will further enhance the positive correlation between saving and asset levels. The dynamic behavior of the liquidity-constrained consumers in this model can be seen by looking at figures 2 through 6 which plot a typical 200-period (year) simulation for, respectively, income (normally distributed white noise), consumption, saving, assets, and the marginal utility of consumption. These simulations correspond to the broken line in figure 1, where the coefficient of relative risk aversion is 3. The simple nonautocorrelated behavior of income in figure 2 results in a much more complex time-series process for consumption in figure 3. Note first that stationariness is preserved, that is, that its stochastic characteristics do not vary with time, again (and essentially) because 6 > r. Assets are accumulated to spread income over time, but since they are costly to hold, a time will always arrive when it is optimal to use them-thatis, to consume Figure 2. Simulated Income Income Time (year) Note: Income is successive random draws from a normal distribution with a mean of NO and standard deviation of 10. 72 Saving in Developing Countries Figure 3. Simulated Optimal Consumption Path Consumption Time (year) Note: Based on simulated income from figure 2, and consumption determined as in figure 1. Figure 4. Simulated Optimal Saving Path Saving 0 20 40 60 80 100 120 140 160 180 200 Time (year) Note: Income plus asset income less consumption is derived from figures 2, 3, and 5. Deaton 73 Figure 5. Simulated Optimal Asset Path End of period asset stock -10 0 20 40 60 80 100 120 140 160 180 200 Time (year) Note: Derived as A(t + 1) = (1 + r) {[A(t) + y(t)]f[A(t)+y(t)]}where A is real nonhuman wealth, P is the fixed real interest rate, and y is real income. Figure 6. Marginal Utility of Consumption Marginal utility ( X 0 20 40 60 80 100 120 140 160 180 200 Time (year) Note: Coefficient of relative risk aversion, p = 3. 74 Saving in Developing Countries everything. If things are bad enough, the consumption value of even the last rupee is more valuable than it is ever expected to be again, and so all income and assets will be used. As soon as this happens, the link between the present and the past is broken, and the consumer begins the next period with no assets and a new income draw. Such "stock-outs" can be made arbitrarily rare if a suitable income process is chosen (for example, one in which with very low probability income is very small), but they must happen eventually. Again, these seem attractive features of the model, at least for the poorest countries. The possibility of extended drought or famine is an important motive for precau- tionary saving, and in such periods, assets will be converted to food, although they may be far from sufficient to avoid serious consequences for consumption and its marginal utility. Figure 5 shows that asset levels are typically low, and even at their peak do not reach one-third of mean income, in spite of the inherent variability in income and the strong precautionary element in preference. Again this accords well with the facts, not only as far as we know them in developing countries, but also in developed countries such as the United States (see, for example, Venti and Wise 1989, who report that-apart from housing and pension wealth, which are not available to buffer consumption-median family wealth in 1985 was only $600). Even so, modest amounts of asset holding can significantly smooth consumption; consumption is much smoother than income (compare figures 2 and 3) and is strongly autocorrelated over time. The literature quoted above also provides overwhelming evidence that consumption is indeed smoothed in developing countries, Paxson's results for Thailand being only the cleanest and most con- vincing example. Note also that the fluctuations in consumption are not symmetric; saving can always prevent consumption from being too high, but when assets are exhausted, nothing can be done to soften the effects of low incomes. Perfect smoothing is neither possible nor optimal, and there is less than perfect protection against bad times-see in particular the downward spike in consumption around period 110, and the much more severe associated spike in marginal utility. Interestingly, although income was low in that period, the effects on consumption and on utility are much more exceptional than appear to be warranted by the behavior of income alone. A bad income draw is not in itself the cause of disaster; much depends on the state of assets before the triggering event. Such extreme responses appear to be characteristic of this sort of nonlinear time series, and they seem to account well for the sort of events labeled "panics,""runs," "manias," and perhaps even for some features of famines. Finally, note the behavior of saving in figure 4, and especially the fact that it is as often negative as positive, as must be the case given the absence of net asset accumulation. The model here predicts the frequently lamented finding from household survey data that "implausibly" large numbers of households appear to dissave. I would not wish to claim that the data are correct; the measurement ~roblemsare real enough. But the fact that half the sample is Deaton 75 dissaving in any given year is not theoretically implausible. Indeed, by choosing an asymmetrical density function for income, with very occasional bumper har- vests, it would presumably be possible to generate time series in which dissaving occurs in nearly every period. But recall that an inability to borrow does not imply that consumers spend their incomes, and so is entirely consistent with frequent dissaving. The results in the figures are consistent with the "excess sensitivity" findings in the U.S. literature. When the change in consumption is regressed on the change in income, the simulation results generate a coefficient of 0.291 with a standard error of 0.016-a coefficient much too large to be accounted for by the per- manent income theory. Similarly, in the regression of the change in consumption on lagged income, the significant negative coefficient ought instead to be zero for an unconstrained, forward-looking consumer. Such a negative correlation exists in the microdata from the "PanelStudy of Income Dynamics in the United States" (Hall and Mishkin 1982). We do not yet know whether similar phe- nomena exist in developing countries, although work on this and other aspects of the model is currently under way using the World Bank's "Living Standards Survey" for C8te d71voire. One aspect of reality that is not consistent with the model is that there should always be aggregate saving in the household sector as a whole. The model is one of income smoothing with assets acting as a buffer stock; there is no motive for accumulation, and over a long enough run of years total household saving would be expected to average to zero. But even if many household surveys show something like this, most observers would allow some credence to the national income accounts, and suppose that there is some saving to be explained. Pop- ulation growth would help explain some accumulation. If there are more house- holds, each will need its own buffer stock of assets, so that increasing population will require positive saving as the stocks are built up. Income growth is harder to handle, since the basic model assumes a stationary income process and is not easily modified to accommodate nonstationariness. One possibility is that each individual faces a stationary income stream for his or her life, but that each new generation faces a higher mean process. Such a model would tend to reinforce the population growth effect as each generation holds a larger stock of assets on average. A more interesting hypothesis is that the preference parameter 6 varies from person to person. Some are patient and willing to wait for higher consumption; others are not. For the majority, for whom (6 > r), the model applies, and they spend their lives smoothing consumption, holding few assets, and always subject to liquidity constraints. The minority, with (6 I r), are also unable to borrow, but their optimal consumption plans, which typically involve saving now to support higher consumption in the future, also have the characteristic that the stochastic process governing their assets is such as to guarantee that assets will exceed any finite level given enough time to do so. This accumulation, which is stochastic, is an almost accidental outcome of optimal consumption plans for 76 Saving in Developing Countries patient consumers who can never borrow. The distribution of preferences thus divides the population into two groups, one of which lives a little better than hand to mouth but never has more than enough to meet emergencies, while the other, as a group, saves and steadily accumulates assets. Such a dichotomy recalls the classic models of saving in which "capitalists" do all the saving (see Lewis 1954; Kaldor 1955-56), but here the capitalists are created willy-nilly as a long- run consequence of their taste for future over present consumption. The pref- erence-based dichotomy also suggests an explanation for the division of con- sumers into two groups-liquidity-constrained or "rule-of-thumb" consumers and "life-cyclers"-that has recently become popular in the U.S. literature (see Hall and Mishkin 1982; Campbell and Mankiw 1989a, 1989b; Flavin 1988). The theory here explains how the two groups arise, why each is behaving op- timally, and why some are liquidity-constrained and others are not. But in its present form it does not do any of this when individual consumers expect their incomes to grow. The Life Cycle, Saving, and Growth One of the most celebrated and most investigated predictions of the life-cycle model is that there should be a relation between aggregate saving and the rates of population and income growth. If saving is hump saving, accumulated during the working years to finance retirement, then population growth provides more savers than dissavers, and positive aggregate saving. Per capita income growth has a similar effect because workers are saving on a larger scale than the retirees are dissaving. For many people, population growth is the issue in economic development, and the relation between population growth and capital accu- mulation is one of the most important of the possible links between population policy and economic welfare (see, in particular, National Academy of Sciences 1986; and Mason 1987, 1988). Even at the theoretical level, however, there are complications. If young con- sumers anticipate a steady growth in income, and can and will borrow against that increase, their dissaving in the early years of the life cycle may induce a negative relation between saving and growth. The standard positive relation works best if each worker experiences a stationary income stream over his or her own life cycle, with taking place between rather than within gen- erations. The effectsof population growth aresimilarly ambiguous. Even if adults would like their own consumption stream to be constant over the life cycle, their expenditures may exceed income, not only during retirement but also when there are children in the household. Population growth expands the ratio of workers to retirees, but also the ratio of children to adults, and saving may be decreased more by the latter than it is increased by the former. The net effect depends on the costs and benefits of children, a balance that may itself change (from net benefit to net cost) with economic growth (see Caldwell 1982). And do house- Deaton 77 holds really want to have flat consumption streams in any case? Cautious young people may not want to borrow against future income growth, even if that gowth is extremely likely. And old people, faced with daunting uncertainties about health and death, may not run down their assets in the prescribed man- ner-a supposition strongly supported by the balance of empirical evidence from developed countries. To these lacunae in the standard argument may be added my own doubts, expressed in the previous section, about the general applicability of the hump-saving concept in developing countries. The cross-country empirical evidence (well reviewed by Gersovitz 1988) gen- erally supports a positive effect of per capita income growth on saving rates, variously defined; however, the results are rarely well-determined and are un- comfortably reliant on the treatment of the simultaneity between saving (in- vestment) and growth, and on the sample of countries selected. The even more ambiguous role of population growth can perhaps be interpreted as negative dependency effects more or less offsetting the positive effects of population growth. Figures 7 and 8 illustrate fairly typical findings for the relation between saving and the total gowth of gross national product (GNP). Both figures show a positive slope, with the saving rate increasing 1-1.5 percentage points for every percentage increase in the growth rate. The t-values are modest but significant at conventional levels. Of course, there are outliers (theyhappen not to have much effect on the results),and the scatter yields plenty of scope for obtaining different results by suitable sample selections or choice of instrumental variables-all this quite apart from whether the countries should be given the same weight, whatever their population size, data reliability, or anything else. The fundamental problem is the direction of causality: from growth to saving (the life-cycle explanation) or from saving to growth? The problem is tackled by several authors with various instrumental variables, but these efforts are hardly convincing in the absence of an adequate theory of growth. Summers and Carroll (1989)have argued that, whatever produces the positive correlation between saving and growth, it cannot be life-cycle saving. They point out that the life-cycle explanation assumes common preferences across countries, but that differences in economic growth generate differences in the relative lifetime economic standing of young and old in different countries. To use Lucas's (1988) graphic example of the power of compound growth: if a grandchild is fifty years younger than its grandfather, using 1965-86 per capita growth rates, then a citizen of the Republic of Korea is 26 times as rich as his or her gandfather, while an Indian is only 2.4 times as rich. These enormous differences should show up in profiles of the relation between age and consumption-if Korea is compared with India, the profile should be relatively higher among the younger cohorts in the faster growing economy. But Summers and Carroll's (1989) age- consumption profiles for Japan, the Unites States, and Canada are essentially identical in spite of the differences in growth rates; the differences predicated by the life-cycle growth effects are simply not present. 78 Saving in Developing Countries Figure 7. Gross Domestic Saving and Growth: World Bank Data Gross domestic saving ratio, 1986 50 Growth rate of GDP, 1980-86 Source: World Bank (1988). Figure 8. Saving and Growth: Summers and Heston data Consumption ratio, 1980 80 - - 60 0 0 0 - 0 40 0 -20 - 0 -40 I 1 I I I I I -2 0 2 4 6 8 10 12 14 Growth rate of GDP, 1965-80 Note: On the vertical axis, the var~ableis one minus the sum of the shares of government and private consumption in total GDP, measured in current international prices. The horizontal axis shows the growth rate of per capita gross national product in constant 1980 international prices multiplied by population. Points for 106 countties are shown. Source: Summers and Heston (1988). Deaton 79 I have been able to obtain data on age-consumption profiles for five countries: C6te d'Ivoire, Hong Kong, Indonesia (rural Java), Korea (cities only), and Thai- land (see figures 9-11). The growth rates of real GNP per capita given in table 1are from both World Bank (1988)and Summers and Heston (1988)data. The right-hand panel shows the ratios of incomes twenty years apart at these growth rates, in which, for example, the numerator could reflect the relative lifetime income of thirty-year-olds and the denominator that for fifty-year-olds. Except for C6te d'Ivoire, where the Summers-Heston growth figure is 0.2 percent a year instead of the 1.2 percent from World Bank (1988),the two sets of estimates agree closely. Both sets of rankings have C6te d'Ivoire growing most slowly, Thailand and Indonesia rapidly, and Hong Kong and Korea very rapidly. The corresponding age-consumption profiles (figures 9-11) show how far the pre- diction is from reality. In slow-growing C6te d'Ivoire the consumption profile is heavily tipped toward the young; in urban and rural Thailand and in Korean cities, where thirty-year-olds are two to four times richer than fifty-year-olds, the profile peaks for households with heads in their fifties. In rural Java, which is growing about as fast as Thailand, the peak is somewhat earlier, in the mid- forties. For Hong Kong, the published data points are sparse, but household expenditure levels seem not to vary very much with age from the late twenties to the late fifties. For Thailand, Korea, and C6te d'Ivoire, where I have data from more than one year, or for Thailand for more than one sector, profiles from the same country are much more similar than profiles across countries. But the differences across countries are not easily explicable by life-cyclegrowth effects. The results should not be overstated. A more sophisticated analysis would take into account the differing family sizes in the different countries as well as the marked declines in fertility in several countries that have meant that older households typically had more children than households of the next generation. Even so, it is hard to see that these considerations could affect the basic point. Note too that the graphs do not show that households do not look ahead when planning their consumption, nor that they do not smooth out short-term fluc- Table 1. Growth Rates and Twenty-Year Growth Factors growth rate 1965-1 986 20-year growth factor Country W D R SH W D R SH percent per year CBte d'lvoire Hong Kong Indonesia Korea (cities) Thailand Note: WDR is World Development Report 1988 (World Bank 1988); SH is Summers and Heston (1988).The SH data are average growth rates from 1965 to 1985. The growth factor is the (1+g)", where g is the growth rate. Sources: World Bank (1988);Summers and Heston (1988). 80 Saving in Developing Countries Figure 9. Age-Consumption Profiles for Java and CGte d'lvoire Expenditure divided by average expenditure 0.75 1 I I I I I I I I 20 25 30 35 40 45 50 55 60 65 Age of household head Source: Calculations based on data from La Direction des Etudes et de la Recherche de ]'Office National de Formation Professionelle, an agency of the government of C6te d'lvoire; and on data from the Indonesian National Socioeconomic Survey, Central Bureau of Statistics, Government of Indonesia. Figure 10. Age-Consumption Profiles: Villages and Urban Areas of Thailand Expenditure divided by average expenditure 1.3 - Urban, 1980 - - - ' Villages, 1986 - - - .:? $ 0.6 I I I I I I I I 20 25 30 35 40 45 50 55 60 65 Age of household head Source: Calculations based on the 1980 and 1986 Socioeconomic Surveys, National Statistical Office, Government of Thailand. Deaton 81 Figure 11. Age-Consumption Profiles for Korean Cities and Hong Kong Expenditure divided by average expenditure 1.3 - - - - - Hong Kong, 1979-80 - 0.6 I I I I I 15 25 35 45 55 65 75 Age of household head Sources: Hong Kong (1980);and Korea (1980, 1983, 1986). tuations in income. But they do show that the relative lifetime economic status of different age groups does not directly determine their current consumption levels. Given this, the standard explanation of life-cycle rate of growth effects, that younger cohorts are saving and spending on a larger scale, simply does not work. Why not is unclear. Even if the life-cycle model is false, there may be strong precautionary motives or constraints that prevent young consumers from borrowing against their expected future incomes. Note finally that all these data show systematic variation of consumption patterns with the age of the household head, so that the model of the infinitely lived demographically stationary house- hold in section I cannot be literally true. Where this leaves us is anyone's guess. My own feeling is that the cross- country correlations exist, if only weakly, but that we have very little idea of why, or at any rate no way of separating out the many possible explanations. But the support our results give to the Summers and Carroll challenge to the life-cycle explanation suggests that a great deal of the literature needs to be rethought. Some Stabilization Issues In developed countries, concern about the nature of the consumption function has centered on its implications for government policy, in particular the extent to which short-term fiscal policy, by manipulating household disposable income, can affect consumption and thus the level of economic activity. If most of 82 Saving in Developing Countries consumption is determined by permanent income, short-term fluctuations in income will have less effect on consumption than if liquidity is constrained for a sizable fraction of consumers. Few developing countries have income tax systems that permit fine-tuning of disposable incomes; nevertheless, fiscal ar- rangements do have important effects on income fluctuations, on the distribution of income, and most likely on the level of national saving. I have in mind various agricultural taxation schemes prevalent in developing countries, particularly in those where there are substantial exports of primary commodities. Prices of primary commodities are extremely volatile, so that the incomes of countries that sell them fluctuate widely. Such fluctuations are generally con- sidered undesirable in themselves, but their undesirable effects would seem to arise from their translation into fluctuations in consumption. If so, developing countries ought to save and dissave in order to ride out the fluctuations in income. My impression is that this particular saving problem has been rather neglected until very recently. (See Gelb 1988 for a discussion of the effects of oil windfalls, and Bevan, Collier, and Gunning 1987 and Balassa 1988 for discussion of the macroeconomic problems created by the windfalls and losses that accompany commodity price fluctuations.) Agricultural pricing and tax policies. Agricultural taxation affects the way income fluctuations are shared between government and farmers, and so deter- mines who must save to smooth consumption. To givesome examples, as detailed by Bevan, Collier, and Gunning (1987), coffee was not taxed in Kenya at the time of the coffee boom in 1975-76, so that the windfall went directly to the coffee farmers, who apparently succeeded in saving a good deal of it. Arrange- ments are almosr the polar opposite in C6te d'Ivoire, where the government, to guarantee a constant real internal price, sets procurement prices for both coffee and cocoa at levels which bear little relation to world prices-a state of affairs common for most export crops in most of Africa (see Gersovitz and Paxson 1989). With such arrangements, all income fluctuations (and hence the respon- sibility for smoothing) accrue to the public sector. Between these extremes, the Thai government has varied the rice "premium" in such a way as to allow the domestic price of rice to fluctuate with the world price, though with a smaller amplitude. The government takes a larger share in tax when the world price is high, and vice versa, so that the smoothing problem is shared between the public and private sectors. The effects of these schemes on total domestic saving depend on how public and private saving differ. One (unlikely)possibility is that there is no difference. Another is that households and farmers do not save, either because of lack of suitable instruments or because they lack foresight. In such a world, the gov- ernment would have a custodial role both as guardian of future generations and as an insurance company, to protect farmers' consumption against the volatility of commodity prices (seeMirrlees 1988).The custodial role for government was prominent in most of the development literature in the 1960s and 1970s, and Deaton 83 it is embedded in most of the standard cost-benefit procedures. A more skeptical attitude toward the ability of governments to handle these problems better than the private sector has rev ailed recently, and although there is undoubtedly an element of fashion in these beliefs, there is plenty of evidence of apparently perverse government responses to temporary income shocks (seeBalassa 1988). Some governments may find it hard to resist pressures to spend in the face of mounting revenues, and equally hard to make cuts when revenuesfall. The result may be that all positive shocks are treated as permanent and all negative shocks as transitory. Even when the temporary revenue gains have been used to finance investment, which is one way of smoothing consumption, there have been prob- lems. Large investment projects are not easy to reverse, and they generate com- mitments that may be hard to meet once the boom is over. A case in point is the C6te d'Ivoire investment program that required overseas borrowing in ad- dition to the sums available from the boom in coffee and cocoa prices. And the quality of investment projects implemented in response to such revenue booms may be questionable. In contrast to this evidence on government behavior, the theory and empirical results quoted in section I, although ambiguous on the applicability of simple versions of the permanent income theory, seem quite unambiguous in their finding that farmers can and do save and dissave so as to smooth consumption over time. Much more work needs to be done, on both private saving behavior and government saving, consumption, and investment patterns, to reach a re- alistic assessment of the effects of pricing schemes, of how to redesign them to avoid the most serious pitfalls, and of the potential value of the sort of inter- national compensatory schemes advocated by Balassa (1988). Fluctuations in commodity prices. How ought consumers and governments to respond to commodity price fluctuations? I shall take the simplest view, that price fluctuations induce income fluctuations, and that consumption, at least in the aggregate, ought to respond to permanent but not transitory innovations. We therefore need a mechanism for sorting out permanent fluctuations from transitory ones. Ideally, it is income that ought to be decomposed into permanent and transitory components, but the relation between prices and incomes is com- plicated and differs from country to country. Instead, I focus on commodity prices themselves. The standard theory of commodity price determination is one of speculative demand for inventories interacting with agricultural supply and demand (see the references on optimal commodity stockpiling given in section I). Typically, the underlying supply and demand conditions are assumed to be stationary, so that, although the theory is consistent with extreme volatility and nonlinearity in the stochastic process which determines prices, the price process is stationary. As a consequence, price booms and slumps are transitory, and price shocks convey no useful information about prices in the far future. But this is theory, not necessarily fact, and although actual commodity prices do indeed go up and 84 Saving in Developing Countries down, it is far from clear that shocks invariably die away and that booms and slumps are always temporary. Indeed, several analysts of commodity prices, for example Labys and Granger (1970) or Ghosh, Gilbert, and Hughes-Hallett (1987),treat the first differenceof commodity prices as stationary (thestandard prescription of the time-series analyst faced with the very high and slowly di- minishing autocorrelations in most commodity prices). Many of these formu- lations imply that shocks are infinitely persistent and that there are no forces that act to bring prices back to some fundamental level. While this position does not seem to be sustainable, it is plausible that some part of commodity price shocks reflects shocks to fundamentals, not just temporary shortages or gluts associated with bad or good harvests and the actions of profit-maximizing spec- ulators. Fortunately, the extent to which price shocks are permanent is a topic ame- nable to empirical investigation. The precise question to be answered is how much of an innovation in the price can be expected to persist indefinitely, and how much to evaporate, at least eventually. Cochrane (1988)and Campbell and Mankiw (1987)have developed an estimate of persistence based on the successive autocorrelations in the first differences of the time series. If the time series is anchored, either to a constant base or a deterministic trend, then innovations must eventually be followed by compensating changes in the opposite directions, and there will eventually be a predominant pattern of negative autocorrelations in the first differences. Consider, for example, a white noise series (p + E,).The first difference is (E,- E,-,), which has autocorrelations at all leads and lags of (... 