sK 17528 T~~~~~~~~~~~~~~~~~~~~ih' V\ P,jolllll'xerl' 1(11 I(, rolloill(i I)cw1op*lcl \vl will \1. CIhoma inn \ \ _ \ t~~~~~~~~~~~~~~~~~~~~1)11z I \ III\ IIIII.I'iis \ 1)11 1)olls slc i c llnils V _ \ <>1~~~~~~~~~~o Ilrop)icll Forcl/sls T N E W O R L D B A N K \ vle'.cililI( i I uI;iI1 \nll Poir\ ill R(isI ix mi_(II i:ng5 II c n'misili,,ll o inl I\sscssillg I nIcpm Icl ill oItllliski * _ - | \ /s \_ ~~~~~~~~~~~~~~~~~olll: lhillmillng\Rctrilillilll> 5 w d L v V ) J * < | ~~~~~~~~~~Prgrl is ill M,:CD Col tItir s I _ r \ \1.i1((11~~~~~~~~~~~~~ll 1,.111f ('.ll(1 I l 1)1 ill \('\\ lc( .kllld , ~~~~\11.lt lschlic k . ~~~~~~~~~~~~~~~~()II \\ 1,w 1,(l1er Dcw1opuill", (z>~~~~~~~~~o ltriclItli('Sl(til(I 1,1'\ .c;1'11d('s Rcoi(ll's j f >~~~111(1 \h1'.1olill G,illis, _~~ ~ ~ ~~ ~~~~~~~~~~~~~~~~~~~~ 7CmI \illi oll Dlor1 il ol l I s .1 1\( 1)1\ (> o Dc1or(-k1lliotl / .~~~~~~~~~~~~111(1 I`ors'l^ slm1....Ids / 1~~~~~~~~~~~~'S'\ 027-)032,,/K TH E W OR L D B AN K Research Observer EDITOR Moshe Syrquin COEDITORS Shantayanan Devarajan, Shahid Yusuf CONSULTING EDITOR Elinor Berg EDITORIAL BOARD Angus Deaton (Princeton University); Barry Eichengreen (University of California, Berkeley); Howard Pack (University of Pennsylvania); Claire Liuksila (International Monetary Fund); Martha De Melo, Gershon Feder, Gregory K. Ingram, George Psacharopoulos, Michael Walton (World Bank) The World Bank Research Observer is intended for anyone who has a professional interest in development. Observer articles are written to be accessible to nonspecialist readers; contributors examine key issues in development economics, survey the literature and the latest World Bank research, and debate issues of development policy. Articles are reviewed by an editorial board drawn from across the Bank and the international community of economists. Inconsistency with Bank policy is not grounds for rejection. The journal welcomes editorial comments and responses, which will be considered for publication to the extent that space permits. On occasion the Observer considers unsolicited contributions. Any reader interested in preparing such an article is invited to submit a proposal of not more than two pages to the editor. Please direct all editorial correspon- dence to the Editor, The World Bank Research Observer, at the address in the copyright notice below. The viewvs and interpretations expressed in this journal are those of the authors and do not necessarily represent the views and policies of the World Bank or of its Executive Directors or the countries they represent. The World Bank does not guarantee the accuracy of the data included in this publication and accepts no responsibility whatsoever for any consequences of their use. When maps are used, the boundaries, denominations, and other information do not imply on the part of the World Bank Group any judgment on the legal status of any territory or the endorsement or acceptance of such boundaries. The World Bank Research Observer is published twice a year in February and August. Single copies may be purchased for US$12.95. Subscription rates are as follows: Individuals Institutions 1-year subscription US$25 US$40 2-year subscription US$45 US$75 3-year subscription US$60 US$100 Orders should be sent to: World Bank Publications, Box 7247-7956, Philadelphia, Pennsylvania 19170-7956, U.S.A. Readers with mailing addresses in developing countries and in socialist economies in transition are eiigible to receive complimentary subscriptions on request. Written confirmation is required every two years to renew such subscriptions. 1998 The International Bank for Reconstruction and Development / THE WORLD BANK 1818 H Street, N.W., Washington, D.C. 20433, U.S.A. All rights reserved Manufactured in the United States of America ISBN 0-8213-4113-8; ISSN 0257-3032 Material in this journal is copyrighted. Requests for permission to reproduce portions of it should be sent to the Office of the Publisher at the address in the copyright notice above. The World Bank encourages dissemination of its work and will normally give permission promptly and, when the reproduction is for noncommercial purposes, without asking a fee. Permission to copy portions for classroom use is granted through the Copyright Clearance Center, Suite 910, 222 Rosewood Drive, Danvers, Massachusetts 01923, U.S.A. The World Bank Research Observer is indexed by the Journal of Economic Literature, the Standard Periodical Directory, ABI/INFORM, the Index to International Statistics, the Public Affairs Information Service, Inc., the Social Sciences Citation Index, and, online, by the Economic Literature Index and DIALOG. It is available in microform through University Microfilms, Inc., 300 North Zeeb Road, Ann Arbor, Michigan 48106, U.S.A. TH E W OR L D B AN K Re search Observer Volume 13 * Number I - February 1998 Creating a Legal Framework for Economic Development Richard A. Posner 1 The Domestic Benefits of Tropical Forests: A Critical Review Kenneth M. Chomitz and Kanta Kumari 13 Poverty in Russia during the Transition: An Overview Jeni Klugman andJeanine Braithwaite 37 How Bad Is Unemployment in Tunisia? Assessing Labor Market Efficiency in a Developing Country Martin Rama 59 Evaluating Retraining Programs in OECD Countries: Lessons Learned Amit Dar and Indermit S. Gill 79 Public Sector Reform in New Zealand and Its Relevance to Developing Countries Malcolm Bale and Tony Dale 103 Why Most Developing Countries Should Not Try New Zealand's Reforms Allen Schick 123 Deforestation and Forest Land Use: A Comment Jeffrey R. Vincent and Malcolm Gillis 133 Deforestation and Forest Land Use: A Reply William F Hyde 141 Creating a Legal Framework for Economic Development Richard A. Posner A modernizing nation ' economic prosperity requires at least a modest legal infrastructure centered on the protection ofproperty and contract rights. The essential legal reforn re- quired to create that infrastructure may be the adoption of a system of relatively precise legal rules, as distinctfrom more open-ended standards or a heavy investment in upgrad- ing the nation sjudiciary. A virtuous cycle can arise in which initially modest expendi- tures on law reform increase the rate of economic growth, in turn generating resources that will enable more ambitious legal reforms to be undertaken in the future. It used to be common in the economic literature on development to enumerate the multitudinous sources of market failure and prescribe complex government inter- ventions to cure them without paying much attention to the equally numerous sources, especially in poor countries, of governmental failure (Stiglitz 1994). A 1995 confer- ence on legal reform, as well as other recent judicial reform initiatives (Rowat, Malik, and Dakolias 1995; Dakolias 1996), attest to the growing awareness that the failure of governments in poor countries to provide the basic framework of a capitalist economy may be an important factor in keeping poor countries poor. Markets are more robust than some market-failure specialists believe. But their vigor may depend on the establishment of an environment in which legal rights, especially property and contractual rights, are enforced and protected-an environment that is taken for granted in wealthy nations (see Gray 1991 and Rausser 1992 for good statements of this essentially Weberian point). The citizens of wealthy countries take this legal machinery for granted because it works well enough most of the time and because it does not cost a great deal. In its ideal form (an important qualification), the machinery consists of competent, ethi- cal, and well-paid professional judges who administer rules that are well designed for the promotion of commercial activity. The judges are insulated from interference by the legislative and executive branches of government. They are advised by The World Bank Research Observer, vol. 13, no. I (February 1998), pp. 1-11 (C 1998 The International Bank for Reconstruction and Development / THE WORLD BANK 1 competent, ethical, and well-paid lawyers. Their decrees are dependably enforced by sheriffs, bailiffs, police, or other functionaries (again, competent, ethical, and well paid). The judges are numerous enough to decide cases without interminable delay, and they operate against a background of rules and practices, such as accounting standards, bureaus of vital statistics, and public registries of land titles and security interests, that enable them to resolve factual issues relating to legal disputes with reasonable accuracy and at reasonable cost to the disputants. Countries differ, however, and even very wide deviations from this capitalist rule- of-law ideal may not seriously compromise economic efficiency. This is an extremely important point for any nation assailed by competing priorities. One need only re- flect on the economic success of certain U.S. states that have elective, highly politi- cized judiciaries of questionable professional competence; on the economic success of East Asian nations such as China and Vietnam in which the rule of law is weak; and on the fact that England, with one of the finest judicial systems in the world, was for many decades among the poorer economic performers in the industrial world. Moreover, India, with a legal system modeled on England's, and with a vast number of lawyers, has underperformed China, which began its economic takeoff in the 1980s when it had only a rudimentary legal system. Legal Systems: Formal and Informal What explains these cases? One possible explanation is that legal reform is too expen- sive, something wealthy nations often do not realize because the relative cost of the institutions that secure the rule of law is small (Posner 1996). Another is that a legal system does more than enforce contract and property rights; it may also enforce bad laws that reduce economic efficiency.' But perhaps the most important explanation is that there are many informal substitutes for the legal enforcement and protection of property and contract rights. These include arbitration, with or without the legal enforcement of the arbitrator's award; reputation, which may be accompanied by retaliation (such as blacklisting people who default on their contracts) (McGrory 1995); merger (so that disputes between independent firms become purely internal); bilateral monopoly, which can provide a substitute for legally enforceable employ- ment contracts (Becker 1993);2 strong-arm tactics, such as those used in illegal mar- kets; and altruism, which enables many family-owned firms to operate effectively outside a legal framework (see Rubin 1994 for a discussion of the situation in Cen- tral and Eastern Europe). The importance of such substitutes is confirmed by the fact that property rights and contract enforcement are methods of coordinating and optimizing economic activity that long predate the state and formal legal institu- tions. Even in the highly litigious culture of the United States, the vast majority of contract and property disputes are resolved informally, and no doubt would be even 2 The World Bank Research Observer, vol. 13, no. I (February 1998) without a threat of resort to law, often an empty threat because of the expense of legal proceedings. Significance for Economic Growth If it is not possible to demonstrate as a matter of theory that a reasonably well- functioning legal system is a necessary condition of a nation's prosperity, there is empirical evidence showing that the rule of law does contribute to a nation's wealth and its rate of economic growth (Posner forthcoming; Barro 1991; Scully 1988). It is plausible, at least, that when law is weak or nonexistent, the enforcement of property and contract rights frequently depends on the threat and sometimes the actuality of violence (Intriligator 1994; McGrory 1995), on family alliances that may be dys- functional in the conditions of a modern economy, and on cumbersome methods of self-protection. These are costly substitutes for legally enforceable rights, as are the discredited "command-and-control" methods used in communist economies. The hidden costs of these substitutes are a bias against new firms, which have no estab- lished reputation to persuade clients that they are reliable, and a bias in favor of simple, simultaneous exchanges over more complex transactions because it is un- likely that legal remedies can be invoked against nonperformance (Hendley, Ickes, and Ryterman 1997). The cumulative costs of doing without law in a modern economy may be enormous. Unfortunately, there may be a chicken and egg problem: a poor country may not be able to afford a good legal system, but without a good legal system it may never become rich enough to afford such a system. Gray (1997) argues that legal and eco- nomic reform should be pursued simultaneously. She notes that without economic reform, demand for legal reform may be weak because the most powerful economic actors will have alternatives to obtain their ends, such as prohibitive tariffs and gov- ernment bailouts. (And, as noted, the absence of law will hurt new firms the most, implying that it may actually help established firms.) Economic reform is thus im- portant on both the demand and supply sides of legal reform: to stimulate the former and to generate the resources necessary for the latter. But this problem must not be exaggerated. As I noted at the outset, economic progress is possible without much-perhaps without any-law. As a modernizing country gradually becomes more prosperous, it will have additional resources for improving its legal system. Given the risk that too heavy an initial investment in legal reform could deprive the productive economy of necessary resources and thus stifle legal and economic reforms, the prudent choice is to defer legal projects that are costly and ambitious and instead to begin modestly. In a discussion of this approach, Hay, Shleifer, and Vishny (1996) examine the possibility of inexpensive legal reforms that may assist in creating preconditions (or at least in facilitating conditions) for the operation of efficient markets. The authors RichardA. Posner 3 point out that it is more costly and time-consuming to create efficient legal institu- tions than to enact efficient rules for the existing inefficient institutions to adminis- ter. The creation and dissemination of a rule involves small fixed costs and (like other information goods) negligible marginal costs, while legal institutions require heavy inputs of high-priced, educated labor. This implies that the rules-first strategy is more advantageous in more populous countries because the average costs are lower. China, the most populous poor country, followed this approach by introducing modern, commercially oriented rules of law at the same time that it liberalized the economy (Potter 1994). RULES FIRST. This strategy can serve as the starting point of several short-term mea- sures for improving legal institutions as well. It is important to emphasize that such measures be both efficient and, less obviously, rules, and to distinguish between substantive and proceduralefficiencies. A rule is substantively efficient if it sets forth a precept that internalizes an externality or otherwise promotes the efficient allocation of resources: a rule forbidding the use of another person's property without consent is an example. A rule is procedurally efficient if it is designed to reduce the cost or increase the accuracy of using the legal system. Examples are: a requirement that contracts be in writing to be legally enforceable; a rule that no claim of infringement of legal rights is enforceable unless filed within three years of the alleged infringement; a requirement that certain disputes, say be- tween employers and employees or between securities brokers and their customers, must be referred to binding arbitration; and a rule entitling the winner of a judg- ment for damages to receive interest on the judgment at the market rate from the date the suit was filed. The first two of these rules are designed to reduce the infor- mation costs of the legal system, the third to reduce the judicial workload by shunt- ing certain disputes to an alternative method of dispute resolution, and the last to enable judges to use delay to cope with a heavy workload without destroying the utility of the legal system to persons whose rights have been infringed. The rule requiring arbitration has additional importance as a method of encour- aging the formation of trade associations and other business groups, useful interme- diaries between the state and the individual or family in a commercial society (Casella 1996). Lawyers tend to ignore such procedural rules, even though they consider themselves experts on procedure, and to emphasize legal doctrine at the expense of methods that actualize it. The procedural rules listed here have received much less attention from lawyers than have the intricacies of doctrine. ADMINISTRATIVE CONSIDERATIONS. I want to make the point that these are "rules" rather than "standards" because determining whether they have been violated is a relatively mechanical, cut-and-dried process rather than one requiring the exercise of discretion or the determination of numerous facts. The trespass rule requires 4 The World Bank Research Ohserver, vol. 13, no. I (February 1998) determining, for instance, where the boundary line is, whether it has been crossed, and whether there was consent; a statute of limitations requires determining only the date on which the alleged infringement occurred and the date on which the suit was brought; and so on. Lawyers being what they are, the actual administra- tion of rules is more complex than I am letting on. But it is simpler than the administration of standards, such as negligence, bad faith, unreasonable restraint of trade, and unconscionability-at least if professional judges rather than lay ju- rors are used to determine violations. This is an important qualification. Stan- dards, to the extent that they are intuitive (such as the concept of due care, which is the heart of the negligence standard), may be easier to understand, accept, and apply than rules, which may be simplified to the point of arbitrariness, as in the case of statutes of limitations. Because understanding and acceptance are impor- tant to achieving voluntary compliance as well as to the sensible decision of cases by lay adjudicators, a mixture of rules and standards is optimal. But in poor coun- tries with weak legal traditions, the tilt should be in favor of rules because they are easier to administer. The relative simplicity of rules has two consequences for the kind of weak judi- ciary that one is apt to find in a poor country. The first is that the application of rules places fewer demands on the time and the competence of the judges and is therefore both cheaper and more likely to be accurate. The accuracy is a little illusory, because it is a property of governance by rules that they never quite fit the complex reality that they govern. But this observation is consistent with their being more efficient than standards if administered by a judiciary that has a limited capability for the kind of nuanced and flexible decisionmaking that standards require. Second, rules facilitate monitoring of the judges and so reduce the likelihood of bribery and the influence of politics in the judicial process. The less discretion a judge has in making decisions, the easier it will be to determine whether a case has been decided contrary to law or whether there is a pattern of favoring one class or group of litigants over another. Adoption of Foreign Lawvs The adoption of such a set of rules is much more easily said than done, but perhaps not so much more easily. There is, to begin with, a long tradition of what is called the "reception" of foreign law. When the American colonies broke away from Great Britain, each new state decided how much of the common law of England would be received into and made a part of the law of the state (Wood 1969). This pattern is typical in former colonies; for example, half a century after independence, the Indian legal system retains the strong imprint of its English origins. But the reception of foreign law is not limited to cases of original imposition. The Japanese and Chinese also borrowed extensively from European legal codes (Ma 1995; Ford 1996). After RichardA. Posner 5 World War II the occupying powers imposed a variety of legal changes on Germany and Japan, for example in competition law and (in Japan) criminal procedure, that have taken root. European Community law is becoming incorporated into the do- mestic law of the member nations. Such grafts do not always take: a notable example was the adoption by most South American countries in the nineteenth century of constitutions modeled on that of the United States. Some of the Western-inspired constitutions in the former com- munist nations of Central and Eastern Europe may encounter the same fate. But constitutional law is a special case; its effectiveness depends on a particularly com- plex cultural and institutional matrix. In other areas, the prospects for transplanting Western law are good. A poor country might do worse than adopt portions of the U.S. Uniform Com- mercial Code (or its European counterpart), a simple and successful set of rules and standards governing primarily sales of goods, negotiable instruments such as checks and letters of credits, and secured transactions such as mortgages. The poor country might want to modify or delete some of the standards in the Code, such as the standard of good faith, and to require (assuming that illiteracy is not widespread) that more contracts be in writing. And it might want to extend the Code, again with various modifications, to transactions that in the United States are left to the com- mon law, such as contracts involving the sale of services rather than goods-for example, insurance and construction contracts. This proposal has, I admit, a somewhat roundabout character. Historically, com- mercial law originated in the customs of merchants enforced privately by arbitration or equivalent informal methods and only later adopted by the courts. The Uniform Commercial Code is in part an attempt to codify commercial practices. To the ex- tent that the business community in a poor country has its own law, it may be better to codify that law than to try to borrow another country's model. But the law may be underdeveloped in a poor country-may in fact be a cause of underdevelopment or a symptom, or both-and the task of codification may require technical skills of drafting and organization that are in short supply. In these circumstances the adop- tion of a foreign code may be the more sensible move. The important point is that both foreign law and the application of local custom as formal law are well-tried methods by which a nation can adopt a legal code without starting from scratch-or needing a Napoleon. It is important, however, to adapt the imported code to the local culture (Rubin 1994), a task for local, not foreign, lawyers, who know something about the country whose law they are borrowing. I do not advise dispatching European or American lawyers to tell a country how to adapt foreign laws to its legal and social institutions and stage of economic development. One approach might be to establish a law re- form commission to rationalize, unify, and modernize national laws, borrowing wherever possible from established foreign models. 6 The World Bank Research Observer, vol. 13, no. I (February 1998) The Judiciary The fundamental tradeoff is between making a rather modest investment in better rules and making a big investment in the judiciary. And it is a tradeoff. If judicial salaries are high enough and tenure sufficiently secure, the judiciary of even a poor country will be able to attract competent and honest lawyers. But highly educated people, who are needed to staff a good court and also to appear before it as advocates, are a very scarce resource in poor countries, making the opportunity costs of a first- rate judiciary and its associated bar of practitioners very high. And if the salaries of one class of officials far exceed those of other officials, it could create a ripple effect throughout the entire civil service, resulting in large fiscal costs and a large drain on the limited talent of the nation. Finally, the political authorities will be reluctant to create a corps of truly independent officials who may constitute a rival center of power; or they may lack the power to protect the judges from private violence if they stand up to powerful interest groups. Moreover, if the political authorities are weak or corrupt, generous compensation may simply increase the value of a judgeship as a patronage plum and result in an actual decrease in judicial quality. The more costly it is to create a high-quality independent judiciary, the more beneficial it is to focus legal reform on the adoption of substantively and procedur- ally efficient rules. In emphasizing this point, however, I do not propose to abandon entirely the task of improving legal institutions. Indeed, if the law's administrative infrastructure is sufficiently weak, even good rules may simply be ignored. This ap- pears to be the case in Russia, which has several modern legal codes on its books. Although a regime of rules reduces the likelihood of financial or political corruption, rules cannot be the complete answer. Countries may also need to alter the structure of judicial salaries. Specifically, the more that judicial compensation is "backloaded" in the form of generous pension rights that are forfeited if the judge is removed from office for incompetence or venality, the greater the incentive of the judge to behave with integrity (Becker and Stigler 1974). If the cost is very great, even if the likeli- hood of being detected is slight, the appointee may be deterred. Another corruption- fighting incremental change worth considering is having judges sit in panels-or with juries-rather than by themselves, to increase the transaction costs of bribery and the likelihood of discovery. Unlike the compensation adjustment, this proposal would cost something because more judges would be needed-even if lay juries were used-since jury trials take longer. Issues of Enforcement Where the suggested approach emphasizing rules over institutional reform is most likely to fall short is in securing people against the threat of government confiscation. It is all very well to have well-defined private property rights determine legal rem- Richard A. Posner 7 edies against the invasion of those rights by private parties and to devise an effective system of contract rights to exchange property among private owners and entrepre- neurs. But these rights may mean little if the state can seize the fruits of successful investment (North and Weingast 1989). Although certain rules, if enforced, will prevent this-rules such as forbidding the state from taking property without just compensation or forbidding discriminatory taxation-their efficacy depends on the willingness of the judges to stand up to government officials. It looks as though we are back to needing the competent, ethical, well paid, and politically independent judiciary so unlikely to be feasible for a poor nation. A solution to this dilemma may be to establish a special court, as in France, whose sole mission is to restrain the government, the Conseil d'tat (Merryman 1996). The judges of this court must be competent, ethical, and well paid, but because the court's jurisdiction will be so circumscribed, the resources required to equip it will be mod- est. If the court is confined to purely economic issues, moreover, the political au- thorities may be willing to tolerate its independence, especially if they understand its importance for the economy. An alternative may be to turn over judicial authority to a regional or international court, although enforcing the decrees of that court against the government of the country may be problematic. Issues of Commitment It should be obvious that effective legal reform depends ultimately on a political will to reform, which in turn is likely to depend on a political will to implement eco- nomic reform. If the dominant political groups in society want economic prosperity and are willing to risk the loss of political control over the economy that modern economic conditions dictate, they will also want legal reform. If they do not want economic reform, the will to adopt legal reform is likely to be absent. I emphasize the importance of modest fiscal outlays in creating a virtuous cycle of legal and economic reform. Remember that economic progress is possible with little- perhaps with no-law and can be stifled by excessive investment in public-sector projects, including legal reform. A small expenditure on law reform can increase the rate of economic growth, which will in turn generate additional resources for more ambitious legal reforms later. Criminal Law I have talked only about property and contract rights, and fleetingly about concepts such as trespass that back up property rights. I have said nothing about the criminal laws or human rights, which are often reciprocals: many basic human rights protect citizens against the overvigorous enforcement of criminal laws. It is sometimes ar- 8 The World Bank Research Observer, vol. 13, no. I (February 1998) gued that economic rights are inseparable from political rights. Dreze and Sen 1995 have pointed out that countries with a free press do not suffer from famines; and the right to vote and the right of free speech place constraints on government that may in turn protect business interests against the risk of confiscatory measures. I am not very confident about the empirical significance of these points, however. Given the role of technology, it is hard to imagine a country concealing a major famine even if its own media are gagged; North Korea has certainly failed. And democracy may not do much for the economy; India has not been economically more progressive than the nondemocratic nations of Asia. As for granting extensive rights to criminals, this is bound to undermine the efficacy of the criminal laws, and by doing so, unsettle property rights. Rights make it harder to convict the guilty as well as the innocent. Sophisticated police forces and prosecutors can apprehend and convict the guilty without trampling on rights; but sophisticated law enforcement is costly. This point is especially salient in countries such as Russia where acquisitive crime is so rampant that it retards economic development. In such countries a strict criminal law and a correspond- ing de-emphasis on the protection of civil liberties may be an important part of legal reform and an important tool for the protection of property and contract rights. So my message is a modest, and perhaps a harsh, one. Legal reform is an impor- tant part of the modernization process of poor countries, but the focus of such re- form should be on creating substantive and procedurally efficient rules of contract and property rather than on creating a first-class judiciary or an extensive system of civil liberties. This is a general prescription, however, and the proper legal structure for an individual country will depend on a host of considerations that I have not attempted to canvas in this overview. Notes Richard A. Posner is chief judge, United States Court ofAppeals for the Seventh Circuit and Senior Lecturer, University of Chicago Law School. This article is based on an address given at the World Bank Workshop on Legal Reform on April 14, 1997. 1 thank Cheryl Gray for inviting me to give the address, Sorin Feiner and Andrew Trask for very helpful research assistance, and Stephen Holmes and three anonymous reviewers for the World Bank Research Observer for many helpful comments on a previous draft. 1. This may explain the negative correlation across countries between the number of lawyers and the rate of economic growth (Murphy, Shleifer, and Vishny 1991). The correlation is misleading because much of the output of lawyers consists of nonmarket goods; but these may not be as impor- tant in poor countries as in wealthy ones. 2. An employee with firm-specific skills is likely to have a better job at this firm than at a different one. Such an employee will be reluctant to quit and the employer will be reluctant to dispense with his or her skills. RichardlA. Posner 9 References The word "processed" describes informally reproduced works that may not be commonly available through library systems. Barro, Robert J. 1991. "Economic Growth in a Cross Section of Countries." Quarterly Journal of Economics 106:407-43. Becker, Gary S. 1993. Human Capital: A Theoretical and EmpiricalAnalysis with Special Reference to Education. 3d ed. Chicago: University of Chicago Press. Becker, Gary S., and George J. Stigler. 1974. "Law Enforcement, Malfeasement, and Compensation of Enforcers." Journal of Legal Studies 3(1): 1-18. Casella, Alessandra. 1996. "On Market Integration and the Development of Institutions: The Case of International Commercial Arbitration." European Economic Review 40(1): 155-86. Dakolias, Maria. 1996. The Judicial Sector in Latin America and the Caribbean: Elements ofReform. Technical Paper 319. Washington, D.C.: World Bank. Dreze, Jean, and Amartya Sen. 1995. India: Economic Development and Social Opportunity. New Delhi: Oxford University Press. Ford, Christopher A. 1996. "The Indigenization of Constitutionalism in the Japanese Experience. Case Western Reserve Journal of International Law 28 (1):3-62. Gray, Cheryl W. 1991. "Legal Process and Economic Development: A Case Study of Indonesia." World Development 19(7):763-78. . 1997. "Reforming Legal Systems in Developing and Transition Economies." Finance and Development 94(3). Hay, Jonathan R., Andrei Shleifer, and Robert W. Vishny. 1996. "Privatization in Transition Econo- mies: Towards a Theory of Legal Reform." European Economic Review 40(315):559-68. Hendley, Kathryn, Barry W. Ickes, and Randi Ryterman. 1997. "Observations on the Use of Law by Russian Enterprises." Post-SovietAffairs 13(1):19-41. Intriligator, Michael D. 1994. "Privatisation in Russia Has Led to Criminalisation." Australian Eco- nomic Review 105 (April-June):4-14. Ma, Herbert H. P. 1995. "The Chinese Concept of the Individual and the Reception of Foreign Law." Journal of Chinese Law 9(2):207-2 1. McGrory, Daniel. 1995. "Civilizing the Russian Underground Economy: Requirements and Pros- pects for Establishing a Civil Economy in Russia." Transnational Law and Contemporary Prob- lems 5(1):65-98. Merryman, John Henry. 1996. "The French Deviation." American Journal of Comparative Law 44:109-16. Murphy, Kevin M., Andrei Shleifer, and Robert W. Vishny. 1991. "The Allocation of Talent: Im- plications for Growth." QuarterlyJournal of Economics 106(2):503-30. North, Douglass C., and Barry R. Weingast. 1989. "Constitutions and Commitment: The Evolu- tion of Institutions Governing Public Choice in Seventeenth-Century England." Journal of Eco- nomic History 49(4):803-32. Posner, Richard A. 1996. "The Cost of Rights: Implications for Central and Eastern Europe-and for the United States." Tulsa Law Journal 32(1):1-19. . Forthcoming. "Equality, Wealth, and Political Stability." Journal of Law, Economics, and Organization. Potter, Pitman P. 1994. "Riding the Tiger: Legitimacy and Legal Culture in Post-Mao China." China Quarterly 138:325-58. 1 0 The World Bank Research Observer, vol. 13, no. I (February 1998) Rausser, Gordon C. 1992. "Lessons for Emerging Market Economies in Eastern Europe." In Chris- topher C. Clague and Gordon C. Rausser, eds. The Emergence of Market Economies in Eastern Europe. Cambridge, MA: Blackwell Publishers. Rowat, Malcolm, Waleed H. Malik, and Maria Dakolias, eds. 1995.Judicial Reform in Latin America and the Caribbean: Proceedings of a World Bank Conference. Technical Paper 280. Washington, D.C.: World Bank. Rubin, Paul H. 1994. "Growing a Legal System in the Post-Communist Economies." In Cornell International Law Journal (27)1:1-47. Scully, Gerald W. 1988. "The Institutional Framework and Economic Development." Journal of Political Economy 96(4):652-62. Stiglitz, Joseph E. 1994. "The Role of the State in Financial Markets." In Proceedings of the World Bank Annual Conference on Development Economics 1993. Washington, D.C.: World Bank. Wood, Gordon S. 1969. The Creation of the American Republic 1776-1787. New York: Norton. RichardA. Posner 1 1 The Domestic Benefits of Topical Forests: A Critical Review Kenneth M. Chomitz * Kanta Kumari Many forest conservation projects seek to preserve biodiversity by protecting habitats from exploitation or degradation. Although such efforts are often motivated byglobal concerns, habitatprotection also yields domestic benefits. Some of these are intangible or difficult to quantify; others, such as watershed protection and the production of nonforest timber products, are immediate and tangible. There are two rationales for quantifying the domestic benefits of habitat conservation. The first is motivational. Host countries capture only a small proportion of the global benefits which stem from biodiversity conservation. Demonstration ofpalpable local ben- efits could help to build supportfor biodiversity-orientedprojects. Second, the magnitude of'domestic benefits could influenceprojectfinancing. Sufficiently large net domestic ben- efits could justify financing of a project on narrow economic grounds, with biodiversity conservation as a by-product. This review finds that the quantifiable benefits offorest preservation in providing hydrologicalservices and nontimberforestproducts are highly variable. Locally important in some situations, these classes ofdomestic benefits may in general be smaller than popu- larly supposed. This underscores the need for financing conservation from the Global Environmental Facility or otherglobal sources rather than placing the burden entirely on domestic resources. This article focuses almost exclusively on forests in the humid tropics and on two of their potentially most important benefits: hydrological benefits such as erosion control and regulation of stream flows; and nontimber forest products, such as rubber, rattan, fruit, and nuts. Hydrological effects are emphasized for three reasons. First, forests are assumed to be economically important for preventing soil erosion and flooding, pro- tecting the water supply, and maintaining rainfall patterns (Botkin and Talbot 1992, p. 51; Myers 1995). Second, these assertions are often made with little supporting The World Bank Research Observer, vol. 13, no. 1 (February 1998), pp. 13-35 (i 1998 The International Bank for Reconstruction and Development / THE WORLD BANK 13 evidence. The claims are often seen as a priori plausible or even obvious, although the scientific literature has been questioning some elements of this received wisdom for at least a decade (Hamilton and King 1983). Only a handful of economists have at- tempted to measure the value of these hydrological functions, and this small literature appears not to have been fully integrated with the scientific literature. Finally, hydro- logical impacts are potentially of great interest for domestic policy because they involve local externalities: upslope actions affect downslope populations. Also of interest are nontimber forest products, which are increasingly seen as a source of domestic benefits. The emergence of a more extensive literature on nontimber forest products offers an opportunity to assess the valuation of these benefits. Two themes guide this review. Although these themes are not novel and in fact appear to be quite obvious, we believe that they have been insufficiently emphasized in the existing literature. First, benefits must be computed relative to an alternative land use. The literature tends to treat the benefits of habitat protection as an absolute number irrespective of alternative uses of the land. This is ecologically and economically invalid. Hydro- logical functions of the land are strongly related to ground cover, as we discuss at length. Hence, the hydrological impact of converting a natural forest to a plantation might be quite different from converting it to annual cropping, and this will affect the value of maintaining the land as forest. More generally, we are interested in protecting areas precisely because they are in current or future danger of being converted to an alternative use. Therefore, to argue that a particular area should remain protected for economic reasons, the benefit stream provided by the forest must be compared with the benefit stream that would result from the likely alternative. In other words, it is necessary to compute the net benefits of forest preservation: the gross benefits under protection less the forgone benefits from the alternative use (opportunity cost). Second, benefit levels are highly location specific and scale dependent. Habitats in general-and forests in particular-are internally quite heterogeneous. Any sizable forest area is likely to contain many varieties and densities of species, types of soil and terrain, and areas that are more or less accessible to markets. This diversity in turn results in a continuous variation over the landscape in both the physical processes underlying forest benefits and in their economic value. For example: * The value of forest products depends on the density of the valuable species and on the cost of transportation from the extraction site to the consumer. * A forest's recreational value depends on its views, accessibility, and species mix. * Its hydrological value depends on the slope, rainfall, type of soil, position in the watershed, and proximity to dams, fisheries, and irrigation systems. * The opportunity costs of forest preservation depend on how accessible the land is to markets, and the suitability of the soil for crops. 