0, 0, 0, 0.5, 1, 0.5, 0, 0, ...) where the 1 is the autocorrelation at lag - - 0, and the -0.5s at lead and lag one. These autocorrelations add to zero, and the series has zero persistence. By contrast, a random walk with drift Ay, = 8 + E, has a first difference which is white noise, and the sum of all its autocor- relations is unity, so that a random walk has a persistence measure of 1; all shocks are permanent. Any stationary series has a persistence measure of zero, but not all series with unit roots are necessarily very persistent. For example, the sum of a random walk and white noise has a persistence measure less than unity, while a unit root series that is positively autocorrelated in first differences has a persistence measure greater than unity. Table 2 lists summary statistics and persistence measures for thirteen com- modity prices important for developing-country exports and incomes. The data come from the World Bank and are monthly from January 1960 through Feb- ruary 1988. The Bank routinely deflates these prices by an index of import prices for developing countries; here I have deflated by the U.S. consumer price index. The differencein the deflator is insignificant compared with the fluctuations in the series. The coefficients of variation show the volatility of the series. The sugar price is by far the most volatile, and the banana price by far the least; neither commodity is traded in anything like a free competitive market. The second and third columns show the autocorrelation coefficients of the prices themselves at lags of one and twelve months. Apart from bananas, which alone Deaton 85 Table 2. Monthly Commodity Prices: Summary Statistics Coefficient of Autocorrelation" Persistence measuresb - Commodity variation 1 month 12 months 60 months 220 months Arabica ( Bananas Cocoa Copper Cotton Iron ore Jute Maize Palm oil Rice Sugar Tea Tin "The first and twelfth autocorrelation coefficients of the deflated series. CampbelliMankiw-Cochrane measures of persistence, that is, the normalized spectral density at frequency zero, estimated using a triangular (Bartlett) window with window widths of 60 and 120 months respectively. Source: Calculations based on World Bank data. among these commodities are perishable and show seasonal price variation, all the first-order autocorrelations are more than 0.95, and only one is less than 0.97; even after twelve months, most of the autocorrelation remains. These kinds of statistics are characteristic of integrated, nonstationary pro- cesses. Nevertheless, the persistence measures show considerable variability from commodity to commodity. Two estimates are shown, corresponding to "window widths" of 60 and 120 months respectively. The need for the window width arises from the impossibility of adding all terms in an infinite series; here I calculate the weighted average of either 60 or 120 autocorrelations with weights declining linearly with the length of the lag. There is a standard tradeoff here between bias and variance. Small window widths are relatively precise but, by omitting higher-order autocorrelations, run the risk of bias, especially if the price reverts only slowly to its base. Wide windows lead to imprecise estimates, and at extreme lengths are biased toward zero. The asymptotic t-values for the two estimates are 2.1 and 1.5 respectively. For most of the commodities, particularly bananas, copper, iron ore, jute, palm oil, sugar, and tea, the persistence estimates are small; it would probably make sense to accept the standard view that shocks are not persistent. For a second group-coffee, cocoa, cotton, maize, and rice-the estimates are a good deal larger, so that there is some evidence that some fraction of actual, historical shocks has been permanent. For one commodity, tin, the persistence estimate is very large, even at the wider bandwidth. A country facing fluctuations in the price of tin would be justified in taking the view not only that price changes are permanent, but that it makes sense to expect further movements in the same 86 Saving in Developing Countries direction. For commodities with low persistence estimates (thefirst group), price booms are times to save, and slumps times to dissave. For those in the second group, price booms would justify at least some increase in consumption levels. For tin, a rise in price would justify a country borrowing so as to spend more than the current increase in income. If commodity income is more than unit persistent, consumption levels should be more volatile than income. The commodity boom in the 1970s appeared to trigger exactly this sort of behavior, with some countries borrowing internationally during the boom. Of course, the evidence here would not support such a strategy in general; indeed the lack of persistence for most of the commodities helps explain why those policies had such disastrous consequences. But it is much easier after the event to assess the transitoriness of a price boom, and the figures in the table show that it is far from obvious in advance how saving ought to respond to fluctuations in commodity prices, even when governments can and will implement the correct policies. The idea that underdevelopment is a problem of too little saving is deeply embedded in the history of development economics. The argument seems simple enough: capital accumulation is a necessary and sufficient condition for growth, and capital accumulation is almost synonymous with saving; the route to de- velopment is then one of raising saving ratios. But since Lewis's (1954)statement of this "central problem of economic development," the argument has been assailed by the standard economist's question of what prevents the market from working without outside interference, and further eroded by questioning of the link between saving and growth. Solow's (1956) model does not generate any relation between saving and growth in long-run equilibrium, although increases in saving will generate increases in growth over a transition path that may be very long-lived. Recently, there has been a renewed interest in "increasing re- turns" models of growth (see particularly Arrow 1962; Romer 1986,1988; and Lucas 1988; reviewed splendidly by Romer forthcoming). But these models emphasize, not so much saving, but the role of human capital formation, so that while such models predict a relation between the willingness to wait and the rate of growth, there is no necessary relation between growth and the rate of physical capital accumulation. Nor has the empirical evidence suggested any straightforward link between growth and either saving or investment. Figures 7 and 8, with the axes reversed, show the same weak relations that, in the previous section, were interpreted the other way round. There is also a very high cross-country correlation between saving and investment, so that the picture would not be different if the latter were substituted for the former. The point is not that there has not been growth, nor that there have not been much higher saving ratios. Even after recent events, postwar growth in the developing world has been very high by all historical standards, and there are now many developing countries with saving rates that would have seemed unimaginably high to the development economists of the 1950s. In 1986, China's gross domestic saving was 36 percent of gross domestic product (GDP), India's 21 percent, Kenya's 26 percent, Thailand's 25 percent, and the People's Republic of the Congo's 30 percent. The poinlt is that across countries there is at best only a very weak relation between savlng and growth, perhaps because it is the productivity of investment that is crucial, not its volume. Nevertheless, there are substantive issues about the "right" amount of saving. If there are positive externalities to saving by each individual, too little of it will be done, an argument formalized in Sen's (1967)"Isolation Paradox." It can also be argued that future generations will not be adequately represented by their currently living ancestors, so that governments must act on their behalf. Alternatively, it might simply be that optimal intertemporal choice under un- certainty is difficult in that it requires a degree of calculation and sophistication that is too difficult for individuals and can reasonably be expected only of a planning agency. Whatever the virtues (and vices) of planning agencies, the evidence that households make good intertemporal allocations is far from over- whelming. The ability of households and farmers to smooth out short-term income fluctuations is well established, but that is not the same as being able to make ideal provision for the long-term future. The empirical literature on life-cycle models, although successful in many respects, has repeatedly failed to observe the sorts of lifetime profiles of consumption and labor supply that would be expected from long-term intertemporal optimization (seeBrowning, Deaton, and Irish 1985; and Mankiw, Rotemberg, and Summers 1985). In opposition to the view that governments need to step in to remedy the deficiencies of private-sector saving is the contention that government interfer- ence, not market failure, is responsible for inadequate saving in many developing countries. The "financial repression" literature (associated with the work of McKinnon 1973 and Shaw 1973; see Fry 1988 for an extensive review and references), argues that governments have reasons for keeping domestic interest rates low and for repressing financial intermediation in general. Low interest rates keep down the cost of domestic borrowing, and the lack of alternative borrowers allows the government to exploit its monopoly as a seller of financial securities. Low interest rates and lack of investment opportunities are then held to be responsible for low domestic saving. According to this view, financial liberalization is the recipe for higher saving ratios and higher growth. These arguments tend to parallel the similar arguments in developed economies that government policies, particularly tax policy, lower the return to saving and hamper capital accumulation. Apart from the connection between saving and growth, which I have already discussed, these arguments take it as axiomatic that saving responds positively to interest rates. Once again, there is no theoretical basis whatsoever for this presumption. Changes in interest rates have both income and substitution effects, and can increase or decrease current consumption depending on the bal~nce 88 Saving in Developing Countries between the two. Higher (real) interest rates do indeed increase the incentive to postpone consumption and tend to make the planned consumption profile grow more rapidly over time, but the current starting point of that profile can move either up or down. There is also an enormous body of research, mostly but not exclusively in developed economies, that has singularly failed to show any em- pirical relation between interest rates and the rate of saving. The empirical work can be divided into two classes: those studies that look for a direct effect of interest rates on saving, and those, following the more recent "Euler equation" approach, that look for a relation between the rate of growth of consumption and the interest rate. Theory predicts nothing in the first case and is consistent with any finding; in the second case, the effect ought - to be positive, at least in the simplest models of infinitely lived consumers. Perhaps the most frequently cited of the first kind of study is Boskin (1978), who finds a very strong positive interest elasticity of saving. However, this study stands almost alone, and it is also notable for the very nonstandard data series that are used. Much more typical is the time-series study by Blinder and Deaton (1985),which finds some interest rate effects in some specifications but whose results are not robust either to changes in the sample period or to the inclusion - - or exclusion of other variables. ~ndeed,the consumption function literature abounds in studies that include, in addition to income, some favorite "exotic" variable, which does well in that particular study. However, attempts to estimate more comprehensive models rarely support the original studies. For developing countries, studies have usually been on pooled cross-section time-series data for a range of countries over some span of years. A number of early studies by Maxwell Fry reported high interest elasticities. Giovannini (1983, 1985),in what appear to be careful studies, was unable to find any positive effects in similar data for the 1970s rather than the 1960s. My reading of Gupta (1987) too is that well-defined robust estimates are very hard to obtain. Fry's (1988) review and update of his earlier studies once again finds positive interest elasticities. However, I find this latest evidence unconvincing, largely because Fry does not give enough information for me to tell how the equations were estimated. The literature as a whole is not very enlightening: the value of these sorts of cross- country studies is in any case dubious, particularly given the data problems, and several of the studies do not reach econometric standards that would allow the reader to take their results at face value. Studies of the second type, linking consumption growth to real interest rates or, better, to expected real interest rates, have been, if anything, even more unsuccessful. In the U.S. data, whether prewar or postwar, there is no relation whatever between consumption growth and real interest rates, whether or not the latter are expected or realized, and however sophisticated or careful the estimation technique. Even at the most obvious level, post-tax real interest rates in the United States have been negative as often as positive, and yet consumption growth has nearly always been positive, a finding that would require a negative rate of time preference (see Deaton 1987). Other, more sophisticated studies, Deaton 89 come to the same conclusion (see Hall 1988; Campbell and Mankiw 1989 and forthcoming; and Hotz, K~dland,and Sedlacek 1988). There have been few recent studies for developing economies, but Giovannini (1985) has examined the effects of expected real interest rates on consumption growth in eighteen developing economies. He finds some nonzero effect in five (India, Jamaica, Greece, Myanmar, and Turkey), and no effects in the other thirteen (Argentina, Brazil, Colombia, Indonesia, Kenya, Korea, Malaysia, Mexico, Philippines, Por- tugal, Singapore, Taiwan, and Thailand). The last two figures, 12 and 13, show the cross-economy scatter of real interest rates and consumption growth. Clearly, the treatment of expectations in these figures could be much improved, but I should still expect any important patterns to show up. An important point to note is what is not on the figures, that is, the set of fourteen economies (thirteen in figure 13) whose real interest rates were less than - 10 percent (Argentina, Bolivia, Ghana, Israel, Madagascar, Mexico, Nicaragua, Sierra Leone, Somalia, Tanzania, Turkey, Uganda, Zambia, and in figure 12, Yugoslavia). Several of these have negative rates of several hundred percent, and their inclusion would dominate the figures, as well as producing regression lines with slope zero. The (insignificant)positive slopes in the two figures should be interpreted with that exclusion in mind. There is certainly no evidence here of any well-defined relation between interest rates and consumption growth, particularly in the Summers-Heston data. If we were to believe that preferences are common across countries, these results would be prima facie evidence against the supposition that consumption is being optimally allocated over time. By the standards of a decade or so ago, such evidence would be taken as favoring some form of state planning. We are nowadays much more skeptical about the ability of planning agencies to solve these problems, but that does not mean that saving is being optimally done. Indeed, the evidence reviewed in this section does not point to any simple policysolution for thesaving problem, if problem it is. Apart from the ambiguity of the empirical results, one of the main difficulties is our lack of an accepted and well-supported theory of eco- nomic growth. IV. CONCLUSIONS My view is that the research priorities for the immediate future lie with the topics covered in sections I and 11.I think the literature has sufficiently belabored the problems of physical accumulation. The issues are certainly important, but I cannot see useful ways forward without major theoretical advances, particularly in the theory of growth. The recent developments reviewed in Romer (1989) hold promise of such an advance, but, if that promise is fulfilled, research is likely to be redirected, perhaps toward a more intensive study of human capital formation. The rather negative results of section 111should at least serve as a warning to those who like to make glib generalizations on the basis of the experiences of 90 Saving in Developing Countries Figure 12. Consumption Growth and Interest Rates: World Bank Data Consumption growth rate, 1980-86 10 - - 8 6 4 2 0 -2 -10 -10 -6 -2 2 6 10 14 18 Real deposit rate, 1980 Note: 60 countries excluding 14 with real deposit rates less than -10 percent. Source: World Bank (1988). Consumption growth is from table 4, column 2, deflated by population growth from table 27, column 2. Interest rates are the nominal deposit rate in 1980 (table 25, column 7) less the actual rate of inflation for 1980-86 (table 25, column 6). Figure 13. Private Consumption Growth and Interest Rates: Summers and Heston Data Consumption growth rate, 1980-85 Real deposit rate, 1980 Note: 62 countries, excluding 13 with real deposit rates less than --I0 percent. Sources: Interest rates: World Bank (1988, table 25);consumption growth: Summers and Heston (1988). Deaton 91 a few carefully selected countries: saving is not only about accumulation, but about consumption smoothing in the face of volatile incomes, and about pro- viding insurance for poor people whose lives are difficult and uncertain. I think that the data exist that would help us understand more about how poor house- holds use saving and assets, and I think we need a more positive understanding of how governments respond to fluctuations in their revenues. Finally, and this is an area in which the World Bank should be taking the lead, we need to know more about the data, what they mean, and how to improve them. Useful work could be done by bringing together national income accountants and survey statisticians in a few countries where there is extensive experience in both areas. We also need experimental household surveys that will track cash flows within households, perhaps in quite small samples more akin to village studies, so that we can learn whether the apparent patterns of saving and dissaving are real, and if not, how to improve the survey questionnaires. Without such studies, and without these data improvements, our understanding is likely to remain precarious. Section I deals with two cases of the Euler equation with a borrowing con- straint. In the first case, which satisfies equation 3, the consumer is spending less than the total of his cash on hand (assets and current income), while in the second, equation 5, everything is being spent, and there may be dissaving. The two branches of the Euler equation can be combined into a single expression by writing In order to derive a solution to equation A-1 and for computational reasons, I will work with the simplest case, in which income is independently and identically distributed over time. I also require an assumption that prevents the marpnal utility of money from becoming infinite in the worst possible case, which is when the individual has no assets and receives the lowest possible value of income. To this end, I assume that the income process is such that y, always falls in the interval (yo,y,), with y, > yo> 0, y, possibly infinite but X(y,) r. Given values of the two pa- rameters, a marginal utility function, and a density function for income, an initial guess is made for f(x,), for example the piecewise linear form where p, = E(y).The guess is substituted into the right-hand side of A-5, and a new function fl(x) is calculated using numerical integration to evaluate the expectation. After the first guess, only numerical solutions are possible, and the function must be evaluated over some suitable grid for x. Deaton and Laroque (1989)show that, provided 6 > r, this procedure defines a contraction mapping from one function to the next, so that the numerical calculations will be con- vergent. Arrow, K. J. 1962. "The Economic Implications of Learning by Doing." Review of Economic Studies 29: 155-73. Balassa, B. 1988. "Temporary Windfalls and Compensation Agreements." World Bank PPR Working Paper 28. Washington, D.C. Processed. Berry, A. 1985. "On Trends in the Gap between Rich and Poor in Less Developed Countries: Why We Know so Little." Review of Income and Wealth 31: 337-54. Betancourt, R. 1971."The Normal Income Hypothesis in Chile."Journalof the American Statistical Association 66: 258-263. Bevan, D. L., P. Collier, and J. W. Gunning. 1987. "Consequences of a Commodity Boom in a Controlled Economy: Accumulation and Redistribution in Kenya 1975- 83." World Bank Economic Review 1 (3): 489-513. Bewley, T. 1977. "The Permanent Income Hypothesis: A Theoretical Formulation." Journal of Economic Theory 16: 252-292. Bhalla, S. S. 1979. "Measurement Errors and the Permanent Income Hypothesis: Evi- dence from Rural India." American Economic Review 69: 295-307. . 1980. "The Measurement of Permanent Income and Its Application to Saving Behavior." Journal of Political Economy 88: 722-43. Blinder, A. S., and A. Deaton. 1985. "The Time Series Consumption Function Revisited." Brookings Papers on Economic Activity 2: 465-511. Boskin, M. J. 1978. "Taxation, Saving and the Rate of Interest." Journal of Political Economy 86: S3-27. Browning, M. J., A. Deaton, and M. J. Irish. 1985. "A Profitable Approach to Labor Supply and Commodity Demands over the Life Cycle." Econometrica 53: 503-43. Caldwell, J. C. 1982. Theory of Fertility Decline. London: Academic Press. Campbell, J. Y., and N. G. Mankiw. 1987. "Are Output Fluctuations Transitory?" Quarterly Journal of Economics 102: 857-80. .1989."Permanent Income, Current Income, and Consumption." Princeton Uni- versity, Woodrow Wilson School of Public and International Affairs. Princeton, N.J. Processed. . Forthcoming. "Consumption, Income, and Interest Rates: the Euler Equation Approach Ten Years Later." NBER Macroeconomics Annual 1989. Cochrane, J. H. 1988. "How Big Is the Random Walk in GNP?" Journal of Political Economy 96: 893-920. Deaton, A. 1987. "Life-Cycle Models of Consumption: Is the Evidence Consistent with the Theory?" In T. F. Bewley, ed., Advances in Econometrics: 5th World Congress, vol. 2. New York: Cambridge University Press. .1989. "Saving and Liquidity Constraints." Princeton University,Woodrow Wil- son School of Public and International Affairs. Princeton, N.J. Processed. Deaton, A., and G. Laroque. 1989. "On the Behavior of Commodity Prices." Princeton University, Research Program in Development Studies. Princeton, N.J. Processed. Flavin, M. 1981. "The Adjustment of Consumption to Changing Expectations about Future Income." Journal of Politzcal Economy 89: 974-1009. . 1988. "The Excess Smoothness of Consumption: Identification and Interpre- tation." NBER Working Paper 2807. National Bureau of Economic Research. Cam- bridge, Mass. Processed. Friedman, M. 1957. A Theory of the Consumption Function. Princeton, N.J.: Princeton University Press. Fry, M. J. 1988. Money, Interest, and Banking in Economic Development. Baltimore, Md.: Johns Hopkins University Press. Gelb, A. 1988. Oil Windfalls: Blessing or Curse? New York: Oxford University Press. Gersovltz, M. 1988. "Saving and Development." In H. Chenery and T. N. Srinivasan, eds., Handbook of Development Economics, vol. 1. Amsterdam: Elsevier. Gersovitz, M., and C. H. Paxson. 1989. "The Economies of Africa and the Prices of Their Exports." Princeton University, Research Program In Development Studies. Princeton, N.J. Processed. Ghosh, S., C. L. Gilbert, and A. J. Hughes-Hallett. 1987. Stabilizing Speculative Cbm- modity Markets. Oxford: Clarendon. 94 Saving in Developing Countries Giovannini, A. 1983. "The Interest Rate Elasticity of Savings in Developing Countries: The Existing Evidence." World Development 11: 601-07. . 1985. "Saving and the Real Interest Rate in LDCS." lournal of Development Economics 18: 197-217. Gupta, K. L. 1987. "Aggregate Savings, Financial Intermediation, and Interest Rate." Review of Economics and Statistics 69: 303-11. Gustafson, R. L. 1958. Carryover Levels for Grains. U.S. Department of Agriculture, Technical Bulletin 1178. Washington, D.C.: U.S. Government Printing Office. Hall, R. E. 1988. "Intertemporal Substitution in Consumption." ]ournal of Political Economy 96: 339-57. Hall, R. E., and F. S. Mishkin. 1982. "The Sensitivity of Consumption to Transitory Income: Estimates from Panel Data on Households." Econometrica 50: 461-81. Hansen, L. P., and K. J. Singleton. 1982. "Generalized Instrumental Variables Estimation of Non-linear Rational Expectations Models." Econometrica 50: 1269-86. Hayashi, F. 1986. "Why Is Japan's Saving Rate So Apparently High?"NBER Macroeco- nomics Annual 1986: 147-210. . 1987. "Tests for Liquidity Constraints: A Critical Survey." In T. Bewley, ed., Advances in Econometrics Fifth World Congress, vol. 2. New York: Cambridge Uni- versity Press. Hong Kong, British Crown Colony of. 1980. Household Expenditure Survey. Hotz, V. J., F. Kydland, and G. Sedlacek. 1988. "Intertemporal Preferences and Labor Supply." Econometrica 56: 335-60. Kaldor, N. 1955-6. "Alternative Theories of Distribution." Review of EconomicStudies 23: 83-100. Kimball, M. S. 1988. "Precautionary Saving and the Marginal Propensity to Consume." University of Michigan. Ann Arbor, Mich. Processed. . Forthcoming. "Precautionary Saving in the Small and in the Large." Econo- metric~. Korea, Government of the Republic of. Various years. Annual Reports on the Family Income and Expenditure Survey. Seoul. Kotlikoff, L. J., and A. Spivak. 1981. "The Family as an Incomplete Annuities Market." ]ournal of Political Economy 89: 372-91. Labys, W., and C. W. J. Granger. 1970. Speculation, Hedging and Forecasts of Com- modity Prices. Lexington, Mass.: Heath. Leland, H. E. 1968. "Savings and Uncertainty: The Precautionary Demand for Saving." Quarterly lournal of Economics 82: 465-73. Lewis, W. A. 1954. "Economic Development with Unlimited Supplies of Labor." Man- chester School 22: 139-91. Lucas, R. E. 1988. "On the Mechanics of Economic Development."]ournal of Monetary Economics 22: 3 4 2 . McKinnon, R. I. 1973. Money and Capital in Economic Development. Washington, D.C.: Brookings. Mankiw, N. G., J. J. Rotemberg, and L. H. Summers. 1985. "Intertemporal Substitution in Macroeconomics." Quarterly Journal of Economics 100: 225-51. Mason, A. 1987. "National Savings Rates and Population Growth: A New Model and New Evidence." In D. G. Johnson and R. Lee, eds., Population Growth and Economic Development: Issues and Evidence. Madison: University of Wisconsin Press. . 1988. "Saving, Economic Growth, and Demographic Change." Population and Development Review 14: 113-44. Mirrlees, J. A. 1988. "Optimal Commodity Price Intervention." Nuffield College, Ox- ford. Processed. Muellbauer,J. 1982. "The Measurement of Long-Run LivingStandards: An Application and Evaluation of the Permanent Income Hypothesis." World Bank Living Standards Measurement Study. Washington, D.C. Processed. Musgrove, P. 1979. "Permanent Household Income and Consumption in Urban South America." American Economic Review 69: 355-68. 1980. "Income Distribution and the Aggregate Consumption Function."Journal of Political Economy 88: 504-25. National Academy of Sciences. 1986. Population Growth and Economic Development: Policy Questions. Washington, D.C.: National Academy Press. Newbery, D. and J. E. Stiglitz. 1981. The Theory of Commodity Price Stabilization: A Study in the Economics of Risk. Oxford: Oxford University Press. .1982. "Optimal Commodity Stockpiling Rules." Oxford Economic Papers 34: 403-27. Paxson, C. H. 1989. "Household Savings in Thailand: Responses to Income Shocks." Princeton University Research Program in Development Studies. Princeton, N.J. Pro- cessed. Rakshit, M. 1982. "Income, Saving, and Capital Formation in India: A Step Towards a Solution of the Saving-Investment Puzzle." Economic and Political Weekly 17 (an- nual). . 1983. "On Assessment and Interpretation of Savings-Investment Estimates in India." Economic and Political Weekly 18. Reserve Bank of India. 1982. Capital Formation and Saving in India 1950-51 to 1979- 80. Report of the Working Group on Savings. Bombay. Romer, P. M. 1986. "Increasing Returns and Long-Run Growth." Journal of Political Economy 94: 1002-37. Romer, P. M. 1988. "Endogenous Technical Change." University of Chicago and Uni- versity of Rochester. Processed. (May). . Forthcoming. "Capital Accumulation in the Theory of Long-Run Growth." In R.J. Barro, ed., Modern Business Cycle Theory. Cambridge, Mass.: Harvard University Press. Samuelson, P. A. 1971. "Stochastic Speculative Price." Proceedings of the National Academy of Sciences 68: 335-37. Schechtman, J. 1976. "An Income Fluctuation Problem." Journal of Economic Theory 12: 218-41. Schechtman, J., and V. Escudero. 1977. "Some Results on 'An Income Fluctuation Problem'." Journal of Economic Theory 16: 151-66. Sen, A. K. 1967. "Isolation, Assurance, and the Social Rate of Discount." Quarterly Journal of Economics 81: 112-24. Shaw, E. S. 1973. Financial Deepening in Economic Development. New York: Oxford University Press. 96 Saving in Developing Countries Sibley, D. S. 1975. "Permanent and Transitory Income Effects on a Model of Optimal Consumption with Wage Income Uncertainty." Journal of Economic Theory 11: 68- 82. Skinner, J. 1988. "Risky Income, Life-Cycle Consumption, and Precautionary Saving." Journal of Monetary Economics 25: 237-55. Solow, R. M. 1956. "A Contribution to the Theory of Economic Growth." Quarterly Journal of Economics 70: 65-94. Summers, L. H., and C. Carroll. 1989. "The Growth-Saving Nexus." Paper presented to National Bureau of Economic Research Conference on Savings, Maui, Hawaii, January. Summers, R., and A. Heston. 1988. "A New Set of International Comparisons of Real Product and Prices for 130 Countries, 1950-1985." Review of Income and Wealth 34: 1-26. Venti, S. F., and D. A. Wise. 1989. "The Saving Effect of Tax-Deferred Retirement Accounts: Evidence from SIPP." Paper presented to National Bureau of Economic Research conference on Savings, Maui, Hawaii, January. Visaria, P., and S. Pal. 1980. Poverty and Living Standards in Asia: An Overview of the Main Results and Lessons of Selected Household Surveys. World Bank Living Stan- dards Measurement Study Working Paper 2. Washington, D.C. Wolpin, K. I. 1982. "A New Test of the Permanent Income Hypothesis: The Impact of Weather on the Income and Consumption of Farm Households in India." International Economic Review 23: 583-94. World Bank. 1988. World Development Report 1988. New York: Oxford University Press. Zeldes, S. 1989a."Consumption and Liquidity Constraints: An Empirical Investigation." Journal of Political Economy 97: 305-46. . 198913. "Optimal Consumption with Stochastic Income: Deviations from Cer- tainty Equivalence." Quarterly Journal of Economics 104: 275-98. P R O C E E D I N G S O F T H E W O R L D B A N K A N N U A L C O N F E R E N C E O N D E V E L O P M E N T E C O N O M I C S 1 9 8 9 Fumio Hayashi Professor Deaton's paper covers a wide range of issues related to saving. Given my background and comparative advantage in empirical studies of saving, it is perhaps appropriate for me to focus on his theoretical and empirical discussions of household saving, the topics to which the author says future research efforts should be directed. The paper starts out by setting out a theoretical model of saving. In this model, the immortal family is the basic decision unit. The two crucial ingredients are, first, that the family is impatient and, second, that it lacks the option of bor- rowing. If borrowing were possible in this infinite-horizon model, it would be almost as good as having access to fair income insurance. With borrowing constraints, the impatient household elects to hold on to assets during high- income periods, not because it wants to shift resources from the present to the future for increased future consumption, but because it wants to insure itself against bad draws of income. Thus the household consumption/saving behavior has two regimes, as described in figure 1 of the paper: the zero-saving regime and the self-insurance regime. I should perhaps note that this type of consump- tion function has been derived (from different settings) by Helpman (1981). The author argues that the model's predictions are consistent with several pieces of empirical evidence, including the nonunitary elasticity of consumption with respect to permanent income, the prevalence of households with negative saving, and the "excess sensitivity" findings in the U.S. literature. The claim that the model does not deliver the standard result of unitary elasticity with respect to permanent income is somewhat misleading. Except for the certainty equivalence case, consumption cannot be related to a single-scalar indicator like permanent income. In the present model, consumption depends on two indicators (or, to be more precise, three, if the variance of income is included) of the household's resources, cash on hand, and mean earnings. As pointed out in the paper, the model does deliver the proportionality result that consumption is homogeneous in those multiple indicators. This homogeneity property of the model is not inconsistent with the empirical finding that rich households tend to save disproportionately more than poor households, if the Fumio Hayashi is a professor of economics at the University of Pennsylvania. O 1990 The International Bank for Reconstruction and Development / THE WORLD BANK. 98 Comment earning process over generations is regressive. In wealthier households, the off- spring are less likely to have as much earning capacity as the current adult generation. If so, the current high-earning generation will save a higher fraction of its wealth than the next generation. Let me also note that the case for the excess sensitivity for the United States is rather shaky. Recent work by Runkle (1989)using the Panel Study of Income Dynamics micro data (compiled by the Survey Research Center of the University of Michigan) finds no evidence for excess sensitivity. My impression is that for most East Asian countries, negative saving is rel- atively rare. For Japan and the Republic of Korea, it is difficult to find a cohort that dissaves (see Collins 1989; Hayashi, Ando, and Ferris 1988). The present model may be applicable to less developed countries, but not to East Asian developing countries. Furthermore, the model has an uneasy prediction that poor countries will be eternally poor. It also cannot explain why saving and growth rates differ between countries. This brings me to the issue addressed in section I1 of the paper: the inability of the alternative theory-the life-cycle hypothesis-to explain the international differences in the age profile of consumption. The paper reinforces the finding in Summers and Carroll (1989) that the international differences in the distri- bution of lifetime resources between generations brought about by the differences in growth rates are not reflected in the age profile of consumption. Although I agree that this finding probably is fundamentally inconsistent with the life-cycle hypothesis, my reaction to the evidence presented in the paper is a cautionary note that the age of the household head does not necessarily correspond to the generation or cohort of that age when the household consists of many genera- tions. To pursue the point, imagine that there are only two generations, young and old, and that the age-earnings profile is flat. Suppose all households are extended families, so that household consumption is the sum of consumption by the two generations. Let w, and w, be earnings of the young and the old living in the same household. Since the average age-earnings profile is flat, the mean of w, equals the mean of w,. Suppose that w, and w, are uncorrelated, so that both the young and the old in the same household have an equal chance to be the head, that is, the main income earner. Then it is easy to see that the age- consumption profile will also be flat irrespective of the growth rate of the economy. Another qualification I would like to make about the Summers-Carroll finding is that the relatively flat age-consumption profile for rapidly growing countries may be (partly) due to government income redistribution programs, such as social security, that shift resources from the young to the old. It must be granted, however, that under the life-cycle hypothesis, it is not clear why people have agreed to such income redistributions. What about the two pieces of evidence presented in the paper against the infinite-horizon model? One is the macroeconomic evidence that there is no international and time-series correlation between consumption growth and the Hayashi 99 real interest rate. The other is systematic variation of saving rates with the age of the household head. But the latter evidence seems to be consistent with the infinite-horizon models. Expenditures on, say, sporting goods depend on the age of the head. One wouldn't take this as evidence against the infinite-horizon models. By the same token, that the aggregate saving rate depends on the age distribution of people (not on resources) seems to be consistent with those models. And, on the lack of correlation between consumption growth and real interest rates, it is not clear that the deposit rate or even the interest rate on short-term government securities is the appropriate interest rate to use, because those rates may reflect transaction services. Collins, S. 1989. "Savings Behavior in Ten Developing Countries." Hamard University Economics Department. Processed. Hayashi, Fumio, A. Ando, and R. Ferris.1988. "Life Cycle and BequestSavings."Journal of the Japanese and International Economies 2, no. 4: 450-91. Helpman, Elhanan. 