14 The World Bank Research Ohserver, vol. 13, no. I (February 1998) Moreover, several hydrological processes are scale dependent: the dynamics of erosion and runoff, for instance, are quite different in 100-, 10,000- , and 1,000,000- hectare watersheds. Scale also affects markets for nontimber forest products: product prices may decline as the supply increases. As a result the values estimated for a small site cannot easily be extrapolated to a large area; simple scaling-up of site-specific estimates will yield inaccurate, and often biased, results. The Hydrological Benefits Conversion of forest land to other uses can disturb the functioning of the forest ecosystem and ultimately its economic value. (This section draws heavily on Bruijnzeel 1990, and Hamilton and King 1983.) Table 1 shows how changes in land use that affect hydrology are linked to economic impacts. The first set of links involves the effect of changes in land use on river sedimenta- tion. Three questions are evaluated: First, under what conditions does deforestation increase erosion? Second, what is the relation between increased erosion and delivery of the resulting sediment to downstream economic activities? And third, what is the relation between sediment delivery and subsequent economic damage to dams, ca- nals, harbors, and fisheries? From Land-Use Change to Sedimentation There are two links here: from land-use change to erosion, and from erosion to sedimentation. The first is relatively well understood from experimentation and ob- servation on relatively small plots, although most attention focuses on surface ero- sion as opposed to erosion that causes gullies and mass wasting (landslides). A review Table 1. Hydrological-Economic Links Possible hydfrological changes Economic impacts Increased sediment delivery Siltation of reservoirs, canals, harbors Damage to fisheries Improved agricultural productivity from downslope soil depletion Erosion Loss of productivity for downslope farmers Increased water yield (runoff and subground flows) Flood damage to crops and settlements Benefits to downstream water consumers Changes in the water table Agricultural productivity and household water consumption Climate change Impacts on agricultural productivity Source: Authors. Kenneth M. Chomitz and Kanta Kumari 1 5 of the best available summary of eighty studies on erosion is quite striking (Wiersum 1984, reproduced in Bruijnzeel 1990, p. 117). Ground cover, rather than canopy, is the chief determinant of erosion. Erosion rates are low in natural forests and in tree gardens, in the fallow phase of slash-and-burn cultivation, and in plantations where weeds and leaf litter are retained. Erosion rates in plots under current slash-and-burn cultivation are ten times as high as in natural forest. In plantations where weeds and litter have been removed, erosion is more than a hundred times as great as in natural forests (table 2). In many cases erosion may result from road construction associated with logging in the forest rather than from a change in land use. For instance, Hodgson and Dixon (1988) find that the rate of erosion in Palawan, the Philippines, increased four times as a result of logging, but the conversion of uncut forest to road surface increased erosion by a factor of 260. Thus, although roads accounted for only 3 percent of the surface area in the area studied, they were estimated to account for 84 percent of the surface erosion. Gully erosion and mass wasting are also important sources of sediment, but these processes are more complex than sheet erosion, and less is known about them. Still, it seems fair to conclude that little sedimentation-related damage results from converting natural forests to appropriately managed plantations, agroforestry, moderate grazing, and shifting cultivation with long rotation periods. Road con- struction, annual cropping, and plantations that remove litter can generate consid- erable erosion, however. To assess the potential damage, we turn to the next ques- tion: Will surface erosion induced by a change in land cover result in major increases in sedimentation? Table 2. Relation between Land Cover and Erosion (tons per hectare per year) Surface erosion Type of land cover Minimum Median Maximum Natural forests 0.03 0.3 6.2 Shifting cultivation, fallow period 0.05 0.2 7.4 Forest plantations, undisturbeda 0.02 0.6 6.2 Multistoried tree gardensb 0.01 0.1 0.15 Tree crops with cover crop/mulch 0.1 0.8 5.6 Shifting cultivation, cropping 0.4 2.8 70 Agricultural intercropping in young forest plantations 0.6 5.2 17.4 Tree crops, clean-weeded 1.2 48 183 Forest plantations, litter removed or burned 5.9 53 105 a. Refers to forests for timber production, as opposed to tree crops. b. A system in which various perennial and sometimes a few annual crops are cultivated simultaneously with trees. Source: Wiersum (1984), reproduced in Bruijnzeel (1990), p. 117. 1 6 The World Bank Research Observer, vol. 13, no. I (February 1998) The answer depends on two factors. First, only a portion of the eroded soil makes its way into rivers and streams; the remainder is trapped (perhaps temporarily) downslope. The amount of sediment deposited varies inversely with the size of the catchment basin; larger basins have more places for the sediment to get caught than do smaller ones. Mahmood (1987) suggests that the sediment delivery ratio-the proportion of eroded material in a watershed that is carried by a stream-declines from almost 100 percent in basins measuring 200 square hectares to about 10 per- cent in basins of a million square kilometers. Sediment delivery ratios tend to be about 0.3 in basins measuring hundreds of square kilometers. And lower ratios are associated with longer sediment transport times, causing a lag between the change in the land cover and the downstream impacts. Second, the induced sedimentation may be large or small relative to existing, or background, sedimentation levels, which vary depending on local geology and the current state of land use in the basin.' Background sedimentation is related to exist- ing agriculture and the configuration of roads within a catchment basin and to un- stable river banks, natural landslides, and commercial dredging for sand and gravel (Enters 1992; Bruijnzeel 1989, 1990). In general, background sedimentation levels are underestimated because sampling rates are too low to capture infrequent but highly erosive episodes (Mahmood 1987; Bruijnzeel 1990). When this bias is not recognized, higher-than-expected siltation rates at new dams are sometimes errone- ously attributed to contemporaneous changes in land use. Given the complexity of erosion and sediment transport processes and their sensi- tivity to biological and geological conditions, is it possible to calibrate the relation between changes in land use and the amount of sediment deposited in a watershed? One approach models erosion and transport over the watershed, using mapped data on precipitation, land cover, and topography. This approach has been used to derive the universal soil loss equation for temperate locations, a simple formula based on land cover, precipitation, and slope. It is generally poorly calibrated, especially for tropical areas, and its use is often criticized. Researchers are trying to build more sophisticated models to represent the physical processes of soil particle detachment, transport, and deposition (Rose 1993). An alternative, purely empirical approach relates changes in the sediment load of a river to changes in land cover in the surrounding watershed. The empirical ap- proach is an essential check on theoretical models, but lack of data usually makes it hard to apply. One exception is a study by Alford (1992) that assembled annual time-series data on sediment transport, streamflow, and precipitation for the Ping River in Northern Thailand from 1958 to 1985. Despite a decline in forest cover from 92 percent in 1973 to 73 percent in 1991 in Chiang Mai Province, sediment concentration in the Ping was approximately constant.2 According to Alford, the near-linear relation between streamflow volume and total sediment transport "im- plies a sediment source within the stream channel rather than erosion from slopes Kenneth M. Chomitz and Kanta Kumari 17 contributing sediment to this channel" (p. 267). This somewhat surprising conclu- sion-that significant deforestation was not accompanied by increased sedimenta- tion-underscores the need for empirical studies to explore the role of local geology and topography in modulating the effects of changes in land use on sediment delivery. Calculating the Extent of Economic Damage to Dams The accumulation of river-borne sediment deposited in dams reduces the active storage volume of the reservoir and by so doing slows the output of irrigation, hydroelectric, and flood control services (Mahmood 1987; Southgate and Macke 1989). Moreover, sedimentation limits the effective life of the dam by advancing the date at which capacity is exhausted (Southgate and Macke 1989). Silt also damages turbines and increases the need for dredging. The total costs of siltation are significant; Mahmood (1987) puts annual global costs of lost reservoir capacity at about $6 billion. Chunhong (1995) reports that sedimentation reduces the stor- age capacity of China's reservoirs by 2.3 percent annually. The relevant question, though, is the marginal effect of deforestation-related sedimentation. The benefits of forest preservation in a dam's catchment area depend not only on the amount of sediment generated but also on the per hectare benefits provided by the dam, whether sediment is directed past the dam, and the timing of sediment-related damage (if not averted). Consider two hydroelectric plants with the same generating capacity, one in a small steep-sided watershed, the second in a broad, shallow, gently sloping one. For- est protection will prevent erosion and sedimentation in both. The benefits of saving a forest hectare will be greater in the narrow watershed not only because erosion rates are higher on hills and sediment delivery faster in steep narrow valleys, but also because the dam benefits per hectare of watershed are higher. - Sediment management. A variety of engineering and operational "fixes" can be used to sluice incoming sediments past dams or to flush accumulated sediments out of reservoirs (Lysne and others 1995; Chunhong 1995). These techniques are not universally applicable and typically have opportunity costs (in downtime or dimin- ished output) in addition to capital costs. They also shift the costs of sedimentation downstream. To the extent that dams can minimize sedimentation using these tech- niques, however, the value of erosion prevention strategies such as forest preserva- tion are reduced. * Timepath ofbenefits. Because siltation is a gradual process, measuring the amount of damage it does depends on how much time elapses before the buildup reduces the dam's effectiveness. Three potential time lags can occur between the initiation of a change in land use and a diminution in the benefits provided by the dam. First, the rate of land-use change matters. Clearcutting or the construction of low-quality log- ging roads could generate substantial amounts of erosion quickly. Conversely, in- I a The World Bank Research Observer, vol. 13, no. I (February 1998) creasing the intensity of shifting cultivation might take decades to make a substantial change in basin-wide erosion. Second, it takes time for sediment to travel. An eroded soil particle works its way down a watershed through a process of redeposition and resuspension. The amount of time between initial erosion and arrival in the reservoir depends on the gradient of the stream as well as on distance. Harden (1993) notes in connection with the Paute watershed in Ecuador that "sediment eroded from agricultural lands in distant, low- gradient tributary catchments may not reach the reservoir in the next half-century, but increased sediment loads in proximal, high-gradient tributary rivers represent an immediate sedimentation hazard" (p. 183). Conversely, sediment may continue to flow into rivers for twenty to thirty years after source erosion stops (Bruijnzeel 1990, Mahmood 1987). The third lag is the time between the arrival of the sediment at a reservoir and the drop in dam output. Although reservoirs are built with dead storage capacity de- signed specifically to catch sediment, significant amounts are deposited in the active storage area, potentially reducing output. Additional sediment would thus be ex- pected to affect dam services immediately. Southgate and Macke (1989) found, how- ever, that earlier retirement of the dam (as opposed to decreased output before retire- ment) accounted for 85 percent of the economic impact of an increase in sedimentation rates. Although these processes are quite complex, a simple numerical example illus- trates the sensitivity of economic impacts to assumptions about the timing of depos- its and discounting of dam benefits. Assume that dam services are constant until the dam is retired and that the effect of watershed damage reduces the expected lifetime of the dam from 100 years to 61 years. Given a 10 percent discount rate, the net present value of watershed protection is about 2.2 percent of the annual flow of dam benefits. A modest increase in the discount rate, to 12 percent, decreases the present value of watershed protection by 75 percent. The introduction of a twenty-year lag between the change in land use and the onset of sediment inflows decreases the present value of protection by a further 95 percent, to just 0.04 percent of the annual benefit flow.3 In short, the benefits of extending the life of a relatively young dam will tend to occur in the distant future and therefore will be highly discounted. Empirical Studies The theoretical bent of this discussion on dams reflects a paucity of empirical stud- ies. Only five estimates of the economic impact of changes in tand use on dam per- formance can be found: Briones (1991); Cruz, Francisco, and Conway (1988); Southgate and Macke (1989); Veloz and others (1985); and Shahwahid and others (1997). The difficulty of gathering primary data on erosion processes is evident. Only the last study is based on empirical analyses relating actual sedimentation to Kenneth M. Chomitz and Kanta Kumari 19 actual changes in land use. Only one study allows for a lag between the project's initiation and the impact of increased sediment. On the economic side, only one attempts to model in detail the process by which sediment reduces the life of a dam, and none incorporates the models used by dam engineers to describe the patterns of sediment buildup in reservoirs. The per hectare benefits differ widely among these five studies. The highest value by far, more than $2,000, refers to a subset of the interventions envisioned for the Dominican Republic's Valdesia watershed management project, namely, the refor- estation of the steepest slopes in the watershed. Some of the assumptions underlying this estimate are open to question, and the 5 percent discount rate elevates the value compared with some of the other studies. Nonetheless, this example illustrates the potential for very high levels of domestic benefits from protecting critical watershed areas. In contrast, Cruz, Francisco, and Conway (1988) found that erosion around the Pantabangan dam in the Philippines resulted in loss of dam services equivalent to about $4 a year per hectare of converted forest. The implication is that forest protec- tion would have provided benefits of that magnitude-somewhat less than $80 per hectare if capitalized at 5 percent.4 These forest benefits, however, are not netted against the opportunity costs of restricting agricultural use, so net domestic benefits are lower. At the same time, Cruz, Francisco, and Conway argue that maintenance of forest cover would also have yielded substantial additional benefits by allowing the dead storage capacity of the dam to be converted to active storage for irrigation. The lowest net value for forest protection is reported by Shahwahid and others (1997) in an analysis of the Hulu Langat Forest Reserve in Malaysia. Here the alter- native to complete forest protection is permitting low-impact logging of the forest. (A noteworthy feature of such logging is that it prohibits cutting within twenty meters of a river or stream; according to the authors, this restriction reduces logging-related sedimentation by 60 percent, even though the stream buffers occupy less than 20 percent of the forest area.) In this case forest protection yielded a gross annual benefit of $44 per hectare. The opportunity cost of prohibiting logging was about $1,400 per hectare, however, reflecting the high density of commercial tree species in Ma- laysian forests. Thus the net benefits of forest protection, relative to low-impact log- ging, are -$1,356. Low-impact logging, as a land use, might have high benefits rela- tive to alternative land uses, however; these have not been evaluated. Impact on Fish and Aquatic Organisms In an analysis of carefully gathered primary data, Hodgson and Dixon (1988) exam- ined the impact of sedimentation on marine life in Palawan, the Philippines. They found that logging in the area had led to the construction of highly erosion-prone roads quite close to the Manlag River, within a few kilometers of Bacuit Bay. The 20 The World Bank Research Observer, vol. 13, no. I (February 1998) consequence was a very large and immediate increase in sediment that was "often more than 1,000 milligrams per liter," while sedimentation levels in a control river "rarely exceeded 10 milligrams per liter." The increased sedimentation destroyed nearly 50 percent of the coral cover on the reef nearest the mouth of the river. Al- though the levels of sediment were not high enough to kill the fish directly, coral mortality severely disrupts the ecosystems on which the fish depend. Hodgson and Dixon impute the per hectare value of forest protection at a high $3,200, which overestimates the social gain because it is based on gross revenues from fisheries and tourism rather than on net profits. It is also worth noting that the authors rule out, as infeasible, interventions to reduce road-generated erosion, even though such interventions may save the loggers money by reducing maintenance costs. Because roads generate the bulk of all sediment, improved road-building tech- niques might make logging, fisheries, and tourism mutually compatible. Erosion and Agricultural Productivity If deforestation causes an increase in on-site erosion and a loss of agricultural pro- ductivity, can that loss be translated directly into forest preservation benefits? The answer is no, if forests and crops are mutually exclusive land uses. Once the forest has been converted to agriculture, erosion diminishes agricultural yields. But that rate of diminution is not the benefit of forest preservation. Forest preservation, however, can yield agricultural productivity benefits through erosion reduction in two situations. First, some woodlands or open forests are used for grazing or cropping. Removing the trees in these areas to intensify production could be self-defeating if erosion increases drastically. Second, deforestation could result in increased runoff and thereby increase erosion on downslope croplands. This seems plausible and may be important in some areas, but we can find no relevant studies. Deforestation might also cause downslope damage from land- slides. Conversely, erosion sometimes delivers valuable soils from uninhabited hill- sides to farmers' fields (Enters 1992). Where that is true, forest preservation im- poses external costs on those farmers. But these effects may be limited to exceptional soil conditions. Impact of Land-Use Changes on Water Yield Popular belief and casual empiricism link deforestation with flooding. If it is true that upslope deforestation threatens downstream cities and croplands with flood damage, the gains to forest preservation might be quite large. In fact, extensive scien- tific evidence links deforestation to annual increases in water yield (that is, the total volume of surface runoff and subsurface flows). As in the case of erosion, the increase depends on how the land is used (Bruijnzeel 1990, pp. 82-92). But increases in the Kenneth M. Chomitz and Kanta Kumari 21 average rate of flow do not necessarily correspond to increases in peak flow or storm flow, which cause floods. Surprisingly, the scientific literature supports a link between deforestation and flooding only at a local level-within a drainage basin of less than about 50,000 hectares (Bruijnzeel and Bremmer 1989). In small watersheds increases in water yield translate directly into increases in storm flow. For larger drainage basins, however, the limited number of studies using long time-series data on floods show no link between deforestation and flooding. Bruijnzeel (1990) cites studies of medium-size drainage basins (up to 1.45 million hectares) in Taiwan (China), and Thailand, which show that extensive deforestation had no effect on flooding, and three studies of India for the period 1871-1980, which show no trend in the frequency of flooding despite massive changes in land use during this time. Bruijnzeel and Bremmer (1989) also argue that there is no relation between changes in land use in the Himalayas and flooding in the Ganges-Brahmaputra basin, although they do not present time-series data. And Anderson, da Franca Ribeiro dos Santos, and Diaz (1993), who analyze eight decades of time-series data on rainfall and storm flow in the Parana-Paraguay river basin, report no structural shift in the relation between intense rainfall and floods, despite the significant conversion of forests over that period to pasture and cropland. At first glance, these results seem paradoxical: How can deforestation cause flood- ing in small basins but not in large basins? The hypothesis is that basin-wide flood- ing depends more on intensity of the rainfall than on the way land is used. Most storms are small and transient. Individual subbasins tend to flood in sequence, as the storm passes over, rather than simultaneously. Local floods are thus averaged out over space and time. Only extremely severe and long-lasting storms affect all the tributaries of a major river at once. Storms of that magnitude would be large enough to saturate the soil's absorptive capacity and cause rapid runoff even if the land were still forested (Hamilton 1987; Bruijnzeel and Bremmer 1989; Bonell and Balek, 1993, pp. 227-28). Impact of Land-Use Change and Dry-Season Flows Since the time of Plato, it has been assumed that deforestation results in lower water tables and reduced flows of water during the dry season (Grimble, Aglionby, and Quan 1994). This belief is still current; Huntoon (1992) links the loss of the "green reservoirs" of hillside forests in South China to severe reductions in the availability of groundwater during the dry season. According to current hydrological science, however, the effects of deforestation on dry-season flows are ambiguous but likely to be counterintuitive. (Bonell and Balek 1993; Bruijnzeel 1990). This is because the conversion of forests to other purposes has two opposing effects on the water table. On the one hand, it increases 22 The World Bank Research Observer, vol. 13, no. I (February 1998) runoff and decreases the absorption of water into the ground. This by itself would lower the water table. On the other hand, trees are highly effective water pumps, removing water from the soil and transpiring it into the air. The replacement of trees by vegetation with shallow roots and lower transpiration rates (such as grass, annual crops, or early stages of secondary regrowth) therefore tends to reduce groundwater loss and raise the water table. Dozens of controlled experiments have been conducted showing that, contrary to expectations, the net immediate effect of tree removal is a rise in the water table, and therefore a probable increase in dry season flows (Hamilton and King 1983). Similar results have been found in studies of actual sites. Nepstad and Schwarzman (1992) compare deep-rooted evergreen forests to an adjacent de- graded pasture in Para, Amazonia. At the end of the dry season, the water in the top eight meters of soil available for plants was 370 millimeters higher in the degraded pasture. In an interesting case study on Thailand, Vincent and others (1995) found that reforestation reduced, rather than increased, dry season flows and imposed costs on downstream users. Starting in 1967 Thai authorities promoted refores- tation and sedentary agriculture in deforested areas of the Mae Theng water- shed. These efforts involved two water-consuming interventions: the construc- tion of irrigation systems and the establishment of pine plantations, which transpire more water than the deciduous forests that originally covered the area. An analysis of monthly stream flow records showed no change in dry season flows from 1952 to 1972 but registered a significant reduction from 1972 to 1991, when annual stream flows slowed by an additional 2.9 million cubic meters each year. These reductions resulted in the seasonal closure of one of Chiang Mai's water treatment plants and forced downstream farmers to switch from rice to soybean cultivation. The marginal costs of these reductions in water availabil- ity ranged from about 1 baht per cubic meter for agriculture to 7 baht for indus- trial users. These results imply that upslope deforestation, while highly undesir- able on many grounds, yielded external benefits rather than costs for downstream water users. Under some circumstances, however, deforestation may indeed reduce water tables. Bruijnzeel (1990) and Bonell and Balek (1993) point out that many processes in- volved in forest conversion compact the soil and cause gullying. Such processes include overgrazing, road construction, and the use of heavy machinery for land clearance. Compaction and gullying, in turn, increase runoff and decrease infiltra- tion. If infiltration is reduced more than transpiration, the water table could drop.5 Hamilton and King (1983) cite Australian studies showing severe reductions in infil- tration following heavy grazing. They also cite a Fiji study finding runoff rates of 90 percent on grassland. They were unable to find analogous results, however, after forests were converted to annual cropping. A different situation is described by Kumari (1995). In this case, selective logging of a peat swamp forest entailed the construc- Kenneth M. Chomitz and Kanta Kumari 23 tion of drainage canals. Expansion of the drainage network could reduce water stor- age sufficiently to imperil dry-season rice production in adjacent fields. Climate Maintenance There is a long-standing belief that deforestation reduces rainfall. Grove (1994) pro- vides a fascinating account of scientific and policy interest in the topic dating to the seventeenth century (see, for example, Halley 1694). To the modern observer, too, it seems intuitively obvious that tropical deforestation reduces rainfall. Because evapo- transpiration from tropical forests makes up between 20 percent (Southeast Asia) and 80 percent (Africa) of incident rainfall (Wilkie and Trexler 1993), it seems logi- cal to expect that forest removal would break this recycling process, resulting in a drier climate. Modern climate theory, however, introduces a host of additional complexities. Changes in land cover introduce not only changes in evapotranspiration, but also in albedo (surface reflectivity) and aerodynamic drag. These changes direcdy affect tem- perature and precipitation and also set off a whole round of positive and negative effects involving changes in cloudiness, air circulation patterns, and even plant tran- spiration. The result is a highly nonlinear, scale-dependent, dynamic system. No longer is it clear a priori that deforestation reduces local rainfall. Eltahir and Bras (1992) suggest, for example, that deforestation on the scale of hundreds of square kilometers increases convection and therefore rainfall, while deforestation on the scale of millions of square kilometers reduces rainfall. In any case the magnitude and spatial distribution of climate effects will be sensitive to local conditions, especially to the nature of the vegetation that replaces the forest. Theoretical analysis of the climatic impact of changes in land use therefore re- quires sophisticated models. During the past ten years, general circulation models of the earth's atmosphere have been used to analyze the effect of large-scale deforesta- tion on global climate. Several exercises have examined the implications of convert- ing the entire Amazon or Southeast Asian rainforests to savanna. In principle, these exercises might be used to evaluate the domestic benefits of forest preservation for large countries such as Brazil or Indonesia. Henderson-Sellers and others (1993) predict that complete deforestation of the Amazon would reduce precipitation in the rainy season by 30 percent, while complete deforestation of Southeast Asia would have no effect on precipitation. Lean and Rowntree (1993) predict that total defor- estation of the Amazon would reduce local rainfall by 14 percent, but increase rain- fall in Eastern Brazil by 20 percent. These results must be interpreted with caution for several reasons. First, the scale and permanence of the deforestation simulated in these exercises is unrealistic. Shukla, Nobre, and Sellers (1990), for instance, assume that the entire Amazon would be reduced to degraded pasture. Many of the deforested areas in the Amazon in fact 24 The World Bank Research Observer, vol. 13, no. I (February 1998) revert to secondary forest (Moran, Mausel, and Wu 1994), whose climatological properties are much closer to primary forest than to pasture. Second, despite their sophistication, general circulation models omit a range of physical processes and rely on a great many assumptions about parameters. How sensitive the results are to these omissions and assumptions is unknown, however. Third, these models divide the planet's surface into a very coarse grid and can only be applied to deforestation pro- cesses at the scale of tens of thousands-or more-of square kilometers. Results at this scale cannot be generalized to deforestation patches of tens or hundreds of square kilometers. Work on more appropriate mesoscale models is still in its infancy. Empirical work in this area is as inconclusive as the theoretical work. Bruijnzeel (1990) reviews the thin literature. There are a great many microstudies of temperature and soil wetness changes in small clearings, but these cannot be generalized to larger scales and are useful mainly to establish the parameters for general circulation and mesoscale models. A limited number of mesoscale empirical studies try to relate changes in forest cover to changes in recorded precipitation (see, for example, Meher-Homji 1988), but Bruijnzeel notes that these are lacking in rigor and data consistency. Current research is just beginning to use remote sensing data to track climatic and land-use changes, resulting in more rigorous studies. Complementing twenty-two years of data from twenty climate stations in the Selva Lancondona region of Chiapas, Mexico, O'Brien (1995) used remote sensing data from 1979 and 1989 to track deforestation around each station. Preliminary analysis indicates that deforestation increases minimum temperature, decreases maximum temperature, and has no sig- nificant effect on precipitation. Cutrim, Martin, and Rabin (1995) use satellite data to show increases in cloudiness (not necessarily implying increased precipitation) following large-scale deforestation in Rondonia, Brazil. In sum, the assumption that deforestation affects local climate is plausible, but the magnitude (and indeed sign) of the effect remains to be demonstrated. The potential economic significance of climatic effects should make them an early priority for research. Commercial Value of Nontimber Forest Products The value of tropical nontimber forest products, such as fruits, nuts, latex, resins, medicines, and animals, has been reviewed by Godoy, Lubowski, and Markandya (1993) and Lampietti and Dixon (1995). In a survey of twenty-four studies, Godoy, Lubowski, and Markandya report per hectare values ranging from $.75 to $422 a year, with a median of about $50. But these studies, as a group, exaggerate the level of benefits that would accrue to the domestic economy from a typical natural, or old growth, forest. They do not distinguish between agroforestry and pure extraction, nor do they take into account the costs of extraction. Furthermore, they do not allow Kenneth M. Chomitz and Kanta Kumari 25 for spatial differences in the densities and therefore quantities of the product or in the rate of extraction. And they fail to allow for competition in the markets for these products in the long term. The Distinction between Agroforestry and Extraction Nontimber forest products can be produced at different levels of intensity, with correspondingly different degrees of disturbance of the original ecosystem. At one extreme are purely extractive systems, where extractors harvest products from an otherwise undisturbed forest. In a slightly more intensive approach, extractors may artificially enrich the forest with a desired plant species. Still more intensive are a range of agroforestry techniques that replace the primary forest with carefully ma- nipulated multispecies plantations to provide raw materials for trade and industry. To give a true picture of the benefits of preserving "natural" forest, nontimber forest product valuation must be based on the profits from purely extractive systems, rather than profits from agroforestry systems. It is tempting to rely on the latter. Agroforestry systems typically generate higher values per hectare because commercial species are planted more densely and extraction and processing costs are lower. But although these systems are attractive for many reasons, including their relatively high degree of biodiversity, they are not the same as the natural ecosystems they replace and therefore should not be used to justify preservation of the natural forest. At the same time, it is worth stressing that agroforestry systems may offer both greater biodiversity and higher economic benefits than other uses of the land. This is true of a Sumatran system in which slash-and-burn farmers create rubber-rich sec- ondary forests. After the rubber trees reach maturity (about ten years), they yield about 600 kilograms per hectare annually of dry-equivalent rubber (van Noordwijk and others 1995, p. 88) with no inputs other than labor (current rubber prices are about $1.60 a kilogram). Extraction can continue for twenty years or more before another cycle of clearing. Because the owners' share in the typical tapping arrange- ment is one-third, per hectare rents are substantial. At the same time, ecological studies show fairly high levels of species richness (Michon and de Foresta 1995; Thiollay 1995). Moreover, these rubber forests should have the same hydrological properties as primary forests. Thus agroforests may in some cases offer both net domestic economic benefits and global or nonmarket benefits relative to competing land uses. Allowing for the Costs of Extraction Conceptually, the value of a hectare of forest for nontimber products is equivalent to the rent that would be paid for the right to harvest that hectare. Clearly that amount is less than the final value of the product in the marketplace because the costs of 26 The World Bank Research Observer, vot 13, no. I (February 1998) extraction and transportation must be netted from the sales price to yield profit or rent. This elementary point has been made many times in the literature and must be applied to the cases presented by Godoy, Lubowski, and Markandya (1993). For instance, they find the highest documented value of a hectare of forest based on actual extraction rates is Chopra's (1993) estimate of $117-$144 a year for fuelwood, fodder, and miscellaneous products from tropical deciduous forests in India. In the absence of market prices for these goods, Chopra values them either by their cost of extraction or by the price of substitute commodities. For instance, he values labor to gather fuelwood at $18.87 to $24.17 a hectare (while its equivalent in softcoke, an alternative fuel, would cost $9.50 to $17.33). Chopra concludes that the value of forests for fuelwood must lie between $9.50 and $24.17 a hectare. But this conclu- sion confuses costs, benefits, and rents. If softcoke is in fact a close substitute for fuelwood, then the data imply that villagers are expending labor worth more than $18.87 to produce fuelwood worth less than $17.33. Clearly the estimates are crude, but the main implication is that the net per hectare value of the forest for firewood production is close to zero. Similarly, labor expenditures to produce miscellaneous goods such as lacquer and dyes amount to $66.67 a hectare. To calculate the value of the forest in producing these goods, it is thus necessary to subtract $66.67 a hectare from the price paid for the lacquer and dyes. Spatial Variation in Density and Extraction Cost Forests tend to be large and heterogeneous. Estimating the value per hectare for a small plot and attributing that value to the forest as a whole (let alone to any other forest) is inappropriate. This point is obvious but is almost universally ignored in practice. Three difficulties arise in generalizing small-plot estimates. First, transport costs are important for some nontimber forest products; acai fruit, for example, spoils within twenty-four hours of harvesting. The high costs of transporting bulky or perishable goods through the forest means that the value at the point of collection will fall off steeply with distance from the road and from the market. Large portions of the forest will not be economically exploitable for this class of commodities and will thus have zero value for that use. At the same time high-value, less perishable commodities, such as rubber, will be economically viable over a larger area. Second, the density of exploitable species can vary dramatically within and be- tween forests. For instance, two of the highest per hectare values cited in Godoy, Lubowski, and Markandya (1993) refer to oligarchic forests, that is, those domi- nated by a few, highly commercial species (Anderson and loris 1992; Anderson and Jardim 1989). Although interesting and locally important, such examples are quite atypical of tropical rainforests, whose hallmark is very high diversity and thus very low densities for any individual species. Kenneth M. Chomitz and Kanta Kumari 27 Third, consumption of some nontimber forest products may tap only a small fraction of the potential supplying area. In Ecuador, for instance, Grimes and others (1994) studied a resin derived from a Protium tree. Harvesting the trees yielded an average potential net return per hectare (after collection, transport, and marketing) of $61 a year in the three forest plots surveyed. The resin, however, is used exclu- sively for finishing local ceramic handicrafts, which are presumably in limited de- mand. We surmise that the total number of hectares being harvested is a small frac- tion of the total area from which the trees could be economically harvested.6 If so, it would be a gross error to apply the $61 value to the entire range of the species. The same situation may apply to valuable medicinal plants. Long-Run Competitive Supply For most commercially attractive products, pure extractive reserves cannot compete with synthetic or domesticated substitutes (Richards 1993; Browder 1989). Many nontimber forest products follow a life cycle in which they start out as extractive products, attain a world market and a high price, and are then domesticated in in- tensive plantation or agroforestry systems. Intensive cultivation reduces labor, land, and capital costs; permits product standardization; facilitates processing; ensures a reliable supply; and takes advantage of scale economies in marketing. As a result, the supply price of the product falls below the viable level for extractive supply. The prime example is rubber. In the early part of this century, Brazilian rubber prices (and extraction volumes) collapsed as Malaysian plantations came on line. In recent years Brazil's extractive reserves have been supported by subsidies. Currently, extractivists in that country's Chico Mendes reserve are being driven out of business by lower-priced latex produced by plantations in Sao Paulo State. Labor productiv- ity in the plantations is about ten times that in the reserve (Brooke 1995). The implication is that lower-cost domesticated or synthetic substitutes greatly reduce-or even eliminate-the rents from extractive reserves. Nontimber forest products provide domestic benefits only when the products are difficult to domesti- cate or duplicate. This gloomy statement requires some qualification, however. First, those forests in which a few commercially valuable species grow may be able to compete with plantations, especially with some small interventions such as pruning (Anderson and Jardim 1989). Second, agroforestry systems such as jungle rubber may in some cases be competitive with plantations while also preserving some biodiversity. Third, and most important, the success of large-scale plantations for nontimber products increases the global urgency of preserving genetic diversity. This may often be best accomplished through preserving natural forests. The implication is that the global benefits of forest preservation may increase even as the domestic benefits decline. 28 The World Bank Research Observer, vol. 13, no. I (February 1998) The Opportunity Costs of Preservation The sustainable benefits associated with forest preservation can be thought of as the gross benefits. The existence of a threat to a forest tract usually implies an economic motive for converting the forest or exploiting it, and these potential benefits are the opportunity costs of preservation. These opportunity costs must be deducted from gross forest benefits to yield the net domestic benefits of preservation. Opportunity costs are highly sensitive to land characteristics. The returns to agri- cultural use of a plot depend on the physical characteristics of the land, current vegetation, market access, and land tenure. Physical characteristics such as slope, drainage, and soil fertility determine the land's relative physical productivity for dif- ferent crops, the need for inputs, and the degree to which output can be sustained over time. The density of commercial tree species also affects the net cost of clearing; in some cases the value of timber may outweigh the benefits from agriculture. Mar- ket access-the cost of transport to the nearest market-determines the potential price paid to the farmer for crops or cattle and their inputs. Land tenure and owner- ship, together with land-related tax and subsidy rules, affect the incentives to invest in land preparation and in perennial crops. The opportunity cost of land-that is, the forgone benefits of conversion-can be obtained through Geographic Information Systems data and techniques. Using such data Chomitz and Gray (1996), for instance, found that the type of soil and the accessibility to roads strongly influence the probability that a forest will be con- verted. Magrath and others (1995) used farm budgets and a land capabilities assess- ment to impute land values in West Kalimantan, Indonesia. The land capabilities assessment divided the province into 1,682 polygons and assessed the suitability of each polygon for a variety of crops. (For lack of data, however, no adjustments were made for the cost of transporting crops to markets or for benefits from timber mar- keting.) The results are quite striking: much of West Kalimantan's land has little value for agricultural production. Of the province's 14.65 million hectares, 3.7 mil- lion have an opportunity cost of less than 20 cents a hectare a year (1991 prices). About 95 percent of the province has an agricultural opportunity cost of less than $2 a hectare annually. Were transport costs factored in, the opportunity costs would be far lower. Summary of Findings The level of net domestic benefits from forest preservation depends on the alterna- tive land use as well as local climatic, biological, geological, and economic circum- stances. When the alternative use is agroforestry or forest plantations (depending on Kenneth M. Chomitz and Kanta Kumari 29 the management system), preservation of the natural forest may not offer net ben- efits from hydrological benefits or the production of minor forest products. At the same time, some agroforests may offer both biodiversity benefits and net domestic economic benefits relative to other land uses. The prospects for economically significant hydrological benefits from forest pres- ervation appear to be smaller than popularly supposed. * Deforestation has not been shown to be associated with large-scale flooding, although it may cause serious local flood damage. * Nor is it associated with diminished dry season flows; on the contrary, it is usually associated with greater flows. * Although it is a priori plausible that deforestation should affect local precipitation, the magnitude and even the direction of the effects are not known, except in the special case of cloud forests, which "harvest" passing moisture. * The link between deforestation and downstream sediment damage is sensitive to the basin topography and geology. Where sediment transport is slow-as in large, low-gradient basins-downstream effects may occur far in the future, so that the net present value of damages is small. Conversely, forest preservation can yield substantial domestic benefits where it averts erosion-generating changes such as road building, annual cropping, or over- grazing; where affected areas impinge directly on streams, reservoirs, coral reefs, or inhabited areas; and where affected watersheds are small, steep, and erosion prone. The measurement of benefits from harvesting minor forest products is still rudi- mentary. Current valuation exercises tend to be specific to particular plots and can- not be generalized to significant forest areas for one of several reasons: a) they are based on inventories of salable products rather than actual extraction; b) their value is based on gross market prices rather than prices net of extraction and transport costs; c) the study describes agroforestry products rather than true extractive prod- ucts; d) the study describes products of unusual forests dominated by a few commer- cial species rather than a typical species-rich rain forest; and e) the value of the forest for extractive production of commercial nontimber forest products is undercut by competition from domesticated or synthetic substitutes. We stress again that our review of benefits is not comprehensive. Potential domes- tic benefits not reviewed here include ecotourism services, sales of bioprospecting rights, and carbon sequestration services, should a carbon offsets market come into existence.7 Moreover forest preservation can yield substantial global or ecological benefits. Also note that these conclusions apply only to tropical moist forests. It may be the case that for other ecosystems, such as wetlands, the links between land use change and economic benefits are both better understood and stronger. Domestic economic benefits provide an uncertain rationale for conservation- and especially for funding forest preservation through market-rate loans. Undoubt- 30 The World Bank Research Observer, vol. 13, no. I (February 1998) edly this rationale is clearly justified for some forest preservation projects, and these should be vigorously pursued. For many projects, however, net domestic benefits either do not exist or cannot be quantified with sufficient rigor to support a market- rate loan or a convincing cost-benefit analysis. Hence the domestic benefits argu- ment-save your forests because they bring palpable economic benefits to your coun- try-cannot be the mainstay of forest preservation in all countries. For many-possibly most-tropical forests, the more compelling rationale for preservation is based on global values. This underscores the need for financing conservation from the Global Environment Facility or other global sources, rather than placing the burden entirely on domestic resources. The hopeful converse, however, is that the net domestic costs of forest preserva- tion may also be small. The argument is: Save your forests because the out-of-pocket costs of doing so are small, and the noneconomic benefits are large. Many biodiverse, carbon-rich forest areas are poorly suited to agriculture because of isolation and poor soils. These areas can be preserved through a three-pronged strategy. First, the op- portunity costs of preservation should be kept low by directing regional develop- ment toward more economically promising districts. Above all, uneconomic road- building should be avoided in these areas. Once roads are in place, the opportunity costs of preservation can increase substantially. Second, where pressures for logging are politically and economically irresistible, low-impact techniques can be required as a condition for access. This approach would entail disabling main access roads after logging was completed. Third, direct and ongoing compensation can be paid, or alternative livelihoods set up, for land users or stakeholders who would otherwise convert the forest to other uses. Notes Kenneth M. Chomitz is with the Development Research Group of the World Bank, and Kanta Kumari is with the Secretariat of the Global Environmental Facility. Preparation of this article was supported by the World Bank's Policy Research and Environment Departments. The authors are grateful to Thomas Enters, William Hyde, Nalin Kishor, and Lini Wollenberg for helpfiul comments. 1. Induced sedimentation may be a concern even when background sedimentation levels are higher if marginal increases in sediment result in further damage and if the costs of averting such sedimen- tation are smaller than those of remedying background sedimentation. 2. Forest cover data supplied by Charles Griffiths, U.S. Environmental Protection Agency. 3. The assumption that dam life with lagged sediment delivery is eighty-one years is very crude but will suffice for illustrative purposes. Also, note that a ratio of 0.04 percent does not necessarily imply that a watershed protection project is uneconomic-the project's value depends on the cost of protecting the watershed, which might be relatively small. 4. Authors' calculations based on Cruz, Francisco, and Conway data. 5. This does not explain Huntoon's reports for South China, because the deforestation resulting from the felling of trees would not be expected to result in soil compaction. 6. What then sustains the net return of $61 a hectare? Why doesn't competition drive these returns toward zero? Three answers are possible: the producing plots are to some extent, perhaps Kenneth M. Chomitz and Kanta Kumari 31 informally, privatized, and the return reflects the opportunity cost of the land; the $61 figure in- cludes returns to the expertise of the collector, who knows how and where to find it; or in fact all economically exploitable areas are being harvested, and the $61 measurement is based on inframar- ginal plots. 7. But see Simpson, Sedjo, and Reid (1996) for a pessimistic view. References The word "processed" describes informally reproduced works that may not be commonly available though library systems. Alcorn, J. B. 1989. "An Economic Analysis of Huastec Mayan Agroforestry Management." In Browder (1989). Alford, Donald. 1992. "Streamflow and Sediment Transport from Mountain Watersheds of the Chao Phraya Basin, Northern Thailand: A Reconnaissance Study." Mountain Research and De- velopment 12(3, August):257-68. Anderson, A. B., and E. M. loris. 1992. "The Logic of Extraction: Resource Management and Income Generation by Extractive Producers in the Amazon Estuary." In K. H. Redford and C. Padoch, eds., Conservation of Neotropical Forests: WTPorking from Traditional Resource Use. New York: Columbia University Press. Anderson, A. B., and M. A. G. Jardim. 1989. "Costs and Benefits of Floodplain Forest Management by Rural Inhabitants in the Amazon Estuary: A Case Study of A,ai Palm Production." In Browder (1989). Anderson, Robert J., Jr., Nelson da Franca Ribeito dos Santos, and Henry F. Diaz. 1993. "An Analysis of Flooding in the Parana/Paraguay River Basin." Dissemination note 5. World Bank, Latin America and the Caribbean Technical Department, Washington, D.C. Processed. Bonell, Michael, and J. Balek. 1993. "Recent Scientific Developments and Research Needs in Hy- drological Processes of the Humid Tropics." In Bonell, Hufschmidt, and Gladwell (1993). Bonell, Michael, Maynard M. Hufschmidt, and John S. Gladwell, eds. 1993. Hydrology and Water Management in the Humid Tropics: Hydrological Research Issues and Strategies for Water Manage- ment. Cambridge, U.K.: Cambridge University Press. Botkin, Daniel B., and Lee M. Talbot. 1992. "Biological Diversity and Forests." In Narendra P. Sharma, ed., Managing the World' Forests: Lookingfor Balance between Conservation and Develop- ment. Dubuque, Iowa: Kendall/Hunt Publishing Co. Briones, N. D. 1991. "Estimating Erosion Costs: A Philippine Case Study in the Lower Agno River Watershed." In Easter, K. W., J. A. Dixon, and M. M. Hufschmidt, eds., Watershed Resources Management: Studies from Asia and the Pacific. Environment Development Series. Singapore: Institute of Southeast Asian Studies. Brooke, James. 1995. "Rubber Trees Being Planted in Brazil, the Source," The New York Times, July 9, p. 14. Browder, John 0., ed. 1989. Fragile Lands in Latin America: The Search for Sustainable Uses. Boul- der, Colo.: Westview Press. Bruijnzeel, L. A. 1989. "Deforestation and Dry Season Flow in the Tropics: A Closer Look." Journal of Tropical Forest Science 1:229-43. . 1990. Hydrology ofMoist Tropical Forests and Effects of Conversion: A State of the Knowl- edge Review. Amsterdam: Netherland Committee for the International Hydrological Programme of UNESCO. 32 The World Bank Research Observer, vol. 13, no. I (February 1998) Bruijnzeel, L. A., with C. N. Bremmer. 1989. Highland-Lowland Interactions in the Ganges Brahmaputra River Basin: A Review of Published Literature. Occasional Paper 11. Kathmandu, Nepal: International Center for Integrated Mountain Development. Chomitz, Kenneth M., and David A. Gray. 1996. "Roads, Land Use, and Deforestation: A Spatial Model Applied to Belize." The World Bank Economic Review 10(3):487-512. Chopra, Kanchan. 1993. "The Value of Nontimber Forest Products: An Estimation for Tropical Deciduous Forests in India." Economic Botany 47(3, July-September):251-58. Chunhong, Hu. 1995. "Controlling Reservoir Sedimentation in China." InternationalJournal on Hydropower and Dams 2(March):50-2. Cruz, Wilfrido, H. A. Francisco, and Z. T. Conway. 1988. "The On-Site and Downstream Costs of Erosion in the Magat and Pantabangan Watershed." Journal of Philippines Development 26(25,1):85-111. Cutrim, Elen, David W. Martin, and Robert Rabin. 1995. "Enhancement of Cumulus Clouds over Deforested Lands in Amazonia." Bulletin of the American Meteorological Society 76(Oc- tober 1):1801. Eltahir, E. A. B., and R. L. Bras. 1992. "On the Response of the Tropical Atmosphere to Large-Scale Deforestation." Quarterly Journal of the Royal Meteorological Society 119:779-93. Enters, Thomas. 1992. "Land Degradation and Resource Conservation in the Highlands of North- ern Thailand: The Limits to Economic Evaluations." Ph.D. dissertation. Australian National University, Canberra. Processed. Godoy, Ricardo, Ruben Lubowski, and Anil Markandya. 1993. "A Method for the Economic Valu- ation of Nontimber Tropical Forest Products." Economic Botany 47(3):220-33. Grimble, R. J., J. Aglionby, and J. Quan. 1994. Tree Resources and Environmental Policy: A Stakeholder Approach. NRI Socioeconomic Series 7. Chatham, U.K.: Natural Resources Institute. Grimes, Alicia, and others. 1994. "Valuing the Rain Forest: The Economic Value of Nontimber Forest Products in Ecuador." Ambio 23(7):405-10. Grove, Richard H. 1994. "A Historical Review of Early Institutional and Conservationist Responses to Fears of Artificially Induced Global Climate Change: The Deforestation-Desiccation Dis- course 1500-1860." Chemosphere29(5):1001-13. Halley, Edmund. 1694. "An Account of the Watery Circulation of the Sea, and of the Cause of Springs." Philosophical Transactions of the Royal Society 142(7):468-72. Hamilton, L. S. 1987. "What Are the Impacts of Himalayan Deforestation on the Ganges-Brahmaputra Lowlands and Delta? Relations between Assumptions and Facts." Moun- tain Research and Development7:256-63. Hamilton, L. S., and P. N. King. 1983. Tropical Forested Watersheds: Hydrologic and Soils Response to Major Uses or Conversions. Boulder, Colo.: Westview Press. Harden, C. P. 1993. "Land Use, Soil Erosion, and Reservoir Sedimentation in an Andean Drainage Basin in Ecuador." Mountain Research and Development 13(2):177-84. Henderson-Sellers, A., R. E. Dickinson, T. B. Durbridge, P. J. Kennedy, K. McGuffie, and A. J. Pittman. 1993. "Tropical Deforestation: Modeling Local to Regional-Scale Climate Change." Journal of Geophysical Research 98:7289-315. Hodgson, Gregor, and John A. Dixon. 1988. Logging versus Fisheries and Tourism in Palawan: An Environmental and Economic Analysis. East-West Environment and Policy Institute Occasional Paper 7. Honolulu: East-West Center. Huntoon, Peter W. 1992. "Hydrogeologic Characteristics and Deforestation of the Stone Forest Karst Aquifers of South China." Ground Water 30(2, March-April):167-76. KenneTh M. Chomitz and Kanta Kumari 33 Kumari, Kanta. 1995. "An Environmental and Economic Assessment of Forest Management Op- tions: A Case Study in Malaysia." Environment Department Paper 026. World Bank, Washing- ton, D.C. Processed. Lampietti, Julian A., and John A. Dixon. 1995. "To See the Forest for the Trees: A Guide to Nontimber Forest Benefits." Environment Department Paper 013. World Bank, Washington, D.C. Processed. Lean, J., and P. R. Rowntree. 1993. "A GCM Simulation of the Impact of Amazonian Deforestation on Climate Using an Improved Canopy Representation." Quarterly Journal of the Royal Meteoro- logical Society 119(April 1):509-30. Lysne, D. K., N. R. B. Olsen, H. Stole, and T. Jacobsen. 1995. "Sediment Control: Recent Devel- opments for Headworks." InternationalJournal on Hydropower and Dams 2(March 1995):46- 49. Magrath, W. B., Charles Peters, Nalin Kishor, and Puneet Kishor. 1995. "The Economic Supply of Biodiversity in West Kalimantan: Preliminary Results." World Bank, Asia Technical Depart- ment, Washington, D.C. Processed. Mahmood, K. 1987. Reservoir Sedimentation: Impact, Extent, and Mitigation. World Bank Techni- cal Paper 71. Washington, D.C.: World Bank. Meher-Homji, V. M. 1988. "Effects of Forests on Precipitation in India." In Evan R. C. Reynolds and Frank B. Thompson, eds., Forests, Climate, and Hydrology: Regional Impacts. Tokyo: United Nations University. Michon, Genevieve, and Hubert de Foresta. 1995. "The Indonesian Agroforest Model: Forest Re- source Management and Biodiversity Conservation." In P. Halladay, and D. A. Gilmour, eds., Conserving Biodiversity outside Protected Areas: The Role of Traditional Agro-ecosystems. Gland- Switzerland: IUCN, and Cambridge-UK. Moran, Emilio F., Eduardo Brondizio, Paul Mausel, and You Wu. 1994. "Integrating Amazonian Vegetation, Land-Use, and Satellite Data." BioScience 44(5):329-38. Myers, Norman. 1995. "The World's Forests: Need for a PolicyAppraisal." Science 268(May 12):82- 83. Nepstad, Daniel C., and Stephan Schwarzman, eds. 1992. Nontimber Products from Tropical Forests: Evaluation ofa Conservation and Development Strategy. Advances in Economic Botany. Vol. 9. New York: The New York Botanical Garden. O'Brien, Karen L. 1995. "Deforestation and Climate Change in the Selva Lacandona of Chiapas, Mexico." Norsk Geografisk Tidsskrift49(3):105-22. Richards, E. M. 1993. Commercialization ofNontimber Forest Products in Amazonia. NRI Socioeco- nomic series 2. Chatham, U.K.: Natural Resources Institute. Rose, C. W. 1993. 'Erosion and Sedimentation." In Bonell, Hufschmidt, and Gladwell 1993. Shahwahid, H. O., and others. 1997. "Economic Benefits of Watershed Protection and the Trade- off with Timber Production: A Case Study in Malaysia." Research Report Series. Economy and Environment Program for Southeast Asia, Singapore. Processed. Shukla, J., C. Nobre, and P. Sellers. 1990. "Amazon Deforestation and Climate Change." Science 247(March 16):1322-25. Simpson, R. David, Roger A. Sedjo, and John W. Reid. 1996. "Valuing Biodiversity for Use in Pharmaceutical Research." Journal of Political Economy 104:163-85. Southgate, Douglas, and Robert Macke. 1989. "The Downstream Benefits of Soil Conservation in Third World Hydroelectric Watersheds." Land Economics 65(1):38-48. Thiollay, Jean-Marc. 1995. "The Role of Traditional Agroforests in the Conservation of Rain Forest Bird Diversity in Sumatra." Conservation Biology 9(2):335-53. 34 The World Bank Research Observer, vol. 13, no. I (February 1998) van Noordwijk, Meine, Thomas P. Tomich, and others, eds. 1995. Alternatives to Slash-and-Burn in Indonesia: Summary Report of Phase 1. ASB-Indonesia Report 4. Bogor, Indonesia: International Centre for Research in Agroforestry-Southeast Asia. Veloz, Alberio, Douglas D. Southgate, F. J. Hitzhusen, and R. D. Macgregor. 1985. "The Econom- ics of Erosion Control in a Subtropical Watershed: A Dominican Case." Land Economics 61:145- 55. Vincent, Jeffrey R., Mingsarn Kaosa-ard, and others. 1995. The Economics of Watershed Protection: A Case Study of the Mae Teng River. Bangkok: Thailand Development Research Institute. Wiersum, K. F. 1984. "Surface Erosion under Various Tropical Agroforestry Systems." In C. L. O'Loughlin and A. J. Pearce, eds., Proceedings, Symposium on Effects ofForest Land Use on Erosion and Slope Stability. International Union of Forestry Research Organization, Vienna, and East-West Centre, Hawaii. Wilkie, David S., and Mark C. Trexler. 1993. "Biogeophysical Setting and Global Climate Change." In CentralAfrica: Global Climate Change andDevelopment. Technical Report. Washington, D.C.: Biodiversity Support Program (a consortium of World Wildlife Fund, The Nature Conservancy, and World Resources Institute). Kenneth M. Chomitz and Kanta Kumari 35 Poverty in Russia during the Transition: An Overview Jeni Klugman K Jeanine Braithwaite It is no surprise that the breakup of the Soviet Union and the overall demise of the planned economy has had a profound effect on the welfare ofthe Russian people. But the absence of reliable survey data has constrained our understanding of the impact that transition has had on the distribution of income. This article draws upon several rounds of a nationally representative household survey to document the sharp increases in the incidence and severity ofpoverty that have occurred during the transition. We investigate the routes by which macroeconomic and structural developments have been transmitted through the labor market and examine the performance ofan increasingly overburdened, unfocused, and inadequate system ofsocialprotection. There is no evidence to suggest that the poor have shared in Russia s emerging economic recovery, and the emergence ofa core group of long-term poor appears to be a distinct possibility. The social and economic changes that swept across the former Soviet Union in the early 1990s were unprecedented in scale, scope, and speed. The collapse of an em- pire that spanned eleven time zones and encompassed more than 250 million people and a multitude of ethnic groups has had far-reaching consequences, both domesti- cally and internationally. The Russian Federation, which covers a vast expanse of Europe and Asia and in- cludes a wide variety of social and cultural groups, naturally looms large in perceptions about the transition from a centrally planned to a market-oriented economic system. For many Russians that transition has been marked by the polarization of a previously egalitarian society and a dramatic increase in the scale of poverty and deprivation. The people who have been most adversely affected by structural changes in the economy were not necessarily so badly off under the old system, suggesting that some conven- tional assumptions about the structure of poverty need to be adjusted. Underlying the sharp increases in the incidence and composition of poverty have been the overall decline in national income and the simultaneous increase in in- The World Bank Research Observer, vol. 13, no. I (February 1998), pp. 37-58 1998 The International Bank for Reconstruction and Development / THE WORLD BANK 37 equality. Aggregate shocks to output and productivity have been unevenly distrib- uted among different groups in the economy. We trace the underlying mechanisms that reflect a shift from a system with controlled wages and incentives to maximize employment to a far less constrained environment, albeit at temporarily much lower overall levels of output and productivity. Adjustment in the labor market has af- fected overall earnings and wage disparities at the same time that the number of employed has declined along with total number of hours worked. Although the gov- ernment social protection system has yet to adapt to the new economic conditions, the fiscal constraints have hindered its ability to offset the increase in poverty. Grow- ing regional disparities associated with economic liberalization have similarly strained the fiscal capacities of local authorities to effect redistribution. As the economy stabilizes and clear signs of economic recovery emerge, the situa- tion of the poor could be expected to stabilize and improve. Recent evidence sug- gests that the worst may be over, although the number of households and individuals below the poverty line continued to rise somewhat in 1995 and 1996. Beyond the direct recessional repercussions of the transition in Russia, the shift toward a market economy will have long-term implications for the poor. Changes in relative prices mean that certain groups (characterized, for example, by skill or location) are likely to be relatively better or worse off than they were before the transition. Inequality is also likely to remain higher than it was in the Soviet period, although the precise pattern of distribution will clearly depend on the government's tax and transfer policies. This article first looks at the macroeconomic trends relevant to household welfare that have characterized Russia's transition. It then reviews the data sources and meth- odology-a critical dimension given the paucity of reliable data-and presents the major findings that have emerged from a recent analysis of poverty. After examining developments in the labor market and the limited effectiveness of public income transfers in alleviating poverty, we present some conclusions about social protection policy and its effect on poverty in Russia. The Economic Backdrop Russia inherited a command economy whose weaknesses had become manifest in the 1980s (Easterly and Fischer 1995). The economic story since 1991 is by now familiar to most readers. As the old political regime crumbled, inflation surged and output fell precipitously, leading to the collapse of internal and external trade and a rapid rise in the fiscal deficit. High inflation, which continued to 1995, severely depressed economic activity throughout much of the economy, and real earnings declined. Rapid consumer price inflation followed market liberalization in early 1992 and had adverse effects both generally, in accentuating the depth and duration of the 38 The World Bank Research Observer, vol. 13, no. I (February 1998) output recession (Popov 1996), and specifically, among those groups whose wages and transfer payments began to lag behind prices. With the collapse of output, gross domestic product (GDP) fell more than 40 per- cent during 1991-96 (EBRD 1997). That decline places Russia among the worst of the formerly planned economies; estimated GDP decline ranged from 11 percent in Poland to 57 percent in Lithuania (de Melo, Denizer, and Gelb 1995). These esti- mates for Russia are problematic, however, because the structure of the decline in output and other measurement problems, as well as the failure to account for output from most informal sector activity, understate post-1991 output. Consistent with the pattern in other transition economies, there has been a relative expansion of the service sector's share of GDP (from 31 to more than 50 percent during 1989-94), especially trade and financial services, and a reduction in the share of investment and military production; part of this 'growth' in services, however, is attributable to changes in classification as well as in relative prices. Unofficial economic activity of both new private firms and privatized state enterprises has been expanding at a robust pace (Gavrilenko and Koen 1994). Industrial restructuring has been associated with a marked shift from secondary manufacturing into raw materials industries. The share of resource industries (energy, steel, and nonferrous metals) in industrial output doubled to 48 percent during 1991-95, while that of secondary manufacturing- particularly machinery, equipment, and light industries-declined. These shifts have been associated with significant changes in the regional and interpersonal distribution of earnings and incomes. The regional concentration of production under central planning magnifies these subsequent distributive effects. Evidence suggests that the regional variation in nominal wages and incomes fell steadily during the two decades preceding the transition in Russia, but that re- gional disparities have increased sharply since 1991. The trend is even sharper when measured in real terms that account for significant geographical price varia- tion (Stewart 1996). Indeed by 1994 income disparities among the oblasts (a po- litical subdivision) of Russia were far greater than those among the states of the United States. The coefficients of variation for income among Russian oblasts and American states were 0.519 and 0.148, respectively. (The coefficient of variation shows the standard deviation divided by the mean. Values closer to one are less equal than values closer to zero.) Measuring the social impact of the transition from a command economy is diffi- cult, not least because the official economic data are unreliable (Koen 1996). Tradi- tional sources of information on production, incomes, and prices, never totally reli- able, have grown weaker, partly because many enterprises now fall outside the reporting network; official GDP figures were revised significantly every year between 1991 and 1994. Even more worrying has been the apparent divergence of time-series data that would normally be positively correlated (such as average money income and GDP). Thus the macroeconomic statistics cited here should be regarded with caution, and Jeni Klugman and Jeanine Braithwaite 39 the revealed trends regarded as indicative. Official poverty and income distribution measurements raise their own set of data problems. Measuring Poverty in Russia: Conceptual and Empirical Issues Assessments of household welfare confront a range of methodological and practical problems, from devising an appropriate definition of poverty to determining the appropriate measure of household welfare (using income, expenditure, or some non- monetary indicator). What Is Poverty? Conceptually, few would dispute that poverty is the inability to sustain some mini- mal level of existence. The standard approach to defining an absolute level of poverty is to price a basket of essential goods and compare an individual's income to the cost of these necessities. But both in the definition of nutritional and other basic needs and in the calculation of corresponding cost, the experts or politicians who decide on a certain method must make value judgments (Hagendaars 1986). Moreover, money income may not be a good measure of real consumption. Dur- ing the Soviet period, most informal sector activities were illegal, and some residual stigma may induce households to under-report such income. Because the goal of poverty analysis is to consider people's real well-being, expenditure can be a better measure than income (Deaton and Muellbauer 1980). The definition of expendi- ture used in the Russian Longitudinal Monitoring Survey results reported below is more akin to consumption than to income, because it includes the imputed value of in-kind goods and services produced in the home, and received (for example, from employers). Whether based on income or expenditures, monetary indicators of welfare can also cause problems for analysts trying to make comparisons across economic re- gimes. High inflation creates additional difficulties in measuring and comparing nominal aggregates. Another complication arises from the shift from a regime that relied on several different allocation mechanisms for goods and services, including queues and rationing with black markets. Searching and queuing imposed signifi- cant costs-although they were not evenly distributed. But price liberalization has largely eliminated scarcities and trade reform has increased the volume and quality of goods available. Consumer surveys reveal that shortages have gradually diminished over the transition, first in the metropolitan areas, and then in provincial and rural parts of the country. It has been argued that the net welfare gains now that queuing is no longer required have been significant, even where real incomes and consump- 40 The World Bank Research Observer, vol. 13, rno. I (February 1998) tion fell (Roberts 1995). This observation, however, does not take into account dis- tributional effects, which are obviously important in measuring poverty. Consumers on low incomes may actually prefer queuing at low prices because their opportunity costs of waiting are low. Even while money has become a much more meaningful indicator of welfare in Russia, comparisons over time are bound to be approximate. On balance the best approach is to base our assessments on household expenditure, mindful of the qualifications noted above. The next question is where to draw the poverty line against which to compare expenditures. Drawing the Poverty Line Drawing the official poverty line is in part a question of judgment and in part a matter of policy and politics. Thus it is important to bear in mind that the extent of measured poverty in the analyses here is sensitive to where the poverty line is drawn.' Under Khruschev, a "minimum consumption budget" was developed (Sarkisyan and Kuznetsova 1967; see also Atkinson and Micklewright 1992), although official estimates of its ruble value as a poverty line were not published.2 By January 1992, when extensive price liberalization was undertaken, the minimum consumption bud- get was manifesdy unusable, because virtually the entire population had incomes below this standard (Mozhina 1992). A revised basket corresponding to World Health Organization guidelines was subsequently adopted, and the assumed high share spent on food (68.3 percent) was consistent with the structure of low-income household expenditures in the early years of transition, although future adjustment will be nec- essary to take account of changing relative prices-especially for services, which de Masi and Koen (1995) find extremely cheap relative to market economy standards. Table 1 compares the nominal value of the official poverty line to the average wages during the 1990s. The poverty line (per capita) has typically been only about half the average wage, suggesting that the threshold is quite austere. It is this poverty line, as calculated by the Ministry of Labor in Moscow, that we use in the analysis below. Data Sources and Problems Official censorship of poverty issues does not mean that information about incomes and expenditures did not exist in the USSR. The Family Budget Survey (FBS), insti- tuted in 1922 (Dmitrichev 1992), is still the most widely cited source of information about poverty in Russia and other former Soviet republics. About 60,000 families made up the survey pool in 1970-85, which widened to 90,000 in 1988-90. In 1995 the Russian sample was approximately 50,000 families. At least through 1996, the survey was completed every quarter. Although the FBS represents a rich potential source of data, the survey's methodol- ogy has been subject to extensive criticism, both internally and externally (Rimashevskaya Jeni Klugman and Jeanine Braithwaite 41 Table 1. Official Poverty Lines and the Average Wage, 1980-96 (rubles) Minimum Official consumption poverty Average Year basket line wage 1980 64.6 1990 93.3 - 1991 190.0 - 770 1992 - 4,282 16,071 1993 - 42,800 141,000 1994 - 145,400 354,000 1995 - 321,000 735,000 1996 - 379,000 944,000 Not available. Source-. Estimates of the Soviet minimum consumption basket are for the whole year, based on press reports (see Braithwaite 1997). The official poverty line is the prozhiochniyy minimum, as calculated by the Ministry of Labor in Moscow. The yearly data are for December. and others 1979; Shenfield 1983; Braithwaite and Heleniak 1989; and Atkinson and Micklewright 1992). A major concern is the sampling frame, which does not encom- pass the entire population but remains based on the so-called branch principle. Work- ers are drawn from enterprise rolls, with large and urban enterprises overrepresented, and workers with seniority much more likely to be included given the material incen- tives to participate. The more workers in a family, the greater the likelihood that the family will be selected. A separate sample was set up for pensioners, but again, those retired from large enterprises are more likely to be included. The net result is that the lower part of the income distribution is not adequately represented. And the exclusion of certain occupational groups, such as senior bureaucrats, KGB officials, and military officers, raises some concerns about the upper end of the distribution (although this is a perennial problem in market economies as well). Nor does the design allow the calcu- lation of sampling errors or confidence intervals, statistical techniques that allow an assessment of the reliability of results obtained from the survey. Further, the interviewing and recording practices are questionable; the interviewer leaves a diary with the household and returns frequently to assist them in filling it out. This practice suggests the possibility of the interviewer simply creating and recording the data. The household is responsible for keeping both expenditure and income data. Before the transition there were severe penalties, including imprisonment, for illegally earned income from informal sources and from private capital or enterprise, and there was thus a very high degree of correspondence between stated household income and expenditures. For example, during the anti-alcohol campaign of the 1980s, households under-reported alcohol purchases to such an extent that the alcohol beverages weight in the consumer price index had to be imputed, using data from retail sales. 42 The World Bank Research Observer, vol. 13, no. 1 (February 1998) There are also serious problems associated with the analytical methodology cur- rently used by the Goskomstat Rossii, the government statistical agency, particularly the calculations of monthly indicators of income and distribution.3 As the figures reported in table 2 suggest, the official monthly estimates of poverty and distribution are not very plausible. Goskomstat's reported poverty head counts fluctuated mark- edly from month to month in 1993 and 1994. Nonetheless, the FBS was the basis for reports and analysis of poverty and distribution published by the Goskomstat. Table 2. Official Trends in Poverty and Distribution of Income in Russia, 1980-96 (percent) Head Poverty Year count gap Severity Gini coefficient 1980 11.3 2.34 0.70 27.60 1985 13.4 2.96 0.94 27.56 1989 11.0 2.24 0.65 26.49 1990 10.1 2.12 0.63 28.45 1991 11.4 2.03 0.56 26.54 1992 January 30.2 7.25 2.45 23.93 June 23.1 6.59 2.68 28.70 Dec. 15.7 4.11 1.49 34.98 1993 March 34.7 10.48 4.45 29.38 June 24.7 7.00 2.75 34.67 1994 January 34.9 - - 40.9 (annual) May 16.4 - - Sept. 20.5 - - 1995 January 33.0 - - 38.1 (annual) July 26.0 - - Dec. 20.0 - 1996 January 25.0 - - 37.5 (annual) July 21.0 - - Dec. 18.0 - - -Not available. Note For 1980-91 the minimum consumption basket was used as the poverty line; for the period since, the subsistence minimum. The head count index shows the percentage of individuals in the population falling below the poverty line. The poverty gap index sums all the poverty gaps in the population, that is, the amount of money needed to bring all the poor up to the poverty line, as a share of GDP. The severity index gives greater weight to those individuals (households) furthest from the poverty line (see Ravallion 1992 for further details). The Gini coefficient is closer to zero when incomes (expenditures) are more equally distributed. Source Authors' estimates based on published Goskomstat sources. Jeni Klugman and Jeanine Braithwaite 43 In July 1992 Goskomstat Rossii implemented a new household survey, the Rus- sian Longitudinal Monitoring Survey (RLMS), with technical and financial assistance from the World Bank and the U.S. Agency for International Development, that ran in parallel to FBS. The first round of the RLMS was nationally representative and involved 6,500 households. It was designed to meet scientific standards for a true probability sample (see Klugman and Braithwaite 1997 for details on the sampling method). The availability of this data has made it possible to undertake detailed analyses of poverty during the transition (World Bank 1995b; Klugman 1997; Glinskaya and Braithwaite forthcoming). The demographics of the original RLMS sample compared favorably with the So- viet census that was carried out four years earlier (Foley 1997). The initial rounds achieved a high response rate (about 90 percent), although participation declined in later rounds. The original panel of households was replaced with a new panel after the fourth round. The household questionnaire collects data on expenditures, in- come, housing conditions, and ownership of land and assets. The individual ques- tionnaire covers employment, use of time, migration, and anthropometry of young children. Trends during the Transition Since 1991, the number of poor households in Russia has risen sharply, and accord- ing to the RLMS, a record 35 percent of the population was living below the official poverty line by the end of 1995 (table 3). The share of individuals in poverty is higher (41 percent in 1995) than the share of households because poor households tend to be larger than average. Beyond the head-count measure, however, indicators suggest that the worst poverty occurred in 1993, when both the share of the popula- Table 3. The Incidence, Depth, and Severity of Poverty among Households, 1992-95 (percent) Unit 1992 1993 1994 1995 Poor householdsa 25.2 31.9 26.8 35.0 Very poor householdsb 8.4 12.0 10.4 10.9 Depth 9.8 13.6 11.7 13.2 Severity 5.4 8.0 7.2 6.9 Head count for individuals 26.8 36.9 30.9 41.1 Note: Percentage of households (unless noted). Expenditure-based. a. Percent of households with expenditure below the poverty line. b. Percent of households with expenditures less than half the poverty line. Source: Rounds 1-6 of the RLMS; Foley (1997); Kolev (1996). 44 The World Bank Research Observer, voL 13, no. I (February 1998) tion classified as very poor (households whose expenditures were less than half of the household-specific poverty line) and the severity of poverty peaked. By late 1995 about 11 percent of households in Russia were very poor, significantly higher than in 1992 although below the peak of 1993. Despite some abatement during 1994, the poverty head count at the end of 1995 was some 10 percentage points higher than it had been in mid-1992, and the depth and severity measures were also markedly higher. These overall trends in poverty result from rapid declines in national income and changes in the distribution of income during the transition. Based on the early years (Milanovic 1995) estimated that an additional 1 percent decline in Russia's national income would lead to only 0.5 percent increment in the poverty head count. De- composition of more recent trends in poverty suggests that the continued lack of significant real growth in 1995-97 accounts for most of the increase in poverty, and that household inequality has been fairly constant during the past two years. There are indications that a group of long-term poor may have emerged (Glinskaya and Braithwaite forthcoming). 