1981. "Optimal Spending and Money Holdings in the Presence of Liquidity Constraints." Econometrics 49, no. 6: 1559-70. Runkle, D. 1989. "Consumption and Liquidity Constraints." Research Department, Federal Reserve Bank of Minneapolis. Processed. Summers, L. H., and C. Carroll. 1989. "The Growth-Saving Nexus." Harvard University Economics Department. Processed. P R O C E E D I N G S O F T H E W O R L D B A N K A N N U A L C O N F E R E N C E O N D E V E L O P M E N T E C O N O M I C S 1 9 8 9 Steven B. Webb Angus Deaton's useful and thought-provoking discussion breaks new ground, as well as reviewing what others have done. He sets out a new and promising model for saving in developing countries. I will briefly comment on this pre- cautionary saving model and suggest how some features of it are also relevant to the more conventional issues raised in the latter parts of the analysis. Deaton's household model for explaining much of the saving by poor agri- cultural households is convincing: the constraint of no net borrowing strikes me as entirely plausible. Net borrowing is uncollateralized lending, which is like lending to sovereign nations-for which a vast literature now attests to the problems in motivating repayment. The model explains many of the stylized facts about the saving and asset- management behavior of common people in developing countries. It will be interesting to see the results of applying Dcaton's model to actual income esti- mates for commodity producers in developing countries, as he plans to do for C6te d'lvoire. Will plausible assumptions of parameter values yield simulations that show a time series of asset accumulation, decumulation, and famines con- sistent with other evidence for the economies examined? One nexus of stylized facts about saving that Deaton's model does not seem to address is the tenacity with which people save for and hold on to their own land and housing. Farmers also own their animals, trees, tools, and seeds. Im- plicitly, this production side of saving behavior is left out of his model. In the face of absolute starvation, people do of course sell off or consume these core assets. But the decision to do so seems a lot more drastic than running down a savings account or a precautionary stockpile of grain. Ownership of core assets, like land and housing, defines one's socioeconomic category. Changing one's wealth status within a category would seem to be much less serious and easier to reverse than changing from one category to another. Some economics is probably also involved in the distinction between holding core assets and other assets. Agency problems in the rental markets imply a higher rate of return on saving for ownership of core assets than on other assets, and higher than the rate of time preference. Saving to acquire core assets is another channel through Steven Webb is on the staff of the Country Economics Department of the World Bank. O 1990 The International Bank for Reconstruction and Development / THE WORLD BANK. 102 Comment which an increase in the number of households would lead to net saving in the aggregate. Deaton points out that some households seem to have a lower rate of time preference, below the interest rate, and therefore accumulate assets and become the capitalist accumulators of the society. If the rate of time preference declines as wealth and income increase, a class of capitalist savers could develop endog- enously from a population with identical basic utility functions. Households that start out saving only for precautionary motives (and for purchase of land and housingfor their own use)might get a sequenceof good draws at the income lottery, which would raise their assetsand current consumption to a point where their rate of time preference would be less than the interest rate. Then they would become capitalists with trend accumulation of assets. Of course, this could work the other way if the economy were hit by a series of negative shocks. More households would be falling out of the capitalist class than were being bumped up into it. This might explain some of the observed correlations between income growth and saving rates. Development involves the long-run expansion of urban sectors, at the expense of agriculture. It would be useful to consider the economics of saving during the process of the urbanization of the poor, who still have a high rate of time preference and need to smooth consumption. For instance, what is the economics of saving when one's income depends heavily on government spending, and when the main available asset for saving-money-is a prime target for taxation in fiscal emergencies that also depress government spending? The third and fourth sections of Deaton's paper mostly address conventional issues on their own terms. He rightly criticizes much of the literature for using inappropriate theories and methods, carried over from the study of industrial economies. Deaton's discussion of the issue of how the government should react to fluc- tuations of commodity export prices does draw on his model of precautionary savingat the household level. If householdsare better than governmentsat saving tosmooth the effectsof fluctuating export commodity prices, as Deaton suggests, then we should think further about how the World Bank's lending programs to commodity exporters might be designed to discourage procyclical government fiscal policy. In the discussionof life-cycle saving, Deaton effectively rebuts the notion that households in developing countries try to smooth and level consumption fully, and aim to run assets down to zero at death. But some of Deaton's model might usefully be carried over to the issueof life-cycle saving-carried over in the sense of using the nonnegative asset constraint and the high rate of time preference. One would need to drop or modify Deaton's assumption of extended-family households,in favor of nuclear households, which become moreprevalent during urbanization and industrialization. No net borrowing would mean that con- sumption by nuclear families (headed by workers) would almost equal average earnings in the parents' younger years, when the presence of children would Webb 103 increase the demand for consumption. They would do only precautionary saving, like Deaton's rural extended families. Nuclear households would start to save for future consumption on a life-cycle scale only after the children left home. Having a rate of time preference above the interest rate would mean that people would tilt consumption toward the present, but would still save some for old age, aiming to have a smooth, albeit declining, consumption path. Deaton notes that efforts to detect the influence of real interest rates on the saving rate have failed to come up with robust results. Strong risk aversion and the predominance of the precautionary motive for saving could help explain why the level of real interest rates is not monotonically related to saving rates. Risk-averse, precautionary savers like those in Deaton's households would worry at least as much about the variance of real interest rates as about the level. Any time the ex post real interest rate is (or has recently been) more than ten points below or above zero, people probably become uneasy about the real interest rate. Firms and wealthy individuals may happily and warily keep their money in the domestic financial system when real interest rates are 25 or 50 percent, because they have the means to move assets abroad quickly. But common people can move assets only slowly from a bank account into farm animals or consumer durables. They will be much more cautious in responding to changes in real interest rates. A real interest rate above 10 percent would probably not be sustainable for long enough to attract such savers into the financial system. Their saving in the informal sector would probably not be well measured. The World Bank tells people in developing countries that they should save in order to grow. And Bank programs try to tell people how to save in ways that will provide more resources for investment. Deaton is telling us that people in developing countries behave as if their primary motive for saving is to protect consumption against disasters. Perhaps the Bank should reprogram its priorities to match theirs. Suppose, for instance, that the first priority in reform of the financial sector were to provide people with more secure and convenient in- struments for precautionary saving. Would precautionary priorities lead to rec- ommendations different from when the priority is to promote investment or close the fiscal gap? Would the design of instruments to suit the desires of precautionary savers raise the amount of saving available for investment? How could the assets of precautionary savers be made available for the most pro- ductive investment while still respecting the priority of security? P R O C E E D I N G S O F T H E W O R L D B A N K A N N U A L C O N F E R E N C E O N D E V E L O P M E N T E C O N O M I C S 1 9 8 9 Deaton agreed partially with Mayashi's (discussant)commentsabout "dynasties" in his model, but he didn't want people to believe that his model was committed to a dynastic, infinitely lived family of consumers. That is a mathematical con- venience. These households often run out of assets altogether. The real horizon is the period until they run out of assets. For a shorter horizon, the behavior would look much the same. More serious for Deaton was the question of whether these people are "doomed"and whether we should feel sorry for them because they are caught forever in this liquidity trap. Deaton mentioned that this was not the case. First, they are doomed partly by assumption, because for simplicity's sake he had assumed that the income process is stationary, that the harvest never gets better, that incomes are constant-and of course if you assume no growth, you never get growth. You need more complicated models to handle that. Second, Deaton said that Webb's (discussant) point that the patient consumers in this model do accumulate wealth and get rich, so that wealth is being accumulated in these societies, was also important. It is just that a lot of the saving is being done by very few people, and most people are saving very little (anidea he still endorsed, though it had gone out of fashion). One participant observed that Deaton, like Irving Fisher, sees poorer house- holds as more impatient than the richer households-unlike Hirofumi Uzawa, who believes that impatience is a property of richer households. Deaton was happy to be associated with Fisher in this view. Commenting on the relation between uncertainty in income and saving, one participant said it is not enough to assume that the households are risk averse. You have to assume more about the nature of the risk aversion before you can say anything about how saving would respond to changes in uncertainty. Another participant congratulated Deaton for formally taking up the impli- cation of covariance of risks in rural areas for at least some kinds of rural behavior. He found the implications to be much broader-to include, for ex- ample, how land markets behave in such economies. Deaton agreed with the participant's observation that the reason this model works well in rural areas is that rural labor markets are tied to harvest outcomes, so when you have a bad harvest both your output and labor markets collapse. The model would not This session was chaired by Yung Chul Park, professor of economics, Korea University. O 1990 The International Bank for Reconstruction and Development / THE WORLD BANK. 106 Floor Discussion of Deaton Paper work well for urban areas, where you can simply make intersectoral switches if one sector has a worse year than another. One participant suspected Deaton's model was also unable to explain saving behavior in urban areas because it did not incorporate technological change in the income function. If family income increased over time because of techno- logical changes, saving behavior might change to serve both precautionary and investment purposes. Deaton's response was that the model could be extended- he was in fact extending it in another paper-but it would look very different when income processes were growing. Deaton's type of model is likely to miss what one participant saw as the most important form of household saving: own-account investment by a rural house- hold that is both a consuming and producing unit. Own-account investment by rural households-in everything from tree crops to buildings-is an important form of saving which is typically not measured at all. Deaton accepted this as a legitimate criticism of his paper, explaining that such saving did not easily fit into his framework. A rural household has both production and consumption sides, and his was a theory of smoothing consumption over time. He agreed that accumulation within the household firm was as important; however, the question remained: why not dump some firm assets when the consumption going gets tough? The answer might be, as Webb had stated earlier, that the conse- quences of doing so were much more serious than shedding your financial assets. Sellingland or capital stock, for instance, might be irreversible. Although Deaton had not tried to build a formal theory that would answer these objections, he agreed that it was important to try to do so. Moving from the household to the aggregate level, Webb suggested that if you are applying this model to agricultural households that save by, say, stock- piling grain, each household operates independently so you don't have to worry about how it all adds up. But if households are saving cash, then when they're on the low side they have to use the cash for purchases, so someone else out there must be holding the stockpiles of grain. Maybe that is the capitalist class that takes advantage of famines or almost-famines once households are holding financial rather than consumable assets-butthen there is an adding-up problem. In a nationwide drought do they import grain, or is someone in the domestic economy holding it?Deaton felt that some of the grain could come from outside and some from other people. He thought of such outcomes as being highly correlated within the village but not so highly correlated across the whole econ- omy. During the discussion, one participant said that he could see the asset-income ratio going down if financial institutions were to develop that allowed for a transfer of funds to less covariant regions. He wondered if Deaton thought the saving rate would also go down. Because the question could be interpreted in different ways, Deaton ended up providing two answers. First, under some circumstances development of the financial system, by increasing the rate of interest and increasing people's options Floor Discussion of Deaton Paper 107 across regions, would increase the saving ratio. This model doesn't produce a high saving ratio, he said, but a higher interest rate would. It is like a commodity model in which people are holding assets as buffers and if the higher interest rate means it is cheaper to hold the assets, people will buffer more. Second, and perhaps more directly addressing the participant's point, would it reduce saving if you opened banks that could invest in other parts of the economy, so that the banks could act as agents spreading risk across the economy?Deaton thought it would, although it would be hard to see them as banks, because only if they could develop instruments for explicitly insuring against bad crops-another issue altogether-would saving decrease and there be less incentive to hold precautionary assets. A participant asked Deaton what moves the aggregate saving rate. If you have a kink at 80 or 100 (see figure 1 in Deaton paper), and behavior to the left of it (essentially precautionary saving) accounts for more saving than behavior to the right of it, whatever determines that kink presumably has a tremendous impact on the overall saving rate. What in the real world corresponds to that point in Deaton's model? According to Deaton that point, or kink, is essentially the mean of the income process or close to it. The participant cited Stephen Zeldes as putting that level somewhere around a low U.S. income, but Deaton obviously puts it at a low "CGte d'Ivoire" income. The question is, what is moving it? Deaton clarified that below the kink people don't save at all and probably dissave, using everything they have to buy food and the like. Above that, they save some. In the aggregate, flashes of temporary income-a good harvest-will produce high aggregate saving and a bad harvest will produce dissaving or a low saving ratio. By itself this model is not suited to explain aggregate saving in countries where saving is positive. For that, said Deaton, you must turn to more traditional explanations such as patient consumers who behave more like standard life-cycle consumers. Another World Bank participant pursued the question, suggesting that if the interest rate affects saving at all, one would expect it would do so after the kink (infigure I). But since Deaton had seen no evidence that the interest rate affected saving, what else would explain the slope after the kink?The investment climate? Deaton said he had no answer to the question of whether there is some point in the income distribution at which saving behavior changes. His figure represents a single household, not aggregate saving, he explained. To the left of the kink, you have no saving at all; to the right of it, some income would be retained for precautionary saving. The individual is behaving in a precautionary manner at all points; there is 110distinction between precautionary and nonprecautionary saving. Everything about that curve-except of course the 45-degree line that stays fixed-is determined by everything else, including the interest rate and the kink point. If the implicit question is whether there is some break in income distribution, below which people don't save and above which they do, the answer might be yes and might be no, but his model could not answer that directly. Do households dissave or not?Deaton felt that data problems make it difficult 108 Floor Discussion of Deaton Paper to answer that question. The Korean evidence Hayashi had cited in his comments was data on Korean cities; household surveys show positive saving in cities in other countries as well, but Deaton's is a theory about agricultural households, and he thought a number of surveys show saving in rural areas to be unrea- sonably low or negative. Deaton mentioned that empirical work is being done on the model; the World Bank is funding a project to examine panel data from C6te d'Ivoire, which is now available for three successive years. The theme of inadequate data on saving came up repeatedly. One participant went so far as to say the data in the World Bank's World Development Report are often not worth the paper they are printed on. Half the tables in the World Development Report, he claimed, give data on income distribution, infant mor- tality, and population for countries for which in living memory there has been no census. Sometimes a bad number is worse than no number at all. He be- moaned the Bank's methodology for computing GDP rates and its reluctance to give ranges for estimates. A Bank participant agreed about the seriousness of the data problem, pointing out, for example, the divergence in saving rates between U.S. flow of funds and national income data. He believed the problem would require vast resources. It would not make sense to put ranges on the estimates; if the Bank didn't know the numbers, it certainly didn't know the ranges. At the outset, Park (chair) had observed that ten years ago a U.S. economist had written a paper about why Koreans save so little-and now people wonder why they save so much. Despite learning a lot from the discussion, Park felt he was still unable to explain why Koreans save too much, if they do. Now we had two competing models to explain saving behavior in developing countries. But we still have a long way to go to understand saving behavior, and there are many data problems. P R O C E E D I N G S O F T H E W O R L D B A N K A N N U A L C O N F E R E N C E O N D E V E L O P M E N T E C O N O M I C S 1 9 8 9 Social Sector Pricing Policy Revisited: A Survey of Some Recent Controversies Emmanuel Jimenez Some developing countries have begun to reevaluate the traditional policy of charging uniformly low (or zero) prices for such social services as education and health. But the reform of social service pricing has been controversial. Some disagreements are ideo- logically based and inherently irreconcilable. Others are about the empirical evidence concerning the impact of raising prices on efficiency and equity. Simulations of be- havioral models show that the price elasticity of demand is generally low and that prices can be raised from present levels without significantly affecting consumption. But this responsiveness varies by price level and income category, underlining the need to protect the very poor. There has been some debate about the sensitivity of the parameters estimated, and the technical, administrative, and economic costs of targetirrg have not been researched. What is needed is more empirical evidence that proposed measures to protect the poor are feasible as well as effective; estimates of the magnitude of the efficiency gain; and systematic evaluations of actual attempts at social sector price reform. Other questions to be considered in further research are: Have alternative forms of financing been adequately compared with price reform? Can and should there be operational rules of thumb about social sector pricing, comparable to the marginal cost approach to public sector pricing? How can noneconomic objectives be incor- porated into pricing policy? Traditionally, the public sector has dominated social services in most countries. Government provides most of these services free, or almost free, and has often restricted the fees private providers can charge users. Subsidized education and health services have dramatically improved basic indicators of well-being, such as literacy and child mortality rates, during the past three decades. But some governments in developing countries, with active support from the World Bank, are beginning to reevaluate the policy of uniform and heavy public subsidization, for three reasons: Emmanuel Jimenez is a senior economist in the Country Economics Department of The World Bank. He is grateful for the useful comments on an earlier version from Nancy Birdsall, Dennis de Tray, Stanley Fischer, Johannes Linn, Benno Ndulu, Javad Khalilzadeh-Shirazi, Nicholas Stern, and several colleagues from the Public Economics Division. John Brondolo and Fiona Mackintosh provided research and editorial assistance. 0 1990 The International Bank for Reconstruction and Development / THE WORLD BANK. 110 Social Sectov Pvicing Policy Revisited First, though more investment in education and health would still be socially profitable, the outlook is bleak for tapping central government resources, which are under more and more strain from recent macroeconomic setbacks. Additional sources of funding have to be found. Second, uniformly low prices for social services mean that expensive services are much more subsidized than cheaper ones. Yet the poorest people often have least access to these high-cost services for a variety of reasons, including the high private cost of consumption and built-in biases when services are rationed. Third, the limited resources devoted to social services are badly used. Too little goes to cheaper, more cost-effective alternatives-partly because there is no pricing mechanism to impose discipline either on users or providers. Out of these problems have come proposals to revamp the traditional pricing structures.' These proposals, and consequently this review, focus on publicly provided education and health services-sectors in which pricing policies have traditionally been ignored. The common link across their recommendations is the idea that policies that differentiate prices by type of service and by type of consumer are generally more efficient and equitable than low and uniform price policies (or slogans, such as free education and health for all). Though few question the diagnosis, many have debated the prescription that better pricing policy will improve the situation. This review addresses the con- troversies under three headings: ideological, empirical, and conceptual. Charging for social services is not a new idea. Both education and health services have, until this century, been provided privately on a fee-for-service basis in many countries (see Roth 1987). None of the papers reviewed here advocated a return to this system. Most analysts recognize that externalities, lack of information about benefits, equity considerations, and failure in com- plementary financial markets mean there is a continuing role for government. What Is the New View? The "new" idea is that governments should reexamine whether, given budg- etary allocations too low to finance socially productive spending in the social sectors, selectively increasing some user fees leads to greater efficiency and eq- ~ i t yThe focus has tended to be on second-best allocation. . ~ The basic premise is that investments in social services, like most other long- term investments, are particularly at risk during periods of adjustment. Yet in many countries where the needs are greatest, social sector subsidies (that is, 1. See, for example, World Bank (1986, 1987); Akin and Birdsall (1987);de Ferranti (198.5);Griffin (1988);Jimenez (1986b, 1987); and Psacharopoulos, Tan, and Jimenez (1989). 2. The word "new" is in quotation marks because the ideas are not original-they are applications of optimal tax models with nonresalable commodities (for theoretical expositions, see Besley 1989; Jimenez 1987; Heady 1989; Katz 1987; and Thobani 1983). public provision at a price less than social cost) are wasted. Revising pricing policy, while not a panacea, may be an effective antidote. The principal recommendations3 are to: Increase prices for some social services that have large private benefits and are consumed mostly by high-income people-to raise revenue, direct implicit subsidies away from the rich, and to assist in demand management. For many countries, these services include higher education and curative care in urban hospitals. Target subsidies toward poor household^.^ To protect the poor's consump- tion of services whose prices are increased, subsidies, in the form of fee ex- emptions as well as outright payments, should be better targeted-if possible, distributed on the basis of income or wealth criteria. Invest revenues in socially profitable services. In education, this often means expanding or improving primary education or expanding secondary and ter- tiary-level disciplines for which skills are in short supply; in health, primary health care, particularly in rural areas, is frequently a priority. Develop credit and insurance markets. In practice, the low level of develop- ment of these markets precludes raising prices, particularly for high-cost serv- ices such as higher education and in-patient hospital care. Expanding these financial markets also serves as an important allocative device because they allow mobilization of resources when payoffs are in the future (education) or involve the pooling of risk (health). Liberalize overly centralized provision. Alternatives to central government providers, such as local communities or private groups, should be allowed to compete with public providers. This will increase private financing as well as encourage supplier efficiency. Some Conflicting Interpretations The view that market-based incentives have a role to play in allocating social sector resources is not universal. Advocacy of price reform has been attacked for its "assumption of essentially selfish individual behavior" (see Gilson 1988, p. 16, on health) or for being "too short-sighted and economistic" (seeDempster 1987 on education.) Similar views are expressed by Bray (1986); Klees (1984); Abel-Smith (1987); Cornia, Jolly, and Stewart (1987); and various commen- tators in Pan American Health Organization (1988); and Neave (1988). The heat of the debate has fueled misunderstanding of the intent, content, and logic of the proposals. Three prevalent misconceptions are that the proposals advocate wholesale cuts in health and education, that they intend full cost recovery, and that they entail drastic downgrading of universities and hospitals. 3. These recommendations are summarized from the works cited in footnote 1. 4. Subsidy 1s defined as public expenditure less private payment. The public expenditure may accrue to an individual through a transfer of purchasing power (for example, a social security payment or a scholarship check) or through access to a public service at a price less than the cost of provision (for example, free health care at a public clinic). 