4 Who are the poor in contemporary Russia? About two-thirds of those living in poverty are in households where the head of the household is employed. They are primarily families with children, the unemployed, and single elderly people living alone. Nearly 85 percent of families with three or more children under 6 years of age were poor in 1994. Poverty, always associated with family size, has become increasingly concentrated in families with children as well as in households with an unemployed person. Thus while the overall head-count index increased by about 8 percentage points between 1994 and late 1995, in the same period poverty in families with one child rose by 14 points, and by more than 19 points in families with children under six years of age. Cross-tabulations show that in late 1995, consistent with the pattern that had emerged earlier in the transition, the inci- dence of poverty was almost 52 percent among children under 18, and an even higher 57.6 percent among children under six, compared with 41 percent for working-age adults and 26.7 percent for the elderly. Controlling for other ob- served factors, the presence of one unemployed person in a household increased the poverty risk of the household by more than 15 percentage points above the national average. Causes of Increased Poverty Dynamic aspects of poverty were investigated between 1992 and 1994 through the RLMS panel. There were significant flows into and out of both poverty and severe poverty, even while the overall incidence of poverty was rising. Jeni Klugman and Jeanine Braithwaite 45 Table 4. Changes in the Gini CoeJjicientfor Money Income, Selected Countries, 1989 and 1994 Gini coefficient Country 1989 1994 Bulgaria 25.0a 35.3 Estonia 27.2 38.6 Lithuania 27.5 37.0 Moldova 26.7 40.0 Poland 26.9 30.4 Romania 26.9 28.4 Russia 25.7 40.9 a. 1990. Source. UNICEF TRANSMONEE, based on data reported by national statistical offices; UNICEF, International Child Development Centre, Florence, Italy. A key element of trends in poverty has been rising inequality. The disruption of the old system with its controlled wages and prices has clearly led to greater differen- tiation in income-earning opportunities. Unlike the output shocks associated with the transition, the disparity in income associated with changes in relative prices is likely to persist. Table 4 suggests that the rise in Russia's Gini coefficient (a widely used measure of income inequality, with higher values indicating less equal distribu- tion) is somewhat greater than that of most Eastern European countries, yet of a similar order of magnitude to several former Soviet countries (including Estonia, Lithuania, and Moldova). Measurements based on Goskomstat (1995) data suggest that the ratio between the highest and lowest income deciles widened dramatically, from 4.5 in 1991 to 15.1 in 1994, then moderated slightly to 13.5 in 1995. Goskomstat's calculated measures of inequality are significantly below other estimates, mainly for reasons related to sampling. RLMS calculations, not surprisingly, reveal higher levels of in- equality both at the outset and through the transition. Given the weaknesses of the FBS in measuring income distribution, this finding seems more plausible than a very sharp rise in inequality within the space of a few months. The RLms-based Gini coefficient stood at 49.2 by mid-1993, then declined slightly to 48.4 by the end of 1995. Trends in the decile distribution based on the RLMS reveal that the top decile's share of total expenditure rose from about 33 to more than 37 percent between 1992 and 1995. The relative share of each of the bottom five deciles also increased, albeit slightly from very low levels, over the same period. Hence in relative terms the con- sumption of the former middle class appears to have been squeezed most. This is consistent with Milanovic's (1995) observation that the income of the former middle class (which includes clerical staff, production workers, teachers, administrators, and doctors) had declined significantly. The changes in relative wages analyzed above help to explain these shifts. 46 The World Bank Research Observer, vol. 13, no. I (February 1998) The Underlying Mechanisms What mechanisms underlie these developments? In this section we attempt to link the risk factors associated with poverty in contemporary Russia to changes in the labor market and in the extent and direction of public transfers. Adjustments in the Labor Market According to Goskomstat data, wages, salaries, and entrepreneurial income amounted to almost 80 percent of total personal income in 1995 (public and private transfers accounted for the remaining 20 percent). Changes in the level as well as in the rela- tive share and distribution of these components significantly affect household wel- fare. Labor market adjustments have been extensive in the wake of the price and wage liberalization involved in the move to a market system (Commander and Yemtsov 1997). Aggregate declines in the demand for labor as a result of the drop in output have led to rising unemployment, albeit at a slower rate and to a lesser extent than originally expected by external observers, including the World Bank. This is partly because significant cuts in real wages have accompanied the decline in demand. Av- erage real earnings during 1992-95 were about half the level of the 1980s (table 5). Various factors appear to determine labor market rewards. Education and skill levels appear to be factors in determining whether someone is poor. The results of the econometric analysis of the RLMS data indicate that higher educational attain- ments are associated with a somewhat lower risk of poverty: for example, the risk is reduced by 9 percent for households headed by university graduates. Households Table 5. Trends in Real Benefits and Wages, 1989-96 Average Minimum Family Minimum Average Year pension pension allowancea wage wage 1989 82 68 - 109 70 1991 73 85 - 85 67 1992 38 43 - 43 36 1993 43 37 - 37 50 1994 55 49 27 27 49 1995 49 32 12 12 35 1996 48 33 20 18 37 -Not available. Note Fourth quarter 1991 equals 100. For the years 1992-96 inclusive, the reference month is January, for 1991 first quarter, and for 1989, annual. a. It is not possible to give a single figure for the period before 1994 because of the variety of benefits (see text). The base was calculated using an estimate based on the minimum wage (to which a number of benefits were linked). Source Braithwaite (1997), table 2.1; Ministry of Labor, Moscow. Jeni Klugman and jeanine Braithwaite 47 that are headed by the unskilled, by clerks, and agricultural workers are at higher risk of poverty. Finally, and perhaps less expected, the number of individuals in a house- hold engaged in primary employment (that is, a main job) does not significantly affect the risk of poverty. Participation in secondary employment (often in the informal sector) does reduce the expected probability of poverty, however, by about 11 percent. The liberalization of wage decisions was expected to affect the distribution of income significantly. Before the transition the share of wages in total personal in- come was almost 75 percent, which is high by international standards. For example, in 1989 wages accounted for only 59 percent of total personal income in the United States (U.S. Department of Commerce 1991). Increasing inequality in the distribu- tion of total income in Russia has been largely driven by the decline in the share of wages in total personal income and the corresponding rise in the contribution of entrepreneurial and capital income (figure 1). The drop in the wage share is quite plausible in the context of the previous regime's ban on virtually all types of entre- preneurial earnings and private capital investments. In 1990 wages and entrepre- neurial income (including capital) contributed 74.1 and 12.9 percent respectively to total monetary income. By 1995 the combined share of entrepreneurial (38.6) and capital (5.2 percent) income outstripped the 39.5 percent share contributed by wage Figure 1. Structure of Personal Income, 1990-95 Percentage of personal income 100 _ 90 Transfers 80 70 60 Entrepreneurial income 50 I 40 30 20 Labor 10 0 1990 1991 1992 1993 1994 1995 Note. Entrepreneurial income includes capital income, which amounted to 5.2 percent in 1995. Source Goskomstat (1996). 48 The World Bank Research Observer, vol. 13, no. I (February 1998) Table 6. Selected Wage Indices Relative to NationalAverage Wage, 1990-95 Sector 1990 1991 1992 1993 1994 1995 Industry 103 111 118 108 104 111 Agriculture 95 84 66 61 50 43 Construction 124 127 134 133 129 131 Transport 115 120 146 151 150 152 Commerce 85 91 91 107 123 117 Science/R&D 113 90 64 68 78 78 Finance 135 180 204 243 208 - Note National average is 100. Source Goskomstat (1996). income. These changes reflect the dramatic erosion in the average real wage as pro- ductivity of employed labor has fallen over the period. As in other transition coun- tries, such as Bulgaria (Milanovic 1995), the growing share of entrepreneurial and capital income has driven the overall shift to a less equal distribution of income. Higher overall wage inequality has been accompanied by sectoral shifts in the economy and changing sectoral wage relativities. Table 6 suggests that workers in the industrial, construction, transport, commerce, and especially, financial, sectors have done relatively better than average. The most marked deterioration occurred in agriculture, in science, and in research and development. The decline in agricultural wages reflects the withdrawal of state subsidies combined with declines in output and only limited adjustment in employment. Additionally, Russia's delay in under- taking comprehensive land privatization likely hurt agricultural laborers. In Arme- nia, by contrast, swift privatization was associated with rates of rural poverty that were significantly less severe than in urban areas (Braithwaite 1995). The official wage figures support the findings of the RLMS that the incidence of poverty among households headed by skilled agricultural workers in late 1995 was the highest among all reported job occupations, at 64.7 percent, even higher than unskilled workers (47.1 percent) and far higher than the national rate of about 35 percent. In the face of inherited overstaffing and falling demand for labor, employers have laid off increasing numbers of workers or reduced their hours of work. By mid-1996 registered unemployment stood at about 4 percent of the labor force, although In- ternational Labour Organization methodology, which includes all unemployed in- dividuals whether or not they are formally registered with the Federal Employment Service, puts it at about 9 percent. Blue-collar workers have borne the brunt of the decline in labor demand. Hours were cut for about 5 million workers in 1994, and another 7.4 million were placed on involuntary leave. During 1993 and 1994 only 40 percent of the work force was paid in full and on time. If wages are paid late and consequently are eroded by inflation or are not received at all, earnings figures can be quite misleading. Multivariate analysis of the RLMS suggests that households report- Jeni Klugman and Jeanine Braithwaite 49 ing wage arrears are 11 percent more likely to be poor. Overall, however, firms ap- pear to prefer to retain workers, a preference that can be traced to insider influence at the firm level, as well as to the limited social safety net for displaced workers. The incidence of poverty among the unemployed is high; in late-1995 about half of households with an unemployed member were poor. An additional unemployed person in the household raised the probability of a family being in poverty by about 9 percentage points. The Federal Employment Service, which is responsible for in- come support and labor market programs for the unemployed, reaches only a seg- ment of its target group. Commander and Yemtsov (1997) found that the majority of the unemployed do in fact have some marginal employment and that, as expected, there was a strong positive link between marginal employment and education, quitting a previous job (as opposed to being laid off), and residence in Moscow or St. Petersburg. Increasing regional disparities-apparently caused by changing relative prices along with reductions in the subsidies to the producer-are a critical dimension of recent labor market trends. Industrial areas in central Russia and the North Caucasus have been especially hard-hit, while regions with abundant natural resources, such as Yakutia, and commercial centers such as Moscow and St. Peters- burg have suffered less. Trade liberalization has had differential impacts because of the sectoral concentration of economic activity. The changing structure of demand has also had an uneven effect across regions, for the same reason. The most obvious example is the impact on areas that were previously dependent on military production. Indeed, there is an inverse relation between industrial output and registered unemployment, with oblasts that did not suffer from high unemployment or wage arrears tending to have experienced either substantial real wage cuts or relatively less industrial decline. Theory and the empirical evidence suggest that regional disparities in unemployment are likely to persist, despite signs of wage flexibility and an emerging inverse association at the regional level between changes to wages and unemployment. Commander and Yemtsov (1997) have identified a geographic mismatch in the distribution of the unemployed and the availability of jobs that is associated with severe constraints to labor mobility (particularly lack of housing). The average duration of unemployment is increasing but is still relatively short. At the end of 1994 labor force surveys suggested that about 60 percent of the unem- ployed found a primary job within six months, while 23 percent were out of work for longer than a year. These relatively short episodes indicate that gross flows in Russia's labor market were large, especially compared with most European transition economies where unemployment was generally higher, while hiring is increasingly concentrated in the more flexible nonstate and informal sectors (Commander and Yemtsov 1997). 50 The World Bank Research Observer, vol. 13, no. I (February 1998) Systems of Social Support The government's social policy response appears to have failed to stem the rise in poverty. The formal system of social protection has become increasingly inadequate for many vulnerable population groups, with the important exception of those who receive old-age pensions. We have examined the impact of public transfers by ana- lyzing the incidence of various benefits at different dates using the RLMS as well as movements in the real official level of benefits. During the Soviet period, the system of social support was based upon low ad- ministered prices for food, rent, household utilities, and other basic goods and ser- vices, along with a guaranteed job. This inherited system of support for households was not easily adapted to the introduction of market principles. Nor was the struc- ture of social protection adequate to deal with the type and scale of needs created by the transition, even though this protection included both social insurance (pensions and unemployment benefits) and social assistance programs (including family allow- ances). Sources of financing have been squeezed because social insurance is largely financed through payroll taxes, which have fallen, whereas social assistance is largely the responsibility of local authorities whose economic fortunes have varied enormously.5 The government allocated about 4.5 percent of total budget expenditures for so- cial transfers in 1995, most of which was provided by regional budgets. The real value of transfers has been seriously eroded, however, as GDP has declined (see table 5). The average pension, which has hovered at about half the level of late 1991, has nevertheless been relatively better protected than both the average wage and other benefits. The minimum pension was also badly affected during the same period, falling to nearly one-third of its level in 1991. During the Soviet period, family allowances were paid to single mothers and to large families. A new system of universal child allowances evolved in the wake of the price liberalization; a complicated system with some 18 different benefits for families with children. Because most of these were set as some proportion of the minimum wage, however, lagged indexation significantly eroded their value (World Bank 1994). In 1994 the system was simplified with the introduction of a unified benefit. But the real value remained very low, at only about one-fifth of the pretransition levels. Increasing regional disparities in poverty rates can be attributed in part to the effects of restructuring in regions with different economic bases, but also to the devo- lution of responsibility for financing social assistance to local authorities. Disparities in resources and incomes have constrained local authorities, who also have responsi- bility for financing and delivering the bulk of education and health services as well as subsidizing housing and domestic utilities. Thus decentralization has contributed to the trend toward regional inequality. Analysis of federal fiscal transfers shows that the impact is not progressive (World Bank 1995a). The introduction of a new fed- Jeni Klugman and Jeanine Braithwaite 51 eral transfer mechanism in 1994 was indeed perversely disequalizing, apparently be- cause the transfer was based on the oblast's own assessment of need derived from the current level of service, which tended to be higher in wealthier areas. Evidence sug- gests that health and education expenditures have been relatively protected against budget cuts but that social assistance has been cut disproportionately. Stewart (1996) found that no correlation between a locality's official poverty head count in 1994 and the amount it spent on social assistance, meaning that areas with greater relative need spend less on social assistance. Because pension payments are closely related to previous employment and levels of income, these transfers are regressive (World Bank 1995b). Pensions appear to be an effective means of intergenerational transfer in that reported poverty among the elderly has been kept consistently below the national average since 1992. This pat- tern mirrors that of an increasing number of western industrialized countries in which public pension systems have signifcantly reduced, and in some countries virtually eliminated, old-age poverty (Scherer 1997). Among pensioners who are not poor (table 7), pensions represent more than half their income. For the elderly living in very poor households, pensions contribute an even larger share of their income. Table 7. Coverage and Significance of Public Transfers, 1994 and 1995 (percent) Very poor Poor Non-poor Share of Share of Share of recipient recipient recipient Receiving household Receiving household Receiving household Type of transfer the benefit income the benefit income the benefit income Family allowance 1994 28.8 23.6 32.4 14.5 25.7 5.9 1995 29.3 16.3 32.3 13.1 17.2 7.1 Pension 1994 40.3 75.0 41.0 66.9 48.7 58.4 1995 36.8 61.5 39.7 54.7 54.0 52.5 Unemployment benefit 1994 0.8 21.7 0.4 17.8 0.3 9.8 1995 2.1 37.6 2.6 19.6 0.9 14.1 Local social assistancea 10.4 9.6 10.4 9.6 14.5 8.1 Housing subsidyb 3.3 7.5 3.6 7.5 6.4 5.2 Scholarship 5.2 17.8 6.2 18.2 6.7 5.2 All transfers 1994 66.8 58.5 70.9 48.4 74.4 42.6 1995 61.6 43.3 67.7 40.3 70.0 43.3 a. Data all for 1994, first quarter. b. Information is for December 1995 and January 1996. Source: Foley and Klugman (1997); Kolev (1996). 52 The World Bank Research Observer, vol. 13, no. I (February 1998) As suggested above, the Federal Employment Service has been largely ineffective in providing social protection for the unemployed. Indeed, table 7 reveals that in late 1995, when about 13 percent of poor households had an unemployed member, less than 3 percent of the poor received unemployment benefits. Moreover, given its minimal level (see table 5), the contribution to income is only significant for the very poor. According to the RLMS, from 1992 to late 1995, public transfers accounted for almost 30 percent of total household income; the average contribution to the in- come of recizpient households was higher-an estimated 42 percent in late 1995. Important trends during this period can be observed regarding the relative impor- tance of different transfers (see table 7). Pensions remain the most widely received benefit. Unemployment benefits, although still very limited in their coverage, are going to an increasing number of the poor and accounting for a growing share of their income. Coverage of family allowances was steady for the poor, but dropped for the nonpoor, probably because families that are not needy may not bother to apply. Such benefits also declined as a share of income for poor and very poor house- holds, given the erosion of their real value and the delays in payment. Although the coverage of public transfers across the population is fairly extensive, there are some perhaps unintended results. For example, most of the poor and very poor are eligible for transfers, and yet they do not receive benefits (table 8). Nearly four out of ten very poor households and one out of three poor households did not receive any transfers. Yet the majority of nonpoor families did receive public trans- fers. Most poor households not revceiving benefits are in fact eligible for them. These errors of exclusion appear to be administrative problems, fiscal ones (that is, local authorities do not have sufficient funds to pay benefits), or both. Inadequate infor- mation about eligibility and lack of access to local offices may contribute to low take- up rates. The number of those receiving benefits to which they are not entitled (er- Table 8. Errors in Targeting of Public Transfers, 1995 (percent) Category Very poor Poor Non-poor Did not receive benefits 39.8 33.0 30.9 Did not receive benefits although eligible 36.4 29.7 22.8 Not eligible to receive benefits 6.2 6.1 13.5 Received benefits although ineligible 0.8 1.5 3.9 Note: Public transfers considered here are limited to pensions, unemployment benefits, scholarships, and family allowances, based on the RLMS. Source. Kolev (1996). Jeni Klugman and jeanine Braithwaite 53 rors of inclusion) appears to be relatively limited. These overall patterns have per- sisted throughout the transition. Only about one in five Russian households rely solely on income from employ- ment and transfers from the formal system of social protection. Private interhousehold transfers have been investigated by Cox, Eser, and Jimenez (1997), who found that private transfers in Russia are large and responsive to the socioeconomic characteris- tics of the household. Private networks are extensive, with four out of ten Russian households acting as donors, recipients, or both. Longitudinal analysis shows that this behavior has persisted throughout the transition, in contrast to Poland, for ex- ample, where private transfers diminished as economic conditions worsened. For the RIMS sample as a whole, private transfers average about 5 percent of household in- come; for net recipients the contribution was much higher, amounting to about 20 percent of household income. Overall, private transfers appear to flow to such vulnerable groups as younger families, female-headed households, and households affected by unemployment. Households participating in private transfers tend to be better off than those who do not participate, although whether such transfers have an equalizing effect on the distribution of income is unclear. Although private transfers tend to be directed from better-off to poorer groups, the probability of receiving a transfer declines only slightly as earned income increases. Empirical investigation reveals that the theoreti- cal concern about public transfers crowding out private transfers are not warranted: private and public transfers appear to be complementary. In light of this evidence, it is perhaps not surprising that overall attitudes toward government-provided social protection are mixed. On the one hand, surveys suggest an increasing degree of self-reliance. The vast majority say they would first rely upon themselves, and then friends and family, for help in time of need. Yet most people still expect the government to make good on the perceived right to employment for every able-bodied person. Moreover, the expressed needs for social support from the government, as measured in nationally representative public opinion surveys, are significant (Zubova and Kovaleva 1997). Conclusion Analysis of household survey data shows that Russia's transition from a command economy has been associated with a significant worsening of household and indi- vidual well-being. Although some indicators suggest that the worst may have passed in 1993, the numbers of individuals and households below the poverty line rose again in 1995, despite signs of economic recovery. Russia is not unique among economies in transition in experiencing worse and more prolonged poverty than many expected at the outset. As elsewhere, aggregate 54 The World Bank Research Observer, vol. 13, no. I (February 1998) trends in poverty result from changes in output and income distribution. In both respects, however, the Russian experience has been relatively adverse. The lack of general consensus on an appropriate economic and social reform strategy delayed sustained stabilization and prolonged the period of high inflation and recession. The associated collapse of public revenue has severely strained the government's capacity to finance social transfers, as well as essential public goods and services The publicly financed system of pensions has kept the rate of poverty among the elderly consistently below the national average. Yet for those outside the formal payroll-based system of social insurance, in particular such families with children, social assistance is ad hoc and limited. The same is true for the increasing number of households affected by unemployment, suggesting that the ongoing European de- bate about the social exclusion of marginalized groups may be relevant to Russia in the longer term. Several factors-low skills, type of employment, regional location, family structure, and benefit rules-appear to be interacting to marginalize some groups in society. At the same time, as elsewhere in transition countries and the West, popular demands for more comprehensive social assistance raise concerns about welfare dependency. The emerging economic recovery does not yet appear to have begun to turn around poverty, according to the aggregate poverty measures. In this sense the experience of Russia might be analogous to that of Poland, where several years of positive growth occurred before poverty began to decline (World Bank 1995b). The extent of pov- erty in Russia-in 1995 some 11 percent of the population was very poor-suggests that the resumption of growth could take longer to reduce poverty. Moreover, re- structuring is needed in several sectors of the economy, such as coal, metallurgy, chemicals, machine-building, and agriculture. Business reorganizations will likely entail a significant number of layoffs, thereby increasing unemployment and possi- bly poverty as well. The interest of the international community in the social impact of Russia's transition lies, at least in part, in its impact on the political sustainability of continued economic and democratic reform. Public opinion surveys are a poten- tially rich source of information in this regard (Zubova and Kovaleva 1997). In Russia, as elsewhere, subjective poverty standards are much higher than the offi- cially established absolute poverty lines, so a much higher share of the popula- tion regards itself as poor. Belief in an egalitarian distribution of income remains persistent and widespread. People tend to attribute poverty to economic and structural causes, such as unemployment and lack of educational opportunities, rather than to individual characteristics. Russian society in general is rather pes- simistic about the future, and the assessments of the poor are especially negative. These results suggest that even if poverty rates improve in the short term, wide- spread dissatisfaction with the process of economic and social policy reform may well persist for some time. Jeni Klugman and Jeanine Braithwaite 55 Notes Jeni Klugman is with the Research School of Social Sciences at the Australian National University and is an associate fellow to the International Child Development Centre of UNICEF, Florence. She is presently on leave from the Europe and Central Asia Regional Office of the World Bank. Jeanine Braithwaite is an economist in the Poverty Reduction and Economic Management Network of the World Bank. 1. The poverty headcounts are not overly sensitive to shifts in the threshold, however. If, in late 1995, the poverty line had been moved up by 10 percent, the poverty incidence rises by about 4 percent; a 10 percent lower threshold reduced the poverty headcount estimate by 5 percent (Kolev 1996). 2. The budget was intended to be an absolute measure of minimally acceptable consumption under a socialist system, rather than an absolute measure of poverty per se (since, after all, poverty did not officially exist). Thus the Soviet standard for protein consumption was twice the level rec- ommended in the United States (Lane, Marston, and Welsh 1987). 3. The published monthly income distribution data are based on the FBS, but are not actually summary totals from that survey. The FBS survey forms are collected and processed quarterly. The monthly income distribution data are synthetically generated by Goskomstat, using a previously tabulated FBS distribution as a historical template for the variance and grossing up the distribution by presumed increases in the mean. Average income is assumed to grow at the rate of such various monthly macroeconomic indicators as average wages or the wage funds of large state-owned enter- prises. Generally, such an estimating methodology would be expected to lead to understatements in measures of income dispersion. 4. A decomposition breaks down changes in the poverty headcount among three components: growth, redistribution, and a residual component that is ill-defined (Ravallion and Datt 1991). 5. The Employment Fund, which is responsible for financing income support for the unem- ployed as well as programs for the employed, faces both problems: the financing base is a 2 percent payroll tax, the bulk of which is retained at the oblast level, limiting scope for any redistribution of funds to finance unemployment benefits in badly hit regions. References The word "processed" describes informally reproduced works that may not be commonly available through library systems. Atkinson, Anthony B., and John Micklewright. 1992. Economic Transformation in Eastern Europe and the Distribution of Income. Cambridge, U.K.: Cambridge University Press. Braithwaite, Jeanine. 1995. "Armenia: A Poverty Profile." Psp Discussion Paper 80. World Bank, Poverty and Social Policy Department, Washington, D.C. Processed. - . 1997. "The Old and New Poor in Russia." In Klugman (1997). Braithwaite, Jeanine, and Timothy E. Heleniak. 1989. "Social Welfare in the USSR: The Income Recipient Distribution." Center for International Research, U.S. Bureau of the Census, Wash- ington, D.C. Processed. Commander, Simon, and R. Yemtsov. 1997. "Russian Unemployment: Its Magnitude, Character- istics, and Regional Dimensions." In Klugman (1997). Cox, Daniel, Zekeriya Eser, and Emmanuel Jimenez. 1997. "Family Safety Nets during Economic Transition: A Study of Inter-Household Transfers in Russia." In Klugman (1997). 56 The World Bank Research Observer, vol. 13, no. I (February 1998) Deaton, Angus, andJohn Muellbauer. 1980. Economics and Consumer Behaviour. Cambridge, U.K.: Cambridge University Press. de Masi, Paula, and Vincent Koen. 1995. "Relative Price Convergence in Russia." IMF Working Paper wp/95/54. International Monetary Fund, Washington, D.C. Processed. de Melo, Martha, Cevded Denizer, and Alan Gelb. 1995. "From Plan to Market: Patterns of Transition." World Bank, Policy Research Department, Washington, D.C. Processed. Dmitrichev, I. I. 1992. Byudzhet semi kak osnova dlya izucheniya urovnya zhizni naseleniya ["The Family Budget as the Basis for the Study of the Level of Living of the Population"; not available in translation]. Moscow: Goskomstat Rossiskoy Federatsii. Easterly, William, and Stanley D. Fischer. 1995. "The Soviet Economic Decline." The World Bank Economic Review 9(3):341-71. EBRD (European Bank for Reconstruction and Development). 1997. "Transition Report." London. Processed. Foley, Mark C. 1997. "Poverty in Russia: Static and Dynamic Analyses." In Klugman (1997). Foley, Mark C., and Jeni Klugman. 1997. "The Impact of Social Support: Errors of Leakage and Exclusion." In Klugman (1997). Gavrilenko, Eugeniy, and Vincent Koen. 1994. "How Large Was the Output Collapse in Russia? Alternative Estimates and Welfare Implications." IMF Working Paper 154. International Mon- etary Fund, Washington, D.C. Processed. Glinskaya, Elena, and Jeanine Braithwaite. Forthcoming. "Poverty, Inequality, and Income Mobil- ity in Russia: 1994-1996." Paper to be presented at the 1998 Annual Meeting of the Population Association of America, Chicago, April 2-4, 1998. Goskomstat. 1995; 1996. "Russian Federation: Report on National Accounts." World Bank, Eu- rope and Central Asia Department, Washington, D.C. Processed. Goskomstat Rossii. 1994-97. Monthly Statistical Bulletins. "Sotsio-ekonomicheskoye polozheniye Rossii" ["Socio-economic Situation of Russia"; not available in translation]. Moscow. Goskomstat Rossii and World Bank. 1996. "Russian Federation: Report on National Accounts." Europe and Central Asia Department, Washington, D.C. Processed. Hagendaars, Aldi. 1986. The Perception of Poverty. Amsterdam: North Holland Publishers. Klugman, Jeni, ed. 1997. Poverty in Russia: Public Policy and Private Responses. EDI Development Study. Washington D.C.: World Bank. Klugman, Jeni, and Jeanine Braithwaite. 1997. "Introduction and Overview." In Klugman (1997). Koen, Vincent. 1996. "Russian Macroeconomic Data: Existence, Access, Interpretation." Commu- nist Economies and Economic Transformation 8:321-33. Kolev, Alexander. 1996. "Poverty Analysis in Russia: What Can We Learn from the RLMS Round 6?" Florence: European University Institute. Lane, Ann M., Ruth M. Marston, and Susan 0. Welsh. 1987. "The Nutrient Content of the Soviet Food Supply and Comparisons with the U.S. Food Supply." In U.S. Congress, Joint Economic Committee, Gorbachev 's Economic Plans, vol. 2, pp. 79-99. Washington: U.S. Government Print- ing Office. Milanovic, Branko. 1995. "Poverty, Inequality, and Social Policy in Transition Economies." Policy Research Department Research Paper Series 9. World Bank, Washington, D.C. Pro- cessed. Mozhina, Marina. 1992. "The Poor: What Is the Boundary Line?" Problems of Economic Transition (October):65-75. Jeni Klugmnan and Jeanine Braithwaite 57 Popov, Vladimir. 1996. "A Russian Puzzle: What Makes the Russian Transformation a Special Case?" Research for Action 29. United Nations University/World Institute for Development Economics Research, Helsinki. Processed. Ravallion, Martin. 1992. Poverty Comparisons: A Guide to Concepts and Methods. LSMS Working Paper 88. Washington D.C.: World Bank. Ravallion, Martin, and Guarav Datt. 1991. Growth and Redistribution Components of Changes in Poverty Measures: A Decomposition with Applications to Brazil and India in the 1980s. LSMS Work- ing Paper 83. Washington D.C.: World Bank. Rimashevskaya, N. M., and others. 1979. Potrebnosti, dokhody, potrebleniye ["Needs, Income, and Consumption"; not available in translation]. Moscow: Nauka. Roberts, Bryan. 1995. "Price Liberalization, Market Power, and Social Welfare in Transition Econo- mies." University of Miami, Miami, Florida. Processed. Sarkisyan, G. P., and N. P. Kuznetsova. 1967. Potrebnosti i dokhodsem 'i: uroven' struktura, perspektivy. ["Needs and Incomes of Families: Level, Structure, and Perspectives"; not available in transla- tion.] Moscow: Nauka. Scherer, Peter. 1997. "The New Social Policy Agenda." Organization for Economic Cooperation and Development, Social Policy Division, Paris. Processed. Shenfield, Stephen. 1983. "A Note on Data Quality in the Soviet Family Budget Survey." Soviet Studies 35(4, October):561-68. Stewart, Kitty. 1996. "Regional Disparities in Social Expenditure in Russia." World Bank, Europe and Central Asia Regional Office, Washington, D.C. Processed. U.S. Department of Commerce, Economics and Statistics Administration. 1991. StatisticalAbstract of the United States, 11th ed. Washington, D.C.: Government Printing Office. World Bank. 1994. "Social Protection during the Transition and Beyond" vol. 1. Report 11748- RU. Europe and Central Asia Department, Washington, D.C. Processed. . 1995a. "Fiscal Management in the Russian Federation." Europe and Central Asia Depart- ment, Washington, D.C. Processed. . 1995b. "Poverty in Russia: An Assessment." Report 1411O-RU. Europe and Central Asia Department, Washington, D. C. Processed. Zubova, Larissa, and Natalia Kovaleva. 1997. "Public Opinion about Social Issues." In Klugman (1997). 58 The World Bank Research Observer, vol. 13, no. I (February 1998) How Bad Is Unemployment in Tunisia? Assessing Labor Market Efficiency in a Developing Country Martin Rama Tunisia's unemployment rate has been among the highest in the world for almost two decades. This article claims that such a high rate reflects measurement problems rather than labor market inefficiency. After discussing the reasons why unemployment rates may not be comparable across countries and reviewing the tools that are available to analyze unemployment in a specific country, the articleprovidesfourpieces ofevidence to substan- tiate its claim. Two of them relate to the criteria used in Tunisia to measure unemploy- ment and the way these criteria have changed over time. Two use records on the number ofactivejob seekers and vacancies as reported to the official employment agency. Together, this body of evidence suggests that unemployment has declined steadily over time and remains an issue for first-time job seekers only. Tunisia's unemployment rate is unusually high. Drawing from the 1989 labor force survey, official estimates put that rate at 15.3 percent. Unofficial estimates for more recent years, based on labor force and employment forecasts, yield rates as high as 16 to 17 percent. Except for economies in transition to a market orientation, only a handful of countries-Algeria, Barbados, Finland, Jordan, the Seychelles, South Af- rica, Spain, and Trinidad and Tobago-report similar or higher rates (ILO, 1996). Whether unemployment is as high in Tunisia as the estimates suggest is an issue that the population census of 1994 could clarify, but it may take some time before the census data are processed, and it is not obvious that the computations will rely on an internationally comparable definition of unemployment. In the meantime, the country confronts major policy reforms, including the implementation of a free trade agree- ment with the European Union and the privatization and downsizing of state-owned enterprises. The problem is that the consequences of these reforms depend on how efficient the labor market actually is. The World Bank Research Observer, vol. 13, no. I (February 1998), pp. 59-77 1998 The International Bank for Reconstruction and Development / THE WORLD BANK 59 If the available unemployment figures are correct, the labor market is very ineffi- cient, which may well force the government to reconsider its policy choices. On the one hand, if it seems likely that those workers who lose their jobs as a consequence of trade liberalization and ownership divestiture will be unable to find new jobs in competitive sectors of the economy, or that they might manage to do so only after a long period of unemployment, the government presumably will hesitate to imple- ment the reforms. On the other hand, the high unemployment rate could indicate significant distortions in the labor market that require a radical reform of labor mar- ket policies and institutions. In general, the almost automatic reaction to a high and stable unemployment rate is to aim at minimum wage and collective bargaining mechanisms that could be pricing workers out of jobs. Assessing the efficiency of the labor market is seldom an easy task, even in indus- trial countries. Conceptually, the issue seems quite simple: the labor market is con- sidered inefficient when a large number of individuals who would be willing to work (or to work more) at the prevailing wage rate are unable to do so. In practice, how- ever, both work and willingness to work can be measured in different ways. For instance, it has been argued that Japan's unemployment rate, which is the lowest among industrial countries, is biased downward because some individuals who are not working are counted as employed, while they would be considered unemployed elsewhere (Hashimoto 1993). Similarly, Spain's unemployment rate, the highest in the industrial world, is said to be overestimated because many employed workers are not registered with the social security system (for conflicting views on the magnitude of this bias, see Franks 1994 and Blanchard and Jimeno 1995). Measurement problems are magnified in the case of developing countries because self-employment and work within the family are much more prevalent than in in- dustrial countries. As a result, the concept of willingness to work "at the prevailing wage rate" loses some of its meaning. For instance, for many women the distinction between working on a family farm (therefore, being a "worker") and taking care of the home (hence being "inactive") is tenuous, as is the distinction between being unemployed and working in the informal sector for many men. Similarly, it is un- clear whether a young high school graduate looking for a job in the formal sector and meanwhile refusing to work in the informal sector should be considered unemployed or rather inactive. When all of these problems are taken into account, it becomes obvious the unemployment rate can be measured in more than one way (World Bank 1995c, chapter 3). It is therefore important to use the same measure when making comparisons across countries or over time for the same country. The credibility of any assessment of labor market efficiency will vary substantially depending on the quality of the data available for the analysis. For example, microeconomic data on living standards may produce a detailed profile of the unem- ployed that could help explain why they are out of work. Data on earnings and individual characteristics of workers (such as education, experience, and the like) 60 The World Bank Research Observer, vol. 13, no. I (February 1998) and their jobs can be used to evaluate whether different sectors pay the same wage to similar workers, which in turn could help identify the sources of labor market ineffi- ciency. But in many cases these data are not available, and governments cannot post- pone key policy choices until they are collected and processed. The challenge is there- fore to assess labor market efficiency based on the kind of data usually published by the statistical offices of developing countries. In this respect, focusing on a country like Tunisia represents an even bigger challenge. Nowhere in the world is the paucity of labor market statistics more acute than in the Middle East and North Africa re- gion (Rama 1997). If a credible assessment of labor market efficiency is feasible in Tunisia, then it should be feasible in other developing countries as well. This article argues that Tunisia's labor market is not nearly as inefficient as the unemployment figures suggest. Four pieces of evidence substantiate this claim. Each of them is partial and subject to criticism, but they all point in the same direction. The first concerns the criteria used to measure the labor force, and the way these criteria have changed over time. The second relates to the methods used to forecast the size of the labor force and total employment since the last labor force survey. The third relies on the number of active job seekers and the number of available vacan- cies, as reported to the official employment agency. The last uses these records to compare the efficiency of the matching process between job seekers and vacancies in Tunisia with that of other countries. The same conclusion cannot be generalized to other developing countries with reportedly high unemployment rates, however. The article highlights the problems surrounding unemployment figures in developing countries, and suggests possible ways to understand what these figures really mean. But it does not claim that the biases observed in Tunisia plague the official figures of other developing countries as well. It only shows that biases of this sort may exist and may lead to inappropriate policy recommendations. Because the nature of any measurement biases is likely to vary from country to country, it is important to make the most of the labor market data available in each country. The four pieces of evidence examined here illustrate this broader proposition. A Conceptual Framework One of the problems in deciding whether an unemployment rate is actually "high" or "low" occurs because the most popular model of the labor market sets the equilib- rium unemployment rate at zero. In practice, however, even the tightest labor mar- ket displays a strictly positive unemployment rate. In the basic model, shown in figure 1, as wages rise (on the vertical axis), more people are willing to seek jobs. This is shown by the increase in the labor supply (on the horizontal axis) along the up- ward-sloping line. Conversely, when wages are extremely low, employers are willing Martin Rama 61 to hire large numbers of workers. But as wages rise, labor demand declines along the downward-sloping line. At the wage level WE the number of jobs offered equals the number of workers seeking employment. Labor market equilibrium is thus charac- terized by no unemployment at all. Within this simple model, the only reason why unemployment could be "high" (strictly speaking, positive) is because wages are misaligned. Assume that a minimum wage or a union-negotiated increase pushes the wage level up (in figure 1, from WE to WA). In this case, LSA individuals would be willing to work, but only LDA would find jobs, because wages would be so high that employers could not afford to hire many workers. This distorted equilibrium is what comes to the minds of many econo- mists when data reveal a high unemployment rate. But in developing countries, usu- ally characterized by a large informal sector, this interpretation is problematic. In- deed, most of the potential sources of wage misalignment are irrelevant in the informal sector: minimum wages are not enforced and unions are almost nonexistent. The Figure 1. A Basic Model of the labor Market Wages W ............... ...... ..Lb........................ .............. ..................L............ /Lab o r supply .. ....................:'' ~~~~~~~~. ....... ....... ..... . J. ./... ..\ Labor demand LI L LD L., LD ,L,,S ADA LS,, Employment LD 4 Source: Author's calculations. 62 The World Bank Research Observer, vol. 13, no. I (February 1998) labor demand curve in figure 1 corresponds to the formal sector only. As wages in this sector rise, some workers lose their jobs. But instead of being forced into unem- ployment, they shift to the informal sector. There are several ways of introducing equilibrium unemployment in simple models of the labor market like the one represented in figure 1. One assumes that workers in the informal sector cannot join the formal sector without previously spending some time unemployed. In the context of development economics, this assumption has been justified on the grounds that the formal sector is mostly urban. Agricultural workers choose between staying in rural areas and migrating to urban centers, where they may either find a job in the modern sector or remain unemployed. The equilibrium unem- ployment rate is the one that equalizes the expected payoffs of these two alternatives (Harris and Todaro 1970). That the informal sector is quite large in urban areas makes this explanation of unemployment unconvincing, however. Rural migrants could in- deed work in the urban informal sector while waiting for openings in the formal sector. A second way of introducing unemployment consists of assuming frictions in the matching process between job seekers and vacancies. The labor market is actually characterized by large flows in and out of employment. There are always individuals who quit their jobs, are fired, retire, or die, while newcomers are constandy joining the labor force. It takes time for job seekers to learn about openings that are suitable to them and for firms to find candidates who meet their needs. So even if the supply of labor were equal to demand at the aggregate level, there would always be some individuals in search of a job and some firms in search of the appropriate workers (Pissarides 1985; Blanchard and Diamond 1989). Because of the frictions in the matching process, actual employment is less than both labor supply and labor demand, as represented by the broken curve in figure 1. For instance, when the wage level is high (equal to W), employment is given by LA and there are simultaneously LDA - LA unfilled vacancies and LSA - LA unsuccessful job seekers. Conversely, when the wage level is low (equal to WB) the employment level is L. and there are simultaneously LDB - LB unfilled vacancies and LS, - LB unsuccessful job seekers. Although these two situations may look alike in terms of employment, they are radically different in terms of their ratios of vacancies to un- employment. Job seekers have better prospects of finding a job in the second situa- tion, so the time during which they are unemployed should be shorter. More gener- ally, the joint behavior of unemployment and vacancies can provide more information on the functioning of the labor market than the mere unemployment rate does. A convenient representation of the joint behavior of unemployment and vacan- cies is provided by the Beveridge curve (figure 2). The curve on the left is based on the diagram in figure 1. Point A in this curve corresponds to the wage level WA, while point B corresponds to the wage level WB. The other curve displays a higher number of unsuccessful job seekers for any number of vacancies, or equivalently, a higher number of vacancies for any number of unemployed persons. These two curves are Martin Rama 63 Figure 2. The Beveridge Curve Unemployment A X A' Vacancies Source: Author's calculations. useful to illustrate the difference between movements along the Beveridge curve and shifts of the Beveridge curve. A movement from point A to B indicates that the labor market has tightened, without changing its overall efficiency. Business cycles are characterized by back-and-forth movements of this kind. A movement from point A to point A', in turn, reflects a more inefficient labor market. That is, the same num- ber of job seekers remain unemployed despite a larger number of unfilled vacancies. For instance, it has been argued that labor market policies and institutions in Europe are such that the temporary increase in unemployment created by the oil shocks (from B to A) has reduced labor market efficiency and become a more permanent situation (from A to A', rather than back to B). Another convenient tool to assess the importance of labor market frictions is pro- vided by matching functions. These functions link the number of new placements, H, to the number of job seekers, U, and the number of vacancies, V In a well- functioning labor market, Hshould increase when more individuals are actively search- ing for a job and when more vacant posts are available as well. In analytical terms, the number of placements is represented as follows: H = h(U, V9 where ho is the matching function. The shape of this function has been investi- 64 The World Bank Research Observer, vol. 13, no. I (February 1998) gated for several industrial countries. The econometric results show that, in gen- eral, if the number of job seekers and the number of offers doubled, placements would roughly double too. In technical terms, the ho function is characterized by constant or near-to-constant returns to scale. The results also show that both argu- ments of the function matter. Even in cases of massive job destruction, as in the transition economies, the stock of the unemployed remains an important determi- nant of the flow of hirings (see Burda 1993, on the former East Germany). This is in sharp contrast with the frictionless labor market shown in figure 1. Indeed, in that model, when the wage level was WA, employment was determined exclusively by labor demand. The analysis in this article proceeds as follows. First the supply of labor in Tunisia is measured according to internationally comparable criteria. The criteria used in Tunisia make the labor supply curve in figure 1 appear to be much farther to the right of the diagram than in other countries. Moreover, these definitions have changed over time in a way that shifts the curve even farther to the right. Second, employ- ment estimates for recent years are examined. The criteria used in Tunisia amount to representing the labor demand curve in figure 1 to the left of its actual location. Third, based on government placement agency records on the number of vacancies and job seekers, a rudimentary Beveridge curve is drawn. This exercise shows a move- ment in the direction of a tighter labor market in recent years, like the one from point A to point B in figure 2. And finally, to assess whether the labor market oper- ates efficiently, a matching function is estimated for Tunisia. Except for first-time job seekers, the results are similar to those observed in several industrial countries over periods characterized by moderate unemployment rates. These results, together with the other evidence, suggest that unemployment is not a serious problem in Tunisia except for first-time job seekers. Defining the Labor Force The criteria used to measure unemployment in Tunisia differ from those commonly applied in other countries in two respects. The most frequently mentioned discrep- ancy concerns job seekers of ages 15 to 17, and 60 and above, who appear on the unemployment rolls in most countries but who are not included in Tunisia's unem- ployment figures. Because these people are available to work, observers believe that official unemployment figures underestimate the magnitude of the problem. The second discrepancy, however, more than offsets this underestimation. Tunisian data count a set of inactive people (mostly housewives) as unemployed, although they would not be counted as such elsewhere. In 1966 the definition of the labor force used in Tunisia was similar to that used in other countries. The labor force thus included all individuals who either held a job or Martin Rama 65 were actively searching for one. In the discussion here, this definition is referred to as the effective labor force. In Tunisia, however, this definition was considered problem- atic because it yielded a higher unemployment rate for men than for women, which is uncommon. Of course, this gender gap may say more about the condition of women at that time than about the appropriateness of the definition. But in practice, the prob- lem led to several endeavors by the statistical office (Institut national de la statistique) to identify potentially discouraged job seekers among inactive women. For this reason the standard definition of the economically active population was gradually extended to include the marginal labor force first, and then the potential labor force. The marginal labor force comprises individuals who do not consider themselves economically active but who did some work in the preceding three months. More specifically, all the interviewees were first asked whether they considered themselves employed, unemployed, housewives, students, in military service, and so on. Then, all the inactive respondents, apart from those in military service, were asked whether they had worked in the three months preceding the survey. Those who did were switched from inactive to marginally active, and then asked whether they worked at least one hour during the week preceding the survey. The answer was overwhelm- ingly negative, which is not surprising given that these individuals reported that they were inactive. As a result, most of the marginally active were counted as unemployed. The potential labor force consists of a subset of housewives who have some free time and would like to have a job. More specifically, those interviewees who classi- fied themselves as housewives and who did not do any other work during the previ- ous three months, not even as family workers, were asked whether their housekeep- ing tasks kept them busy all day. If so, these women are considered inactive. But if not, they were asked whether they would be willing to take a job if it were offered. An affirmative answer shifts them from inactive to potentially active. Not surpris- ingly, most of them were reported unemployed when asked whether they worked at least one hour during the week preceding the survey. A consistent presentation of labor force and employment data based on existing population censuses and labor force surveys is shown in table 1. In this presentation all job seekers aged 15 to 17 or 60 and above are counted as unemployed, so the data are comparable to those of other countries. The presentation also uses the question- naires of the censuses and surveys to disentangle whether the figures include the marginal and the potential labor force. The table should be interpreted with caution though, because it is based on published data and not on the original information. In particular, when the same presentation used to construct the table is applied to much narrower groups of individuals (for example, men aged 60 and above) certain incon- sistencies emerge that highlight the limits of this approach. At an aggregate level, however, table 1 yields two insights. The first insight is that unemployment has declined over time. This is not the story line that emerges from official unemployment figures, which, after adding young 66 The World Bank Research Observer, vol. 13, no. I (February 1998) and old job seekers, have fluctuated around 15-16 percent from 1966 to 1989. (In 1980 the official unemployment rate was lower, but that year appears to be an out- lier). If comparable unemployment rates are considered, instead of the official ones, the story is one of decreasing unemployment. For instance, if data on the effective labor force are used, the unemployment rate declines from 15.2 percent in 1966 to 12.1 percent in 1980 to 11.2 percent in 1989. And the trend is similar when using broader definitions of the labor force. (Again, 1980 was the only exception, a year that may be seen as idiosyncratic). The second insight is that most of the unemployment problem is associated with new entrants to the labor force. When measured according to standard criteria, the unemployment rate was around 11.2 percent in 1989. But when first-time job seek- Table 1. Unemployment Rates in Tunisia Unemployment data 1966 1975 1980 1984 1989 1990 Number of persons (thousands) A. Employed, ages 15+ 927.3 1,366.5 1,576.9 1,786.4 1,978.8 - B. Effective actives, ages 15+ 1,093.7 - 1,793.3 - 2,229.1 C. Marginal actives, ages 15+ - - 16.5 - 89.1 - D. Effective + marginal,ages 15+ - 1,621.8 1,809.8 - 2,318.2 - E. Potential actives, ages 15+ - - - - 42.4 F. Total actives, ages 15+ - - - 2,137.2 2,360.6 - G. First-time job seekers, ages 18-59 - - - - 135.2 - H. First-time job seekers, ages 15-17 - - - - 54.8 I. First-time job seekers, ages 15+ 40.8 119.1 89.8 - 190.0 - Official unemployment rates (percent) Total 15.21 15.74 12.87 16.41 16.17 - Household heads 11.74 - 7.23 - - 1.40 Comparable unemployment rates (percent) (B-A)/B 15.21 - 12.07 - 11.23 - (D-A)/D - 15.74 12.87 - 14.64 - (F-A)/F - - 16.41 16.17 - Rates excluding first-time job seekers (percent) (B - A - I)/(B - I) 11.92 - 7.43 - 2.95 - (D - A - I)/(D - I) - 9.06 8.32 - 7.02 - (F - A - I)/(F - I) - - - 8.84 - -Not available. Source: Recensementge'ne'ral de la population et des logements, 1966; Recensementgeneral de la population et des logements, 1975; Enquete population-emploi, 1980; Recensement ge'ne'ral de la population et de l'habitat, 1984; Enquete nationale population-emploi, 19891, Enquete nationale sur le budget et la consommation des menages, 1990 (tables produced for the World Bank); Institut national de la statistique, Tunis. Martin Rama 67 ers are excluded, the "core" rate was less than 3 percent. Moreover, the unemploy- ment rate of heads of households was as low as 1.4 percent in 1990. Although these figures refer to a subset of the unemployed, they are more informative about the efficiency of the labor market than the consolidated unemployment rates. Many first-time job seekers can be considered "voluntarily" unemployed because they take advantage of family support to wait for the right job opening, while rejecting existing work opportunities that are attractive. Voluntary unemployment is less likely among those who had a job and lost it. Methods of Forecasting Unemployment Estimates of the unemployment rate for years after 1989 are based on a comparison of employment forecasts and labor force projections. The problem here is that the method used to forecast employment leads to a downward bias, while the method used to forecast the labor force leads to an upward bias. By underestimating the number of people at work and overestimating the number of people available for work, these two biases lead to an unrealistically high unemployment rate. The employment forecasts are produced by the planning ministry (Ministere du developpement economique), which updates the 1989 employment figures to ac- count for the creation of new jobs and the destruction of old ones. New jobs are identified from employers' reports to the investment promotion agency (Agence de promotion des investissements). Job destruction is estimated based on social security records and on data from the labor inspection of the ministry of labor. Based on this approach, 2.13 million people would have been at work in Tunisia in 1993, a figure used here to illustrate the bias resulting from current forecasts. But this approach has two shortcomings: the estimates of job creation and job destruction refer only to the formal sector; and they probably underestimate job creation more sharply than job destruction. One reason why job creation is underestimated is that the declaration of new jobs by employers is not compulsory. In the past, employers had to report their investment projects and the associated new jobs to the Agence de promotion des investissements in order to get subsidies; but reporting is now a mere formality. Moreover, employers are only asked to dedare new permanent jobs, whereas employment growth in recent years has tended to rely heavily on temporary contracts, even in the formal sector. Last but not least, there is an evasion problem. Firms tend to declare only the personnel they register with the social security system. And they do not register all their personnel because they prefer to escape payment of social security contributions. The coverage of job destruction, which is based on social security records and data from the labor inspection of the ministry of labor, is better than the coverage of job creation. It still refers only to the formal sector but covers both permanent and tem- 68 The World Bank Research Observer, vol. 13, no. I (February 1998) porary workers. As a result of these differences, the planning ministry underesti- mates employment growth in the formal sector. Moreover, the methodology used by the ministry implicitly assumes zero employment growth in the informal sector of the economy, although most observers agree that this sector has expanded since 1989. The underestimation of employment growth in both the formal and (to a much larger extent) the informal sectors suggests that employment forecasts are likely to suffer from a substantial downward bias. Forecasts of the labor force, in turn, extrapolate from the growth rate observed in previous years. Taken at face value, the data from population censuses and labor force surveys imply the Tunisian labor force grew at a rate of roughly 3 percent a year be- tween 1980 and 1989. Part of this growth is fictitious, however, because it results from the change in the definition of the labor force. As noted earlier, the 1989 survey in- cluded the potential labor force among the economically active, but the 1980 labor force did not. A more reliable estimate can be obtained by comparing the effective labor force in both years, which yields an annual growth rate close to 2.5 percent. The comparison between the total labor force (including marginal and potential actives) in 1984 and 1989 yields an even lower annual growth rate of about 2 percent. A different picture of unemployment emerges depending on which of these three growth rates is used. If the 3 percent annual growth rate of the labor force is applied to the 1989 effective labor force, the corresponding figure in 1993 would have been 2.51 million. When the 2.5 percent growth rate is used instead, the effective labor force forecast is 2.46 million. At a 2 percent growth rate, it is 2.41 million. Assume for a moment that the planning ministry's employment forecast is not biased, so that total employment in 1993 can be estimated at around 2.13 million. Depending on whether the labor force actually grew at a rate of 3, 2.5, or 2 percent a year, the unemployment rate in 1993 would thus be 15.1, 13.1, or 11.7, respectively. Note that this last figure is close to the one observed in 1989, which suggests some stability of the unemployment rate in recent years. But in fact total employment is underesti- mated, so that the "true" unemployment rate for 1993 was likely to have been lower. Unemployment versus Vacancies The efficiency of the Tunisian labor market can also be assessed based on trends in the number of unfilled vacancies as well as in unemployment. For this measurement, the data collected by the government's placement agency (Agence Tunisienne pour l'Emploi, or ATE for short) turns out to be particularly valuable. When Tunisia was a control economy, publicly owned firms had to report all of their vacancies to the ATE, which was in charge of filling them based on its own records of job seekers. As the economy became more market oriented, the obligation of reporting vacancies to the ATE remained, but the share of hirings directly processed by firms increased. Martin Rama 69 Table 2. UnemploymentandJob Vacancies in Tunisia 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 Number ofpersons Unemployed registered 189,529 173,922 189,152 206,858 194,697 209,737 237,302 174,193 152,153 133,072 136,885 142,212 160,224 Skilled 51,931 50,292 57,496 72,379 64,940 68,215 73,482 45,554 27,363 23,960 24,035 29,184 34,692 Unskilled 82,213 67,976 73,731 80,556 76,954 77,738 84,924 60,681 65,400 57,008 53,594 57,012 57,816 First job 55,385 55,654 57,925 53,923 52,803 63,784 78,896 67,958 59,390 52,104 59,256 56,016 67,716 Vacancies reported 49,678 45,345 53,057 62,616 56,421 61,992 62,619 67,642 80,419 68,620 65,292 68,100 81,653 Skilled 18,560 18,567 23,269 25,975 21,338 24,624 26,981 29,585 34,332 29,428 31,045 34,956 44,522 Unskilled 31,118 28,778 29,788 36,641 35,083 37,368 35,638 38,057 46,087 39,192 34,247 33,144 37,131 Total placements 43,187 41,730 46,937 56,813 53,348 58,055 57,827 62,886 75,418 64,325 61,415 65,484 76,559 Skilled 12,076 12,508 15,387 19,621 17,489 19,243 19,368 20,915 16,244 15,023 17,439 18,036 20,898 Unskilled 27,313 25,655 26,125 32,259 31,861 34,508 33,104 34,425 42,239 37,805 32,379 31,848 36,845 First job 3,798 3,568 5,425 4,933 3,998 4,304 5,355 7,546 16,935 11,497 11,597 15,600 18,816 Percentage of effective labor force Unemployed registered 9.99 8.86 9.31 9.99 9.23 9.76 10.85 7.81 6.70 5.75 5.81 5.92 6.55 Skilled 2.74 2.56 2.83 3.50 3.08 3.18 3.36 2.04 1.21 1.04 1.02 1.22 1.42 Unskilled + first job 7.26 6.30 6.48 6.50 6.15 6.59 7.49 5.77 5.50 4.72 4.79 4.71 5.13 Vacancies reported 2.62 2.31 2.61 3.03 2.68 2.89 2.86 3.03 3.54 2.97 2.77 2.84 3.34 Skilled 0.98 0.95 1.15 1.26 1.01 1.15 1.23 1.33 1.51 1.27 1.32 1.46 1.82 Unskilled 1.64 1.47 1.47 1.77 1.66 1.74 1.63 1.71 2.03 1.69 1.45 1.38 1.52 Total placements 2.28 2.13 2.31 2.75 2.53 2.70 2.64 2.82 3.32 2.78 2.61 2.73 3.13 Skilled 0.64 0.64 0.76 0.95 0.83 0.90 0.89 0.94 0.72 0.65 0.74 0.75 0.85 Unskilled + first job 1.64 1.49 1.55 1.80 1.70 1.81 1.76 1.88 2.61 2.13 1.87 1.98 2.28 Source: Agence Tunisienne pour l'Emploi, and author's estimates of the effective labor force. Despite the shift from a control to a market economy, however, the ATE still is an important labor market intermediary. The absolute numbers of job seekers registered with, and vacancies reported to, the ATE from 1982 through 1994 is shown in table 2. The table also presents the data on job seekers and vacancies in percentage of the effective labor force, with the latter estimated based on comparable observations from population censuses and labor force surveys under the assumption of a constant growth rate between observations. The relative stability in the number of placements by the ATE is striking. If anything, the ratio between these placements and the effective labor force has gone up over time. At the beginning of the period considered, total annual placements amounted to slightly more than 2 percent of the effective labor force; at the end of the period the figure was close to 3 percent. These figures suggest that ATE records may still reflect the main trends in unemployment and vacancies, despite the shift to a market economy. The picture that emerges from these records is one of an increasingly tight labor market. This interpretation is straightforward when the data are displayed in a Beveridge curve format, as in figure 3. Both unemployment and vacancies are mea- sured in this figure as percentages of the effective labor force, which controls for the increasing size of the labor market. The clusters of points show that there were more vacancies and fewer job seekers in the early 1990s than in the 1980s. The similarity between the two groups highlighted in figure 3 and points A and B in figure 2 is striking and suggests that Tunisia experienced a movement along the Beveridge curve in recent years, in the direction of a tighter labor market. Given the intermediate position occupied by 1989, the year of the most recent labor force survey, it is safe to conclude that the unemployment rate must be lower now than it was at that time. The movement along the Beveridge curve is even more pronounced when only skilled workers and vacancies are considered. At the other end of the spectrum, first- time job seekers may face higher unemployment rates despite an increase in the overall number of vacancies available. Several explanations for this trend are pos- sible. The first is that sustained growth has created increased affluence. Families can now afford to support their young members for longer periods of unemployment, thus allowing them to find better jobs. Another possible explanation is that increased education raises the expectations of these new entrants to the labor market, making them reluctant to take the jobs that are actually available. Evaluating the empirical relevance of these two hypotheses is beyond the reach of this article, however. Placements and the Matching Process New jobs are created when employers find job seekers who are suitable to fill their vacancies and who are in turn interested by those vacancies. Limited job creation can Martin Rama 71 Figure 3. Unemployment and Vacancies in Tunisia as a Percentage of the Labor Force Unemployment , - 1988 7' 1982 1985.' 10 ,' U , 1984 1987,' 9 , 1983 1986 , _,.. ..... 8 8 ~~~~1989 7 - 1990 ,' ~~ / 1994 , 1993 e \\ 1992 *1991 , ~1992- 5- 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 Vacancies Source: See table 2. therefore be expected when few vacancies are available or when the process of match- ing vacancies and job seekers is very inefficient. The Beveridge curve analysis in the previous section suggests that there is no shortage of vacancies in Tunisia. According to the ATE records, many more vacancies were reported in the early 1990s than in the 1980s, even though firms are no longer obliged to report vacancies to the ATE. There- fore, if the unemployment problem were as serious as the official unemployment rates indicate, the matching process would be very inefficient in Tunisia. To assess whether this is so, matching functions estimated from ATE records can be compared to matching functions estimated in industrial countries during periods of moderate unemployment. Whereas the ATE records distinguish between skilled and unskilled vacancies, they consider a third category of unemployed workers, namely first-time job seekers (who are not considered skilled by employers). Therefore, two separate matching func- tions are estimated. One explains new placements of skilled workers as a function of the number of skilled vacancies and of skilled job seekers. The other matching func- 72 The World Bank Research Observer, vol. 13, no. I (February 1998) tion explains new placements of unskilled workers as a function of the number of unskilled vacancies and of unskilled and first-time job seekers. Matching functions are usually estimated under the assumption that a 1 percent increase in the number of vacancies leads to the same percentage change in the num- ber of new placements, regardless of the number of job seekers. Conversely, a 1 percent increase in the number of job seekers is expected to lead to the same percent- age change in the number of new placements, regardless of the number of vacancies available. In more technical terms, a Cobb-Douglas specification is used. Also, a time trend is often included in the specification. If, other things equal, the number of new placements increases over time, then it is safe to conclude that the matching process is becoming more efficient. More specifically, matching functions usually take the form: log Hit = a, + (xlog U* + j3log Vi, + & + Ei, where the subindex t corresponds to the period, and the subindex i to the district or geographic area to which the data refer. Depending on the countries, the t or the i dimension may be missing. Parameter oc measures the percentage change in place- ments when the number of job seekers increases by 1 percent, and parameter ,B the corresponding change when the number of vacancies increases by 1 percent. The sum oc + 1 indicates the returns to scale of the matching function. If this sum is close to one, doubling the number of vacancies and job seekers roughly doubles the num- ber of new placements. Parameter 6, in turn, shows whether the efficiency of the matching process increases (for 6 positive) or decreases (6 negative) over time. Fi- nally, E is an error term. The results obtained when estimating matching functions on Tunisian data are reported in table 3. Note that the results probably represent a lower bound for labor Table 3. The Matching Function in Tunisia Time Unem- Returns Adjusted Sample Data Constant trend ployment Vacancies to scale R2 Skilled job Yearly 0.684 0.0040 0.213 0.655 0.87 0.850 seekers and data, (0.525) (0.832) (3.785) (5.866) vacancies 1970-94 Unskilled job Yearly 0.271 0.0261 0.014 0.930 0.94 0.912 seekers and data, (0.220) (2.899) (0.159) (9.669) vacancies 1970-94 Source: Author's estimates, using data from the Agence Tunisienne pour l'Emploi. The dependent variable is the log of the yearly number of placements. Independent variables are the log of registered unemployed and the log of reported vacancies. Coefficients were estimated using the AR(1) method. These coefficients are statistically significant when the corresponding t-statistics (in parentheses) are higher than 2. The Durbin- Watson statistics were 2.03 for the first equation and 2.06 for the second one, which implies that there are no auto-correlation problems. Martin Rama 73 market efficiency. The data are from a government placement agency, and it would not be surprising if paperwork, red tape, and relatively low levels of effort by civil servants slowed down the placement process. Private sector matchmaking, operating through personal networks, press ads, and other informal mechanisms, is likely to be more efficient than matchmaking by the ATE. This distinction may not be that cru- cial when comparing the matching function in Tunisia and in industrial countries, given that the data used to estimate the latter are from public placement agencies like the ATE. Available estimates for industrial countries are reported in table A- 1. The estimates for Tunisia show a striking difference between a relatively efficient matching process for skilled workers, and a clearly inefficient one for unskilled and first-time job seekers. For the skilled segment of the labor market, these estimates are comparable to those in industrial countries. To some extent, this should not be sur- prising, given that a skilled job in Tunisia may be similar to an average job in an industrial country. Still, it is worth noting that the coefficients for Tunisia are almost indistinguishable from those for England and Wales, Sweden, or the United States in periods where the unemployment rate of these countries was well below 10 per- cent. In all of these cases, a 1 percent increase in the number of vacancies leads to larger variation in the number of placements than a 1 percent increase in the number of job seekers. In terms of the equation above, 13 is larger than oc. Also, returns to scale (the sum oc + 1 are close to one, or slightly below. But while the efficiency of the matching process declines over time in all of the industrial countries for which there are estimates (that is, parameter 6 is negative), there is no such downward trend in the Tunisian data. The picture is different when it comes to unskilled workers and first-time job seekers. The coefficients of the matching function in this case are not comparable to those observed in industrial countries. In particular, hirings appear to be driven by the stock of vacancies only, without any noticeable effect on the stock of job seekers. Note that this result aggregates over both first-time job seekers and unskilled unem- ployed, so that the role played by each of these two groups cannot be disentangled. But the lack of significance of the stock of job seekers could actually indicate that people in any or both of these two groups are not willing to take any job they are offered. This is most likely to be true in the case of first-time job seekers. Conclusion Tunisia's unemployment problem appears to have been greatly overstated. The use of comparable labor force definitions shows that unemployment has declined quite steadily over the years, which is hardly surprising given the remarkable performance of the Tunisian economy. The extent of the decline is partly hidden by the increase in the number of individuals seeking their first job. But the unemployment rate 74 The World Bank Research Observer, vol. 13, no. I (February 1998) excluding first-time job seekers is low by any standard, and the unemployment rate of household heads is low too. Trends in the number of vacancies relative to the number of job seekers indicate that, if anything, the unemployment rate has de- clined since the last available estimation. A critical evaluation of the methods used to forecast unemployment rates and an econometric analysis of the data on vacancies and unemployment also suggest a relatively efficient labor market. Although the diagnosis in this paper is based on piecemeal evidence, the 1994 population census offers a valuable opportunity to refine it. Data from this census should be processed using the standard international definition of unemployment, in addition to the one used in Tunisia, to allow more accurate comparisons with other countries. Processing the available information on first-time job seekers would improve our understanding of the reasons for their long unemployment spells, and help in designing employment policies targeted at this specific group. Trade liberalization and public ownership divestiture will of course entail a sig- nificant reallocation of labor across firms and activities. But it is not obvious that these reforms will lead to a surge of unemployment in the short run because first- time job seekers, who constitute the bulk of the unemployed, will be largely unaf- fected. The unemployment spells of first-time job seekers are usually determined by mismatch between skills and jobs, by expectations about public sector job openings, by the availability of family support during the job search, and the like. The decline and eventual increase in blue-collar jobs resulting from trade liberalization and privatization is probably irrelevant in this respect. Therefore, the level of unemploy- ment rates in Tunisia does not justify delaying much needed structural reforms. From a broader perspective, this paper highlights some caveats that may be useful when interpreting unemployment data in developing countries, as well as some short- cuts that may help assess labor market efficiency based on these data. If piecemeal evidence can be put together in a coherent way in a country such as Tunisia, where labor market statistics are partial and scattered, it should also be feasible in other developing regions of the world, where data tend to be better on average. This kind of critical evaluation of the available statistics is particularly recommended before undertaking labor market reforms, such as changing the wage-setting mechanisms or implementing safety nets. In this respect, no analytical model of the labor market can replace a good understanding of the facts. Martin Rama 75 Table A.1. The Matching Function in Selected Countries Sample Time Unem- Returns Adjusted and source Data Constant trend ployment Vacancies to scale R2 Austria Monthly -1.30 -0.0172 0.76 0.41 1.17 0.66 (Christl 1992) data, (2.5) (6.8) (13.2) (4.4) 1974-89 Former Federal Daily data, 0.88 0.11 0.99 0.93 Republic of by district, (42.7) (9.2) Germany 1990-92 (Burda 1993) England and Data -1.19 0.34 0.66 1.00 0.91 Wales (Coles by district, (-7.2) (10.6) (16.0) and Smith for 1987 1996) Israel Monthly -1.00 -0.0063 0.45 0.49 0.94 0.63 (Berman data, (-0.7) (3.4) (4.9) (5.5) 1997) 1978-90 Sweden Monthly Not -0.0005 0.23 0.56 0.79 0.91 (Edin and data, reported (-4.91) (6.76) (23.52) Holmlund 1970-88 1991) United States Monthly 0.52 -0.0015 0.35 0.54 0.89 0.47 (Blanchard and data, (7.50) (-2.40) (3.90) (6.90) Diamond 1989) 1968-81 Note: The dependent variable is the log of placements, referrals, or flows out of unemployment. Indepen- dent variables are the log of the registered unemployed and the log of the reported vacancies. t-statistics are in parentheses. A blank indicates that the corresponding variable was not included in the econometric analysis. Notes Martin Rama is senior economist with the Development Research Group of the World Bank. He would like to thank Nasser Jharsalli and Ben Zarthi for helpful conversations. 1. Note that the paper does not aim at delivering a comprehensive study of the Tunisian labor market. An overall picture of the main trends in this market at different points in time is provided by Morrisson (1986) and Zouari-Bouattour (1994). Also, several papers have dealt with specific labor market issues in Tunisia, including the determinants of earnings at the individual level (Abdennadher, Karaa, and Plassard 1994), the links between poverty and employment status (World Bank 1995b), the nature of social security programs for workers (World Bank 1993), the importance of intra- household transfers (IREP 1994), and the links between wage increases and productivity gains (Azam 1995, World Bank 1995a). The interested reader can fruitfully refer to those studies. References The word "processed" describes informally reproduced works that may not be commonly available through library systems. 76 The World Bank Research Observer. vol. 13, no. I (February 1998) Abdennadher, Chokri, Adel Karaa, and Jean-Michel Plassard. 1994. "Tester la dualite du marche du travail: l'exemple de la Tunisie." Revue d'Economie du De'veloppement 2Uune):39-63, June. Azam, Jean-Paul. 1995. "Labor Market Institutions." CERDI, Centre d'Etudes et de Recherches sur le Developpement International, Clermont-Ferrand. Processed. Berman, Eli. 1997. "Help Wanted, Job Needed: Estimates of a Matching Function from Employ- ment Service Data." Journal of Labor Economics 15(1, part 2, January):S251-92. Blanchard, Olivier, and Peter Diamond. 1989. "The Beveridge Curve." Brookings Papers on Eco- nomic Activity 1: 1-60. Blanchard, Olivier, and Juan F. Jimeno. 1995. "Structural Unemployment: Spain versus Portugal." American Economic Review 85(2) (Papers and Proceedings): 212-18. Burda, Michael. 1993. "Modeling Exits from Unemployment in Eastern Germany: A Matching Function Approach." CEPR Discussion Paper Series 800. Centre for Economic Policy Research, London, U.K. Processed. Christl, Josef. 1992. The Unemployment/Vacancy Curve: Theoretical Foundation and Empirial Rel- evance. New York: Springer-Verlag. Coles, Melvyn, and Eric Smith. 1996. "Cross-Section Estimation of the Matching Function: Evi- dence from England and Wales." Economica 63(252):589-97. Edin, Per Anders, and Bertil Holmlund. 