112 Social Sector Pricing Policy Revisited Do the proposals advocate wholesale cuts in health and education? The pro- posed policy packages are clear that prices should be raised only selectively. Services with fewer "public good" characteristics, such as positive externalities, are first on the list for price increases. These tend to be the services consumed least by the poor. The proposals generally exclude charges for primary education and basic rural health care or some types of preventive care from the proposed increases-and in fact call for increased subsidies for these basic services. Only when it is clear that central governments are unable or unwilling to direct adequate resources to social priorities should prices for basic services be in- creased. Then, the money raised should be spent on basic supplies that are often the first to be cut when budgets are tight, such as school books or medicines. Experience has shown that many people, even the relatively poor, are willing to pay for a service if they know that their access or the quality will be conse- quently improved. Do the proposals advocate full cost recovery?' For university education or hospital care generally not; the recommendation generally differs in this respect from the traditional pricing suggestions for infrastructure sectors, like electricity, water, and housing (seeJulius and Alicbusan 1988 for a review of their imple- mentation record). Even services not characterized by severe externalities or not otherwise considered true public goods cannot necessarily be bound by standard public sector pricing rules because of imperfections in credit and insurance markets. The proponents of user charges are careful to make pricing recom- mendations contingent on the success of complementary markets. One possible source of confusion may be that the papers on pricing policies did not develop operationally oriented rules of thumb (for example, what percentage of costs to recover) to correspond with marginal cost pricing in other sectors. Are the policy packages meant to cut back university education and the pro- vision of hospital care? Concern over this issue has sometimes meant dismissal of the whole policy package. But, although the fee advocates do argue that some subsidies can be cut, they stipulate that this should be done only when present subsidies produce few benefits, consumption is not going to be severely affected, and the poor can be adequately protected. For example, in many countries there is currently excess demand for student places at the same time that graduates are finding it hard to get jobs. Moreover, current rationing schemes do not encourage use by the poor. Rather than indiscriminately dispensing allowances, low tuition, and grants across all activities, these and other subsidies should be directed toward high-return activities and toward the poor. Primary education is an obvious priority, but some activities in higher education research or even teaching may also have high returns. 5. Cost recovery is defined as the amount of public expenditure that is financed by user payments. It generallydoes not implysetting priceequivalent to marginal cost. (SeeJimenez1987 for a fuller discussion of average versus marginal cost pricing in this connection.) Jimenez 113 The empirical evidence that uniformly high subsidies across all social services meet neither equity nor efficiency objectives is fairly strong. But can user charges improve the situation? The principal empirical bones of contention are: the impact of user fees on the poor's access to public subsidies; the impact of user fees on efficiency and allocation; and the administrative and political feasibility of implementing major reforms. Do User Fees Impede the Poor's Access to Social Services? In many countries, poor consumers do not generally have good access to publicly subsidized social services, despite uniformly low fee levels. But would increased user fees help? Present subsidies are not directed toward the poor. Free or almost free pro- vision does not ensure that the poor will get more or even their proportionate share of the subsidies. Uniformly low prices mean that high-cost services are subsidized more heavily than low-cost services. The relatively rich consume more of the high-cost services, for several reasons. First, high subsidies are usually financed from general revenues. When budgets are constrained, services often have to be rationed. Unless carefully and delib- erately targeted, rationing often favors high-income categories. For example, rationing of school places, nominally by ability, may in fact favor richer house- holds which can afford to pay for tutoring or to repeat a grade.6The same may be true of rationing through providing the service only in urban areas. Population densities are high in urban areas and the services are cheaper to provide, but urban incomes are also likely to be higher than rural incomes. Second, even with a zero or very low price, social services are costly to use. The private cost includes payments for transport, materials (medicinesor school books), and opportunity cost. For education, the opportunity cost alone can amount to 25-50 percent of the total social (privateplus public)cost of education (Jimenez 1987, p. 19). Even though poor adults generally have a lower oppor- tunity cost of time than richer adults, this relationship may be reversed for children. Poor children tend to work at home and in the marketplace and may find it more difficult to attend school. Also, other costs (particularly transport) may be higher for the poor. Table 1 shows the distribution of education and health subsidies in several countries. Students from the highest quartile of the income distribution profile in Chile, Colombia, Indonesia, and Malaysia receive between 51 and 83 percent 6. Rationing through waiting for a service such as health care can favor the relatively poor if the price of a wait is higher to those with a higher opportunity cost of time. But the benefits from redistributing resources in this way (as opposed to the benefits of improved access but at a higher tax) may not exceed the costs if the government can redistribute income effectively by direct transfers that can be used to finance higher prices (see Bucovetsky 1984); other allocation mechanisms are available (Sah 1987); or the benefits are unevenly distributed among those who wait. 114 Social Sector Pricing Policy Revisited Table 1. Who Gets Social Sector Subsidies? -- Percentage of government subsidy received by income group Year of Lower Middle upper Country and sector survey 40 percent 40 percent 20 percent ~~~~~ All education Argentina 1983 48 35 17 Chile 1983 48 34 17 Colombia 1974 40 39 21 Costa Rica 1983 42 38 20 Dominican Republic 1976-77 24 43 14 Uruguay 1983 52 34 14 Indonesia 1978 46 25" 29" Malaysia 1974 41 41 18 Higher education Argentina 1983 17 45 38 Chile 1983 12 34 54 Colombia 1974 6 35 60 Costa Rica 1983 17 41 42 Dominican Republic 1976-77 2 22 76 Uruguay 1980 14 52 34 Indonesia 1978 7 10" 83 Malaysia 1974 10 38 51 Public health Argentina 1980 69 27 4 Colombia 1974 42 40 20 Costa Rica 1983 49 38 13 Chile 1983 51 47 11 Dominican Republic 1984 57 44 9 Uruguay 1983 64 25 12 Indonesia 1978 19 36" 45 Iran 1977 51 37 13 Malaysia 1974 47 37 17 Philippines 1975 27 33 40 Sri Lanka 1978 46 39 14 Hospitals Colombia 1974 23 53 23 Malaysia 1974 36 34 20 a. These figures are for the middle 30 percent. b. These figures are for the upper 30 percent. Sources: Jimenez (1987); Petrei (1987);World Bank (1986). of all public expenditures on higher education, whereas those from the lowest 40 percent receive between 6 and 15 percent. This effect is only partly coun- terbalanced by the concentration of primary education subsidies among poor families, which have most of a country's younger school-age children. The net result is a distribution of overall educational subsidies roughly proportional to each income group's population share, with the exception of the Dominican Republic, where the poor's share is still less. The income bias is less for health. Health subsidies for Colombia and Malaysia are roughly proportional to each income group's population share. But in In- donesia, the poorest 40 percent capture only about 19 percent from public health centers and hospitals. These figures need to be replicated for a wider group of countries and to be updated, particularly for African countries. But conditions are likely to be even worse for the poorest countries, where greater scarcity means greater rationing. For example, with lower primary and secondary enrollment rates than in other countries, African university systems have a much smaller applicant pool from which to draw. Poor income groups would have self-selected themselves out before applying. Subsidies may be targeted through differential pricing o f services. Can a more efficient pricing policy-one that generates enough revenue and encourages appropriate use-also ensure access for all? The popular method for ensuring access has been to provide all social services at a heavy per unit subsidy through uniformly low or zero prices, accompanied by quantity rationing when budgets are tight. This has not worked. The "new"view of social sector pricing advocates targeting primarily by differential pricing: increase prices for those services con- sumed by the rich, and increase subsidies (lower the private cost) for those consumed by the poor. The estimates of the potential effectiveness of this policy are often dramatic. For example, if those engaged in higher education paid its public cost, and the resources thus freed were used to expand the lower levels of education, the Gini coefficient measuring the distribution of public subsidies would move from 0.60 to 0.27 (Mingat and Tan 1985, p. 306). Similar distributional benefits are predicted for health services (World Bank 1987, p. 26). With a price-differentiating strategy, would the poor (or other vulnerable groups) have to be protected?Would the poor's use be affected by raising prices for higher levels of education and curative care? It depends, first, on whether the poor use any of the service whose price is to be raised. In some countries, access may already be so restricted that few if any poor people use the service; prices of zero may be insufficient to overcome other obstacles, such as high private costs or a bias in the rationing scheme. Some urban-based specialty hospitals, for instance, cater solely to diseases that afflict the relatively wealthy. In Africa the steep education pyramid means that many children have already been selected out at lower levels, so that by the tertiary level the elite form a majority of all applicants (World Bank 1987). In other countries, however, the poor may represent a minor but still significant portion of students, particularly at primary and secondary levels and for mod- erately priced services. For example, evidence indicates that they make substan- tial use of publicly provided health services (see table 1). A second determinant of the impact of differential pricing on the poor's use 116 Social Sector Pricing Policy Revisited of services is the set of behavioral parameters, such as the elasticity of demand. Studies on this topic differ in methodology and in measurement, particularly of the price variable. Nevertheless, there are some consistent findings. The demand for education and medical care is generally responsive to price changes. But the magnitude of this response varies, although most studies indicate that the elas- ticity, at current prices, is significantly less than one for almost all services (table 2). The implication is that a 1.0 percent increase in price, say, will amount to a decline in consumption of less than 1.0 percent, leading to a rise in revenue. Table 2. Price Elasticities of Demand for Health and Education Services Price elasticity by income group Year of Price range Lowest Highest data Country and service (US. dollars) Overall quartile quartile - - ~~~~~~ Health 1985 CBte d'lvoirea Clinic Free-$. 11 n.a. -0.61 -0.38 $.11-$.22 n.a. - 1.16 -0.05 Hospital Free-$. 11 n.a. -0.47 -0.29 $.11-$.22 n.a. -0.86 -0.51 1985 Perua Private doctor Free-$1.56 -0.14 -0.20 -0.06 $1.56-$3.12 -0.29 -0.44 -0.12 Hospital Free-$1.56 -0.41 -0.67 -0.33 $1.56-$3.12 - 0.64 -1.18 -0.05 Clinic Free-$1.56 - 0.46 -0.76 -0.03 $1.56-$3.12 - 0.68 - 1.28 -0.06 1975 Malaysia Outpatient visits n.a. -0.01 n.a. n.a. 1981 Philippines Prenatal child care n.a. - 0.01 n.a. n.a. 1980-81 Kenya $5.80 inelasticb n.a. n.a. 1985 Ethiopia n.a. Outpatient care -0.05--0.50 n.a. n.a. 1986 Sudan n.a. Outpatient services n.a. -0.37 n.a. n.a. Education 1985 Perua Secondary education Free-$1.56 -0.14 -0.18 -0.20 1982 Mali' n.a. - 0.98 n.a. n.a. 1983 Malawid n.a. - 0.52 n.a. n.a. ppppp- n.a. Not available. a. These are arc price elasticities, which is why price ranges are relevant. b. Magnitude could not be computed from available data. c. Uses distance as price variable. d. Uses household enrollment ratio as dependent variable. Sorrrces: Health in C6te d'lvoire and Peru: Gertler and van der Gaag (1988).Education in Peru: Gertler and Glewwe (1989). Philippines: Akin and others (1986). Kenya: Mwabu (1986). Ethiopia: Dunlop (1987).Mali: Birdsall and others (1983).Malawi: Tan, Lee, and Mingat (1984).Sudan: Schwabe (n.d.). Jimenez 117 Equally important for these point estimates, especially for large price changes, are assumptions about the structure and shape of the demand curve over the entire range of feasible consumption levels. For example, if the elasticity of demand increases as price rises or as consumption falls (asit would for a straight line demand curve), then the magnitude of the price response depends on where elasticity is computed along that curve.' Specification tests are also important, yet most analysts are content with imposing a predetermined structure on the curve (seeJimenez 1987 for a review). Moreover, demand elasticity may vary by income with important conse- quences for the consumption of the poor. A series of recent microeconometric studies (Gertler, Locay, and Sanderson 1987; Gertler and van der Gaag 1988; Gertler and Glewwe 1989) show that, in C8te d'Ivoire and Peru, the price elasticity of demand rises with falling income (table 2). Thus the consumption of poor people falls proportionately more than that of richer people for a given price increase. One piece of corroborative evidence from,experience is in Ghana, where urban-based health stations that raised fees saw an initial drop in at- tendance but recovered quickly, whereas attendance at small, rural-based health stations was much lower after two and a half years (Ennimayew, cited in Gertler and van der Gaag 1988). These research results support the view that special consideration ought to be given to poor people when prices are raised by relatively large increments (in this connection, see World Bank 1986, p. 24, on loan and credit schemes for education; and World Bank 1987, pp. 31-32, on community insurance and differential pricing for health).The critical question is whether schemes to protect the poor can be implemented effectively. Subsidies may be targeted by consumer characteristics. Governments have to evaluate fundamental tradeoffs in choosing: whether to target at all; to target through differential pricing (orper unit subsidies)of services; or to target through direct identification of consumer characteristics. Subsidy schemes that are less costly to administer, such as a general subsidy of commodities or services that the poor are likely to consume, may be more costly in terms of "leakage." Although it is possible to design self-targeted schemes in which poor income groups are given a greater incentive to participate than others, such schemes may be socially unacceptable. In the Philippines, for example, a plan to subsidize low-quality four-year public colleges, where poor students are more heavily represented, while further restricting access to the prestigious national university, ran into opposition because of the fear that it would institutionalize a two-tiered educational system (World Bank 1988b). Unfortunately, there are few systematic evaluations in developing countries of targeted health or education programs-countries are only beginning to im- 7. This point is controversial. Several authors (Dor, Gertler, and van der Gaag 1986; Gertler, Locay, and Sanderson 1987; and Gertler and van der Gaag 1988) argue that demand for health services is very responsive over large price increases. Akin and others (1986) argue that small changes in prices have no impact. Both sides can be technically right but have different policy recommendations. 118 Social Sector Pricing Policy Revisited plement such policies, and documenting such experiences is a priority task. However, preliminary reviews of analagous experiences in other sectors, such as food (Namor 1987),and from developed countries (Sawhill1988) show that promising targeting alternatives exist. Some important lessons emerging from this literature are: 1. Program size matters: the effectiveness and administrative cost of targeting by recognizable income group depends on the size of the group. This in turn is partly determined by the inequality of the income distribution. A highly skewed income distribution means that a large proportion of the population may be classified as "poor" and that, within that group, incomes are roughly similar. In this case, it may be easier to target by pricing specific services because the administrative costs of collecting information on a large proportion of the pop- ulation are large and the tastes of the targeted group are likely to vary from the untargeted group. The size of the targeted group also varies according to the program goals- how different members of the targeted group are to be treated; whether all members of the targeted group should be raised above a certain threshold, and so on (seeBesley and Kanbur 1988). Food subsidies are usually meant for broad groups: the bottom x-percentile of the income distribution, oral1 those who fail to meet a certain threshold nutrition level. However, targeted scholarship schemes in universities are for a much smaller pool-those who apply from among the graduates of secondary education. This number, and the consequent administrative cost, is generally small for the poorest developing countries. Such a scheme is being tried in the University of the Philippines in its "Socialized Tuition and Financial Assistance Program." Fees are being raised overall, but certain income categories are being exempted because a university study found that, despite very low tuition charges and merit scholarships, "the higher a student's annual family income, the greater are his chances of being admitted into the University" (University of the Philippines, n.d., p. 8). Program size and consequently the cost of targeting will change over time. It is essential to monitor the program's growth, and how this affects the ability of government to use means-testing devices to target effectively. 2. When income criteria are inappropriate, proxies can be used. Using income criteria has been cost-effective in countries where beneficiaries are reasonably well educated and are concentrated in urban areas, and where income taxes are successfully collected. For example, in the United States the administrative cost of most poverty programs ranges from 5 to 10 percent of program benefits. Overpayments to ineligible applicants were about 3 percent of program benefits for both Medicaid and supplemental social insurance (Mackintosh1989). There are few studies of the effectiveness and cost of using income criteria in other circumstances. The costs are bound to be higher. However, there are proven alternatives to using income criteria that approximate their benefits without incurring the costs. Jimenez 119 Studies of food subsidies have found that physical criteria, such as nutrition level, are generally more effective in reaching target populations and in keeping out others than income criteria (Timmons and others 1983; Edirisinghe 1987; Garcia and Pinstrup-Andersen 1987; Namor 1987). In education and health, these kinds of data would probably be costly to collect. Two criteria that have been applied in developing countries (see Griffin 1989, who mentions twelve countries; and Pfeffermann and Griffin 1989) are health status (or incidence of certain diseases) or adult literacy (children of illiterate parents are given full scholarship). A drawback of the health criterion would be that it would rule out subsidies for preventive care where broad population groups are at risk. Other proxies for income are already widely used in education and health. In Chile, education levels are used to identify potential program recipients (Chile 1988). But again these measures are generally effective only in middle-income countries that have adequate record keeping, relatively high literacy levels, and a high rate of participation in formal sector jobs. For others, coverage will likely be limited; Zaire, for instance, has a sliding fee scale for hospitals by broad occupational category, with students and the handicapped paying the lowest rates and enterprise employees the highest (Griffin 1988, p. 65). An alternative to targeting by individual or household characteristics is to target by geographic location. The simplest division, by rural and urban areas, is used in many developing countries in the distribution of food subsidies. For example, food subsidies distributed in ration shops have been judged to be more effective and less costly than vouchers in Mexico and Sri Lanka (Namor 1987; Edirisinghe 1987). The experience in health and education is more limited (see Griffin 1989), but the rule of exempting institutions from fees in urban slums or poor rural areas would probably be easy to implement. The one drawback is that, if the locations with different rates are close enough, people may seek the cheapest alternative. 3. Welfare stigma may be a cost-individuals do not like to be classified as needy and may even turn down some subsidies as a result. Recent evidence, however, indicates that in the United States, although the decision not to participate in a welfare program may be affected by stigma, the probability varies with the size of the potential benefit. Nor does the stigma seem to affect the amount of the benefit, once someone has decided to be on welfare (Moffit 1983). Moreover, the United Kingdom experience suggests that the importance of stigma may reflect prevailing political trends. In the 1940s "the strength of popular objection to any kind of means test" (LordBeveridge quoted in Ahmed 1989) pushed the United Kingdom from targeted programs to an untargeted social security system, but this trend was reversed in the 1980s with Prime Minister Margaret Thatcher's reform of major U.K. social programs. These issues can be studied in more depth for developing countries. 4. Incentive effects of subsidies may affect incidence and efficiency. The full economic costs of the alternative subsidy scheme mentioned above have yet to 120 Social Sector Pricing Policy Revisited be assessed. Subsidy schemes may exact incentive costs. Do those who do not initially belong in the targeted population distort their behavior so as to obtain the benefit? What disincentive effects are there for the potential beneficiaries? Another research question is: can the benefit be traded? Some benefits of targeted subsidized services, such as education or medical care, cannot generally be traded by poor beneficiaries. But medicine or school books can. It is probably impossible to prevent the sale and resale of education or health vouchers-such bartering has been evident in housing programs (Mayo and Gross 1987). Al- though such sales are not necessarily bad since targeted recipients receive the proceeds, questions must be raised about how the subsidies are shared. There may be more efficient means of distributing the subsidies. Can Increased User Fees Make People Better Off? The conceptual argument that selectively increasing user fees can improve welfare is compelling. The combination of tighter budgets and low or zero prices for publicly provided services has led to rationing or a deterioration in quality. Increased prices will generate more revenue. This will, in turn, improve welfare by financing the expansion of a previously rationed service, or an improvement in quality of a deteriorated service, or the expansion of another service with higher social returns (see Besley 1989b,Jimenez 1987, or Katz 1987 for a more detailed and rigorous treatment). In addition, increased fees may induce con- sumers to use the service more carefully, thus freeing up capacity for those who value the service the most. Are these conceptual arguments supported by facts? The empirical record is uneven. User fees, revenues, and service availability: actual experience. Few developing countries have dramatically changed their financing of social services in the past twenty years. Many newly independent countries in the 1950s and 1960s na- tionalized schools and health systems, but beforelafter comparisons have little value for policy judgments now because the preindependence systems were gen- erally so limited and colonial in nature. Large-scale subsidized expansions oc- curred when systems were small and budgets were not as tight. Reforms of social sector pricing structures beginning in some countries are mostly too recent to generate much information.' One country where evidence is beginning to be collected is Chile. During the early 1980s,Chile moved from a centrally directed social service system, heavily subsidized and untargeted, to a decentralized system more reliant for its finance on users and local communities. Preliminary evaluations are that this system has allowed Chile to maintain its impressive record in improving a variety of social sector indicators even while economic growth has lagged. The nutrition and housing programs in particular have been cited in recent World Bank reports 8. Between 1984 and 1988, thirteen education and six health projects have incorporated new financing ideas in World Bank operations. A list is available from the author. Jimenez 121 as possible models for other countries on how to target scarce spending. Eval- uations of the education reforms, which include selective scholarship schemes at universities, were not available for this analysis (see Castaneda 1986 and Chile 1988 for program descriptions). Chile's experience with health financing has been less encouraging. The 1981 reforms were meant to shift some of the burden from the general budget to user fees, employee payroll deductions, and the sale of vouchers to patients. Recent World Bank reports indicate that although central government budgetary al- locations have indeed fallen, the other financing sources have not filled the gap. The reasons appear to be that fee structures for vouchers and direct use have not been adjusted to keep up with rising costs; high-income employees have been opting out of the payroll-payment plan and transferring to private provid- ers; and "eligibility creep" has meant a tremendous expansion in the number of beneficiaries exempted from voucher or fee payments. This expansion has arisen from underreporting of income and ambiguities about the definition of what constitutes a household. Another way of drawing on experience is to make cross-section comparisons. Some countries rely more heavily on user financed social services than others. In the 1960s, the Tanzanian government reduced the share of public spending allocated to secondary schooling and, until recently, also restricted the estab- lishment of private and community schools. In contrast, the Kenyan government encouraged the growth of both public and self-financing private secondary ed- ucation. By 1980, secondary school enrollment ratios were six times greater in Kenya than in Tanzania. A World Bank study has concluded that the overly restrictive and rationed educational system in Tanzania contributed to its rel- atively poor economic performance (Knight and Sabot 1986). User fees, revenues, and service availability: simulations. In the absence of the data afforded by actual experience, a strong case for pricing policy reform has often been based on predictions of its probable impact on revenue mobili- zation, use, and even welfare. The impact of increased user fees on revenue depends on the price elasticity of demand for the service and whether, at current prices, there is excess demand for the service. On the evidence cited earlier-that, on average, the demand for many education and health services is relatively price inelastic-an increase in price will cause the quantity demanded to drop, but less than proporti~nat~el~. So revenue can be expected to rise, at least with modest price increases. If the choice is between a service that is low-priced but heavily rationed or low-quality and one that is inaccessible, evidence indicates that many households are willing to pay for the service, even in poor areas. Some innovative studies showing this result have been done in CGte dYIvoire and Peru, where high-quality household data are available (Gertler, Locay, and Sanderson 1987; Gertler and van der Gaag 1988; Gertler and Glewwe 1989). The methodology follows the logic of revealed preference. The estimated pa- rameters of a utility function are used to calculate the welfare of a hypothetical 222 Social Sector Pricing Policy Revisited nonuser, if that nonuser were then given the option of an increased access at increased prices. (Access is measured by travel time to the public facility.) The price level that makes the consumer indifferent between a low price-low acces- sibility scenario and a high price-high probability of access scenario is then compared with the marginal cost of providing the service. This price can be interpreted as the marginal "willingness to pay" for a service. The results for Peru and CGte d'Ivoire (table 3) indicate that, although the estimated willingness to pay is less than the estimated marginal cost price, the amount that can be recovered is still substantial. For example, in Peru, if the school is one hour away, the average student's willingness to pay for a place in a secondary school at his or her own village would be roughly equal to about half of the cost of providing that place (assuming constant marginal cost and a class size of about sixty students). The willingness to pay for education and health might be even greater if the users felt that the funds were going to be used to improve quality, such as in the provision of school materials and medicines. Birdsall and others (1983)found this to be true for households in Mali; Mwabu (1986)found similar results for Kenya. Fee-generated revenues can be used to make rationed services more readily available. For services with low costs and relatively large benefits, the impact can be dramatic. A report on financing education (World Bank 1986) estimated that merely eliminating living allowances to university students in several African countries could finance an 18 percent average expansion in the yearly primary education budget (table 4). In countries where higher education enrollment and Table 3. Individual Willingness to Pay for Access to Social Services and the Cost of Provision (1985 U.S. dollars) Willingness to pay for obtaining access to a nearby facility when the alternative is: Country and service One hour away Two hours away Marginal cost CBte d'lvoire West Forest Health Clinic" 0.10 Savannah Health Clinica 0.03 Peru Coastal Health Clinic" Sierra Health Clinica Secondary Schoolh a. Dollars per visit. b. Dollars per academic year. Sources: Health: Gertler and van der Gaag (1988, tables 5 and 6). Education: Gertler and Glewwe (1989). Jimenez 123 Table 4. Potential Impact on Primary Education in Africa of Increasing Fees in Higher Education (percent) - - Increase in primary education Potential primary budget if higher education students enrollment ratio bear entire cost of: Current with full cost Living Operating enrollment recovery in higher Countrv exDenses cost Both ratio educationa Benin Burkina Faso Central African Republic Congo C6te d'lvoire Malawi Mali Niger Senegal Sudan Tanzania Togo Average n.a. Not available. a. Full cost recovery refers to the elimination of university student allowances and introductiori of tuition payments to cover operating costs. Source:World Bank (1986, pp. 22, 23). unit subsidies are large, such as C6te d'Ivoire, Malawi, Senegal, Sudan, Tanzania, and Togo, the scope for expansion is even greater. The approach taken by the studies is very promising but has some significant limitations. First, the data requirements are substantial. The present studies, for example, have only a rudimentary data baseon the quality of the various schools. An integrated