1991. "Unemployment, Vacancies, and Labour Market Programmes: Swedish Evidence." In Fiorella Padoa Schioppa, ed., Mismatch and Labour Mobil- ity. Cambridge, U.K.: Cambridge University Press. Franks, Jeffrey. 1994. "Explaining Unemployment in Spain: Structural Change, Cyclical Fluctua- tions, and Labor Market Rigidities." IMF Working Paper 94/102. International Monetary Fund, Washington, D.C. Processed. Harris, John, and Michael Todaro. 1970. "Migration, Unemployment, and Development: A Two- Sector Analysis." American Economic Review 60(1):126-43. Hashimoto, Mansanori. 1993. "Aspects of Labor Market Adjustment in Japan." Journal of Labor Economics 11(1):136-61. ILO (International Labour Office). 1996. Yearbook of Labour Statistics. Geneva. tREP (Institut de recherches et d'etudes sur la population). 1994. L'entraide familiale, cahier 11. Tunis: Ministere du developpement economique. Morrisson, Christian. 1986. "Evolution of the Tunisian Labor Market." DRD Discussion Paper 279. World Bank, Development Research Department. Washington, D.C. Processed. Pissarides, Christopher. 1985. "Short-Run Equilibrium Dynamics of Unemployment, Vacancies, and Real Wages." American Economic Review 75:676-90. Rama, Martin. 1997. "A Labor Market Cross-Country Database." World Bank, Development Re- search Department, Washington, D.C. Processed. World Bank. 1993. "Republic of Tunisia: The Social Protection System." Middle East and North Africa Department, Washington, D.C. Processed. . 1995a. "Toward the 21st Century." Country Economic Memorandum of Tunisia. Middle East and North Africa Department, Washington, D.C. Processed. . 1995b. "Poverty Alleviation: Preserving Progress while Preparing for the Future." Poverty Assessment of Tunisia. Middle East and North Africa Department, Washington, D.C. Processed. 1995c. World Development Report 1995: Workers in an Integrating World. New York: Ox- ford University Press. Zouari-Bouattour, Salma. 1994 "Evolution du marche du travail Tunisien, 1983-1993." Universite de Sfax, Tunisia. Processed. Martin Rama 77 Evaluating Retraining Programs in OECD Countries: Lessons Learned Amit Dar * Indermit S. Gill Are retraining programs for the unemployed more effective than job search assistance? Governments of the member countries of the Organisation for Economic Co-operation and Development have considerable experience with retraining programs in a variety of industrial settings. Evaluations of these programs show that the results are disappointing, however. This article discusses the factors associated with retraining programs for two types of workers: those laid off en masse and the long-term unemployed. Evaluations indi- cate poor results for both groups: retrainingprograms are generally no more effective than job search assistance in increasing either reemployment probabilities orpostintervention earnings, and they are between two and four times more expensive than job search assistance. Industrial countries spend sizable amounts on labor programs for the unemployed. In 1992 member countries of the Organisation for Economic Co-operation and Development (OECD) spent between 0.1 percent Japan) and 2.6 percent (Sweden) of gross domestic product on labor programs. In several countries this training is "the largest category of active programs, and is often perceived as the principal alter- native to regular unemployment benefits" (OECD 1994a). Countries in Eastern Europe-where the role of active labor programs is a topic of current debate-also spent between 0.2 and 3 percent of gross domestic product on these programs (OECD 1 994a). Yet possibly because such assistance is viewed almost as a fundamental right in Western Europe, these programs are rarely evaluated outside the United States. This article surveys evaluations of retraining programs in the OECD countries, high- lighting the shortcomings of such schemes and illustrating the payoff to investing in rigorous evaluations. We examine only the evaluations of retraining programs, so our focus is largely on adults with previous work experience rather than on unemployed school-leavers. We distinguish here between retraining schemes for workers displaced by plant closures The World Bank Research Observer, vol. 13, no. I (February 1998), pp. 79-101 C) 1998 The International Bank for Reconstruction and Development ! THE WORLD BANK 79 or restructuring and programs aimed at the long-term unemployed. These groups differ in important ways. First, the long-term unemployed are a relatively hetero- genous group of individuals compared with those laid off en masse from a single plant or firm. Second, although some programs are targeted at specific regions, the long-term unemployed are generally more dispersed geographically. Third, the du- ration of unemployment is, almost by definition, greater for the long-term unem- ployed. Finally, retraining programs for the long-term unemployed are generally a mix of classroom and in-plant training, while programs for displaced workers are usually confined to the classroom. Evaluation Techniques Techniques for evaluating the effectiveness of retraining programs can be broadly classified into two categories: scientific and nonscientific. Scientific evaluations can be further divided into experimental and quasi-experimental. Experimental, or clas- sically designed, evaluations require selecting "treatment" and "control" groups be- fore the intervention: the treatment group receives the assistance, and the control group does not. If large numbers of individuals are randomly assigned to each group, the average characteristics of the two groups should not differ significantly, so any difference in outcomes can be attributed to program participation. In quasi- experimental studies treatment and control groups are selected after the interven- tion. To compute the program's effects, statistical techniques are used to correct for differences in characteristics between the two groups. Nonscientific techniques do not use control groups to evaluate the effect of interventions but instead rely on statistics compiled by a program's administrators. These evaluations are of little use; without a control group, it is difficult to attribute the success or failure of the partici- pants to the intervention, because the changes in individuals' behavior might have resulted from other factors, such as worker-specific attributes or economywide changes (Grossman 1994). Classically Designed (Randomized) Experiments This technique, which was originally developed to test drug effectiveness, identifies and randomly assigns individuals to either a treatment or a control group before the intervention. Its main appeal lies in the simplicity of interpreting results-the effec- tiveness of the program is computed as the simple difference in the outcome between the participants in the treatment group and the nonparticipants in the control group (Newman, Rawlings, and Gertler 1994). The main pitfalls of this method are a failure to select individuals through random assignment, changes in behavior as a result of their assignment to either group (for instance, enrolling in private programs or intensi- 80 The World Bank Research Observer, vol. 13, no. I (February 1998) fying their job search), high costs because of the number of participants in the sample, and ethical questions about excluding a group of people from the intervention. Although randomization is thought to eliminate selection bias in determining who will participate, proponents of this methodology make an important assump- tion: that random assignment does not alter the behavior that is being studied. This may not be the case and, in fact, the bias may be quite strong (Heckman 1992). For example, ambitious individuals who would have taken a training course in any case will not apply to the program if they stand a chance of being in the nontreatment group. Such individuals, who might have enrolled in a nonrandomized regime, may make plans anticipating enrollment in a training program. With randomization they may alter their decision to apply or undertake activities complementary to training. Thus risk-averse persons will tend to be eliminated from the program. Quasi-Experimental Techniques In quasi-experimental methods the treatment and control groups are selected after the intervention. Econometric techniques are used to correct for the differences in characteristics between the two groups. The main appeal of this method lies in its relatively low cost and the ability to undertake the evaluation at any time. The main drawback is that these techniques-if done properly-are statistically complex. The techniques used to adjust for differences in observable attributes (for example, sex, age, region of residence, education) are relatively straightforward but subject to speci- fication errors; correcting for unobservable characteristics (for example, motivation, family connections) requires a convoluted procedure that can yield wildly different results. Quasi-experimental evaluations fall into three types: regression adjusted for observable characteristics; regression adjusted for observable and unobservable char- acteristics (selectivity corrected); and matched pairs. REGRESSION ADJUSTED FOR OBSERVED CHARACTERISTICS. This technique is used to assess the impact of participation in a program when the observable characteristics of the participants and the comparison group are different. This method is appropriate for estimating the effect of a program when the difference between participants and nonparticipants can be explained by these observable characteristics. For example, if better-educated workers are more successful in finding work regardless of whether or not they had special training, then controlling for the effect of education (using regression techniques) will provide more reliable estimates than would a simple com- parison of the reemployment probabilities of the control and treatment groups. REGRESSION ADJUSTED FOR OBSERVED AND UNOBSERVED VARIABLES (SELECTIVITY CORRECTED). When selection into programs is not random, and participation in a program is due to both observable and unobservable characteristics, the above tech- Amit Dar and Indermit S. Gill 81 nique, which corrects for observed characteristics, is likely to be biased. Even if par- ticipants and nonparticipants have similar observable characteristics, unobservable characteristics (such as innate ability) would lead to incorrect inferences about non- participants. This technique uses a method developed by Heckman (1979) called "sample selectivity correction" to try to control for these unobservables. MATCHED PAIRS. Because the observed characteristics of individuals in the control and treatment groups are bound to be different to some degree, these groups are likely to have different success rates in finding employment even in the absence of active labor market programs. To control for these differences, synthetic control groups are constructed using a matched pairs approach. The synthetic control group, which is a subset of the entire control group, is composed of individuals whose ob- servable characteristics most closely match those of the treatment group. Relative Strengths of Techniques Estimating the effect of an employment program on the earnings of trainees, using randomized and quasi-experimental techniques, LaLonde (1986) found that ran- domized experiments yielded results significantly different from those that relied on quasi-experimental techniques. Policymakers should be aware that available nonexperimental evaluations of training programs may contain large biases. While randomized experimentation is theoretically the best technique to estimate the ef- fects of interventions, quasi-experimental techniques may be superior in practice. The main weakness of randomized experiments is their inability to ensure that individuals in the control group do not alter their behavior in a way that contami- nates the experiment. For example, individuals denied public job training might enroll in private programs, which would bias the results of any evaluation of public programs.' It may also be difficult to ensure that assignment is truly random. For example, applicants may be selected into the program because of nepotism, or pro- gram administrators may deliberately exclude high-risk applicants to achieve results that reflect well on the program. A third problem concerns ethical questions about treating individuals as subjects in an experiment. Finally, experimental evaluations are possible only for future programs, because the control and treatment groups have to be selected before the program is initiated. Using the dual criteria of rigor and feasibility then, randomized experiments are not necessarily superior to quasi-experimental techniques. Because the decision to evaluate labor market programs often occurs after the programs are in place and because the costs of setting up the experiments are high, randomized evaluations should perhaps be the last alternative. Within quasi-experimental techniques, selec- tivity correction may not add much, especially when information is available for a considerable number of observable individual and labor-market characteristics (such 82 The World Bank Research Observer, vol. 13, no. I (February 1998) as education, age, sex, household wealth, and region of residence). In addition to being cumbersome and somewhat counterintuitive, this method often gives arbi- trary results depending on the selectivity-correction specification that is used. This leaves the matched pairs and regression-adjusted techniques. Of the two, the matched pairs technique is preferred because the procedure is less arbitrary. Because the observed differences between the treatment and comparison group are mini- mized, the exact specification of the model becomes less important. And because the program measures the simple difference in the variables that policymakers want an- swered (reemployment probabilities and wages) between the control and treatment groups, the results are easier for nonstatisticians to interpret. One weakness shared by both the scientific and nonscientific evaluations is that they do not take into account the displacement that may result from the retraining program. For example, in countries where demand for labor is constrained, retrain- ees may "bump," or displace, employed workers, so aggregate unemployment may not change despite the size of the program. In general, displacement implies that the social benefits from reemployment attributable to the retraining program are lower than indicated by the evaluation, however well done. The Importance of Costs For the purposes of informing policy decisions, an evaluation is not complete until the costs of both the retraining program and its alternatives are considered. For ex- ample, if retraining is twice as costly as job search assistance but no more effective in finding people jobs and increasing their wages, job search assistance is twice as cost- effective. At least at the margin, such a finding would constitute a case for reallocat- ing resources from retraining to job search programs. Unfortunately, costs appear to be the least analyzed aspect of these programs in OECD countries. Even the most careful evaluations of retraining programs cannot be used for social cost-benefit analysis because of the displacement effects discussed above. But when done correctly, evaluations are good guides for cost-benefit analysis of private train- ing programs, which policymakers can use to institute cost recovery in public pro- grams and to promote private provision. Evaluations may also help in deciding whether retraining programs reduce budgetary expenditures by moving people off unem- ployment benefits into productive employment or whether the programs are a drain on the budget despite being effective. Evaluating Retraining in OECD Countries Retraining programs in OECD countries have been designed primarily to assist three categories of workers: those laid off en masse; those who have lost their jobs because Amit Dar and Indernmit S. Gill 83 of plant closures; and those who have been unemployed over the long term. This review of the evaluations of eleven retraining programs classifies the specific situa- tion that the program was designed to address, describes the policy intervention that was selected and the type of evaluation used, and reports the main results. Because the success of retraining programs depends on aggregate or regional labor market conditions, such as unemployment rates and the state of the leading industry, these indicators are reported as well. We examine the type of retraining provided, whether in classrooms or on-the-job, and whether it was accompanied by-or in lieu of-other measures, such as job search assistance. The evaluations are classified as experimental, quasi-experimental, and nonscientific, and we do not consider inferences drawn from the nonscientific evaluations to be reliable. We then look at the effects of the program on reemploy- ment and wages, both for subgroups of retrainees and type of intervention. The costs of the program are included when they are reported. Training Programs Instituted as a Result of Mass Layoffs and Plant Closures The results of eleven programs (three in the United States, four in Sweden, and one each in Australia, Canada, Denmark, and France) examining the effectiveness of retraining programs for workers displaced through mass layoffs and plant closures have been reviewed (Table A.1). Five of the evaluations were nonscientific and five were quasi-experimental. One study relied on more than one technique to evaluate the impact of the program; no study used experimental evaluation techniques. The retraining programs were undertaken to assist workers in the steel, pulp, min- ing, shipbuilding, and automotive industries. The number of workers who lost their jobs varied from about 500 to 3,000 per plant. The rationale underlying the pro- grams was apparently to assist the affected workers in any way possible. Generally, these programs were instituted during periods of high or rising aggregate unemploy- ment or during a contraction in certain manufacturing industries. For example, the evaluations in the United States and Canada covered primarily the auto and steel industries, which were battered by competition from Japan. Between 1978 and 1980, auto production in the two countries declined 25 percent, precipitating layoffs and plant closures in the early 1980s. In Europe and Australia the retraining programs seem to have been instituted during periods of high or rising rates of unemployment. Most of the retraining programs were classroom based, and accompanied by job search assistance. With only one exception, on-the-job training was not provided or facilitated. In France retraining was accompanied by financial incentives to regional firms that hired trainees, so the full costs of the program were likely high (OECD 1993a). 84 The World Bank Research Observer, vol. 13, no. I (February 1998) Quasi-experimental and nonscientific techniques were used in evaluating all of these retraining programs. None of the studies was longitudinal, so they provide at best a snapshot of the labor market benefits of the program. The longer-term ben- efits of retraining were not evaluated even by the scientific evaluations. Although nonscientific evaluations indicate that these programs are very effective in placing high numbers of male workers in wage-earning jobs or in self-employment and women in business start-ups, more reliable quasi-experimental evaluations indicate that re- training programs are generally no more effective than job search assistance in in- creasing either reemployment probabilities or postintervention earnings. Some re- training programs resulted in modest gains in reemployment probabilities, but the wage changes were sometimes negative. Interestingly, evaluations of three retraining programs for U.S. auto workers showed the contrast between scientific and non- scientific techniques: in San Jose, California, a nonscientific evaluation indicated high placement rates, while in Buffalo, New York, and Michigan-during the same period-scientific evaluations showed that these programs were ineffective (OECD 1993a; Corson, Long, and Maynard 1985; Leigh 1992). The costs, when they are known, varied between $3,500 and $25,000 a person. Evaluations seldom report the full costs of retraining or job search programs, how- ever, so determining cost-effectiveness is difficult. Retraining programs appear to be between two and four times more expensive than job search programs: for example, in Buffalo, job search services cost $850 a participant, while retraining cost $3,300 (Corson, Long, and Maynard 1985). If, as the findings indicate, both programs have roughly the same success, job search assistance may be more cost-effective than re- training in assisting displaced workers get jobs. Training Programs for the Long-Term Unemployed There is no reason to assume that the impact of retraining on the long-term unem- ployed is the same as it is on workers laid off en masse. The results of studies exam- ining the effectiveness of retraining programs for the long-term unemployed are shown in Table A.2. Of the eleven evaluations (four in the United States, three in Germany, two in the Netherlands, and one each in Canada and Britain), four were nonscien- tific, four were quasi-experimental, and three were experimental. The clientele of retraining programs for the long-term unemployed is relatively heterogeneous. Because these individuals are displaced from various sectors and some have never worked, they are likely to be more varied in age, skills, and education than laid-off workers. New Jersey's retraining program in 1986-87 included work- ers whose previous jobs were in manufacturing, trade, and services (Anderson, Corson, and Decker 1991); many were more than 55 years old. In contrast, half the workers in Germany's retraining program for the long-term unemployed were less than 35 years old (Johanson 1994). Amit Dar anad indermsit S. Gill 8 5 These programs are generally instituted during improving conditions in industry or in aggregate employment. Interventions at this stage, if appropriately designed, are thought to enable the long-term unemployed to obtain some of the jobs that are being created. The programs are more comprehensive than are those for people laid off en masse and generally provide a mix of classroom or workshop training, on-the- job training, and job search assistance. Some experimental evaluations were conducted in the United States, and quasi- experimental techniques were used in both the United States and Europe. Some of these studies were longitudinal, providing an indication of the medium-term labor market benefits of the retraining program (see Johanson 1994 and OECD 1993b for Europe; and Leigh 1992, and Anderson, Corson, and Decker 1991 for the United States). But many of the evaluations in Europe were nonscientific. Again, the results of nonscientific evaluations, which were encouraging, were not confirmed by evaluations based on scientific techniques. The effectiveness of pro- grams for the long-term unemployed, while not high, was better than that of pro- grams for those laid off en masse. A few programs did result, or were thought to have resulted, in gains in either reemployment probabilities or wages; some evaluations also indicated that these programs were more effective in helping women. But where participants did record gains in employment, longitudinal studies generally found that the effects of retraining dissipated within a couple of years. In this group, too, retraining programs were generally no more effective than job search assistance in increasing either reemployment probabilities or postintervention earnings. For ex- ample, evaluations of the Texas Worker Adjustment Demonstration program indi- cated that participants were likely to be employed more rapidly than nonpartici- pants, but that over time, the employment opportunities of male participants were no better than those of nonparticipants or of those who only had job search assis- tance (Bloom 1990). The costs, when known, varied between $900 and $12,000 a person, about twice as much as job search services, but the lack of data makes it difficult to determine the absolute cost-effectiveness of these programs. Still, job search assistance programs appear to be somewhat more cost-effective than retrain- ing programs in finding jobs for the long-term unemployed. Evaluation Results for Hungary Rising unemployment and falling real wages are a vivid and costly aspect of countries in transition from controlled to market economies. Long-run unemployment is par- ticularly pernicious, and many countries have mounted active labor programs (for instance, public retraining programs and public service employment) to deal with this problem. With the exception of Hungary, where the World Bank has sponsored rigorous evaluations and collections of data on costs, little reliable information exists 86 The World Bank Research Observer, vol. 13, no. I (February 1998) on their effectiveness. As an early reformer, Hungary provides valuable lessons for other transition economies. Since 1989 sharp declines in the country's gross domes- tic product have been accompanied by rising unemployment and falling real wages. The unemployment rate appears to have stabilized since 1993, but employment continues to decline, reflecting a continuing withdrawal of workers from the formal labor market. Since 1990 the government has offered workers retraining programs financed by a national employment fund. Trainees, who are either currently unem- ployed or are working but expect to become unemployed, include participants in public works programs. Instruction is largely classroom based. Evaluations using different quasi-experimental techniques-matched pairs, regres- sion adjusted, and selectivity corrected-yielded different results (O'Leary 1995). Before adjusting for differences in the characteristics of program participants and control groups, estimates of effectiveness indicated that retraining significantly raised the probability of reemployment. But when the observable characteristics of treat- ment and control groups were taken into account, retraining (and training, because about 40 percent of trainees had not worked previously) was only marginally suc- cessful at best, increasing the probability of finding employment by 6 percent. Fur- ther controls for unobservable attributes led to ambiguous results. Retraining was not at all successful in raising earnings. Preliminary analysis indicates that retraining was a substitute for attributes that lead to higher reemployment probabilities in the absence of any intervention, such as being younger, better educated, and from more dynamic regions. That is, the program's value-added (in terms of improving labor market outcomes) is greater for relatively disadvantaged job-seekers. This finding may be country specific, and other countries or regions should determine whether it applies to them before implement- ing similar large-scale programs. For the programs evaluated in Hungary, focusing more on job-seekers who lack these attributes would appear to serve both equity and efficiency objectives better than simply ensuring support for programs whose re- trainees have high probabilities of reemployment and gains in earnings. The analysis implies that public retraining programs should target older men from backward regions.2 This finding also highlights the usefulness of rigorous impact evaluations, which net out the effects of such attributes in determining whether a program is effective. And it underscores the need to agree upon a reliable, feasible, and easily interpreted technique to evaluate the efficiency and equity effects of labor programs. Because of its analytical rigor and feasibility, we argue that the preferred evaluation technique is matched pairs analysis, where trainees are compared with a subset of the control group whose characteristics most resemble their own. Private cost-benefit analysis of the costs of retraining (an average of approximately $900 per trainee in 1994, according to Pulay 1995) and the level of gains in reem- ployment reveal that based on reasonable assumptions about the durability of Amit Dar and Indermit S. Gill 87 Table 1. Summary of OECD Retraining Programs and Evaluations Displaced Long-term Item workers unemployed Previous sector of employment Manufacturing Heterogeneous Labor market conditions Generally deteriorating Generally improving Training venue Mainly classroom Classroom and OJT Other measures provided JSA JSA Type of evaluation Nonscientific 6 4 Quasi-experimental 5 4 Experimental 0 3 (all U.S.) Effectiveness Nonscientific techniques Positive Generally positive Scientific techniques Negative Generally negative; some groups benefit Relative to JSA No more effective No more effective Costs Retraining Data are not available Data are not available Relative to JSA At least twice as costly At least twice as costly Note: JSA, job search assistance; OJT, on-the-job training. program effects and the amount and duration of unemployment benefits, it would take more than 30 years to recover costs of the program (Gill and Dar 1995). It seems difficult to justify retraining programs based on economic considerations alone, even in a country such as Hungary where the government provides benefits for un- employed workers. Summary and Conclusion The paucity of rigorous evidence on the costs and effectiveness of retraining pro- grams does not permit a definitive conclusion on whether such interventions can be justified economically (Table 1). The scattered evidence does not appear to justify the indiscriminate expansion of retraining programs to cover more of the unem- ployed. These conclusions are consistent with the findings in the OECD Employment Outlook (1 993b), which concluded that "for the broadly targeted sub-group of pro- grams, the overall impression is most troubling. Available evidence does not permit strong conclusions, but it gives remarkably meager support of a hypothesis that such programs are effective." From this discussion it appears that transition and other restructuring economies can draw the following lessons from experience in the OECD countries: 88 The World Bank Research Observer, vol. 13, no. I (February 1998) First, sound techniques should be used to evaluate retraining programs (and other public interventions). Although the nonscientific evaluations of retraining programs present a rosy picture based on placement rates and other informal evidence, scien- tific evaluations are quite discouraging. Relying on nonscientific evaluations may lead to incorrect policy conclusions. Second, rigorous evaluations, although not necessarily allowing a complete social cost-benefit analysis, can be useful for policymakers in allocating public expenditure on labor programs. Reviews of evaluations find, for example, that job search assis- tance measures-which cost less than retraining but appear equally effective-may be a more cost-effective device in assisting displaced workers. Third, OECD experience of retraining programs for workers displaced en masse may be useful in designing assistance programs in transition countries such as the former Soviet Union and liberalizing economies that expect labor shedding in the manufacturing and mining sectors. Principally, these economies should recognize that retraining should not be the main form of assistance. Fourth, transition and developing economies that are beginning to experience long-term unemployment can learn from OECD experience, which indicates that re- training programs for the long-term unemployed are more beneficial for some groups than for others within this relatively heterogeneous group of job-seekers. Which group will benefit most from retraining is difficult to predict, however. Principally, these results call for using modest untargeted pilot programs, evaluating them rigor- ously, and then tightly targeting public retraining programs to those for whom they are found most cost-effective. Notes Amit Dar is in the Social Protection Group of the World Bank's Human Development Network, and Indermit S. Gill is senior economist in the World Bank's Country Management Unit in Brasilia. The authors are grateful to Jane Armitage, Kathie Krumm, Ana Revenga, Michal Rutkowski, and anonymous referees for helpful suggestions. 1. Heckman (1992) documents other limitations of this technique when applied to social experi- ments; these limitations derive from selectivity biases. 2. These findings contrast with those of Revenga, Riboud, and Tan (1994) for retraining pro- grams in Mexico. Using quasi-experimental techniques, they find that effectiveness can be improved if programs better target relatively educated job-seekers of both sexes with previous work experience. They also find generally more encouraging results than those for OECD countries and Hungary. Amit Dar and Indermit S. Gill 89 O Table A. 1. Overview of Studies Evaluating Training Programs for Workers Displaced by Plant Closures and Mass Layoffi Labor market problem Relevant indicators Intervention design Type of evaluation Result Comments Ford Motor Co. plant A 25 percent decline in Basic skills training as Nonscientific High rate of success in Perceived success was closure in San Jose, auto production be- well as targeted voca- placing workers in due to adequate re- Calif. (1982). Some tween 1978 and 1980. tional training in mar- jobs. source base ($6,000 2,400 workers affected Unemployment rates ketable skills. grant per worker) (OECD 1993a). rose from 7.5 percent and a high degree in 1981 to 9.5 percent of coordination in 1982 and 1983, and provided by Ford and manufacturing employ- by government. ment declined by 5 percent during 1981- 83. Automotive plant A 25 percent decline in Both JSA and classroom Quasi- Classroom training did Earnings estimates vary closures in Michigan auto production be- training were provided experimental not significantly im- (ranging from negative (1980-83). More than tween 1978 and 1980. promptly after plant prove the participants' to significantly posi- 3,000 workers laid off, High unemployment closures. Retraining was reemployment rate. tive). Training does mostly experienced, rates in 1981-82 na- offered in occupations Trainees did no better not seem to have been highly paid blue-collar tionally (9.5 percent), in which there was than those receiving very effective, espe- male employees (Leigh which fell to 7.5 per- growing demand, JSA. cially in light of the 1992; 1994). cent by 1984. Manu- including blue-collar fact that training cost facturing employment trades. twice as much as JSA. rose by 5 percent be- tween 1983 and 1984. Auto and steel plant High unemployment Displaced workers Quasi- JSA services had a fairly Classroom training and closures in Buffalo, rates in 1981-82 na- were provided with experimental large impact on earn- OJT were ineffective. N.Y. (1982-83). Mass tionally (9.5 percent), both JSA and either ings for the first six These programs cost layoffs of highly paid, which fell to 7.5 per- classroom training or post-program months. about four times as experienced blue-collar cent by 1984. OJT. Program services However, there is no much as JSA, implying male workers (Corson, Manufacturing were provided after a evidence that either that JSA is potentially Long, and Maynard employment rose by 5 fairly lengthy period of classroom training or the only cost-effective 1985). percent between 1983 unemployment. OJT had any incremen- program. No evidence and 1984. tal effect. was provided on the participants' employ- ability. Plant closures at 13 Unemployment rates JSA and training pro- Quasi- Participants had a 7 Training seems to have steel factories and rose between 1980 and grams were provided. experimental percent higher rate of had a greater impact mines in Canada in the 1983. Employment in employment than than JSA. No informa- 1980s (Leigh 1992). nonagricultural activi- comparably displaced tion was available on ties fell slightly during workers who did not the cost of training. Job the period. participate in the pro- counseling has little gram. This impact was impact in Canada, attributed to training in where such services are which 28 percent of routinely provided to workers participated. the unemployed by the But at two mining sites, Public Employment the program had no Service. impact. Steel and coal plant Steel sector contraction Workers received 70 Nonscientific High success rate in No evidence that re- closures in Creusot- resulted in the loss of percent of their former finding jobs for partici- training programs Loire, France (1984). 6,000 jobs during salaries for 10 months pants. produced long-term 1,230 people affected 1984. Unemployment while enrolled in re- employment benefits. (OECD 1993a), rates rose from 8.1 training and job-search percent in 1982 to activities. Trainees (in 10.2 percent by 1985. engineering, plastic In mining employment molding, refrigeration, fell by almost 40 per- etc.) were promised cent between 1980 and reemployment in the 1990. Manufacturing region. Career counsel- employment fell about ing and short courses in e 6 percent from 1983 to work skills, production 1985. methods for small firms, (Table continues on thefollowingpages.) Table A.1. (continued) b Labor mnarket problem Relevant indicators Intervention dlesign Type of evaluation Result Comments and job search tech- niques were also pro- vided. Firms were given financial incentives to hire these workers. Pulp plant closure in Unemployment rates Participants were pro- Quasi- Participants received No information pro- Kramfors, Sweden in Sweden rose from vided with classroom experimental lower weekly wages vided on reemploy- (1977) (OECD 1991). 1.6 percent in 1976 to training. than those who did not ment rates or costs. 2.2 percent by 1978. receive training. Drop Benefits from retrain- Employment in manu- in earnings was espe- ing program were facturing declined by 8 cially significant in the insignificant. percent between 1977 first year; there were and 1979. no appreciable long- term gains. About 2,000 workers Volvo Co. set up a A significant number of Nonscientific BY November 1987, Two factors accounted laid off at the Udde- manufacturing plant in workers joined retrain- more than 90 percent for the success of the valla shipyard in Swe- Uddevalla. Unemploy- ing programs several of the workers who training program: den (1985) (Alfthan ment rates were declin- months before they had completed training economic and labor and Janzon 1994). ing in this period while were laid off. Courses had found jobs or market conditions manufacturing em- of varying duration become self-employed were buoyant in the ployment was fairly were offered in weld- -most of them in region, and manage- steady. ing, engineering, and occupations for which ment, employment control engineering. they had trained. offices, and training These retraining pro- agencies worked in grams were provided by close cooperation. the state-owned train- ing board, municipal education institutions, and other adult educa- tion services. Volvo Co. plant at Volvo planned to A retraining program Nonscientific Program has not yet Program costs are G6teborg, Sweden, recruit 800 workers to was designed to help the been evaluated. expected to be about announced that 1,000 manufacture the new existing work force $25 million ($25,000 workers would be laid line of cars. Unem- manage the change per person), close to off in 1992 as it phased ployment rates were without job losses. The half of which will be off in 1992 as it phased ~~~~~~~~~~~~~~~~~~~~paid for by the govern- out an old model and rising sharply-from company accepted the paid In by the tooled up for a new 3.2 percent in 1991 to proposal under the ment. In judging the one (Alfthan and 5.9 percent in 1992. stipulation that the cost cost-effectiveness of Janzon 1994). Manufacturing em- be shared by the govern- this program, it should ployment dropped by ment. The program was be noted that the un- 9 percent in 1992. a broad competency- employment benefits Total employment fell raising exercise, with the government would by 4 percent. specific training to have had to pay prepare the participants woul million. for the production of $6.5 million. the new automobiles. Shipyard closure in High unemployment This program (1986- Nonscientific Fifty-one businesses The program did not Storstrom county, rates in region, espe- 89) trained women were set up in 1989. seem to be successful Denmark (1986). cially among women. entrepreneurs to help Less than a third of the and was apparently 2,000 people lost their National unemploy- them start their own participants opened a quite costly (reliable jobs (OEcD 1993a). ment rates for men in businesses. A total of full-time business, and cost estimates are not 1986 and 1987 were 200 hours of specific few hired any employ- available). 6.1 and 6.4 percent, business-oriented ees, thus the total respectively, while the training was provided. additional employment corresponding num- generated was low. bers for women were 10 and 9.6 percent. Employment in manu- facturing was stagnant from the mid-1980s. 4 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~(Table continues on the following page.) 1 Table A.1. (continued) Labor market problem Relevant indicators Intervention design Type of evaluation Result Comments Sweden, 1980s and Unemployment rose Various types Various types Retraining programs The cost-effectiveness 1990s: general evalua- steadily over the have become less effec- of these programs tion of public retrain- period. tive over time-espe- declined both because ing programs (OECD cially since the they were less success- 1991). economy has begun ful and because the deteriorating. Partici- costs increased. pants have had more trouble finding jobs than other unem- ployed workers. Automobile plant Unemployment rates Labor Adjustment Quasi- Over a nine-month There were no data closure in Australia in rose steadily in Austra- Training Arrangement experimental period, driver training showing the cost- 1984. 445 workers laid lia, reaching a high of offered classroom increased the probabil- effectiveness of differ- off (Leigh 1992). 9 percent in 1984 training (average ity of reemployment. ent types of training before falling to 8 length 19 weeks). However, fewer train- courses. Self-selection percent for the next Main distinction be- ees of other training was a problem because few years. Between tween courses was courses found jobs. individuals who chose 1980 and 1984 em- whether they provided not to participate were ployment in manu- driver training or not. in the control group. facturing fell by 4 The impact of longer percent. training was negative. Note: JSA, job search assistance; OJT, on-the-job training. Table A.2. Overview of Studies Evaluating Training Programs for the Long- Term Unemployed Labor market problem Relevant indicators Intervention design Type of evaluation Result Comments Training for workers A decline in petro- Texas WAD project. Experimental One year after the Despite the high costs eligible for JPTA Title chemical industry led Displaced workers were program's completion, of classroom training III program (displaced to layoff. U.S. unem- provided with JSA or a participants registered (twice as much as JSA), worker) in Houston, ployment rates de- mix of JsA with class- no additional earning no additional gains Texas (1983-85). clined from 9.5 per- room training. gain when compared accrued from this type Eligible individuals cent in 1983 to 7.4 with the sample of of training. were unemployed and percent in 1985 (and workers who had been had a low probability of maintained this trend given only JSA. returning to their previ- until the late 1980s). ous occupation or Employment in this industry (Bloom 1990). industry increased slightly between 1983 and 1985. Training for workers U.S. unemployment Texas WAD project. Experimental One year after the The increase in earn- eligible for JPTA Title rates declined from 9.5 Displaced workers program's completion, ings of female partici- III program (displaced percent in 1983 to 7.4 provided with JSA alone this program recorded pants was slightly more worker) in El Paso, percent in 1985 (and or combined with no effect on male than the program's Texas (1983-85). maintained this trend classroom training. employees' earnings costs. Male employees (Criteria same as until the late 1980s). but did increase the reported no beneficial above.) Workers laid Employment in manu- earnings of women. effects. off from light manu- facturing fell by close facturing plants to 2 percent between (Bloom 1990). 1982 and 1983 but rose to about 5 percent in 1985. (Table continues on the followingpages.) U, Table A.2. (continued) Labor market problem Relevant indicators Intervention design Type of evaluation Result Comments Assistance aimed at the Unemployment rates New Jersey UI Reem- Experimental More than 10 quarters While the results of long-term unemployed in the U.S. fell from ployment Demonstra- after the program training were positive, in New Jersey (1986- 7.1 percent in 1985 to tion project. This ended, both CT and the following caveats 87). In general, work- 6.1 percent by 1987. program provided OJT significantly in- apply: the trainees were ers were laid off from While employment in displaced workers with creased earnings over self-selected (thus these manufacturing, trade, manufacturing re- JSA alone, or followed those who received JSA results may not apply and services (Corson mained fairly steady by OJT, classroom only. These individuals for a random group of and others 1989). between 1985 and training (CT), or a were also employed for claimants); only 15 1988, trade employ- reemployment bonus. longer periods of time percent of those of- ment increased 7 per- in each quarter than fered training accepted cent and service-sector the JsA-only group. it; and training ben- employment rose efited those who al- about 5 percent. ready had marketable skills. Finally, analysis showed that the costs exceeded the expected benefits. Long-term unem- Unemployment rates The Trade Adjustment Quasi- Individuals who re- The analysis was lim- ployed workers in the in the U.S. fell from Assistance Program experimental ceived training began ited to the manufactur- U.S. (1988), primarily 7.1 percent in 1985 to was intended to de- earning significantly ing sector. Long-term resulting from plant 6.1 percent by 1987. velop workers' skills in more than those who investments in training closures (Leigh 1992). While employment in new occupations. Most received extended may be effective in manufacturing re- of these skills were income-maintenance increasing earnings, mained fairly steady supplied by a voca- benefits by the 6th but the training is between 1985 and tional college or local quarter and continued costly (each trainee was 1988, trade employ- community college in to earn more until the given a $12,000 train- ment increased 7 per- courses that were more 12th (last) quarter, ing voucher). cent and that in ser- than a year long. reaching a level of vices about 5 percent. $500 a quarter. Long-term unem- Compared with the The Job Development Quasi- Employability of In view of the high ployed workers in U.S., unemployment Program provided experimental women rose while that program costs (around Canada (Goss and rates have been fairly classroom training and of men declined. $9,300 per partici- associates 1989). high in Canada. How- OJT. Wages and direct Weekly earnings of pant), this training is ever they dedined from costs of classroom women were insignifi- not cost-effective, 11.2 percent in the mid- training were subsi- cant relative to that of especially for men. 1980s to 7.5 percent in dized. The wage sub- the control group, 1989. Spurred by sidy helped employers while they were lower growth in commerce cover the cost of OJT. for males. and services, employ- ment grew about 3 percent annually over the time period. High regional unem- In 23 of the 142 re- Among other interven- Nonscientific Training reduced In light of the extremely ployment levels in gions in Germany, tions, training and unemployment some- high cost (around $500 Germany in the late unemployment levels retraining of unem- what, but it is esti- million for training and 1970s (OECD 1991). were above 6 percent. ployed in firms. The mated that more than other interventions), the Some firms were also firms who employed 40 percent of these results were very disap- facing serious problems these trainees received hard-to-place individu- pointing. No informa- in adjusting to eco- a subsidy of 90 percent als had already left tion was provided on nomic changes. of the employees' their jobs by 1981. wages. wages for two years. Individuals at risk of Unemployment rates This is an evaluation of Quasi- No type of training No information on being unemployed in were steady between four programs, two of experimental had any significant wages or costs of 1987-88 in Germany 1986 and 1988. They which trained workers impact on the flows training. Uohanson 1994). fell somewhat by 1990. for new occupations out of short- or long- Employment grew at (one offered further term unemployment, slightly more than 1 training for employed or on the flows into percent during this and unemployed indi- unemployment. period. viduals, while the other e offered retraining for the unemployed). (Table continues on the following pages.) Table A.2. (continued) Labor market pro6bem Relevant indicators Intervention design Type of evaluation Result Comments Individuals who were Unemployment rates Training prepared Nonscientific The overall success No data on cost were either at risk of becom- remained fairly steady workers for certifica- rate, measured in terms available. The number ing unemployed or at just over 10 percent tion in one of the 375 of retention rates, was of workers who left were already unem- in the early 1980s. apprenticeable trades. 54 percent. Employ- their jobs was fairly ployed in Germany in The program required ability varied by age- high: two years after the early 1980s up to two years to less than half of those completing training Uohanson 1994). complete. Contents over 45 years old who only 60 percent of the and specification of were unemployed men and 66 percent of retraining correspond more than one year the women were still to those of initial voca- found jobs, while 86 employed. tional training. percent of 25- to 35- year-olds did. Help for long-term High unemployment This community-based Nonscientific Little impact on The cost-effectiveness unemployed and dis- rates (around 10-11 program provided flows from long-term of these programs is placed workers in percent) persisted vocational training by unemployment. likely to be negative. England in the 1980s through much of the local authorities in (Addison and Siebert 1980s. Between 1983 conjunction with local 1994). and 1990 employment colleges, central gov- rose by around 1.5 ernment, and volun- percent. tary organizations. Little private sector involvement. Improve skills and Unemployment rates Training provided Nonscientific In 1991, of the 82 Training was quite provide better access to fell from about 13 hands-on experience individuals enrolled in expensive: annual jobs for the long-term percent in the early (through a simulated the course, 52 com- funding of the pro- unemployed in 1980s to below 10 workshop). The empha- pleted it and 41 found gram was about $7.5 Tilburg, the Nether- percent by the end of sis was on technical jobs or went on for million, or $10,000 lands (OECD 1 993a). the decade. skills (metalworking, further education. The per trainee. woodworking, installa- most successful training tion techniques, and was in metalworking. apparel trades) as well as on good work habits. Courses ran from 4 to 10 months. Job training programs Unemployment rates Centers for adult voca- Quasi- Unemployed people No data were available for unemployed and fell from about 13 tional training. experimental who were not in the on wages and costs, displaced workers in percent in the early program found jobs as but training seems to the Netherlands (OECD 1980s to below 10 quickly as those who have been ineffective in 1 994b). percent by the end of were. Even two years terms of increasing the decade. later, the employment the employability of situation for the two participants. groups was not signifi- cantly different. Note: JSA, job search assistance; OJT, on-the-job training. References The word "processed" describes informally reproduced works that may not be commonly available through library systems. Addison, J. T., and W. S. Siebert. 1994. "Vocational Training and the European Community." Oxford Economic Papers 46(4):696-724. Alfthan, Torkel, and B. Janzon. 1994. RetrainingAdult Workers in Sweden, Training Policy Study 3. Geneva: International Labour Office. Anderson, Patricia, Walter Corson, and Paul Decker. 1991. "The New Jersey Unemployment In- surance Reemployment Demonstration Project: Follow up Report." U.S. Department of Labor Unemployment Insurance Survey. U.S. Department of Labor, Washington, D.C., Processed. Bloom, Howard. 1990. "Back to Work: Testing Reemployment Services for Displaced Workers." W. E. Upjohn Institute for Employment Research, Kalamazoo, Mich. Processed. Corson, Walter, Stephen Long, and Rebecca Maynard. 1985. "An Impact Evaluation of the Buffalo Dislocated Worker Demonstration Program." Mathematica Policy Research. New Jersey. Pro- cessed. Cotson, Walter, S. Dunstan, Paul Decker, and A. Gordon. 1989. "New Jersey Unemployment Insurance Reemployment Demonstration Project." Unemployment Insurance Occasional Paper 89-3. U.S. Department of Labor, Washington, D.C. Processed. Gill, Indermit S., and Amit Dar. 1995. "Costs and Effectiveness of Retraining in Hungary." Inter- nal Discussion Paper IDP-1 55. World Bank, Europe and Central Asia Region, Washington, D.C. Processed. Goss, Gilroy, and Associates. 1989. "Evaluation of the Job Development Program: Final Report." Report prepared for the Program Evaluation Branch, Canadian Employment and Immigration Office, Ottawa. Processed. Grossman, jean B. 1994. "Evaluating Social Policies: Principles and U.S. Experience." The World Bank Research Observer 9(2):159-80. Heckman, JamesJ. 1979. "Sample Selection Bias as a Specification Error." Econometrica47(1):153- 62. . 1992. "Randomization and Social Policy Evaluation." In Charles Manski and Irwin Garfinkel, eds, Evaluating Welfare and Training Programs. Cambridge, Mass.: Harvard Univer- sity Press. Johanson, Richard. 1994. RetrainingAdults in Germany. Training Policy Study 4. Geneva: Interna- tional Labour Office. LaLonde, Robert. 1986. "Evaluating the Econometric Evaluations of Training Programs with Ex- perimental Data." American Econometric Review 76(4):604-20. Leigh, Duane. 1992. "Retraining Displaced Workers: What Can Developing Nations Learn from OECD Nations?" Policy Research Working Paper 946. World Bank, Population and Human Re- sources Department, Washington, D.C. Processed. . 1994. RetrainingDisplaced Workers: The U.S. Experience. Training Policy Study 1. Geneva: International Labour Office. Newman, John, Laura Rawlings, and Paul Gertler. 1994. "Using Randomized Control Designs in Evaluating Social Sector Programs in Developing Countries." The World Bank Research Observer 9(2):181-201. OECD (Organisation for Economic Co-operation and Development). 1991. Evaluating LaborMar- ket and Social Programs: The State ofa Complex Art. Paris. 100 The World Bank Research Observer, voL 13, no. I (February 1998) . 1993a. Partnerships: The Key to Job Creation. Experiences from OECD Countries. Paris. 1993b. Employment Outlook. Paris. 1994a. The OECD Jobs Study: Evidence and Explanations. Paris. . 1994b. Vocational Training in the Netherlands: Reform and Innovation. Paris. O'Leary, Christopher. 1995. 'An Impact Analysis of Labor Market Programs in Hungary." W. E. Upjohn Institute for Employment Research, Kalamazoo, Mich. Processed. Pulay, Gyula. 1995. "Measuring the Efficiency of Employment Programs in Hungary." Paper pre- pared for a seminar on Evaluating Active Labor Programs. World Bank, Eastern Europe Depart- ment, Washington, D.C. Processed. Revenga, Ana, Michelle Riboud, and Hong Tan. 1994. "The Impact of Mexico's Retraining Pro- gram on Employment and Wages." The World Bank Economic Review 8(2):247-77. World Bank. 1995. World Development Report 1995: Workers in an Integrating World. New York: Oxford University Press. AmitRDarand lndermitS. Gill 101 Public Sector Reform in New Zealand and Its Relevance to Developing Countries Malcolm Bale . Tony Dale Does New Zealand's success story have lessons for developing countries contemplating public sector reform? That question usually elicits one of two reactions, both inadvisable in the authors' view. Thefirst reaction is to he impressed with the efficacy of the reforms and conclude that they should be adopted uncritically in other countries. The second reaction is that the special conditions existing in New Zealand are such that none of its reform experience is relevant to others. The authors take a middle position, maintaining that poorer countries can indeed extrapolate from the experience of their higher income neighbor despite the different conditions under which they have to operate. New Zealand's comprehensive overhaul of its public sector affords both general principles and specific elements relevant to countries looking to improve the quality, efficiency, and cost effective- ness of their public service sectors, and a careful analysis of those reforms can ascertain what might be transferable and what principles might apply. In 1984 New Zealand's newly elected Labor Government took over an economy characterized by comprehensive controls on the financial sector, extensive subsidies to farmers and exporters, and a highly sheltered private sector. Its deficit was a high 9 percent of gross domestic product (GDP), and public debt, at 60 percent of GDP, was rising. High underlying inflation and slow economic growth had reduced per capita income from one of the highest in the Organisation for Economic Co-operation and Development to one of the lowest (OECD 1983, 1984). The new government put in place a macroeconomic stabilization plan and broad structural reforms to correct the core problems; the design, implementation, and outcome of the reform program are widely cited in the economic literature and busi- ness press and will not be elaborated here (OECD 1990-94; Bollard 1992; Evans, Grimes, and Wilkinson 1996). But government expenditures still accounted for about 40 percent of GDP, which, policymakers came to believe, meant that the improved The World Bank Research Observer, vol 13, no. I (February 1998), pp. 103-21 C 1998 The International Bank for Reconstruction and Development / THE WORLD BANK 103 performance of the economy as a whole might be limited by the large public sector. Government departments were viewed as bloated, inefficient, and poorly managed. The bureaucracy behaved in ways that stereotypically undermine the capability of any government, and service delivery was poor. Departments habitually exceeded their budgets; unused balances were spent in end-of-year shopping sprees; creative accounting was used to give the appearance of good performance; and the manage- ment of assets and cash was haphazard. Most important, politicians felt that some core ministries had their own policy agendas and could override or outlast the wishes of elected officials. Thus the objec- tive of the new government was to create an efficient public sector that was also responsive to the strategic policy direction of the government. The first step was to decide which activities the government should provide and which could be divested or spun off to the private sector. The second step was to undertake structural and management reforms in the remaining "core" departments- those concerned with broad, cross-cutting, nonsectoral issues such as Treasury, the ministries of Defense and Commerce, and the Inland Revenue Department (for details on the privatization program, see Boston and others 1991; Scott and Gorringe 1991). To paraphrase Holmes and Shand (1994), the government wanted the ethic of value for money and customer service to take its place alongside the ethic of pro- bity and stewardship in the public sector. Reform of Government Commercial Enterprises In 1984 the government owned much of the economic infrastructure, including banking, postal, and telecommunications services; a steel mill; a shipping company; production forests; electric power; and a large highway construction business. Most of these activities were being run by departments that also had policy advice func- tions; nearly all ran at a loss and required taxpayer support. To reform the provision of these services, the government decided to "corporatize" these activities. Corporatization involved forming government-owned enterprises with clear com- mercial objectives, a neutral policy environment, managerial flexibility and author- ity over decisions, performance monitoring, and explicit transfers for noncommer- cial activities required by government to meet social objectives such as keeping open small, unprofitable post offices or branch railway lines. To meet these objectives, the enterprises were established under the normal com- mercial company legislation common in all countries. The only difference from a regular commercial company was that the stock of the state enterprises was not pub- licly traded but held by government ministers. These firms were required to operate profitably, to take on normal commercial levels of debt from the financial market, and to pay taxes and dividends at commercial rates. Regulatory reform proceeded in 104 The World Bank Research Observer, vol. 13, no. I (February 1998) parallel with these reforms to ensure, wherever possible, competition in both the input and output markets. The companies were given the opportunity to succeed or fail based on their commercial performance. Although corporatization was very successful, it had several drawbacks: the new corporations tied up capital that could be used to repay public debt; they were not subject to the ultimate discipline of the market because of the perception that an implicit government guarantee existed; the large amounts of capital needed to de- velop the businesses typically came at the expense of government investments in social infrastructure, such as health and education; and the government was exposed to commercial risk if the business suffered losses. Thus successive governments have moved to sell their interest in commercial ac- tivities, a relatively straightforward procedure under commercial law. The govern- ment had a single criterion in each sale-to maximize the value to the taxpayers by selling the enterprise in a competitive environment. By mid-1997,35 companies had been sold, at a value of approximately $15 billion. Most of the proceeds have been used to retire sovereign debt. The privatized entities have prospered, benefiting from the additional capital available from the private sector. Telecom New Zealand, for instance, went from one of the most inefficient telephone systems in the world to a position of international leadership in telecommunications services (Duncan and Bollard 1992). All the remaining government-owned companies now run as profitable businesses and pay substantial dividends and taxes. For example, the quality of services pro- vided by New Zealand Post has improved greatly. Although almost 80 percent of its revenue in 1994 came from activities that are competitive, it paid $22 million in taxes, paid dividends of $55 million to the government, and posted net profits of $46 million. In 1995 it lowered the cost of first-class postage from 29 cents to 25 cents and in 1996 offered users a "postage free" day. In 1997 its exclusive franchise over first class mail was removed so that its business is now fully contestable. Reforming the "Core" Public Sector To reform core activities, planners took their inspiration from public choice theory, principal-agent theory, transaction-cost theory, and the new public management literature (a brief explanation of these theories and a bibliography are provided in the appendix.) Agency theory was perhaps the most powerful of these concepts. This literature addresses the nature of contracts between two parties: the principal (or government) and the agent (or bureaucrats). In New Zealand planners believed that the problem was not with the bureaucracy-the civil service had many well-qualified and capable managers and staff who were responding in a rational way to the set of incentives they faced-but with the incentives themselves. The reforms were there- Malcolm Bale and Tony Dale 105 fore intended to replicate, as closely as possible, the types of incentive structures for performance that might be found in a well-functioning private-sector concern, while taking into account the distinctive character of public services. The approach had five characteristics: * Establishing clear lines of accountability between government ministers and their departments * Defining performance in an unambiguous and measurable way * Delegating authority to chief executives * Establishing incentives that reward or punish results relative to the agreed outcome * Reporting and monitoring performance. Two laws were passed to effect these changes: the State Sector Act of 1988, and the Public Finance Act of 1989. Accountability and Employee Relations The emphasis on a strong accountability framework implied a new relationship be- tween government ministers and permanent heads of departments. First, department heads lost their permanent tenure and were appointed to fixed terms up to five years, renewable for a further three years depending on performance. Now known as chief executives, department heads work under specified, performance-based contracts that they negotiate with the responsible minister. The State Services Commission monitors performance on behalf of the minister. The chief executives are free to run their depart- ments in the ways they deem best suited to meet the performance goals. It is said that chief executives in the New Zealand public service used to have full tenure and limited authority; now they have limited tenure and full authority (Ball 1994). Second, public sector employees were placed under the same rules and regulations as the private sector. That is, there are no public service employment regulations; all sectors, public or private, in New Zealand must comply with the same general em- ployment regulations. The chief executive of each department is free to operate un- der more or less the same conditions as a chief executive officer of a private company, making all decisions on the number of employees needed and the skill mix of per- sonnel as well as all appointments and terminations (based on performance-based contracts for managers and staff). Defining Performance Because vague or unachievable performance specifications undermine good account- ability arrangements, it was necessary to define the performance that departmental chief executives were expected to deliver. Four elements of departmental performance were considered. 106 The World Bank Research Observer, vol. 13, no. I (February 1998) THE OUTPUT-OUTCOME DISTINCTION. In New Zealand accountability between min- isters and their departments is based on the conventional distinction between out- puts (goods and services produced) and outcomes (the effect of those outputs on the community). Chief executives are responsible for specified outputs from their de- partments, while the minister chooses which outputs will be purchased to achieve certain outcomes. That is, the minister, not the department, is responsible for the outcome. The distinction is important. Governments are interested in achieving outcomes and would like to contract for them if it makes sense to do so. Outcomes are often not within the control of the chief executive, however, and he or she cannot be held accountable for them. But chief executives can be held accountable for out- puts, which can be relatively well-defined and are within the executive's control. Say, for example, the police commissioner contracts with the minister of police to provide a certain level of policing services, patrols, community security programs, road safety commercials, and so forth. These are clearly outputs. The commissioner does not contract to lower the crime rate. That outcome may be forthcoming, but the crime rate is affected by many variables beyond the control of the commissioner, such as the level of unemployment, immigration policy, social policy, perhaps even the result of certain sporting events. Thus, holding the commissioner accountable for outcomes is not operationally useful; this distinction between who is accountable for outputs and who for outcomes is common in the private sector as well. Accountability processes are much more effective if the outputs to be delivered are well specified in advance. Vague specifications allow managers to determine exactly what it is their organization will produce. For this reason, the minister and chief executive prepare annual purchase agreements, or contracts, that set out in reason- able detail the outputs to be delivered. The quality and robustness of this system depend on the careful specification of outputs in the agreement. Many ministers have purchase advisers to assist them with this task. A common concern is that this approach emphasizes outputs at the expense of outcomes. Critics maintain that unless departments are held accountable for out- comes, they will not focus on these issues and therefore are less likely to achieve them. That is not necessarily the case. Greater clarity over what is being produced (outputs) can increase the attention to outcomes. Since 1990 the government, as part of the annual budget process, has reviewed all the outputs produced by depart- ments against the criterion: How does this help achieve the outcomes the govern- ment wishes to pursue? This focus makes clear that policy advice should be about the relationship between interventions (including outputs) and outcomes, about what the government is trying to achieve and alternative ways to achieve it. This frame- work has helped departments understand that, just as in the private sector, their survival is dependent upon meeting the needs of their customer. Because their cus- tomer is interested in outcomes, departments, given sufficient competitive pressure, will strive to design and provide public services to help achieve those outcomes. Malcolm Bale and Tony Dale 107 THE OWNER-PURCHASER DISTINCTION. In addition to purchasing most of a department's outputs, the government is also the department's owner. As the pur- chaser, the government wishes to obtain goods and services of a specified quantity and quality for the lowest price, whether it is buying from the private or public sector. To enhance price competition, all ministers are free to purchase their outputs from nondepartmental sources. The finance minister, for example, may purchase economic projections and advice from Treasury officials or from any private domes- tic or international economic consulting firm. As the owner of the Treasury, the government wants to obtain the best possible return on its investments in it. If the Treasury fails to record a positive rate of return, public production lowers the wealth of society. It would better in that case to close that department and contract out all of its work (outputs). The New Zealand approach recognizes these two perspectives as two different dimensions of departmental performance. Performance agreements with each chief executive separately specify each dimension, and both are monitored. The fact that governments purchase goods and services from their departments and also own those departments could give rise to conflicting objectives. This conflict is solved by pric- ing departmental outputs at a price equivalent to that charged by the private sector. The department's performance as a business can then be fairly assessed using normal business evaluations that enable governments to determine whether the department is operating efficiently and whether owning the business is advantageous. Where the private sector produces the same output, this model does not pose a problem. But where there is no comparable private supplier, an alternative bench- mark must be established. In some cases, prices charged by different public sector entities can be compared, as in a 1993 study of the unit price of policy advice out- puts provided by departments. Where no endogenous benchmark can be established, officials are experimenting with establishing an initial price that applies some pricing rule to costs (such as cost minus x). Although this method will not establish an efficient market price, it does create some separation between cost and price. Ministries regard the minister as a customer rather than as the recipient of a ser- vice. Ministries are accountable only to their minister, not to taxpayers and service recipients. Ministers are accountable to taxpayers and service recipients. This con- trasts with the citizen charter approach used in the United Kingdom, which creates a dual accountability for departments: to their minister and to the recipients of their services. A difficulty can arise if those accountabilities are inconsistent-for example, if the quantity or timeliness of services demanded by recipients is incompatible with the amount ministers are prepared to pay for the services. The approach used in New Zealand resolves this problem by having ministers make the tradeoff among quan- tity, quality, time, place, and cost of service delivery; that is, between common inter- est and individual interest. This does not mean that departments are not concerned about whether customers are satisfied. Such a response will inevitably become a 108 The W'orid Bank Research Observer, vol. 13, no. I (February 1998) concern of the minister, whose political antennae should be well attuned to such feedback. THE GOVERNMENT-DEPARTMENT DISTINCTION. Chief executives make all input de- cisions, including capital investment decisions (within a defined capital base). In- deed, such authority is necessary if they are to be held accountable for producing outputs in the most efficient manner. Some departments, however, manage inputs over which they do not have full control, that are not used to produce their outputs, or for which the government does not wish to delegate authority. As a result, activi- ties and assets have been divided into two categories: "the Crown" (ministers); and the departments. For Crown activities or assets, accountability remains with the gov- ernment. For example, the Department of Social Welfare is responsible for adminis- tering the social support scheme, but the minister is responsible for the size and number of welfare payments. Understandably, the chief executive cannot be held accountable for, say, the number of people unemployed or the number of single- parent children. Similarly, the national parks are managed by the Department of Conservation for the government but are owned by the government and listed as a Crown asset. POLICY ADVICE-SERVICE DELIVERY DISTINCTION. Where an agency provides both policy advice and service delivery, a potential conflict of interest arises between the two functions. Separating them reduces the potential for policy advice bias. Under the reforms, policy advice is an output provided by departments in much the same way as consulting advice is an output provided by a consulting firm. Good policy advisers must be able to evaluate the tradeoffs between different outcomes and identify the nexus between outputs and outcomes. They must evaluate spending proposals against all alternative interventions that could pro- duce the same outcome. If, for example, the government decides it wants to reduce road accidents, it may purchase outputs such as highway patrols, road repairs, and vehicle inspection checks. Alternatively, it may intervene legisla- tively with speed limits, higher alcohol taxes, different speed sanctions, or com- pulsory driver-education programs. Viewed from that perspective, policy advice is a specialized business that is inher- ently different from service delivery. In most cases, it has been decided that policy advice is most effectively and efficiently produced by a department dedicated to its production rather than by one with distinctly different lines of business. For ex- ample, in the environmental area, policy advice is the responsibility of the Ministry for the Environment, while another agency, the Department of Conservation, is responsible for delivering services such as operating the national parks. Because the ministry does not deliver services, its advice about appropriate interventions can be independent of the business implications for the department. Malcolm Bale and Tony Dale 109 A related reason for separating these functions is to reduce the tendency for special interest groups to "capture" the agency that regulates them (Posner 1974). This rent-seeking behavior can be reduced if regulatory policy is designed in one agency and enforced in another. Reporting, Monitoring, and Coordination Improved reporting and monitoring of departmental performance was the quidpro quo for enhancing the chief executive's autonomy. This required upgrading financial management systems and skills. Ex POST REPORTING. Defining and monitoring purchase and ownership performance requires information about the full resource cost, including the consumption of as- sets and the opportunity cost of capital, and about assets and liabilities, their utiliza- tion, and the return being generated. For this reason, all government entities are required to report financial performance on an accrual accounting basis, using the same generally accepted accounting practices as does the private sector.' Each department must provide a full set of financial statements to its minister and to the Treasury on a monthly basis. In addition, departments must produce and submit an audited annual Statement of Service Performance, outlining the outputs produced versus the outputs agreed and giving information about purchase perfor- mance in the same way that private companies produce annual reports showing their financial statements and performance. As a result the government can simply sum all of the accounts to produce national financial statements on an accrual basis. These are published monthly, and annual audited financial statements are presented to the Parliament within three months of the close of the fiscal year. Thus government accounts look similar to an annual report produced by a private company. They show, among other things, the net worth of the government. Ex ANTE BUDGETING. The reforms also changed the budget and appropriations sys- tems in two ways to fit with the performance management system. First, appropria- tions for departmental outputs (not inputs) are now made on an accrual accounting basis, and managers are free to acquire their inputs from any provider. Second, only capital injections into departments are appropriated (not capital expenditures), re- flecting the chief executive's authority to manage assets within a defined capital base. Budgetary reporting at both departmental and national levels mirrors ex post report- ing. Financial forecasts are prepared using generally accepted accounting practices and are identical in form to financial statements produced by private corporations. CAPITAL CHARGE. Twice a year departments pay a capital charge, calculated on the basis of their net assets, for the cost of the capital the government has invested in 1 10 The World Bank Research Observer, vol 13, no. I (February 1998) them. This charge has several benefits. First, it ensures that the cost of capital is reflected in output prices because a department's total cost must be allocated to its outputs to ensure comparability with nongovernment producers. Second, the charge encourages departments to manage their balance sheets carefully and to divest sur- plus or redundant assets. Third, it encourages management to consider the mix of assets needed to produce services efficiently. If a manager finds it more efficient to purchase more computers and sell some cars, he or she is free to make this decision. If the sale of assets reduces the overall capital charge, the savings can be applied to other expenses. STRATEGIC POLICY COORDINATION. Three means were devised to coordinate strate- gic policy, which was a key objective of the reforms. Ministers were required to specify and publish the outcomes the government wished to achieve. Coordinating committees were established, made up of senior government officials from depart- ments with an interest in broad areas of government activity, such as social policy, education and training, or environment. Their role is to ensure that policy options are developed in a coordinated way across the government. The third method of coordination is created by the policy-delivery split. Because ministers want advice about policies that can be effectively implemented, and they want output proposals that meet their objectives, policy and delivery departments have a relatively strong incentive to coordinate with each other. Rather than weaken the horizontal linkages between departments, these arrangements have served to strengthen them. The use of the private sector profit center approach combined with market pric- ing techniques has also solved one of the common public sector problems: transfer pricing-the price of a good or service for a transaction within an organization. Because there is no market price, the organization can set the price at any level, including zero. Modern business practices require that transfer prices approximate market prices so that accurate profit-loss data can be established by each unit of an organization. Because each department is regarded as a profit center and all transac- tions between departments are treated as arms-length transactions, the transfer pric- ing problem has effectively disappeared. Results The core sector reforms have succeeded in improving both service delivery and effi- ciency. The system is widely supported by departmental managers, although the effect of the reforms on managerial behavior has varied depending on the quality of leadership and the levels of efficiency prior to the reforms. In general, performance has improved in tandem with the development of wage scales linked to performance. Savings from improved cash management have been substantial (enough to pay for all the system costs of the reforms), and unappropriated expenditures, which were Malcolm Bale and Tony Dale 1I1 quite extensive, have now all but disappeared. Cost per unit of output has declined, in some cases quite markedly (Deloitte, Ross, Tohmatsu 1990). Human resource management has improved measurably, and explicit attention is paid to issues such as succession and the development of management skills. The strategic gains are more difficult to assess, although some improvement is evident. Economic Policy Reform The reforms designed to determine the government's overall economic policy are embodied in two pieces of legislation: the Reserve Bank Act of 1989, and the Fiscal Responsibility Act of 1994. The former is intended to make monetary policy deci- sions transparent. The independent governor of the central bank manages the money supply and interest rates to achieve a target rate of inflation established in an agree- ment with the government. Apart from periodically determining the target inflation range (currently 0-3 percent annually), the act prohibits the government from in- volvement in monetary policy. The Fiscal Responsibility Act is designed to provide a similar political discipline on fiscal policy, although clearly fiscal policy cannot be made at arm's-length. In budgetary decisionmaking, officials have a political incen- tive to trade off the government's aggregate fiscal position against increased expendi- tures. In countries around the world-and formerly in New Zealand-the negative economic impact of fiscal deficits can be easily pushed off for future generations to deal with. The act uses transparency and accountability mechanisms to put the onus on governments to act in a fiscally responsible manner to control the deficit.2 This approach has two notable features. First, the fiscal aggregates (debt, net worth, operating surplus or deficit, operating expenses, and operating revenue) are defined in accrual terms. As a result, the budget forecasts must include all noncash expenses, such as depreciation and the unfunded liability to the government employee's retire- ment fund, and exclude capital transactions, such as the proceeds from asset sales. This requirement significantly reduces the government's ability to manipulate the aggregates between years and prevents "off-budget" manipulations. It