household-institution data base is required. Second, the revealed preference approach assumes that households, not in- stitutions, choose to obtain access. In many cases, a model in which institutions, as well as households, choose who obtains access is more appropriate. These choices are unlikely to be independent of one another. For example, students who believe they have very little chance of passing stringent entry requirements may be discouraged from even applying. Third, analysis of the supply side is simplistic. The impact on the provisil~n of social services depends on the cost elasticity of supply. Cost estimates are very rough, and marginal cost is assumed to equal average cost for the syste.m. The assumption is restrictive, but often necessary, because there are very few studies of the level and structure of provider cost in developing countries, either in education or health (see Jimenez 1986b for one such on education, Jimenez 124 Social Sector Pricing Policy Revisited 1987 and Vitialano 1987 for reviews of the sparse studies that exist). There is little consensus, for example, on whether or not educational and health insti- tutions currently operate at a size at which scale economies are still possible. The empirical work is complicated by: the need to see providers as multiproduct facilities in which some inputs can be used for a variety of outputs; the need to measure and to control for quality; the need to have specifications consistent with behavioral assumptions, such as cost-minimization; and the need to control for incentives by looking at private sector observations. Demand-side efficiency. Demand-side efficiency involves allocation of a fixed quantity among competing uses. Increased user fees will depress demand, more for those who value the service least, in this way releasing greater access to those who value the service more, for example, the rich. For those who do obtain access, the increased price will make them more wary of wasted consumption. But equally, increased user fees might discourage those who value the service but who are confronted by a liquidity constraint. This would be particularly true for expensive services, such as higher education and hospital care. Subsidized access may be more efficient with an appropriate rationing scheme, such as scholarships or queuing-but these too have their drawbacks, as discussed above. So the efficiency gain on the demand side is an empirical issue. It remains to be rigorously evaluated. Although most of the estimated models have found demand for many social services to be relatively price inelastic (see table 2), these models generally use, as the dependent variable, the decision to use a service at all or not. The decision of how much to consume may be more sensitive to price changes. Supply-side efficiency. Supply-side efficiencyinvolves the provision of a service at least cost. Providers who have to rely more heavily on fee revenue for their survival are more likely to be subject to pressures from their clients to be efficient. This might lower unit costs. This argument hinges on the effect of competitive forces in making providers accountable for their actions. Fees alone may not bring greater pressures from consumers; the users have to have alternative services to which they can choose to transfer. The limited evidence available indicates that institutions confronted by com- petition are more efficient. For example, some recent research shows that, at the secondary level, with background and selection held constant, students in private unsubsidized schools generally score better than their publicschool coun- terparts. At the same time, average (per student) costs in private schools are generally less (table5).There is also some evidence in the Philippines that, among public schools, those that rely more heavily on community contributions have lower costs, given average achievement levels within those schools (Jimenez, Paqueo, and Lourdes de Vera 1988). Cost estimates in these and other com- parisons need further refinement. Table 5. Efficiency of Private Unsubsidized Secondary Schools Relative to Subsidized Public Schools -- Achievement indicator: Corrected private score Private cost as Year of average score on as proportion of public proportion of Countrv data standardized test scorea bublic cost Colombia 1981 Mathematics and verbal 1.12 0.69 Philippines 1983 Mathematics 0.99 0.55 Verbal 1.20 Tanzania 1981 Mathematics and verbal 1.77 0.69 Thailand 1981-82 Mathematics 2.63 0.39 - - a. Proportional gain (loss)in achievement score if a randomly selected student, with the characteristics of the average public school student, attends private rather than public school, holding constant for that student's background. Source: Computed from Jimenez, Lockheed, and Paqueo (1988) and recent World Bank data. Are Increased User Fees Feasible Administratively and Politically? Many critics claim that the administrative and political costs of increasing user fees are prohibitive. But countries are already beginning to implement some of these policies. This suggests that, important as they are, these factors can be assimilated in the design of policy changes. Administrative and collection costs. The public administration of many de- veloping countries is weak: personnel are poorly paid, trained, and motivated. Will this lead to poor collection and administration of fees? The few case studies indicate that, with some exceptions, the administrative cost of implementing user fees is not excessive. For example, in Senegal, the direct cost of collecting fees for primary health care, which includes incentive payments to ticket clerks, amounted to less than 10 percent of collections in 1979. In larger hospitals, the administration netted about 75 percent of fees. In the Sudan, the annual cost of collecting user fees as a percentage of revenues for clinics, pharmacies, and hospital services amounted to 5 percent or less (see Griffin 1988, p. 78.) Many countries in eastern and southern Africa already collect fees or community levies to finance the building of public primary alnd secondary schools (World Bank 1988a). These costs are probably lower than the estimated collection and economic cost of the alternative method of gen- erating revenue-general taxes (World Bank 1988~).Precise estimates would be useful. One general lesson on collection is that, to give the institution an incentive to collect the fees and to limit the cost of handling the funds, the revenue shollld remain with that institution. In Senegal, for example, a fee structure, in place since 1968, was not enforced until 1982, when newly created hospitals were allowed to retain proceeds collected from nonindigent users (Vogel 1988, p. 166). Retaining revenue has the added advantage that users may be more will,,ng 126 Social Sector Pricing Policy Revisited to pay if they know how their funds are being spent, particularly when small amounts are frequently being collected for a variety of services and materials, such as drugs. In Ghana, for example, the government is linking increased university fees with a program to improve the availability of books and supplies (World Bank 1988~). Another important problem is the administration of effective educational credit and health insurance. The unit costs are particularly large for smaller schemes. In Latin America, the only region in the developing world where student loan schemes are widespread, administration amounted to 12-23 percent of total outlay, compared with 1.5-2.0 percent in larger schemes in developed countries (World Bank 1986, p. 28). But the financial viability of many student loan institutions is also at risk because of overly generous terms. Subsidized student loans may be necessary to make the schemes initially viable but, if the institutions are not allowed to be self-financing they become just another way to subsidize consumption-and a particularly inefficient way, as the eventual recipients are probably not those who should be targeted for such subsidies. If subsidies are necessary, they should be given directly. The same situation holds for health. Latin America is also the region in which social insurance, funded mostly by payroll taxes, has most been used to finance health care. The cost of these schemes is beginning to be a burden on governments, as their traditional sources of finance have either been poached by other sectors or have been insufficient to cover costs because no copayments are collected (World Bank 1987, pp. 36-7). In addition, coverage tends to be limited to relatively well-off workers in the formal sector (Zschock 1986). The economic and administrative issues are beyond the scope of the present paper. However, most of the proponents of the "new view" of social sector financing recognize that, until countries successfully implement credit and in- surance schemes, full cost recovery for higher education and hospital care, even for those who can afford it, cannot be implemented. But, in the meantime, this does not preclude some partial cost recovery scheme, along with more research on workable ways to introduce student loans and health insurance to the system. Political economy. When there are losers as well as winners from a policy change, even if the net social benefit is positive, problems can arise. Winners will not necessarily compensate losers, and losers, if they happen to be powerful, may block the policy change. The conceptual problems involved are discussed more fully in the next section. What Have We Learned and What More Do We Need to Know? Despite some obvious gaps, the empirical research shows solid support for the following conclusions: Free or heavily subsidized provision of social services has not led to adequate access for the poor. Subsidies tend to be available across the board for all types of social services, even when they are being rationed ineffectively and people are willing to pay for some of them. a Moderate increases in fees will raise revenue because, as most studies shosw, demand for many social services is relatively price inelastic. But these fees should not be set to recover full costs, particularly of expensive services snch as tertiary education and inpatient hospital care, because of: positive exter- nalities (for example, research at universities); the lack of adequate credit or insurance markets; little evidence on the cost-structure of facilities; and fees so low that cost recovery would require unsustainable price increases. a Demand elasticities are lower for the rich than for the poor. Thus, the poor have to be ~rotectedwhen prices are raised. a Practical schemes have been tried to target subsidies toward the poor. Their benefits often exceed the high costs of implementation. At the same time, much more work needs to be done. First on the research agenda are the following: a The robustness of the above results has to be tested. The distribution of social sector subsidies has yet to be calculated for many countries, particularly in Africa, along the lines of Meerman's (1979) study for Malaysia, Rodriguez- Grossi's (1985) study for Chile, and Selowsky's (1979) for Colombia. The innovative models of demand for social services should also be applied to other countries, and they should be extended to study the demand for the intensity of use, rather than just participation. a Little is known about the cost structure of social service provision. Most studies simply assume that marginal costs are constant. Cost functions that are consistent with economic theory should be estimated. The supply side should not be ignored because practical advice on how much fees should be raised depends partly on what determines cost: scale, scope, joint production, and so on. a Given that the ability to protect the poor is a central theme in the literature, social experiments on alternative means-testing schemes that apply specifically to education and health services should also be evaluated. Actual administra- tive costs should be estimated, as well as economic costs in terms of incentives. a More work on the economics of educational credit and health insurance has to be undertaken. Under what conditions can these institutions be made to work in developing countries? The literature has yet to address systematically, from the conceptual as well as the empirical standpoint, such topics as: ways other than price rises or general taxation of providing additional resources; and improving implementation through consideration of political constraints. 128 Social Sector Pricing Policy Revisited Increasing Budgetary Allocations Most of the recent literature on pricing education and health services takes a sectoral perspective: given tight budgets, how should these sectors cope? This view is undoubtedly valid for many countries. But a broader approach that considers both budgetary allocations and prices as flexible policy levers is some- times more appropriate. Even when the budget resources are scarce, governments may maximize social welfare by making resources available from other sectors. Or governments could try to alleviate tight budgetary constraints by raising more nonprice revenue without resorting to traditional general taxes. In either case, a more complicated analytical model must underlie the pricing arguments. Intersectoral subsidization. Even when the fiscal position is poor, funds can be made available for social services, either by cutting back subsidies destined for other sectors or, for those sectors that make net budgetary contributions, by charging more if demand is relatively price inelastic. The traditional way of choosing among spending alternatives would be to use social rates of return. The techniques can be adapted to evaluate recurrent spending, which is more important in social spending than, say, in infrastructure expenditures that are heavily weighted by new investments. This is sound advice in theory. In practice, although there are a growing number of studies that estimate the rate of return to different levels of education (see Psa~haro~oulos 1985 for a review), they are not available for many countries, are of variable quality, and generally ignore some high cost-services, such as alternative disci- plines in higher education. The difficulty of evaluating benefits in health services has forced analysts to turn to cost-effective analyses which do not allow inter- sectoral comparisons of spending options. These drawbacks, combined with the lingering view that "social" services cannot inherently be compared with "pro- ductive" services, have meant that rate-of-return comparisons are not used to guide spending decisions in the social sectors. Even among those sectors where such calculations are routine, the computation of social rates of return is an inexact ~cience.~Thus, how to provide sound, practical guidance on intersectoral spending choices remains an important item on the conceptual agenda. Cross-sectoral pricing can, in principle, be tackled using the modern theory of optimal taxation (see Atkinson and Stiglitz 1980; Newbery and Stern 1987; Besley 1989a).The problem would be to derive a set of optimal subsidies, which can be interpreted simply as negative taxes. Given costs, these would imply a set of optimal prices for all public services. These would presumably depend upon the relative elasticities of demand for the range of goods and services. If, for example, demand for transport services were less elastic than that for some health service, then cross-subsidies from transport to health might be in order. If work on the price elasticities of different public services continues to be fruitful, this could be a promising avenue for future research. 9. A recent internal report of the Bank reviewing public expenditure-public investment studies shows that most fail to derive a set of objective criteria to make investment decisions. Jimenez 129 Other means of raising revenue. Community financnng and the payroll tax have been prominent in discussion of ways of financing services other than general taxation or user fees. Many local communities may already be doing more than aggregate data indicate. For example, in education, communities have a heavy hand in funding capital and sometimes recurrent spending in eastern and southern Africa and in Asia (Bray and Lillis 1988). But the experience varies within and across countries. An important research topic would be to use this variation to explain the ability of some communities to finance and provide social services. Communities that finance and provide their own social services have incentives similar to those for individuals who pay for the use of services. They can provide the quality of service they prefer and have incentives to do so at least cost.1° But in societies that want to provide a homogeneous type of service, particularly in education, and to equalize opportunities across communities, the central gov- ernment must play a regulatory role. The tradeoffs in the balance between lacal and central authorities have yet to be fully addressed (see Winkler 1989 for a review of the education cases). Payroll taxes are already being used to finance many health programs in Latin America through the social security system, and to finance training programs. Should they be more widely used? The answer probably varies from country to country. Poor countries with small urban labor forces may find it difficult to levy payroll taxes, whereas low-middle-income countries find them attractive because they are relatively easy to collect and rates can be targeted by sector. Their efficiency effects depend on coverage of the labor force and on the incidence of the tax. Most studies of developed countries assume that employers are able to pass on the burden of the tax's effects to labor (Whalley and Ziderman 1989). Whether this is likely to hold for developing countries remains a research ques- tion. Most evaluations of health insurance schemes in Latin America also find that the tax does not obviate the need for pricing through copayments and other means to raise revenue and gain efficiency (Zschock 1986). Other schemes include poll taxes (leviedon graduates) and schemes to generate income by having students work. Data on the impacts on efficiency or equity of these schemes or whether they are generally effective are limited at present. Making Service Provision More Efficient The key question in making service provision more efficient is how to lower unit costs without sacrificing quality, particularly for the most expensive services such as universities and big-city hospitals. One way is to improve input mixes, technology, managerial procedures, and technical skills among service providers in developing countries. This work is already the focus of much foreign assistance to developing countries. But the expertise, the new technology, and the skills 10. A study in the Philippines shows that schools that rely more heavily on community-level financing have lower costs than those that rely more on central government financing (Jimenez and others 1938). 130 Social Sector Pricing Policy Revisited will not solve efficiency problems if providers are not also given appropriate incentives. Do we really know what input mixes are more efficient? Production-function studies of education and health services have produced conflicting results (see Hanushek 1986 for a review of the education production-function literature in developed countries, and Fuller 1987 for developing countries.) Researchers have begun to see that the parameter values and ensuing policy recommendations are likely to differ for countries that have differentcharacteristics (Fuller 1987). Even with trustworthy estimates of the effectiveness of various public inter- ventions, costs also have to be evaluated. Lack of data for solid empirical analysis partly accounts for the paucity of research on these costs, but part of the problem is conceptual: estimating a cost function means assuming that minimizing costs is the goal. If public providers are guided by other motives, those other motives would have to be modeled. Otherwise, the analysis would have to be conducted on a sample of private (and presumably competitive) providers. If the providers are heavily regulated or restricted to serve only a narrow market, these com- plications would also have to be addressed. Incentives to be efficient. Showing educational and medical personnel more effective ways of operating may not improve efficiency unless incentives are also changed. In most instances competitive pressures to produce the required high- quality services at least cost simply do not exist. More work needs to be done on finding out which incentives work. A greater reliance on self-finance through pricing may improve efficiency if administrators are held accountable for their actions. But in centralized systems, accountability is not transparent. And when consumers are uninformed about benefits and do not have to pay each time they use a service, they have an incentive to go along with expensive interventions prescribed by the more knowl- edgeable provider. In such instances, decentralization may be appropriate. But the proper balance of responsibility between traditional central authorities and decentralized institutions is difficult to determine. Models of Political Economy Even those governments in favor of pricing reform may be deterred from taking the necessary steps. They may simply not have the necessary information or skills to make the necessary budgetary allocations and intersectoral tradeoffs. Alternatively, those in government may be acting not in the public interest but in their own interest-they may not want to make the necessary policy correc- tions if they themselves stand to lose by them (see Mills 1986; Birdsall 1988). Another possibility is that some governments may be unable to withstand pres- sure from powerful minority interests. This is what Birdsall (1988) calls "the Robin Hood failurem-taking from the poor and giving to the rich. Such pressures might explain the Chilean government's reluctance to adjust the fee structure in their health reform program. They could also account for the findings of Mingat and Tan (1986)that white collar workers in francophone Africa (6 percent of the population) appropriate over a third of all education subsidies, while farmers (three-quarters of the population) get less than half- a discrepancy due primarily to the heavy subsidy of tertiary education. The white collar workers live in urban areas, and many of them are in government. So their influence on government is greater than their proportionate numbers. The influence of pressure groups may also partly explain the systematic un- derfunding of certain services. For example, social and political pressures have made schools more accessible, but often at the expense of quality. Quality is particularly easy to neglect because it is difficult to detect small changes and citizens have little information about it. For similar reasons, policymakers may not only resist reforms but may tatke measures to counteract reforms already under way. For example, user fee rev- enues will not be of much use to social services if they are outweighed by c~uts in budgetary allocations. More work is thus needed on countervailing measures. Two possibilities are to identify and compensate losers, and to replace govern- ment discretionary authority with rules. Compensating the losers and mobilizing the winners. At the very least, when social sector reforms are proposed, potential winners and losers have to be identified by documenting who benefits from social sector subsidies. A more complex incidence analysis of the proposed program would be based on this information (as yet available for only a few countries; see table I).Who is likely to benefit? Can benefits be passed on to others in the free market? The next step is to assess whether and how to compensate the losers. Losers could be directly compensated or even protected by "grandfathering" a fee increase. Or the support of the winners could be mobilized. If winners can "outmu~cle"the losers, then compensation may not be needed. For example, in support of extensive educational reforms which will hit subsidized tertiary levels of education hardest, the government of Ghana is mounting a massive public education campaign to alert the population to the inequities of the present distribution of subsidies (World Bank 1988~).This is being accompanied by a media campaign which stresses that the funds now spent on boarding privileged students could provide many more places for others. Public opinion matters, even in countries that do not have Western-style democracies. Finally the complicated issue of designing and implementing compensation systems must be tackled. Even if it were possible to devise a compensation scheime that would make everyone better off, the winners and losers might not be able to agree on the amount of compensation. This problem is addressed conceptually by the vast literature on the design of incentive compatible mechanisms (see Groves and Ledyard 1985). Rules versus discretion. Political pressure works best on "discretionary" spending items. It is often harder to change laws, rules, or institutions. When the portion allocated to nondiscretionary spending-for instance, interest pay- ments on foreign debt-has to increase, pressure on the discretionary portlon correspondingly heightens. For developing countries (excluding highly indebted 132 Social Sector Pricing Policy Revisited ones) interest payments on foreign debt have been increasing from an average of about 1 percent of gross domestic product in 1975 to 3 percent in 1985. For the seventeen highly indebted countries (seeWorld Bank 1988, p. 11)this average percentage has risen from about 1 to over 8 percent over the same time period. Discretionary budget items that have few political backers, such as spending on nonsalary operations and maintenance of schools and health facilities, are most at risk. When discretionary public policy does not work to improve social welfare, rules and institutions may. Earmarking-assigning revenues from specific sources, by statute or by constitutional fiat, to finance some areas of government activity-may be an effective way of ensuring that some funds continue to go to some specified services and that fee revenue stays with the collecting insti- tution. But the net effect on the collecting institution could be ambiguous. If - earmarked revenues must be supplemented by some significant amounts from discretionary funds, governments could still manipulate those funds to coun- teract the nondiscretionary portion. Earmarking also incurs costs in terms of the lack of budgetary control and inflexibility of budgets-costs that fall most heavily on governments that are able, for the most part, to behave in the public interest. But this mechanism may be fruitful for tackling the problem of regimes overly influenced by minority private interest groups, particularly if the incidence of the tax falls heavily on the beneficiaries of the expenditures for which revenues are earmarked (Mc- Cleary 1988). other "rules" can be equally effective in reducing the discretionary power of governments that do not act in the public interest-for instance freezing ex- penditures on such items of chronic overspending as expensive hospital care or university teaching. But all such action is likely to have only short-term or medium-term effects. In the long run, the laws, rules, or constitutions themselves may be subject to the same pressures that confront discretionary budgets. The Agenda for Conceptual Research The literature to date has assumed a fixed budgetary allocation to the social sector; has focused on user fees as a solution to the financing, efficiency, and equity problems; and has used a "public interest" welfare economics model of analysis. This section has discussed how these assumptions may be relaxed in order to address broader questions. They could be ordered in ascending order of difficulty, as follows: increasing efficiency as an alternative to raising prices; accounting for flexible budgetary allocations in the model; and taking the "pri- vate interest" view into account. One area of work, not purely research, deserves some mention. The propo- nents of social sector pricing have so far been writing sweeping policy documents. Their objective has been to change a way of thinking, away from social services as entitlements and toward reaching social objectives with efficient policy in- struments. They have not tried to furnish detailed guidelines on what practitio- Jimenez 133 ners should be doing in the field. Although it is probably not feasible or even perhaps useful to codify guidelines for the education and health sectors (see Julius and Alicbusan 1988 for an account of how public pricing rules have been implemented in other sectors) it may be time for researchers to address mlore detailed issues, specifically: Under what conditions can a specific price level be computed and recommended (most recommendations are on the direction and not the magnitude of change)?How can different types of prices (for example, copayments versus premiums in insurance or registration versus tuition fees in education) relate to one another? IV. CONCL SIONS U Recent proposals that traditional low- and uniform-price policies for education and health services be replaced by policies that differentiate prices by type of service and by type of consumer have generated considerable debate. Some of the disagreements are ideological and have led to misunderstandings about the extent and scope of cost recovery being proposed. A substantial part of the debate has also focused on the empirical evidence. There is strong evidence that free or highly subsidized social services have not led to adequate access for the poor. But because countries are only beginning the process of reform, rigorous evaluations of the experience of policy reform have yet to be made. The evidence that selective user charges will improve efficiency and equity is partly based on simulations-the price elasticity of ide- mand is generally low and prices can be raised without significantly affecting consumption. The responsiveness does vary by price level and income category, underlining the need to couple price increases with measures to protect the poorest income groups. The critical missing link in such research is assessmlent of the technical, administrative, and economic costs of targeting education and health subsidies in developing countries. Lessons can be learned from other sectors and developed countries on questions such as the costs of means-testing and the alternatives to using income as a criterion for targeting. Research should also look at the unresolved conceptual problems. These in- clude seeking ways other than price reform to direct funds to the social services (such as reallocating funds among sectors and raising revenues through means other than pricing) and ways to increase the efficiency of current provision, incorporating incentive effects. And the broader perspective encompassing po- litical-economy arguments, particularly the "private interest" view of govern- ment, might offer ways to ensure that policy reform is undertaken and sustainied. Finally, more work should be devoted to the details of price reform. Most of the debate has been about generalities-the direction, rather than the magnitude, of change-and rightly so, for the tradition of free or highly subsidized provision has been so firmly entrenched. Detailed work is now needed to inform opera- tional activities that are trying to put these ideas into practice. Such policy work 134 Social Sector Pricing Policy Revisited should distinguish clearly among groups of countries with different character- istics and tailor the recommendations accordingly. In sum, there is much merit in the proposals to increase user charges selectively. Aside from ideologically based challenges, the conceptual framework has not been undermined. There is also strong empirical evidence to support many of the arguments, particularly the present distribution of public subsidies and the raising of revenues through pricing reforms for high-cost social services. But there are still some weak links. Research is needed on the feasibilityof measures to protect the poor, on estimates of the magnitude of the efficiency gain, and on the cost structure of service provision. More systematic evaluations of actual attempts at social sector price reform are also crucial. Unless these links are forged soon, the momentum for price reform in the social sectors may falter. Abel-Smith, Brian. 1987. "Book Review" of Financing Health Services in Developing Countries: An Agenda for Reform [World Bank 19871. Health Policy and Planning 2, no. 4: 355-58. Ahmed, Ehtisham. 1989. "Social Security and Poverty Alleviation: Issues for Developing Countries." World Bank Country Economics Department. Washington, D.C. Proc- essed. Ainsworth, Martha. 1984. "User Chargesfor Cost Recovery in theSocialSectors: Current Practices." World Bank Country Policy Department Discussion Paper 1984-6. Wash- ington, D.C. Processed. Akin, John C., and Nancy Birdsall. 1987. "Financing Health Services in LDCS." Finance and Development (June): 40-43. Akin, John C., Charles Griffin, David K. Guilkey, and Barry M. Popkin. 1986. "The Demand for Primary Health Care Services in the Bicol Region of the Philippines." Economic Development and Cultural Change 34: 755-82. Atkinson, A. B., and Joseph E. Stiglitz. 1980. Lectures on Public Economics. London: McGraw-Hill. Besley, Timothy. 1989a."Reforming Public Expenditures: Some Methodological Issues." World Bank Country Economics Department. Washington, D.C. Processed. . 1989b."Welfare-ImprovingUser Chargesfor PubliclyProvided PrivateGoods." Woodrow Wilson School of Public and International Affairs, Princeton University. Processed. Besley, Timothy, and Ravi Kanbur. 1988. "Food Subsidies and Poverty Alleviation." Economic lournal 98 (September): 701-19. Birdsall, Nancy. 1988. "Pragmatism, Robin Hood, and Other Themes: Good Govern- ment and Social Well-Being in Developing Countries." World Bank and Rockefeller Foundation. Washington, D.C. Processed. Birdsall, Nancy, Fransois Orivel, Martha Ainsworth, and Punam Chuhan. 1983. "Three Studies on Cost Recovery in Social Projects." World Bank Country Policy Department Discussion Paper 1983-8. Washington, D.C. Processed. Bray, Mark. 1986. "Book Review" of Financing Education in Developing Countries: An Exploration of Policy Options [World Bank 19861. International]ournal of Ed- ucational Development 7, no. 2: 147-48. Bray, Mark, and Kevin Lillis, eds. 1988. Community Financing of Education: Issues and Policy Implications in Less Developed Countries. Oxford: Pergamon Press. Bucovetsky, Sam. 1984. "On the Use of Distributional Waits." Canadian Journal of Economics 11, no. 4: 699-717. Castaneda, Tarsicio. 1986. "Innovations in the Financing of Education: The Case of Chile." World Bank Education and Training Department Discussion Paper 35. Wash- ington, D.C. Processed. Chile, Government of. 1988. "Evolucion de la Extrema Pobreza en Chile." Santiago. Cornia, G~ovanniA., Richard Jolly, and Frances Stewart. 1987. Adjustment with a Human Face. Oxford: Clarendon Press. de Ferranti, David. 1985. Paying for Health Services in Developing Countries: An Overview. World Bank Staff Working Paper 721. Washington, D.C. Dempster, Carolyn. 1987. "African Heads Attack World Bank Plan." The Times (Lon- don) Higher Education Supplement. April 24. Dor, Avi, Paul Gertler, and Jacques van der Gaag. 1987. "Non-price Rationing and Medical Care Provides Choice in Rural CGte d'Ivoire." Journal of Health Economics 6: 291-304. Dunlop, David. 1987. "A Study of Health Financing: Issues and Options, Ethiopia." World Bank Sector Review. Washington, D.C. Processed. Edirisinghe, Neville. 1987. "The Food Stamp Program in Sri Lanka: Costs, Benefits, and Policy Options." International Food Policy Research Institute Research Report 58. Washington, D.C. Fuller, Bruce. 1987. Raising School Quality in Developing Countries. World Bank Dis- cussion Paper 2. Washington, D.C. Garcia, Marito, and Per Pinstrup-Andersen. 1987. "The Pilot Food Price Subsidy Scheme in the Philippines: Its Impact on Income, Food Consumption, and Nutrition Status." International Food Policy Research Institute Research Report 61. Washington, D.C. Processed. Gertler, Paul, and Paul Glewwe. 1989. The Willingness to Pay for Education in Devel- oping Countries. Living Standards Measurement Study Working Paper 54. Washing- ton, D.C.: World Bank. Gertler, Paul, Luis Locay, and Warren Sanderson. 1987. "Are User Fees Regressive?: The Welfare Implications of Health Care Financing Proposals In Peru." Journal' of Econometrics 33: 67-88. Gertler, Paul, and Jacques van der Gaag. 1988. "The Willingness to Pay for Med~cal Care: Evidence from Two DevelopingCountries." World Bank Population and Human Resources Department. Washington, D.C. Processed. Gilson, Lucy. 1988. Government Health Care Charges: Is Equity Being Abandoned?- A Discussion Paper. EPC Publication 15. London: School of Hygiene and Trop~cal Medicine. Griffin, Charles C. 1988. User Charges for Health Care in Principle and Practice. Eco- nomic Development Institute Seminar Paper 37. Washington, D.C.: World Bank. .1989. "Means Testing in Developing Countries." Eugene: Department of Eco- nomics, University of Oregon. Processed. Groves, T., and Ledyard, J. 0. 1985. "Incentive Compatibility Ten Years Later." De- 136 Social Sector Pricing Policy Revisited partment of Economics, Northwestern University Discussion Paper 648. Evanston, Ill. Hanushek, Eric A.. 1986. "The Economics of Schooling: Production and Efficiency in the Public Schools." Journal of Economic Literature 24, no. 3: 1141-77. Heady, Christopher. 1989. "Public Sector Pricing in a Fiscal Context." World Bank Policy, Planning, and Research Working Paper 179. Washington, D.C. Processed. Jimenez, Emmanuel. 1986a. "The Public Subsidization of Education and Health in De- veloping Countries: The Impact of Equity and Efficiency." World Bank Research Observer 1, no. 1: 111-29. . 1986b."The Structure of Educational Costs: Multiproduct Cost Functions for Primary and Secondary Schools in Latin America." Economics of Education Review 5: 25-40. . 1987. Pricing Policy in the Social Sectors: Cost Recovery for Education and Health in Developing Countries. Baltimore, Md.: Johns Hopkins University Press. Jimenez, Emmanuel, Marlaine Lockheed, and Vicente Paqueo. 1988. "The Relative Efficiencyof Public and Private Schoolsin Developing Countries." World Bank Policy, Planning, and Research Working Paper 72. Washington D.C. Processed. Jimenez, Emmanuel, Vicente Paqueo, and M. Lourdes de Vera. 1988. "Does Local Financing Make Primary Schools More Efficient?:The Philippine Case." World Bank Policy, Planning, and Research Working Paper 69. Washington D.C. Processed. Julius, DeAnne, and A. P. Alicbusan. 1988. "Public Sector Pricing Policies: A Review of Bank Policy and Practice." World Bank Policy, Planning, and Research Working Paper 49. Washington, D.C. Processed. Katz, Michael. 1987. "Pricing Publicly Supplied Goods and Services."In David Newbery and Nicholas Stern, eds., The Theory of Taxation for Developing Countries. New York: Oxford University Press. Klees, Steven J. 1984. "The Need for a Political Economy of Educational Finance: A Response to Thobani." Comparative Education Review 28, no. 3: 424-40. Knight, John B., and Richard H. Sabot. 1986. "Overview of Educational Expansion, Productivity, and Inequality: A Comparative Analysis of the East African Natural Experiment." World Bank Education and Training Department Discussion Paper 48. Washington, D.C. Processed. Mackintosh, Fiona, J. 1989. "Notes on Means-Tested Poverty Programs in the U.S." World Bank Country Economics Department. Washington, D.C. Processed. Mayo, Stephen K., and David J. Gross. 1987. "Sites and Services-and Subsidies: The Economics of Low-Cost Housing in Developing Countries." World Bank Economic Review 1, no. 2: 301-35. McCleary, William A. 1988. "Notes on the Principles and Practice of Earmarking." World Bank Country Economics Department. Washington, D.C. Processed. Meerman, Jacob. 1979. Public Expenditure in Malaysia: Who Benefits and Why. New York: Oxford University Press. Mills, Edwin S. 1986. The Burdens of Government. Stanford, Calif.: Hoover Institution Press. Mingat, Alain, and Jee-Peng Tan. 1985. "On Equity in Education Again: An International Comparison." Journal of Human Resources 20, no. 2: 298-308. . 1986. "Who Profits from the Public Funding of Education? A Comparison by World Regions." Comparative Education Review 30: 260-70. Jirnenez 137 Moffit, Robert. 1983. "An Economic Model of Welfare Stigma." American Econo~nic Review 73, no. 5 (December): 1023-35. Mwabu, Gerrnano. 1986. "Health Care Financing in Kenya: A Simulation of Welfare Effects of User Fees." Social Science and Medicine 22, no. 7: 763-67. Namor, Eugenio. 1987. Issues in the Targeting of Food Subsidies for the Poor: A Survey of the Literature. International Monetary Fund Working Paper WP/87/75. Washing- ton, D.C. Neave, Guy, ed. 1988. "Higher Education and Development: A Reappraisal. Higher Education Policy (special issue)1, no. 1 (March): 7-36. Newbery, David, and Nicholas Stern. 1987. The Theory of Taxation in Developing Countries. New York: Oxford University Press. Pan American Health Organization. 1988. "Financing Health Services in Developing Countries: An Agenda for Reform." Bulletin 22, no. 4. Petrei, A. Humberto. 1987. El Gasto Publico Social y sus Efectos Distributivos. Rio de Janeiro: Estudios Conjuntos de Integracao Economia da America Latina. Pfeffermann, Guy, and Charles C. Griffin. 1989. Nutrition and Health Programs in L,atin America: Targeting Social Expenditures. Washington, D.C., and Panama City: World Bank and International Center for Economic Growth. Psacharopoulos, George. 1985. "Returns to Education: A Further International Update and Implications." Journal of Human Resources (Fall):583-604. Psacharopoulos, George, Jee-Peng Tan, and Emmanuel Jimenez. 1989. "Financing Ed- ucation in Developing Countries." IDS Bulletin 20, no. 1 (January):59-71. Psacharopoulos, George, and Maureen Woodhall. 1985. Education for Development: An Analysis of Investment Choices. New York: Oxford University Press. Rodriguez-Grossi, Jorge. 1985. La Distribuci6n del Ingreso y el Gasto Social en Chile, 1983. Santiago: Instituto Latinoamtricano de Estudios Sociales. Roth, Gabriel. 1987. The Private Provision of Public Services in Developing Countries. New York: Oxford University Press. Sah, Raaj Kumar. 1987. "Queues, Rations, and Markets." American Economic Review 77, no. 1 (March): 69-77. Sawhill, Isabel V. 1988. "Poverty in the U.S.: Why Is It So Persistent?" Journal of Economic Literature 26, no. 3 (September): 1073-119. Schwabe, Christopher. Nondated. "The Demand for Curative Health Services in Juba, Sudan: An Empirical Investigation into the Potential for Introducing User Fees." Grad- uate Research Paper, Syracuse University. Syracuse, N.Y. Selowsky, Marcelo. 1979. Who Benefits from Government Expenditure? A Case S;tudy of Colombia. New York: Oxford University Press. Tan, Jee-Peng, Kiong Hock Lee, and Alain Mingat. 1984. User Charges for Education: The Ability and Willingness to Pay in Malawi. World Bank Staff Working Paper 1561. Washington, D.C. Thobani, Mateen. 1983. Charging User Fees for Social Services: The Malawi Education Case. World Bank Staff Working Paper 572. Washington, D.C. Tirnmons, Robert J., Roy I. Miller, and William D. Drake. 1983. "Targeting: A M'eans to Better Intervention." Report to U.S. Agency for International Development on nutrition programs in Brazil, Colombia, India, Sri Lanka, and Thailand. Washington, D.C. Processed. 138 Social Sector Pricing Policy Revisited University of the Philippines. Nondated. "Democratization of Admissions." Office of the Chancellor, Diliman, Quezon City, Philippines. Vitaliano, Donald F. 1987. "On the Estimation of Hospital Cost Functions." Journal of Health Economics 6: 305-18. Vogel, Ronald J. 1988. Cost Recovery in the Health Care Sector: Selected Country Studies in West Africa. World Bank Technical Paper 82. Washington, D.C. Whalley, John, and Adrian Ziderman. 1989. "Payroll Taxes for Financing Training in Developing Countries." World Bank Policy, Planning, and Research Working Paper 141. Washington, D.C. Processed. Winkler, Donald. 1989. "Decentralization in Education: An Economic Perspective." World Bank Policy, Planning, and Research Working Paper 143. Washington D.C. Processed. World Bank. 1986. Financing Education in Developing Countries: An Exploration of Policy Options. Washington D.C.: World Bank. . 1987. Financing Health Services in Developing Countries: An Agenda for Re- form. A World Bank Policy Study. Washington, D.C. .1988a. Educationin Sub-SaharanAfrica:Policies for Adjustment, Revitalization, and Expansion. A World Bank Policy Study. Washington, D.C. . 1988b. The Philippines Education Sector Study. Report 7473-PH. Washington, D.C. Processed. . 1988c.World Development Report 1988. New York: Oxford University Press. Zschock, Dieter. 1986. "Medical Care Under Social Insurance in Latin America." Latin American Research Review 21, no. 1: 99-122. P R O C E E D I N G S O F T H E W O R L D B A N K A N N U A L C O N F E R E N C E O N D E V E L O P M E N T E C O N O M I C S 1 9 8 9 Benno J. Ndulu To look at the debate about social sector pricing clearly and objectively, it is imperative to set the issue squarely in its policy context. To do this, I need to revisit the nature of the problem confronting the providers of social services in developing countries. The gap between objective growth in demand forsocial services and resources available to meet this demand is growing. This is a long-term trend to which a combination of rapid population growth and youthful structure of population is fundamental. Not only are requirements for health and education services expanding; there are also proportionately fewer people in the population to carry the growing burden. In the short to medium term these pressures have been exacerbated by economic crises. Two major sources of pressure have been: (1)erosion of the public revenue base as a result partly of a decline in real economic activity and partly, for many countries, of a reduction in the share of formal (taxable) incomes as various groups in the economy cushion decline in real incomes through tax evasion and (2)that the resultant reduction in both quantity and quality of social services has given rise to rationing, to the dis- advantage of the more vulnerable groups. To create a sustainable delivery system for social services to meet growing needs requires policies both to manage demand and to augment supply. At the same time, in the short to medium term, solutions to the problem must take into account the broader stress on macroeconomic performance brought by the impacts of economic crises. The policy action required and the adjustment costs involved are clearly more intense and complex currently than before the crises. The overlap of pressure fromlong-term trends and pressures from crises demands that measures to restore normalcy be carefully separated from those tackling the sustainability problem in the long term. Benno Ndulu is research coordinator for the African Economic Research Consortium, Nairobi, Kenya. O 1990 The International Bank for Reconstruction and Development / THE WORLD BANK. 140 Comment The aggregate demand side for social services, dominated as it is by population pressures and state guarantees of some socially desirable minimum provision of services, seems to leave much less room for policy maneuvering. Demand struc- ture and supply augmentation seem more susceptible to policy manipulation. For demand structure, there is some latitude for internalization of benefits and for distinguishing between basic and nonbasic services. But to the extent that most services are still rudimentary, and social returns from their provision are still relatively high, the room for policy maneuvering is still very limited. Pricing reforms to influence demand structure face a particular constraint in the case of education. The relatively slower growth in absorptive capacity, especially for higher education output, may reduce prospective private benefits relative to private investment. This may dampen private demand, further constricting the effective room for modifying the structure of demand. As is apparent from the set of studies Jimenez reviews, resource gaps are large at the aggregate level. Price measures alone will probably generate insufficient revenues to fill the gap. Other measures-includinga stopgap inflow of external resources-may have to be regarded as complementary, rather than as alter- native. The political calculus of adjustment cost incidence across various interest groups is important for implementing reforms. As the paper correctly points out, the poor suffer most under relative scarcity, while those currently benefiting from the status quo command significant political resources to maintain it. The process of reform then inevitably becomes political and has to be resolved in- ternally. Well-grounded studies for each particular situation may facilitate, but can never replace, this process. The costs of reforms are more visible than the benefits because of the empirical weaknesses of the studies reviewed. Uncertainty about benefits could hinder the adoption of the reforms. The expected magnitude of the impacts of policy meas- ures matters at least as much for implementation as their direction. Jimenez's discussion assigns a threefold role to pricing in tackling the problem described above: resource mobilization, promotion of efficiency, and equity. Resource Mobilization The case for pricing as an effective instrument for reducing revenue gaps is perhaps the most convincing. The basic hypothesis is for cost-sharing in line with the proportions of private to social benefits derived from services. The workability of this proposal hinges on resolving the liquidity constraints asso- ciated with the nonexistence or inadequacy of capital markets, especially in Africa. The prevalence of such constraints may reduce levels of supply to socially inefficient levels. Empirical information on what level of consumption and pri- vate demand would be best for the society is required before pricing can be determined. Scarcities have already brought implicit (albeit tacit) implementation of user charges in some countries. Where public supply has fallen below private demand, private enterprise has tended to fill the gaps. Whether through increased use of private services, or through supplementing public services (for example, pur- chasing medicines after consultation with public medical personnel or supple- menting shortfalls of educational materials in public schools), costs of social services over and above the normal access costs are partly privately borne. Measuring the extent to which private expenditures have cushioned the drop in quantity and quality of public social services would furnish some indication of private willingness and ability to contribute toward costs. This would provide information on the scope for pricing action that brings into the open implicit charges and necessary targeting for those without the ability to pay. To estimate revenue yields from pricing it is critical to obtain relevant demand estimates that incorporate intensity of service utilization in addition to partici- pation ratios (for example, attendance or enrollment). The wide variations of characteristics across income groups and types of services make this imperative. Methods of generating revenue other than pricing are referred to passim, but in my opinion more detailed work on the comparative efficacy of the various instruments is needed. Because operational pricing strongly suggests budgetary earmarking, contrary to current broad budgetary practice, the strong case for relative efficacy will have to be a good one. Promotion of Efficiency The case for pricing as an instrument of allocative efficiency is perhaps the weakest. Prevalence of externalities, current inadequacy of services supplied, the issue of institutional legitimacy and complementary requirements (competitive- ness, capital market efficiency, and efficiency of sectors competing for access to the same resources) make this function close to untenable. Several "second-best" scenarios have been suggested, but none of them is definitive. This is not by any means to say that efficiency of the social service delivery system is unimportant. Cost minimization (in this case not necessarily the "stan- dard dual") is important. Research should emphasize methods for making de- livery systems more cost-effective to alleviate the pressure of budget constraint. Here, the paper lacks review of relevant cost studies. Equity Pricing addresses the equity issue indirectly through revenue generation, which reduces relative scarcity and hence the impetus for potentially inequitable ra- tioning. To the extent that a general increase in access reduces the need for rationing, inequality of accessibility is reduced, if we accept the author's argu- ments about unequal access via rationing. But, as argued earlier, because any 142 Comment measures to ease the budgetary constraint would have the same impacts, pricing has no intrinsic superiority, from the equity point of view, over any other method. On the contrary-increased prices, given income distribution, are as likely to have a negative as a positive impact on equity by constricting access for the poorer segment of the population. Price rises flanked by programs targeted to offset disadvantages to the poor need to be carefully compared with other schemes for their effectiveness in generating revenue while preserving equity. The paper by Jimenez has no doubt raised some important issues concerning the role of pricing in the provision of social services for long-term sustainability and in response to short-term resource shortfalls. But significant empirical gaps remain. The superiority of pricing over other instruments to achieve the com- bined targets of revenue generation, cost effectiveness, and equity needs to be established. In view of the wide variations in characteristics across different types of social services and countries, the prevalence of externalities, and the inade- quacy of complementary capital markets, the need for such empirical studies cannot be overemphasized. P R O C E E D I N G S O F T H E W O R L D B A N K A N N U A L C O N F E R E N C E O N D E V E L O P M E N T E C O N O M I C S 1 9 8 9 Nicholas Stern My comments concern, first, theory; second, empirical questions (though the latter are guided by and suggest further theoretical questions); and third, further work. Early World Bank manuals apparently recommended that public sector pricing be based on "efficiency prices" modified appropriately for "non-economic ob- jectives." The health and education sectors were specifically exempted from this prescription. The "new view" analyzed in Jimenez's paper may be seen as in part an attempt to bring health and education into the fold of standard public sector pricing. Thus one way of thinking about the subject is to examine whether we should bring them under that rubric. In other words, we ask what is different about health and education from the point of view of pricing policy. But before doing this we must note in passing that the correct price for public sector pricing is not marginal cost. In an economy that is revenue constrained in the sense that lump-sum taxes are not possible (and this applies to all real economies) the correct public sector price is marginal cost plus an element of indirect taxation to contribute to revenue. What precisely that extra contribution should be will depend on the revenue requirement, the distributional pattern of consumption of the good, value judgments about distribution, and the pattern of demand elasticities. However crudely the problems of calculation are resolved, the central question of revenue cannot be overlooked, and it is simply wrong to say that economic theory points to marginal cost pricing in this context. It does not, and the World Bank would be making a serious error if it encouraged countries with real revenue difficulties to overlook revenue generation in public sector pricing. To return to the ways in which health and education differ from the standard publicly supplied private goods: the first of the different features of health and education is that they cannot easily be resold. This means that the public sector can charge different prices to different individuals. Such a scheme would not Nicholas Stern is a professor of economics and chairman of the Suntory-Toyota International Center for Economics and Related Disciplines at the London School of Economics and Political Science. 01990 The International Bank for Reconstruction and Development 1THE WOR D ANK. L B 144 Comment work with standard private goods because reselling would mean that people who were charged a higher price would simply buy from others who had been charged a lower price rather than from the public sector. If, however, women are charged lower prices than men for chest X rays, then it makes little sense for me, observing this low price, to ask my wife to go and have the chest X ray for me. It is easy to show that, if personalized prices are to replace uniform prices, welfare would be improved if the reform is in the direction which, for equal revenue, raises prices to the richer person and lowers prices to the poorer person (provided that demand patterns and other taxes are such that a small price increase to richer households does not result in sharp revenue losses else- where). Whatever the method used to calculate the personalized prices, the central issues to be resolved-revenue, the ability to charge personalized prices, and a lower valuation on increases in income for the rich relative to increases for the poor-would combine to call for a solution where the rich pay more than the poor. The second difference between health, education, and ordinary private goods concerns externalities. It may be our view that better education and better health confer positive externalities on other members of the community. These positive externalities might lead to the suggestion that the goods be subsidized, for the standard reasons, and that the measurement of these externalities would thus be useful for applied policy analysis. I doubt that the measuring can be done with enough precision to act as a direct input into policy, but the consideration does point unambiguously toward a subsidy. A third peculiarity of the health and education services concerns the consum- er's lack of knowledge about quality. (This was a central feature of Arrow's justly famous article on the economics of medical care; Arrow 1963). Lack of knowledge might better be addressed through government programs to dissem- inate information and regulate quality rather than through a subsidy. But there might still be a case for a subsidy if consumers systematically underestimate quality despite information and regulation. Fourth, in education at least, the decisionmaker-the parent-is not neces- sarily the only relevant consumer. Children consume education too, and the interests of parent and child may not always exactly coincide. Indeed, in most countries, the state is seen to have a role, in extremis, in protecting children from the parent. Such considerations lie in part behind the idea of compulsory education up to a certain age. The level of prices charged may have serious effects on the ability to enforce such regulation, and this should be taken into account in considering the appropriate pricing policy. Fifth, the government may consider that education and health are what Mus- grave (1959) terms "merit goods'-in other words, the consumer is deemed 9 unable or unwilling to understand what is good for her or him, and the gov- ernment might charge lower prices to offset this partial self-neglect. Finally, and related to the previous two points, the government may see that it has a responsibility to provide universal access to health and education of a certain quality, and that individuals have rights to such access. This may not accord directly with our standard welfare economic approach, which is based largely on individual sovereignty and individual judgment of own best interest. But the view that policy should take into account responsibilities and rights should not be lightly dismissed. For many aspects of economic policy this con- sideration may not be central, but in this case we have to consider the possibility that it might be. The first proposition arising from the empirical work discussed by Jimenez is that existing cheap provisioning with rationing can lead to substantial use of the ~rovidedservices by the middle and upper classes. The finding--often re- ferred to as "middle class capturew-is well documented in both developed and developing countries. The common inference that the service is not particularly egalitarian or well targeted is often seen as a partial indictment of the system itself. The charge is not necessarily appropriate for two reasons. First, from the standpoint of rights and responsibilities, all should consume regardless of in- come. Second, the participation of middle and upper classes in the service can often improve its quality for everyone because they may be better capable of translating complaints into action than less privileged groups-though they might, by the same token, use their articulateness and influence to divert re- sources to themselves. The first proposition seems more firmly based than the other five, which follow. The second stylized fact discussed is that total supply may be higher in an unrestricted private market. Jimenez uses the examples of Kenya and Tanzania here. We do not know, however, whether thisis a reliable empirical phenomenon, and if so how much it matters. Higher supply strongly biased toward more privileged groups may be worse than lower supply. Third, the cost of supply may be too high. As is pointed out in the paper, this may be hard to judge because the outputs of health and education systems are multidimensional. Moreover, the tests of output may be biased. For example, the ability to pass certain tests into a higher level of education may be an exaggerated focus of attention for a private school, whereas a public school may have broader objectives that are less measurable and do not appear in output: measures. Nevertheless, one does have the feeling that cost minimization does encourage the husbanding of resources, and that it may be only desultorily applied in the public sector. Fourth, what has been supplied may not be well allocated. This idea raises interesting theoretical and empirical questions that deserve further research. What do we mean by resources being allocated well in the health and education sector? What criteria should we use and how can we measure them? No doubt these issues have been the central concern of education and health economists, 146 Comment but one does not have the impression that workable yardsticks are easy to come by Fifth, the wrong things may be supplied by the system. Again, this raises an interesting question of what should be supplied and raises similar problems of how to formulate criteria. Sixth, it seems that price-elasticities of demand may decrease with income. The paper's conclusion on the policy consequences of this observation is wrong according to standard applied welfare economic theory. The author argues that the observed decrease of price-elasticities with income implies that the poor are hit more by the price increase. According to standard theory, the higher the elasticity of demand, the less I am hurt by a unit price increase, becauseessentially I can substitute toward other goods. The paper's divergencefrom standard theory here is revealing: it shows that the commentator is concerned with the distri- bution of the good itself, which in turn indicates that something more than simply individual perception of welfare is involved. We would, I think, attach no special welfare significance to the distribution of the consumption of crois- sants or chewing gum, whereas it appears from the inference in the paper that we do attach significance to the distribution of health and education. The last two or three of the differences between health and education indicated in the preceding theory section do indeed seem to be relevant, at least for the author of the paper. The research agenda provided in the paper is sensible, and it need not be repeated here. Similarly, I need not belabor the theoretical issues mentioned in the first section of my remarks, since the research agenda arising from them is self-explanatory. What I want to bring out here is the need for more precisely formulated and researched policy proposals that embody some of the ideas of the "new view." It is simply not enough to attack the existing state of affairs as a shambles, nor to denounce opposition to the new view as consisting merely of slogans. Furthermore, to protest that one does care about the poor is not persuasive unless the protestation is supported by specific and workable pro- grams that will, demonstrably, not only meet goals described in the new view but at the same time protect the poor. As with the standard analysis of tax and price reform, we should show who will be the gainers and who the losers from a policy change. This information is critical not only to our standard procedures of applied welfare economics, but also to any analysis of political economy that seeks to identify which groups would try to lobby for or oppose which policy. Prediction of winners and losers is a critical issue, not just for this particular problem but for almost all problems of policy. And it is the most difficult aspect of most applied policy research. The severe difficulties of researching and designing workable policies bring with them the temptation to assume problems away-to say, for example, that problems of credit and insurance should be resolved in those markets and not in the market for health and education. The problems of credit and insurance markets cannot simply be assumed away; they are deeply rooted both theoret- ically and empirically and must be taken as central to those of health and education. Similarly, we should ask whether we can really target effectively without running into dangers of both fraud and disincentive. And, aside from those dangers, there are major difficulties in defining eligibility. Income-based tests bring all the problems of defining and measuring income. But if tests are not income-based, then are they well targeted?Further, means-testing can cause real personal difficulties of self-esteem, and doubtless contributes to poor take- up. Measuring success or failure for policies should not be confined to assessing gainers and losers simply in monetary terms. We must ask directly about out- comes in terms of literacy and standard of health, rather than simply looking at who consumes what service and at what price. Thus we must ask of any health and education delivery systems how they have affected, for example, literacy and infant mortality in different parts of the community. Some important examples exist of countries, such as China, that have followed policies very different from those suggested by the new view yet seem to have achieved impressive advances in increasing literacy and reducing infant mortality in vul- nerable communities. Social sector pricing is a fundamental area of policy, presenting difficult in- tellectual and empirical challenges. This is surely a priority for research at the World Bank, which should mount the appropriate research programs to meet and respond to these challenges. Without further research, and in particular the testing of the effectsof specificprograms, we should be cautious about widescale advocacy of a particular line of reform. Arrow, Kenneth J. 1963. "Uncertainty and the Welfare Economics of Medical Care." American Economic Review 53, no. 5 (December):941-73. Musgrave, R. A. 1959. The Theory of Public Finance. New York: McGraw-Hill. P R O C E E D I N G S O F T H E W O R L D B A N K A N N U A L C O N F E R E N C E O N D E V E L O P M E N T E C O N O M I C S 1 9 8 9 One participant felt that the paper may be posing the wrong question by em- phasizing pricing of social services rather than first examining the allocative decisions and framework-therelative merits of, say, spending money on defense or on health and education-within which pricing decisions are made. You need to conclude that public funds are allocated efficiently; only then do you address the question: Within a sector, how do you compensate the losers-the poor- with higher price elasticities of demand? A participant pointed out that China had performed better than India in providing social services and, put very simply, China had not used pricing, whereas India had. Is rationing better than pricing as a way of allocating social services? The answer, he thought, depended not just on price elasticities but on the distribution of needs that society places a value on satisfying. Jimenez agreed with the speaker, that governments want to provide some services universally, but was not sure that many governments would want to provide equal access, regardless of economic status, to all social services. For example, would gov- ernments in Africa currently want to provide universal higher education? One member of the audience asked if the purpose of making user-cost pricing an element of conditionality wasn't deficit reduction more than effective resource allocation. And if so, in the typical developing country, where so few people participate in postsecondary education, would not the potential savings be small? Jimenez responded that potential revenues from pricing schemes vary by country. Studies in Africa show, for example, that if you raise user fees in universities and use those funds to increase subsidies in primary education, the potential impact on primary school enrollment can vary a lot depending on the number; of higher education students in the system and the relative costs for the two levels of education. Another participant agreed that the fiscal implica- tions of delivering social services are exaggerated-that social sectors account for only about 5 percent of most government budgets. Subsidies for higher education were introduced in many developing countries because the expected returns to education were such that if you priced these services at the notional marginal cost plus, you would drive higher education out of business. The World Bank, he felt, needs to do more research on the implications of different ways of arriving at an optimal pricing strategy. A participant noted that the treatment of these issues seemed to be quite This session was chaired by Nancy Birdsall, chief, Population and Human Resources Division, Brazil Country Department, World Bank. O 1990 The International Bank for Reconstruction and Development I THE WORLD BANK. 1SO Floor Discussion of Jirnenez Paper different, at least in the public finance literature, for the countries that belong to the Organisation for Economic Co-operation and Development (OECD) than in the development literature discussed in Jimenez's paper. The OECD-oriented literature rationalizes certain types of activity in the public sector in very tra- ditional public goods terms, including rationalizing zero- or low-cost pricing for education, health-care, urban transit, and so on, on a case-by-case basis. For example, there is a lot written on education as a quasi-public good, where you enforce uniform consumption but there is a cost of exclusion. The participant felt that the development literature, on the other hand, seems to be taking a very broad approach of recommending pricing of all social services on a user- cost basis.Jimenez responded that some of the literature approaches the question of public goods from the point of view that perhaps some levels of education or some services within education are "more public than others," and similarly in health. A participant emphasized the weakness of the conceptual and empirical foun- dation on which social sector price reform was building. In his opinion, the agency that wants to spread this gospel should be responsible for stating what is needed, so that analysts can reliably estimate the critical parameters on which price reforms would proceed, and evaluate the consequences. He emphasized certain problems. For price reform to work, we have to identify those parts of the population for which withdrawing a subsidy and charging a higher price would not discourage desirable behavior. Rural-urban and regional differences may make for viable market segmentation, but he doubted that separating people by income testing was viable in many parts of the world. Then, we need to see if such differential pricing schemes can be administered. Finally we need to understand much more about the political difficulty of charging certain parts of the population for services that are provided free to other parts of the pop- ulation. Intuitively, he doubted these sorts of problems were as easily dealt with as the original World Bank publications implied, and deserved intensive study. A Bank participant suggested that although it is difficult to estimate the "exact"price to charge forsocial sector services,there is empiricalevidence that something should be charged. In a systematic study, the Rand Corporation had examined the effect on demand for health insurance of different co-insurance rates. Rand found that the biggest change in demand for services came at the step from a zero rate to something, not the step from a positive rate to something higher. Charging a little made a big difference in efficiency. In most public sector goods and services-and primary education may be an exception-itmay make sense to charge a little simply to make people aware that resources are not free. So he firmly believed that there were significant pricing, accountability, com- petition, and related issues that could be discussed without getting into ideo- logical arguments. Ndulu (discussant)pointed out that in some economies during periods of economic crises when the public supply has declined in quality or quantity, some individuals have been willing and able to pay extra for the free services they had gotten used to. Floor Discussion of Jimenez Paper 151 In designing policy on user charges, it is important to research both equity issues and political costs, commented one participant. It is not enough to just show what proportion of a subsidy goes to what income groups-say, for instance, that the bottom 40 percent of the population gets only 10 to 15 percent of the total subsidy. Targeting is important because that 10 to 15 percent may be a large part of household income for the poor and may make a big difference to them. As for political costs: we always distinguish too simplistically between the poor and the nonpoor, but the urban lower middle class is a politically important beneficiary of services. It is important to find measures that are po- litically acceptable to this group. Stern (discussant) argued for a practical approach. He did not see the point of computing optimal prices. We are interested in reform and we can see the direction we want reform to take-with the rich paying a bit more for these services. What we want to be sure about is that we have schemes that actually work. We should look at the impact on, say, literacy and infant mortality, of different schemes-not just at who pays more or less or consumes more or less. Stern found the public goods argument made by a participant earlier in the discussion inappropriate in this context. These were private goods, on the whole, with externalities-and price elasticities were probably not as important as the size of the externalities, in his view. A Bank participant noted that we should first deal with the problem of fi- nancial discipline and financial incentives and then talk about introducing user fees, rather than the other way around, as the paper did. Without fiscal discipline, evasion, avoidance, and leakage from user-fee revenues would be substantial. He recollected that a study of hospital finances in Malaysia, which has a system of user fees, had found that revenues amounted to only 7 percent of what was collectible. The people who do not end up paying the user feesare predominantly the rich and the influential-the very people from whom you are trying tocapture the fees. With financial discipline and the ability to collect revenue, we can make sure that any user fees that are introduced actually work. Wrapping up the discussion, Birdsall (chair)observed that, not surprisingly, there had been some agreement on the notion that because of equity concerns and externalities associated with providing these services, policymakers must look beyond the market mechanism and treat social services as more than private goods. Disagreement about whether pricing reform is worth it, or would work, seemed to be related to confusion and disagreement about the primary purpose of pricing reform. Is it to achieve fiscal gains, efficiency gains, or better equity? In her opinion, the discussion did not address an important point raised in the ~aper-that you can probably improve equity with pricing reform, particularly if you focus on selective, not across-the-board, user charges. She observed that the participants agreed that more research was needed about reforms now being undertaken-their outcomes, use of services, and issues of political economy; and about which groups lose, which gain, and how it is that some pricing reforms go ahead and others do not. P R O C E E D I N G S O F T H E W O R L D B A N K A N N U A L C O N F E R E N C E O N D E V E L O P M E N T E C O N O M I C S 1 9 8 9 The Role of Institutions in Development Brian Van Arkadie Economists interpret"institutions" in at least two ways. Institutions can be the"rules of the game" (which provide the context-such as markets-in which actors make decisions), or they can be organizations (typically, systems of nonmarket relations). What is the role of public policy in influencing the rules of the game, and what is its role in public sector economic organizations?Informal conventions and informal rules of the game make institutions function differently than their formal structure might lead us to expect. Governments can intervene and influence the rules of the game, but their interventions should be based on an adequate perception of existing formal and informal arrangements and on processes of adaptation that were under way before intervention was initiated. The literature does not adequately explore the normative arguments for public economic organizations, nor has there been enough rigorous analysis of the causes of public sector failure. Many institutions are dysfunctional in terms of development, however, not because they are inefficient but because their intended purposes conflict with the requirementsof economic growth. Diagnosing the causes of good or bad performance by public organizations could provide the basis for exploring a rational public sector organizational strategy. Which government institutions and institutional interventions seem essential to growth, which can be helpful, and which are likely to obstruct growth?Such questions cannot be answered definitively, and answers cannot be applied uni- versally: the observed failure or success of an institution in one national setting may reveal little about its potential performance elsewhere. But short of a check- list of "good" and "bad" institutions, a good deal of light can be thrown on what should be considered in appraising them. Strong institutional assumptions are often implicit in the design of programs and projects. The logical and em- pirical basis for such assumptions should be explored, even if the result is far from a cook book of recipes for effective institutions. Two different, although related, meanings are given to the term "institutions" in discussions of development. The first is as rules of the game. The second is as organizations. Brian Van Arkadie is a professor of economics at the University of Dar es Salaam, Tanzania. O 1990 The International Bank for Reconstruction and Development 1THE WOKLD BANK. 154 The Role of Institutions Rules of the Game The general meaning that informs much of the academic discussion of insti- tutions is captured by Ruttan and Hayami (1984): Institutions are the rules of a society or of organizations that facilitate co- ordination among people by helping them form expectations which each person can reasonably hold in dealing with each other. They reflect the conventions that have evolved in different societies regarding the behavior of individuals and groups relative to their own behavior and the behavior of others. In the area of economic relations they have a crucial role in establishing expectations about the rights to use resources in economic activities and about the partitioning of income streams resulting from eco- nomic activity. Similarly, Feeny (1988) quotes Douglas North (1981): Institutions are a set of rules, compliance procedures, and moral and ethical behavioral norms designed to constrain the behavior of individuals. In this definition "institutions" encompass the fundamental rules of the game within which the economic system operates. Economic historians have been particularly concerned with the broad changes in these rules that accompanied the emergence of modern capitalism-for example, the change in land tenure systems, the decline of serfdom, and the emergence of "free" labor. Such broad issues might seem to have little operational interest for contem- porary policymakers. But in fact the institutional context determining access to land, and relationships between farmers, landlords, workers, traders, and mon- eylenders has been long recognized as critical for rural project design. And the current fashionable concern with an "enabling environment" in which nongov- ernment actors can contribute to development emphasizes the importance of institutional factors in this sense (see Aga Khan Foundation 1987). Organizations Donor agencies tend to define institutions somewhat more narrowly, essen- tially as synonyms for organizations-government departments, state enter- prises, banks, armies, hospitals, and the like (seevan Rennin and Waisfisz 1988). Discussions of "institutional development" and "institution building" are there- fore typically concerned with how to make organizations work. The difference between the concepts has some interest from an analytical as well as a semantic point of view. Institutions in the "rules of the game" sense provide the context in which markets operate-influencing both their efficiency and distributive impact. Institutions define the terms under which the various actors in the market confront each other, molding their expectations and defining their rights. In this sense, economists should be very much at home with insti- tutional questions, relating as they do to the central concern of neoclassical economics: the operation of markets. Van Arkadie 155 The institution as organization is another matter. From the point of view of an economist an organization can be usefully defined as an area of activity within which the market does not coordinate the activities of the participants. An organization may be externally constrained by market factors, but its internal arrangements are coordinated by nonmarket instruments. Internal decision rules and incentive systems may be chosen to simulate the market, but even then the relationship between the actors is inherently different from that between inde- pendent actors in the marketplace. The principal strengths of neoclassical economics derive from propositions of widespread applicability about markets. Not only is the underlying assumption of maximizing behavior plausible, but also the pressures of competitive markets force actors to perform in ways approximating such behavior as a condition for survival. The tools of neoclassical economics cannot, however, be readily used to ex- plain behavior within organizations. Bureaucratic (or organization) man is not differentin inherent motivation or psychology, but the incentives and constraints he faces differ from those facing the actor in the market. (Hence, Israel 1987 seeks a prescription for improving institutional performance by creating a "com- petitive" institutional character.) This being so, it is not surprising that early models used in the development of the study of business administration were borrowed from engineering (Turn- erism), the military (line and staff organizational structures), and the social psychology of nonmarket behavior (Elton Mayo) rather than from economics, and that subsequent explorations by economists took account of nonmaximizing managerial behavior within the large firm. The Influence of Informal Factors In either definition of institutions there are both formal and informal char- acteristics to be considered. Access to land, tenant-landlord arrangements, and so on may be subject to legal provisions, but they may also incorporate informal conventions and understandings either not incorporated in the law or even at variance with formal legal provisions. And the behavior of organizations with the same formal structure (chain of command, job descriptions, and so on) may vary enormously depending on the informal environment. Those of us who had the experience of working in the postcolonial twilight know that the operational characteristics of systems changed often when the formal structure did not because informal assumptions altered. Colonial civil servants were bound together by an "old boy" network from"home,"informally exchanged information in the club, were little dependent on status in the local society, and had no family responsibilities in the local community. The local bureaucracy were connected by a different "old boy" network (although often a copy of the metropolitan model),were able to exchangeinformation informally in a different social setting, were dependent on local society for status, and were faced with a network of extended family obligations. 156 The Role of Institutions Assumptions or expectations of behavior, in general currency but informally based, influence organizational behavior. In a system with little or no corruption, it will be equally difficult to ask for or to offer a bribe; in one where corruption is the norm it becomes routine to ask for a bribe and impossible to conduct business without offering one. The influence of informal factors makes the transfer of organizational models from one setting to another risky. A particular "success" may be based on a successful congruence between the formal model and the informal setting, or it may even result from particularly propitious informal circumstances actually transcending inadequacies in the formal structure (for instance, charismatic lead- ership sustaining a flawed organization). The same formal model may mean very different things in different settings. Cooperatives, for example, with a fairly standard legal constitution, perform quite differently in terms of operational efficiency and distributional conse- quences not only between countries but even within the same country. This suggests that the design of programs on the basis of successful models should be approached with caution, be it the Kenya Tea Development Authority, the Grameen Bank, the Uganda Development Corporation (of the early 1960s)-all of which at one time or another have been proposed or used as organizational models. This caveat is particularly relevant for itinerant advisers, who may be well informed about "success stories" but not about the specific origin of the observed success or the relevant informal characteristics of the setting to which the model is to be transferred. The absence of any single driving force (analogous to the market) determining organizational behavior plus the impact of informal factors on formal structures makes generalization difficult. It is therefore, as Nellis (1980, p. 413) notes, not surprising that early efforts to create a general theory of development admin- istration failed. The best line of attack, therefore, is probably not to seek out a grand theory of the role of institutions but to sort out the questions about that role relevant to the design and implementation of development programs. This is done here by first considering the general approach of economists to the role of institutions (section 11).A discussion of institutions in the two senses defined above follows, with consideration of the rules of the game that influence and constrain the behavior of the various actors in the economy (section 111)and of the effectiveness of government organizations (section IV).The discussion concludes with a com- mentary on the political economy of government approaches to institutional issues (section V). Broadly, economists' questions about institutions cluster around four themes. The first concerns the origins of institutions. How far can economists offer an endogenous explanation of the origins of institutions? Explanation of origin may Van Arkadie 157 in turn relate to the identification of current function, although there is no necessary connection. Institutions may in their actual operation serve purposes quite different from those intended by their creators. Moreover, their role may change in response to a changing environment. Economic institutions as rules of the game can be seen as defining the terrain over which economic actors maneuver. The second theme concerns how eco- nomic actors behave in a particular institutional setting, and how their behavior, as affected by that setting, determines the economic outcomes. The third theme is normative. How does one evaluate the economic effec- tivenessof institutions?Astraightforward response would be to apply the normal criteria of welfare economics. However, insofar as some institutions exist to meet social objectives which transcend, or are separate from individual welfare maximization, evaluation against individualist, utilitarian criteria is not uncon- troversial. The fourth theme relates to the practical concerns of policymalzing. How can institutions be changed to perform better according to chosen criteria of eval- uation? Exploration of the first three issues-the origins of institutions, the motivation and behavior of economic actors in the existing institutional setting, and institutional effectiveness-provides a foundation on which to tackle the fourth: institutional engineering. The initial brief for this conference paper asked what light the various par- adigms in economics throw on the role of institutions in development. An ad- equate answer to that question would demand a long essay on the history of ideas outside the competence of this author, but there is room to locate the issues in the context of at least some economists' ideas. Because, in a Kuhnian sense (Kuhn 1970),most of the literature usually identified in the World Bank milieu as "economics" falls within the neoclassical paradigm, approaches from within that paradigm are emphasizkd. The issue of "institutions" was raised by the U.S. institutionalists, following an earlier tradition of the German historical school, arguing that propositions of economic theory (that is, of mainstream neoclassical economics) were highly - . relative and were based upon unstated institutional assumptions much less uni- versal than implied by the theorizing. "Institutionalism" therefore challenged economic theory in its dominant form. The Neoclassical Positions In contrast to the institutionalist view, the "purist" defense of neoclassical economics, set out in great clarity over fifty years ago by Lionel Robbins (1931), defined economics as the study of the interrelation between means and ends through theories of choice of general applicability-potentially useful in ex- ploring economic decisionmaking in all institutional settings. Within that vision, the concept of "institutions" can be applied to aspects of human organization that may be accepted as important but are not the subject of economic analysis as such. Like technology, tastes, or resource endowments, they were seen as 158 The Role of Institutions exogenous to the economic model-data that influence parameters, constraints, or the definition of a social welfare function without themselves being a subject of study by the economist. No economist would have claimed that political institutions or social attitudes are unimportant; the argument was rather that in the intellectual division of labor they were not the appropriate subject of study for the economist. The approach might be seen either as modest in the limited scope it claims for economics, or arrogant in its view of "economic analysis" as being sufficient to handle the "hard" issues. There continues to be a great deal of work-even some practical analysis of policy and projects-within that "purist" tradition,' but neoclassical economics has also vigorously developed beyond the overly restrictive boundaries envisaged by Robbins, through what Jack Hirshleifer (1985) has described as "imperialist economics" (an unintended double entendre in this context). The rational choice approach has thrust neoclassical economics in the direction of issues once considered noneconomic. Thus Hirshleifer quotes Gary Becker (1976): The combined assumptions of maximizing behavior, market equilibrium and stable preferences, used relentlessly and unflinchingly, form the heart of the economic approach and goes on to add: It is this approach that has powered the imperialist expansion of economics into the traditional domains of sociology, political science, anthropology, law and social biology-with more to come. This approach has both a positive and a normative element. Positively, it attempts to explain the behavior of a wide range of institutions in terms of the maximizing pursuit of self-interest by the actors in those institutional settings. Normatively, in light of such behavior, institutional performance is evaluated against the individualistic and utilitarian objectives that underpin neoclassical welfare economics. In the development literature, the rational choice approach has informed the discussion of rent seeking in general and the analysis of state agricultural mar- keting in particular, and it has been extended to the analysis of household behavior, and so on. Although its adherents are often drawn from the other end of the political spectrum, rational choice explanations of political and social behavior share many of the strengths and weaknesses of Marxist analysis. In particular, the attempt to endogenize institutional factors through explanations based on a 1. An important part of the thrust of Gunnar Myrdal's critique of the work of economists in Asia was concerned with economists' neglect of institutional questions. It is not difficult to find work equally subject to that earlier criticism. Van Arkadie 159 strong element of economic determinism are shared characteristics of rational choice and Marxist analysis. At its best, this sort of analysis provides powerful insights into the ways economic actors pursue their interests both through the market and through nonmarket institutions, and how this in turn affects the institutional context in which markets operate. It may, however, claim too much, taking too little account of factors other than the pursuit of economic self-interest that determine political and social behavior. Another contribution within the neoclassical tradition is the market-failure- transactions-cost approach, which sees the existence and behavior of a number of economic institutions as resulting not from the effective operation of markets but from market failures that require nonmarket institutions (and sometimes nonmaximizing behavior)for their resolution. (On this point, see the discussion of institutions below and Bardhan 1988.) This approach was developedin the context of the theory of the firm, following R. H. Coase, and has been used to explore such issues as vertical integration and antitrust policy in the context of the U.S. economy (see Williamson 1985, 1986). It has also provided the intellectual underpinnings for a burst of recent literature on rural institutions. Some economists faced with institutional questions have called on other dis- ciplines for a joint attack on the problem. In some heroic cases, the economist has taken on the task of acquiring the skills of those other professions (see, for instance, Hill 1986). Whether, in practice, multidisciplinary work has typically meant cross-fertilization or cross-sterilization of ideas can be debated. But it would take a very robust imperialist of economics to believe that the economist's tools are uniquely appropriate for analyzing all important development insti- tutions. Structuralist and Marxist Challenges In the development debate, the earlier institutionalist criticisms of neoclassical economics were echoed in the structuralist2 critique. Two observations were par- ticularly influential: (1)that "Western" economics had many implicit assump- tions regarding institutions, among other things, that undermined its applica- bility to economies in the developing world; and (2) that insofar as certain institutional factors constrained the effectiveness of policies, such rigidities had to be either incorporated into the analysis of policy or tackled by political or social reform. Some aspects of the structural approach have similarities with, or even roots in, Keynesian economics. The structuralist differsfrom the strict neoclassicist in 2. This imprecise designation covers an enormous potential range of literature. Meier identifies Lewis, Myrdal, Prebisch, Singer, and Rosenstein-Rodan-six of the ten authors covered in Pioneers in Devel- opment (Meier and Seers 1984) as "introducing elements of structural analysis into their work." See also Furtado (1964)and Seers (1963). 160 The Role of Institutions regarding what the latter sees as transitory "market distortions" as fundamental characteristics of the system, for which neoclassical solutions are impracticable, at least in the absence of profound structural change (for a neoclassicist's view, see Little 1982). Structuralists have been concerned with the persistence of archaic rural in- stitutions (forinstance, in the Latin American literature of the 1960s a backward rural institutional structure was seen to inhibit response to price incentives and thus contribute to inflation), which suggests the need for active land reform; with skewed distribution of wealth, which requires active redistributive policies; and with weak entrepreneurial institutions, which justify strong state interven- tion and protectionism. Central to the structuralist policy agenda was the need for deep-seated insti- tutional reform. The reforms in question demanded explicit examination of the political conditions in which they would be implemented. But when conducted, such analyses often revealed a pessimistic outlook for the reform agenda. (See de Janvry 1981, p. 146, for a succinct criticism of the structuralist thesis on agricultural stagnation in Latin America.) Evidently the term "structuralist" as used here covers a wide range of ideas, united more by a common deviation from tenets of neoclassical economics than by a shared theoretical approach. In such terms, much early development eco- nomics (say 1945-65) could be described as structuralist (see Meier and Seers 1984). Since the mid-1960s there has been a revival and consolidation of a neoclassical approach to the subject (see Little 1982). 'The other main challenge to the neoclassical paradigm is from Marxist analy- sis. Its strength lies precisely in its holistic ambitions, which have never allowed it to accept the boundaries traditionally defined for neoclassical economics. Four strands of Marxist thought throw light on the q~estion:~ (1)the emphasis on the historical setting and the historical process of change; (2)the exploration of the interaction between institutions (superstructure), economic forces (forces of production), and social relations (the relations of production); (3) the focus on class interests and class struggles (rather than individual interest and welfare ' maximization through exchange); and (4) explicit consideration of the role of ideology. Yet there is no more unanimity in the Marxist than in the neoclassical par- adigm on any of the basic issues facing development economists. Thus Marxism, as much as neoclassicism, can inform both an activist and a deeply skeptical view of the potential role of the state in the reality of the developing world (see Baran 1957). In this regard, the neoconservative critique of the rent-seeking bureaucracy resembles that of Marxists who emphasize the corruption of the 3. The quick characterization of paradigms demanded by limits of space runs the obvious risk of vulgarization. The only way to do justice to each of the paradigms is to illustrate their use in tackling the problems discussed here: for instance, de Janvry's (1981) study of the agrarian questions in Latin America demonstrates the sensitive use of Marxist concepts. Van Arkadie 161 neocolonialstate: the pessimismof both is based on an economicallydetermined explanation of political behavior, although for one the mainspring is individual or pressure group interest, and for the other the pursuit of class interest and the international context. It is by now generally accepted that a critical task of government, neglected in earlier models of economic planning, is to provide the "right" incentives to stimulate and guide the various actors in the economy-thoughviews may differ as to what the right incentives are. A corollary to this view is that development policy must focus on the institutional arrangements that provide the setting in which the incentive system will operate. Rules of the game which stand out as susceptible to government influence include: (1)rules defining private property rights and their allocation; (2)other rules and conventions governing the relationships between participants in the economic process (for example, those affecting the operation of the labor market); (3)rules and conventions defining the economic role of a hierarchy of social institutions (for example, local governments, trade unions, cooperatives and peasant associations, and churches);and (4)rulesand conventions restricting participation in the economic process. Potential government influence on the rules of the game is illustrated by reference to three important areas: property rights and rural development, in- stitutional pluralism in decisionmaking, and the promotion of entrepreneurship. Property Rights and Rural Development Property rights (land rights and their interrelations with other institutional arrangements) have figured prominently in the discussions of agricultural de- velopment (Bardhan1988).Changes in rural property rights, encompassingsuch issues as the decline (orsurvival)of serfdom and the enclosure of common land, have been central concerns for the historian, whether the neoclassicistexploring responsesto changing economic opportunities and factorscarcitiesor the Marx- ist for whom relations of production in the rural sector are seen as defining the mode of production. Two questions particularly intrigue economists: (1)Why do differing property rights emerge to provide the framework within which economic decisions are made in rural areas? (2)What are the relative merits of alternative systems on both efficiency and distributive grounds? One view is that, on the one hand, economic development has meant the displacement of one set of rural property rights by others more adapted to the emergingeconomic requirements, whereas, on the other hand, existing arrange- ments can and should be improved both in efficiency and distribution. This view has fueled the argument for rural institutional (land) reform. In reaction, the rationale of existing institutions has been explored to question whether they are as atavistic, inefficient, and disequalizing, as the reform agenda implies. 162 The Role of Institutions Property rights determining access to land are part of a complex of institutional arrangements providing the context for rural economic activity. These include institutional arrangements determining labor contracts and access to land, credit, and commodity markets; that is, the set of rules and conventions that define the relations between the various actors in the rural economy. A critical analytical concern is with the interlinkages between markets. Thus neoclassical economists have explored such institutions as sharecropping as part of a more general attempt to understand the role of property rights in light of transactions costs, or alternatively under conditions of imperfect information in interlinked factor markets. For the Marxists, the critical characteristic of the changing labor mar- ket, the emergence of "free" labor, is related in turn to changes in access to land and the expansion of commodity production. There is a body of empirical literature which provides useful insights into the effectiveness of rural institutions. But the neoclassical theoretical literature has recently become so complex that it yields little policy guidance for the applied economist, other than that existing institutional arrangements are likely to be a good deal more complex, and their efficiency and distributional implications less straightforward, than at first sight. It is particularly tricky to establish the actual status of rural institutions (see Feder and Noronha 1987). The concept of property rights in land is far from clear cut: often the requirements of rural economic activity lead to the useful development of a range of differing property rights relating to differing attributes of land; de facto practice may vary greatly from the de jure system; and the national system of statutory laws may coexist with local law and customs. The likely impact of statutory reform is therefore particularly hard to predict, since it is both difficult to delineate the existing structure to be replaced and even harder to forecast the de facto outcome of de jure proposals. Thus while there has been a widespread presumption that a movement toward securely held registered title is economically desirable, recognition of the prob- lems involved has been slower to emerge. For example, conferring land tenure rights on sedentary agriculturalists (already apparent occupants) may disrupt a system of pastoral seminomadism, based on rights of passage and of dry season grazing, which represented a sophisticated response to a particular ecological problem. Likewise, the allocation of land rights to an individual may misinterpret and disrupt arrangements within the family (displacingthe rights of women, for example). Views on what the best institutional structure may be for fostering growth vary and change; indeed, research on rural institutions seems to be widening the range of debate, as institutions earlier seen as barriers to progress are re- habilitated in light of a more sophisticated interpretation of their function. Theories of endogenous institutional change, which argue that the institutions that ~rovidethe framework for the market tend to adapt themselves to changing market conditions, such as changing factor scarcity and relative factor produc- tivity, suggest that informal institutional arrangements are likely to adjust to Van Arkadie 163 new conditions even without official sanction. This suggests that one official task should be to adjust formal rules to accommodate changes under way so as to improve the certainty of transactions by making them official. But there are dangers: formalizing informal arrangements-for instance, through land regis- tration-will have its own effects, by shifting the access of different economic actors to the formal system-for instance, through land grabbing. Institutional Pluralism The rules of the game, besides defining property rights and acceptable con- tractual relationships, also define how and by whom the rules themselves are made. A recent collection of essays on institutional analysis and development (Ostrum, Feeny, and Picht 1988) argues that the institutional choice should be presented not as a dichotomy of state or market but as a pluralism of possible contractual relationships, both explicit and implicit. To quote (p. 456): Systems of governance can be constituted by conceptually simple but so- cially complex configurations of implicit or explicit contractual relation- ships. There is no theoretical reason why there must be a single centre that has exclusive authority to formulate and enforce rules in a society. The volume is in the tradition of U.S. pluralism, favoring a diversity of social institutions as the desirable base for a democratic order, with a place for ini- tiatives from many levels of society and not providing central government with undue concentration of power. Systems of decentralized decisionmaking and control might bring to bear user pressures as stimuli to performance not possible in a centralized bureaucracy. Skepticism about the concentration of power in centralized institutions is not confined to neoconservatives: on the left, popular participation, workers' and peasants' movements, and other forms of grass-roots mobilization are seen both as virtues in themselves and as mechanisms to make institutions responsive to local conditions and needs. Of course, while decentralized institutions may be more susceptible to local participation and control, by the same token they may be captive to the local structure of power. The drawbacks of centralized power-insensitivity to local needs and ignorance of local capacities-havetheir counterparts in the skewed access that can arise from local systems of power and prejudice. The Promotion of Entrepreneurship Another area to explore is the possible impact of government molding of the rules of the game on entrepreneurial initiatives. While not all would share the Schumpeterian view that entrepreneurship is central to capitalist development, entrepreneurial capacity is clearly important, and particular development suc- cesses seem to be associated with a concentration of entrepreneurial flair or capitalist "animal spirits." Given their pivotal role in capitalism, entrepreneurship and entrepreneurial 164 The Role of Institutions institutions would be expected to be central in neoclassical treatments of de- velopment. Yet entrepreneurship has found a more prominent place in the work of economic and business historians (see, for example, Iliffe 1983) than in the neoclassical theoretical tradition. This lack leaves at least some free market advocacy with a strange lacuna: the market sets the stage for successful private initiatives, but it is unclear from where the cast of characters is to be supplied. The lack of an endogenous economic theory of entrepreneurship has made this an area in which economists have often been willing to traverse the mul- tidisciplinary route. And in practice folk wisdom, which identifies particular social groups as being "entrepreneurial" in culture, is often influential. But folk characterization, and even learned identification of entrepreneurial groups, tends to be based on ex post recognition of performance. Stereotypes change very fast with changes in performance (see, for instance, Elkan's (1988)contrast of Taw- ney's negative characterization of Chinese entrepreneurship in 1932 with current perceptions). What is lacking is a good predictive model of entrepreneurial capacity to bring out ways of enhancing entrepreneurship through policy meas- ures. Theories that emphasize deep-seated psychological or cultural characteristics (see,for instance, McClelland 1961) do not seem particularly relevant for policy, as these variables are unlikely to be subject to policy influence over any useful time horizon (thoughMcClelland's psychological interpretation of entrepreneur- ial capacity was incorporated in business training in India and elsewhere). Are there specific ways in which government can influence the institutional context of entrepreneurship (otherthan the obvious panoply of economic policies and commercial laws that determine the broad incentives and constraints)?Elkan (1988) broadly concludes that there is none: the best policy is laissez-faire. An acceptable role for private entrepreneurs is defined both by formal rules and informal signals. Here, not only the stated intentions of governments but the outcomes expected by the entrepreneurs are important. One interesting his- torical insight relates to the entrepreneurial role played by cultural, ethnic, or religious minorities, whose entry into certain entrepreneurial areas may be re- stricted but who, by the same token, in being politically excluded from easier avenues of advancement, are sometimes stimulated into a risk-taking innovative r01e.~ These considerations could lead to a bizarre (dialectical)conclusion about the virtues of discriminating against potentially entrepreneurial groups. More seri- ously, it would be interesting to know what conditions would be conducive to a nice balance in which a minority, pushed into entrepreneurship, is nevertheless 4. The East African Asians, for example, have had repeated experience as a minority community, first under British colonial rule, then in independent Africa, and now in Britain and North America, and have shown a high degree of entrepreneurial ingenuity in all three situations. At the same time, of course, the culture of a minority group may be more or less consistent with taking on a successful entrepreneurial role. Van Arkadie 165 sufficiently accepted to be allowed to play the part effectively, and secure enough to take a long view of investment opportunities-as against a common situation of insecurity in which suspicion engenders precisely the short-term, capital- exporting behavior that reinforces the initial suspicion. The view that entrepreneurship is likely to appear when other avenues of advance are blocked relates to two other observations. The first is that entre- preneurship often seems not to be highly correlated with formal educational achievement, in part because the able but uneducated are excluded from more secure avenues of social advance. Second, crises that erode bureaucratic incomes and frustrate the expectations of the educated elite may call forth considerable entrepreneurial response. This seems to be the case in the recent experience of a number of African countries. Strangely, Marxist and related literature has tackled the long-term institutional issue of capitalist development more explicitly than neoclassical writing. The main thrust of the Marxist (anddependency) literature is to see the development of national capital and multinational enterprise as essentially antagonistic-with indigenous capital trapped in a stunted, essentially comprador role (see Baran 1957). In some Marxist writing this comprador role is extended to the state sector (Shivji1973).But the opposite view has also been expressed in the Marxist literature (see Warren 1980; Sender and Smith 1986). It has been suggested, for example, that multinational business supported indigenous capital in the colonial period (as compared to settler colonialism; see Cowen 1979). The propositions of neoclassical economics are most strongly developed in relation to the virtues of liberal international commodity markets. While no doubt for most neoc~assicaleconomists, the commitment to freely operating markets would extend to capital markets, there are different issues involved from those related to free trade in commodities. Albert Hirschman (1969) made the interesting point that quite different issues are at stake when a country specializes in the supply of one factor of production, labor, than when it specializes in producing a particular commodity. The assumption implicit in much pro-free-market analysis is that there is a potentially complementary and supportive relation between multinational busi- ness and indigenous private enterprises, while in contrast the relation between state and private enterprise is likely to be competitive and antagonistic. Hence an institutional choice must be made between state and private enterprise. Such a view is not, however, necessarily consistent with rent-seeking inter- pretations of political processes, which could equally well suggest that public corporations are likely to be responsive to private interests either bringing pres- sure through the political or administrative process, or engaging in commercial manipulation to gain access to potential rents. A large scale public sector, even if conceived in the context of a transition to socialism, can as well provide one path toward capitalism, depending on the role of the state system vis-a-vis incipient national capital. 166 The Role of Institutions IV. GOVERNMENT ECONOMIC ORGANIZATIONS The discussion in World Development Report 1983 (World Bank 1953) on "Management in Development" described the role of the state as follows: Some economic activities are universally recognized as the sole responsi- bility of the state; others, it is widely agreed, are best left to private initiative. Between these extremes governments have tended to expand their sphere of activity for a variety of reasons. This statement implies some consensus about the state role in a considerable range of activities. Such consensus may exist from time to time, and in particular countries, but it often changes when big political shifts occur. In Britain, for example, there have been two major shifts since World War 11. Even activities on most "commonsense" lists of state responsibilities may in practice be far from universal state monopolies. Cases in point are police services, supplemented in many countries by booming private security services, and postal services, virtually displaced in some lines of business by private courier services. On the other hand, international experience yields examples of government involvement in virtually all areas of economic activity. The origins of public economic organizations are extraordinarily diverse. Some spring from a systematic initiative to extend state influence; some emerge to handle problems of private bankruptcy, or as vehicles for political patronage; and aid donors themselves sometimes generate new institutions as a conduit for external funding. Many have probably been created for conventional reasons-in the colonial period, for example, reflecting metropolitan practice as much as local requirements or local interests. And perhaps as important as their origin is the tenacity with which public institutions, once created, defend their own survival (on this topic, see Bernard Schaffer, quoted in Lamb 1985). Thus there is a wide diversity of government economic organizations. Among the reasons for this diversity are political factors, both ideological and, more prosaically, those that respond to the day-to-day needs of the political system. It is surely a mistake to avoid recognizing this openly, by presenting the issue as largely technocratic-particularly since shifts in the international climate of opinion on this issue reflect shifting political winds in the industrial countries. Nevertheless, the question of the appropriate form and role for government economic organizations cannot be sidestepped as simply political. Good or bad performance by government organizations is critical to growth, and particularly important for aid donors, because these organizations are the preponderant channel for aid. Despite the difficulties, the task of improving the performance of public economic organizations must be confronted. The messy array of issues is tackled here by addressing three sets of questions: (1)What are the arguments for the existence of public economic organizations? (2) How should aid donors relate to institutional performance? (3) What are the factors that influence performance? Van Arkadie 167 Arguments for Public Economic Organization The most widely accepted role for public economic organizations is that of supplying goods-such as road services, public health, and agricultural exten- sion-that the market cannot supply efficiently (because of externalities, the "free rider" ~roblem,natural monopoly, and the like). A second function, argued for specifically in relation to development, is for government enterprise to supplement an inadequate supply of private entrepre- neurship: to demonstrate the viability of an activity so as to encourage involve- ment by nongovernment actors. Government enterprise here is seen as necessary to reduce the risk, lower the cost, and increase the profitability of private ventures in the context of economic backwardness. A third view sees government organizations as an instrument for the transfer of surplus-in effect to tax or subsidize (for example, through agricultural mar- keting boards or state cigarette, match, and liquor monopolies). Here the jus- tification may be administrative-that without the involvement of government organizations, the tax system would be inadequate to achieve the desired result. More cynically, it is easier to tax agriculture under the guise of providing a service, or stabilizing prices, than through overt levies. A fourth function of public sector economic institutions is to further some policy intended to affect patterns of property holding in a society--either to encourage a general commitment to social ownership or to shift the balance of property ownership between different national or ethnic groups. There are also regional arguments for public economic institutions, in the two different senses of promoting the development of backward regions or promoting cooperation between different nation-states in a regional grouping. Other arguments could be listed, but these five encompass the main economic justifications for government economic organizations, and as such provide a basis for judging what is to be done about the complex array of them that exist. Where there is a need that cannot be met by the market, the task of achieving efficient nonmarket economic management solutions must be confronted. That the state should undertake an activity does not guarantee that it will be able to do so effectively. Where the objective is to change the balance of property ownership or the distribution of income, differentlines of reasoning are possible. One possibility is that public economic organizations be subject to the market, along the lines of the Lange-Lerner model of market socialism: public managers having to respond to the market on the same terms as private managers, the public interest being that of a shareholder, and any deviation from a "pure" market solution being handled by explicit public subsidies or taxes. Alternatively, the option of government intervention to achieve the intended result without the existence of a public economic organization could be ex- plored-taxing, subsidizing, licensing, private ventures-or where the objective is wealth redistribution a once-and-for-all reallocation among private economic actors might be seen as an alternative (as in some land reforms). 168 The Role of Institutions Donors and Institutional Performance Some governments command public economic organizations capable of im- plementing the government's intentions, others not. Bureaucracy may be rea- sonably effective in one ministry, parastatal, or sector, and quite ineffective in others. Why? And what determined changes in performance over time? For an institution such as the World Bank, acting as it does primarily through govern- ment institutions, appropriate answers to these questions are crucial, and errors of judgment in assessing administrative capacity are probably as important a cause of project failure as mistakes of economic calculus or technical design. Israel (1987),analyzing the evidence on institutional performance from the World Bank's own project experience, found that institutional performance varied more across sectors and activities than among countries. He summarized the record as follows (p. 4): The most successful were found in industry, telecommunications, utilities and finance; the least successful in agriculture, education and services. Within institutions, technical and financial activities fared the best, while maintenance, personnel issues, and co-ordination were the least successful. The study recounts the piecemeal explanations of differential performance found in the Bank's own operational evaluations and, finding them wanting, identifies two factors which he argues are decisive in explaining institutional performance in practice: "specificity"and "competition." Specificity is defined in terms of concreteness of objectives, means, and rewards, and the immediacy and transparency of the effects of an activity. Competition covers the influences that impel an organization to improve its performance. These are seen as in- cluding market pressure but also encompassing pressure from clients, benefici- aries, or suppliers, from the political and bureaucratic establishment, and from internal managerial measures to stimulate a competitive atmosphere within an organization. Israel uses the empirical record of Bank projects to draw some plausible conclusions about ways to improve the performance of most institutions. But there are problems with using this evidence to judge the appropriateness of public institutions. The first is that the evidence seems to suggest that institutions that work best more or less approximate to a modern industrial plant. Thus plantation projects in the Bank portfolio have worked better in Africa than projects that support smallholder agriculture. Unfortunately, this does not tell us whether plantations or smallholder agri- culture are the more reasonable vehicle for agricultural development in Africa. Smallholder agriculture might be (and, indeed, I believe is) superior on both efficiency and distributional grounds, although it does not respond readily to direct public institutional interventions. The fact that"modern institutions" have the characteristics of specificity and competitiveness does not mean that the best Van Arkadie 169 growth path is found by emphasizing such institutions, any more than the fact that wealthy countries are industrialized means that at a particular moment industrial investment should be emphasized. A second problem is that the identified characteristics of effective institutions and the characteristics of institutions that would be considered appropriate for public sector attention often do not match-for instance, it is in the nature of some public goods that their supply lacks specificity. One is left with the awk- ward conclusion that the public sector is in part the residual legatee of activities that do not lend themselves to effective institutional performance. But this need not be so alarming, as some of these activities may be quite satisfactorily handled by governments, even if they do not respond to systematic efforts to improve productivity, nor are plausible candidates for external assis- tance. Even though primary education is a low-specificity, low-competition ac- tivity, in most countries it is done quite well, sometimes in extremely difficult conditions. But there is no very good reason for external funding to be mobilized for primary education (except for textbook production-which can be organized in ways to meet the Israel criteria). This brings out an important point that is sometimes overlooked: there is a difference between a worthwhile organization or activity and one that is an appropriate object for external assistance. The institutions that can productively use external support form a specialized subset of the group of viable or potentially successful institutions. Analysis is required at two stages. Is a public economic institution desirable, or likely to be effective?If so, is external assistance required, and does a candidate-funding institution have the competence (comparative advantage) to supply that assistance? Evaluating the institutional performanceof aided projectsis tricky. Elaborately designed externally funded projects sometimes fail because they demand an unattainable level of managerial performance. And those that succeed might do so because they divert scarce managerial talent from other activities. If so, there are hidden costs, particularly where the response to a weak implementation structure is to use external funds to set up parallel institutions. A Summary of Factors Affecting Performance A bureaucracy can only be as effective as the tasks delegated to it allow: if it is called upon to undertake economic management tasks that are incoherent or inconsistent, and generate considerable potential rents, it is not surprising if its effectiveness or honesty is impaired. Declining institutional capacity can result when the state's effortsto control economic activity are overextended. Admin- istrative performance is also likely to be correlated with a realistic system of material incentives, and with a decision system within the administrative struc- ture that is able to locate responsibility and reward effective performance. However, the capacity, loyalty, and honesty of bureaucracies vary for reasons that are not simply to be explained by material incentives or the availability of 170 The Role of Institutions <