also forces the government of the day to interpret the requirements of the law in the context of prevailing economic conditions and to justify those interpretations to Parliament and the public. This approach is designed to be more politically and economically sustainable than legislated targets. The second feature calls for regular reporting of fiscal intentions, forecasts, and results. Transparency in a democracy is a powerful discipline. On the forecasting side, the act requires the government to give Parliament a statement specifying its strategic budget priorities, three-year fiscal intentions, and long-term policy objec- tives; a fiscal strategy report showing whether the budget forecasts are consistent with the budget policy statement (and if not, why those intentions have changed); 1 12 The World Bank Research Observer, vol. 13, no. I (February 1998) three-year economic and fiscal forecasts at the half-year point and within two to six weeks before a general election; and a current year fiscal forecast with the supple- mentary estimates that are normally tabled in the last quarter. On the reporting side, aggregate financial statements of the Crown must be released within five weeks after the end of the month. The existing requirement for audited, annual Crown financial statements remains. All aggregate fiscal reporting, including fiscal forecasts, must comply with generally accepted accounting practices. The Reserve Bank Act is regarded as highly successful and is credited with New Zealand's good record on inflation in recent years. The value of the Fiscal Responsi- bility Act has yet to be fully tested. It was enacted after the major fiscal correction was completed and has yet to apply during a period of economic downturn. It has, however, provided a frame of reference for public debate over the Crown's fiscal position, and most political parties have cast their economic policy with its prin- ciples in mind. The quality and quantity of fiscal reporting has been welcomed by the financial markets and rating agencies. Relevance to Developing Countries It is not difficult to understand how dysfunctional public sector agencies impair devel- opment and perpetuate poverty. The most common solutions frequently involve me- chanics, such as streamlining procedures, banning dual employment, increasing civil service salaries, fighting corruption, upgrading training, and decentralizing govern- ment services. These ideas are useful, but they do not appear to offer even a partial solution to the deeper problems encountered by public sector agencies in many devel- oping countries that have been seriously depleted by years of neglect and corruption. In some cases public organizations are no longer able to perform the task for which they were created. Thus civil service reform may require more than minor changes; it may need to be more fundamental and based on a more finely tuned understanding of the causes of the malfunction. Difficulties in the reform of New Zealand's health sector illustrate some of the potential problems that can arise (Box 1). What lessons can be drawn from New Zealand's experience that may be applicable to developing countries? The Value of a Consistent, Comprehensive Conceptual Model The one aspect that sets New Zealand apart from other public sector reforms is its underlying conceptual framework. That framework, which was based on identifi- able theoretical constructs, proved valuable in several ways. First, it helped ensure that the reform was developed from a broad, systemwide perspective that focuses on the causes, not the symptoms, of dysfunctionality. For example, some involved in writing the reforms considered issues such as financial waste, excessive rules, Malcolm Bale and Tony Dale 113 Box 1. Reforming Health Care: Anatomy of Limited Success Historically, New Zealand's health care system has been predominantly publicly operated and funded through general taxes. Before 1993, when the health care system was reformed, secondary and tertiary (hospital) services were the responsibility of area health boards, with funding provided according to population. The new system (which is similar to the United Kingdom's) is based on a Ministry of Health (a policy advice agency), four regional health authorities (who purchase primary, secondary, and tertiary health services for specified regional populations), and a series of Crown health enterprises (government-owned hospitals run on commercial business practices). Primary services continue to be provided by private practitioners (who are heavily subsidized). The regional health authorities determine the mix of health services to be provided to their populations within a given funding level. They are then responsible for purchasing those services and monitoring delivery, with oversight from boards of directors made up of health, community, and commercial representatives. The results have been mixed, but generally the public health system has failed to achieve the performance gains expected, and considerable political and public debate continues about the desirability of the arrangements. Although the quality of financial management and the degree of transparency among the Crown health enterprises has improved significantly, efficiency gains have been slow to emerge, and government funding has been required to keep many enterprises solvent. The public remains uneasy with the notion of for-profit hospitals. The explicit division of policy and purchasing responsibilities has been less successful than anticipated, health officials maintain significant control over purchasing decisions, and there is a continuing demand for additional funding. As a result of political compromises, most of the reforms to improve the delivery of social services such as health and education have not been implemented as designed. The reforms were predicated on the existence of market competition (on both the supply and demand side), which has not developed. The quasi-commercial and social objectives of providers have diluted the focus on performance. Political issues were also underestimated. Social policy concerns the public rationing of private goods, and removing these concerns from the political arena has proved to be much more difficult than anticipated. This has resulted in much more direct ministerial involvement (some would say interference), which has undermined the authority of the various providers. The contrast in performance between the purer commercial model versus the political-interference model provides a salient lesson. If quantum efficiency gains are desired and if commercial objectives are to be achieved, it is necessary to divorce political influence from commercial decisionmaking. and poor staff performance to be the problem. They were, in fact, only the symp- toms; the real problem was the lack of management incentives. Second, the frame- work provided consistency for the multiple layers of decisions required in the de- sign and implementation of the reform. These decisions, such as the nature of the accounting system, the approach to budgeting, and the degree of personnel del- egations, were all made in a consistent manner. Third, it focused attention on a comprehensive approach. As a result the reform addressed all aspects of public sector management and all parts of the public sector (departments, government corporations, local governments). Fourth, the framework guided the sequencing and implementation of the reforms. These decisions were based on what was most important from a top-down perspective rather than what took the fancy of depart- 1 14 The World Bank Research Observer, vol. 13, no. 1 (February 1998) ments. Fifth, it aided in marketing the reforms to departments and the public. The coherence and comprehensiveness reduced fears that the reform was just another ad hoc initiative. The lesson for other countries is clear. Basing reforms on an analytically rigorous conceptual framework appropriate for the jurisdiction con- cerned and having the framework apply to the entire public sector is likely to improve significantly the chances that the reform will be successful. The Importance of a Clear Performance Definition The second lesson relates to performance specification. The reforms were based on identifying the various principal-agent relationships; specifying and reporting per- formance in a clear and unambiguous manner; and ensuring that managerial author- ity matched the department's responsibilities. Providing state-owned enterprises with a clear profit-maximizing objective is an example of this expectation. So, too, are the features described in the discussion of performance, which are designed to clarify what performance the principal expects of the agent in a way that holds the agent unambiguously accountable. However performance is defined, the agent must have control over it; for this reason, the outputs approach has great merit. The existence of several principals (as in the U.S. separation of powers) does not necessarily invalidate the principal-agent approach. Rather, it highlights the need to be clear about the different aspects of performance each principal requires of the agency, and for the principals to reconcile any mutually exclusive conflicts in those requirements. New Zealand sought to ensure that managers faced incentives congruent with the performance expected of them. This basic notion is often overlooked in the design of public sector reforms. Performance incentives are much more subtle and pervasive than the "personal bonus or reward" incentive commonly mentioned. Other aspects need to be considered, including * Whether the institutional and organizational arrangements encourage the performance required. * Whether agencies can be made more efficient and responsive to their customers' needs, either by increasing competition or by direct customer purchasing of services (rather than the minister purchasing services on behalf of consumers). The evidence is that they can. * Whether all aspects of the management system are sending the same signals to managers. Often budget and accounting systems define and measure per- formance in totally different ways. * Whether the actual performance of all the actors in the systems is transparent. * Whether the system is empowering or controlling. * Whether the systems encourage managers to actively manage all their resources. The purchase of cleaning services, motor vehicles, or even personnel by central Malcolm Bale and Tony Dale 115 agencies at no cost to operating departments discourages those departments from efficiently managing their costs. 'Whether the personnel system encourages good performance. Focusing on What Government Does Best It is important to focus the government's resources on areas in which government can add value, such as establishing an appropriate regulatory framework and eco- nomic environment in which the private sector can thrive. Wherever possible, New Zealand turned over businesses to the private sector because governments do not have the ability to manage and monitor enterprises in the same thorough way as does the private sector. Commercial objectives and decisions are easily compromised by political and social ones. In addition, state enterprises compete for scarce govern- ment financial and management resources; thus they tend to be undercapitalized and undermanaged. Governments should concentrate on what they do best-estab- lishing the regulatory and economic environment, financing public infrastructure, and ensuring the delivery of public goods. And these activities can be operated along commercial business lines to a much greater extent than is commonly believed. A public sector management system is a means; not an end. It will not deliver better fiscal performance on its own, but instead needs to work inside an overarching political economy framework that sets clear macroeconomic goals and has the politi- cal resolve to achieve them. This point is clear from the New Zealand experience. All of the reforms were driven by the same central goal: to improve macroeconomic performance by reducing the negative impact of the public sector. A good public sector management system will provide politicians with the tools they need to achieve those targets; a poor one will make a politician's task more difficult. It is more difficult to draw conclusions about whether specific practices adopted in New Zealand are applicable to developing countries. If the cultural and political environment is too dissimilar, the applicability of these practices may be limited. The following precedents formed the basis for the reforms adopted in New Zealand: a tradition of a politically neutral, relatively competent civil service; little concern about corruption or nepotism; a consistent and well-enforced legal code, including contract law; a well-functioning political market; and a competent, but suppressed, private sector. The right reform mix for any developing country must reflect any major differ- ences in these preconditions; New Zealand's reforms cannot simply be transplanted. For example, because reducing corruption or nepotism and increasing democratic participation were not objectives in New Zealand, the reforms did not address these issues. If financial performance and service delivery are concerns, however, then spe- cific New Zealand techniques may be appropriate. Subject to this caveat, the follow- ing techniques are likely to find broad applicability in developing countries. I 16 The World Benk Research Observer, vol. 13, no. I (February 1998) * If the performance of government commercial activities is an issue, developing countries would find it useful to separate trading activities from core departmental functions, apply a "level playing field" regulatory regime, and appoint independent boards with business-rather than political-expertise. The success of such reforms is dependent on acquiring the necessary governance and managerial skills from the private sector. Many developing countries have this expertise. * If the concern is about users of the service heavily influencing policy advice, then separating policy advice and service delivery functions into different agencies is likely to lead to both better advice and better service delivery because conflicting objectives have been removed. If this route is pursued, the institutional design should put more emphasis on processes to coordinate advice. These can operate at the political level, the bureaucratic level, or both. * If service delivery is a concern within core government agencies, a management system focused on outputs would be relevant to developing countries. Politicians and managers would need to make a significant paradigm shift in the way the government operates, not least in securing the political commitment to such a change. Publishing the agreed outputs would be a first step in transparency and has the potential to improve accountability. To be fully effective, the planning, budgeting, appropriation, performance assessment, and reporting systems all need to be based on outputs. This does not preclude accountability for input management as well, if that is considered desirable. If financial performance is an issue, improved measurement of financial performance (through the use of accrual accounting) and explicit attention to this information in budgeting and reporting systems is important, although skill shortages may make implementation difficult for some countries. Significant gains can be achieved through improved financial management and financial control systems even without accrual accounting. The establishment of an effective cash accounting system may be considered a precursor to accrual accounting. * If personnel performance is a concern, performance-based personnel systems are likely to improve the staffs incentives to perform. Such contracts may be politically difficult, but they have been a central part of institutional reform in New Zealand, are viewed as very effective, and were the key to changing the public service culture. Even if this model is not fully acceptable in developing countries, changing the appointment, appraisal, and promotion systems so that they are based on performance is likely to be possible and beneficial. * Transparency is an important incentive device. Providing that voters are able to replace their politicians if they do not make decisions in the public interest, developing countries can require improved specification and reporting of performance to government and to the public. Malcolm Bale and Tony Dale 1 17 * The performance of civil service managers is often impeded by a lack of author- ity. New Zealand's approach of delegating authority for input management to managers is likely to be useful in many developing countries providing that a meaningful output accountability system is implemented. In cases where del- egating process decisions, including personnel and capital matters, may not be appropriate, separating the role of politicians and managers may be useful. The former can determine the output to be produced, while the latter determine the method of production. Transplanting the system and structures in one country unchanged into an- other is seldom possible because the efficacy of a system depends so much on the complementary structures. At the same time, countries commonly study and adapt systems and structures from other countries to fit their particular circumstance. In the last decade, many industrial societies have moved to narrow the government's role in the economy and insist on a more sharply focused and less intrusive con- duct. Part of that process has been not only a questioning of what governments do, but also a reexamination of how they do it. Thus even if developing countries must adapt the reform agenda to their own circumstances, policymakers can learn from what may well be a best practice and draw lessons and principles from New Zealand's experience. Appendix. The Theoretical Underpinning of the New Zealand Reforms The theoretical concepts that influenced New Zealand's reform program included public choice theory, agency theory or principle-agent theory, transaction-cost theory, and new public managerialism. This discussion draws heavily on Boston and others (1991). Public Choice Theory Closely associated with the work of Buchanan and Tullock (1962), Tullock (1965), Olson (1965), and Niskanen (1971), this theory seeks to explain how voters, politicians, bureaucrats, and lobbyists will behave in different institu- tional settings with different incentive rules. It is based on the idea that human behavior is dominated by self-interest. Thus, government officials will attempt to enlarge their department budgets, say, without regard to the overall govern- ment budget. Similarly, public and private groups will undertake rent-seeking activities to the disadvantage of the broader society, and politicians may pur- sue their own objectives at the expense of many of their constituents. As a 1 18 The World Bank Research Observer, vol. 13, no. I (February 1998) result powerful interest groups may capture a disproportionate share of na- tional income, and politicians may misuse their power. Democracy is thus undermined. Principal-Agent Theory At the core of this theory is the idea that interchange between parties can be charac- terized as a series of contracts where one party, the principal, enters into agreements with another party, the agent, who agrees to perform tasks on behalf of the principal in return for compensation (Moe 1984, 1990; Pratt and Zeckhauser 1985; Bendor 1988). Moe (1984:765) notes that politics can be seen as a series of principal-agent relationships from citizen to politician to senior bureaucrat to subordinate bureau- crat to service providers. Agency theory assumes rational, utility-maximizing behavior by individuals. Hence conflicts will arise between principals and agents as their self-interests differ. Add to this asymmetric or incomplete information, the difficulty of observing and monitor- ing agents' behavior, and the imperfect mapping of agents' outputs and the out- comes desired by the principal, and an even larger ground for conflict will exist. Principal-agent theory is concerned with the best way to construct and monitor contracts so that these kinds of conflicts are minimized. The theory is useful in ana- lyzing the selection of agents, designing incentives and pay systems, and choosing between in-house or outside contractors. Transaction-Cost Theory This approach compares the costs of planning, adapting, and monitoring under al- ternative governance structures. Decisionmakers wish to minimize their aggregate costs of production and transaction. But like agency theory, this approach assumes that principals and agents will act in their own self-interest and thus may be unreli- able parties to a contract. The literature on transaction costs indicates that some transactions are better suited to market-type arrangements, while others are better suited to hierarchical or rule- driven organizations. For example, contracting out is likely to be desirable where the supply of a good is contestable, quality and quantity can be easily measured and specified, and suppliers are numerous. In-house provision is likely to be more effi- cient when the opposite conditions exist. When transactions occur frequently, are associated with uncertainty, and involve specific assets or skills, hierarchical organi- zation tends to be more efficient. Where the supply is competitive and transactions costs are "average," the preferred organization is less clear. But generally, in-house provision is likely to be more efficient where there is a high risk of self-interest, conflicts of interest, substantial uncertainty, Malcolm Bale and Tony Dale 1 19 and recurrent, complex transactions. According to Williamson (1985), these factors explain the concentration of production in some sectors in a few large firms. Thus direct provision may be preferable when maintaining quality is critical and opportun- ism poses a serious threat. It is for these reasons that governments are hesitant to con- tract out the gathering of military intelligence and the collection of taxes. The New Public Management This approach (see Aucion 1990 and Caiden 1988) centers on the presumption that a distinct activity called "management" can be applied to public and private busi- nesses alike, and that it includes the following elements: a move away from input controls, rules, and procedures toward output measurement and performance tar- gets-the "accountability" framework; the devolution of management control with improved reporting and monitoring mechanisms; a preference for private owner- ship, contestable provision, and contracting-out of publicly funded services; the adop- tion of private-sector management practices in the public sector, such as short-term labor contracts, performance-linked remuneration schemes, the development of a mission statement, greater concern with corporate image, and the development of a corporate strategy and action plan; an emphasis on efficiency, often referred to as "value for money" (Hood and Jackson 1991). Notes Malcolm Bale is principal economist, East Asia and Pacific Region, the World Bank, and Tony Dale is a principal of Public Sector Performance (NZ) Ltd., a public management consulting business. 1. Accrual accounting is an accounting method that recognizes transactions and other events when they occur and not as cash transactions or their equivalent. The events are recorded in the accounting period and reported in the financial statements in the periods to which they relate. 2. The five principles of responsible fiscal management are (1) reducing total Crown debt to prudent levels by ensuring that total operating expenses for the Crown are less than total operating revenues in the same financial year; (2) maintaining prudent debt levels, once they have been achieved, by ensuring that on average over a reasonable period of time, Crown operating expenses do not exceed Crown operating revenues; (3) achieving and maintaining Crown net worth at a level that provides a buffer against future adverse events; (4) managing prudently the fiscal risks facing the Crown; and (5) pursuing policies that are consistent with reasonable predictability about the level and stability of tax rates. References The word "processed" describes informally reproduced works that may not be commonly available through library systems. Aucoin, Peter. 1990. "Administrative Reform in Public Management: Paradigms, Principles, Para- doxes, and Pendulums." Governance 3:115-37. 120 The World Bank Research Observer, vol. 13, no. I (February 1998) Ball, Ian. 1994. "Reinventing Government: Lessons Learned from the New Zealand Treasury." The Government Accountants Journal (Fall): 19-28. Bendor, Jonathan. 1988. "Formal Models of Bureaucracy." British Journal of Political Science 18(3):353-95. Bollard, Alan. 1992. New Zealand Reforms, 1984-91. San Francisco: ICs Press. Boston, Jonathan, John Martin, June Pallot, and Pat Walsh, eds. 1991. Reshaping the State: New Zealand's Bureaucratic Revolution. Auckland: Oxford University Press. Buchanan, James, and Gordon Tullock. 1962. The Calculus of Consent: Logical Foundations of Constitutional Democracy. Ann Arbor: University of Michigan Press. Caiden, G. 1988. "The Vitality of Administrative Reform." International Review ofAdministrative Sciences 54(3):331-58. Deloitte, Ross, Tohmatsu. 1990. "Post-Implementation Review of Cash Management Reform." Paper prepared for and available from the New Zealand Treasury, Auckland. Processed. Duncan, Ian, and Alan Bollard. 1992. Corporatization and Privatization: Lessons from New Zealand. Auckland: Oxford University Press. Evans, Lewis, Arthur Grimes, and Boyce Wilkinson. 1996. "Economic Reform in New Zealand, 1984-95: The Pursuit of Efficiency." Journal of Economic Literature 34(December): 1791-902. Holmes, Malcolm, and David Shand. 1994. "Management Reform: Some Practitioner Perspectives on the Past Ten Years." In Ten Years of Change: Proceedings of the SOG Ten-Year Reunion. Hood, Christopher, and Michael Jackson. 1991. Administrative Argument. Aldershot, U.K.: Dartmouth Publishing. Moe, Terry M. 1984. "The New Economics of Organization." Americanjournal of Political Science 28:739-75. . 1990. "Political Institutions: The Neglected Side of the Story." Journal ofLaw, Economics, and Organizations 6:213-53. Niskanen, William. 1971. Bureaucracy and Representative Government. Chicago: Rand McNally. OECD (Organisation for Economic Co-operation and Development). 1983, 1984, 1990-94. OECD Economic Surveys: New Zealand. Paris. Olson, Mancur. 1965. The Logic of Collective Action. Cambridge, Mass.: Harvard University Press. Posner, Richard A. 1974. "Theories of Economic Regulation." BellJournal of Economics and Man- agement Science 5(2):335-58. Pratt, John, and Richard Zeckhauser, eds. 1985. Principals and Agents: The Structure of Business. Boston: Harvard Business School Press. Scott, Graham, and Peter Gorringe. 1989. "Reform of the Core Public Sector: The New Zealand Experience." Australian Journal of Public Administration 48(1) :81-92. Tullock, Gordon. 1965. The Politics ofBureaucracy. Washington, D.C.: Public Affairs Press. Williamson, Oliver. 1985. The Economic Institutions of Capitalism: Firms, Markets, Relational Con- tracting. New York: Free Press. Malcolm Bale and Tony Dale 121 Why Most Developing Countries Should Not Try New Zealand's Reforms Allen Schick During the past decade New Zealand has introduced far-reaching reforms in the struc- ture and operation of government departments and agencies. This model has attracted interest in developing countries because it promises significant gains in operational effi- ciency. But developing countries, which are dominated by informal markets, are risky candidatesfor applying the New Zealand model. The author suggests that basic reforms to strengthen rule-based government andpave the way for robust markets should be under- taken first. Developing and transitional countries have an understandable desire to accelerate public sector reform by adopting the most advanced innovations devised by indus- trial countries. This interest has been stimulated by the New Zealand model, which gives public managers broad discretion to operate within an accountability frame- work that specifies the results to be achieved and closely monitors performance. During the past decade, dozens of countries have sent delegations to New Zealand to observe its avant garde management practices and to interview government officials on how the new systems and procedures have affected the cost and delivery of public services. The World Bank and other international organizations have showcased New Zealand's reforms at various conferences, and some of the architects of the reforms have crisscrossed the globe extolling the virtues and portability of their country's version of results-oriented public management. Despite the interest and the sales efforts, only a few developed countries (such as Iceland and Singapore) have adopted selected features of the model; others (such as Sweden and the United Kingdom) have embraced a managerial ethic without sub- scribing to the hard-edged contractualism that differentiates New Zealand's reforms from those tried elsewhere. To this writer's knowledge, however, not a single devel- oping or transitional country has installed the fill New Zealand model, although quite a few have been enchanted by the prospect of leapfrogging to the front ranks in the international reform sweepstakes. A few countries (such as Mongolia) are in the The World Bank Research Observer, vol. 13, no. I (February 1998), pp. 123-31 1998 The International Bank for Reconstruction and Development / THE WORLD BANK 123 early stage of adopting selected features of the system, but it is much too early to gauge how far they will go in embracing its basic tenets. On the whole, industrial and developing countries have not implemented such reforms because the reforms are beyond their reach or do not fit their current needs. I draw this conclusion despite New Zealand's enormous contribution to the theory and practice of public management. Not only has the menu of reform possibilities been greatly expanded, but New Zealand has brought its public management much more closely into line with institutional economics and with contemporary business practices. And the rigor with which the model has been applied is impressive. This was not a case in which reformers selected discrete entries from a large menu of reforms. The change agenda was driven by ideas that have only recently entered mainstream economics, and the ideas were applied with full fidelity to their internal logic. But I do not accept the view that New Zealand offers practical guidance on how developing countries should surmount deficiencies in public management. I believe that there are important preconditions for successfully implementing the new public management approach and that these should not be ignored by countries striving to correct decades of mismanagement. In contrast to those who take the position that managerial deficiencies should be the driving factor in determining the suitability of these type reforms, I argue that they should be deterring factors. The greater the shortcomings in a country's established management practices, the less suitable the reforms. The New Zealand Model: Government by Contract Since 1988, New Zealand has implemented an enormous number of management reforms that add up to an integrated concept of how government should work. That concept is expressed in the heading to this section. Virtually every element of reform has been designed to establish or strengthen contract-like relationships between the government and ministers as purchasers of goods and services and departments and other entities as suppliers. Hundreds of contracts are formally negotiated each year; the typical contract specifies the resources that one side will provide and the perfor- mance the other side will produce. Ministers are always on the resource-providing side of the relationship; chief executives can be on either side, depending on the role they are playing. A chief executive provides resources in negotiating employment contracts with managers but promises results in negotiating purchase agreements with ministers and performance agreements with the State Services Commissioner. This "new contractualism" replaces the implicit or relational contracts that char- acterize traditional public administration.' Contracts convert the budget from an understanding between government and parliament on the amounts to be raised and spent into an explicit statement of what will be done with the resources to be made 124 The World Bank Research Observer, vol. 13, no. I (Februaiy 1998) available. In a similar vein, performance agreements displace the old civil service ethic of trust and responsibility with accountability for the results expected from each chief executive. New Zealand has gone to extraordinary lengths to create conditions under which formal contracts are negotiated and enforced. It has restructured many departments to decouple policymaking functions from the delivery of services. (For example, the national defense organization was split into two separate entities: the Ministry of Defense, which provides policy advice to government; and the Defense Forces, which carry out assigned operations). Under the new system, ministers can purchase ser- vices from government departments or from any alternative public or private sup- plier. Appropriations are on an accrual basis, so that the full cost of the goods and services is incorporated in the purchase price. In fact, to promote competition among suppliers, appropriations include an amount for the tax on goods and services that would be paid in a private transaction. In a similar vein, a capital charge is levied on the net worth (as shown on its balance sheet) of each department. This charge re- flects the opportunity cost of money and resembles the internal rate of return ex- pected by firms from their operating units. Contract-like arrangements have been extended to policy advice as well, so that ministers can opt to obtain information and ideas from consultancies and other external sources. To put alternative suppliers on an equal footing, the gov- ernment accords the chief executives of departments the same operating flexibil- ity that is enjoyed by executives in nongovernmental organizations. Chief execu- tives are given a block of resources for each class of outputs they contract to purchase, and they have discretion to select the mix of inputs used in producing the outputs. These outputs are specified in detail so that the government can have reasonable assurance that departments are producing the outputs contracted for. Under the present system, when the budget is tabled in Parliament, each department publishes a departmental forecast report that, among other things, specifies the outputs it will produce in the next financial year. Shortly before the start of the year, the outputs are specified in greater detail in purchase agree- ments signed by the chief executive and the minister purchasing the services. Multiple purchase agreements are written when more than one minister pur- chases services from the same department. After the year is done, each depart- ment publishes an annual report that specifies the outputs actually produced, thereby enabling the government to determine whether the terms of various con- tracts have been fulfilled. Before assessing whether the New Zealand model is appropriate for developing countries, it is necessary to consider its effectiveness at home. In my view, organiza- tional performance has been significantly enhanced. But this favorable assessment carries certain caveats, some of which were discussed in my report commissioned by the New Zealand Government (Schick 1996). Allen Schick 125 * The New Zealand model emphasizes matters that can be specified in contracts, such as the purchase of outputs, but gives inadequate attention to outcomes and the government's ownership interest because they do not fit easily into the contracting framework. * Robust contracting depends on voluntary, self-interested action. Sometimes, however, self-interest defeats the government's collective interest. In the early years of reform, for example, efforts to establish a senior executive service were undermined by managers who preferred to contract on an individual basis. * Contractualism may weaken traditional values of public service, personal responsibility, and professionalism. It can induce managers to take a checklist approach to accountability-"if it's not specified, it's not my responsibility." * Contract-like arrangements do not themselves create arms-length relationships in the public sector, nor do they enable the government to toughen its insistence on performance. In most cases, government has little choice but to contract with internal suppliers, typically its own departments. If these fail to perform, the government can sack the chief executive and apply some pressure. But it rarely has the exit option that is essential to the effectiveness and enforcement of private contracts. * Chiefexecutives, senior managers, and others attribute most of the improvement in government performance to the discretion given to managers rather than to formal contracts. Managers differ on how much value is added by contracts, but few think that they have been the main contributor to higher operational efficiency. * Contracting is not costless. Negotiating and enforcing contracts entails enor- mous transaction costs that have not been systematically studied, although they take a deep bite out of operating budgets, especially those of small departments. These concerns point to the unfinished business of public sector reform in New Zealand. There is much more to be accomplished before a final assessment can be made. At this early stage, one is justified in acknowledging that the country has vasdy enlarged the stockpile of public management ideas and practices. In promot- ing internal markets within government, it has devised creative alternatives to privatization while carrying the pursuit of operational efficiency well beyond stan- dard market-type mechanisms such as user charges. Yet one should not lose sight of the fact that these are not real markets and that they do not operate with real contracts. Rather, the contracts are between public entities- the owner and the owned. The government has weak redress when its own organiza- tions fail to perform, and it may be subject to as much capture in negotiating and enforcing its contracts as it was under pre-reform management. My own sense is that while some gain may come from mimicking markets, anything less than the real thing denies government the fill benefits of vigorous competition and economic redress. 126 The World Bank Research Observer, vol. 13, no. I (February 1998) The Informal Public Sector In New Zealand, formal contracts and internal markets were feasible because the country had a robust market sector and established mechanisms for enforcing con- tracts-conditions that are often absent in developing countries, which tend to have an informal economy with relatively weak specification of property rights and other formal processes to regulate economic activity. Informality is not a new concept; shortly before New Zealand embarked on its reforms, de Soto (1989, p. 12) emphasized informality as the distinctive condition of the Peruvian economy. Characterizing it "as a gray area which has a long frontier with the legal world and in which individuals take refuge when the cost of obeying the law outweighs the benefits," de Soto found that the informal economy supplied more housing to Peruvians than did the government, was the main source of public transportation, and enabled entrepreneurs to start businesses when they were blocked by government regulations. He also concluded that the informal economy was inef- ficient, bred corruption, denied home and business owners access to capital, and retarded economic development. De Soto and others have focused on informality in the market economy. I believe that informality is as pronounced in the culture of government as it is in the market- place. In fact, the parallel incidence of informality in the public and private sectors is not happenstance. Norms, practices, and ideas migrate from one sector to the other, as does the dead hand of overregulation and the eagerness of government officials to look the other way in exchange for favors. The emergence of open, robust markets is as much a precondition for modernizing the public sector as it is for developing the private economy. It is highly unlikely that government will operate by the book when rules and regulations are routinely breached in private transactions. If New Zealand-style contracts are at one end of the spectrum, then informality is at the other end. And if contracts and the rule of law are underdeveloped in business rela- tions, it is highly improbable that they can be effectively applied in the conduct of the government's business. It would be foolhardy to entrust public managers with complete freedom over resources when they have not yet internalized the habit of spending public money according to prescribed rules. Many developing countries have formal management control systems that prescribe how government should operate. These systems are overseen by powerful central agencies such as the finance ministry, the civil service board, and the procurement agency. On paper everything is done according to rule. The civil service system is based on a detailed classification of positions and ranks, each with its own job descriptions, skill and experience re- quirements, eligibility rules, and pay scale. In this formal control process, operating units must obtain advance approval from the civil service agency (and sometimes from the finance ministry as well) before they can fill vacant positions. Formal rules dictate every step in the hiring process: announcing the position, establishing eligi- Allen Schick 127 bility qualifications, processing applications, appointing the winning candidate, and setting each employee's pay and grade levels. Each step is monitored by a central agency to assure compliance with the rules. Where informality flourishes, however, this is not the way many civil servants get their jobs. They are hired because they know the right person or have contributed to some organization or cause. Because official pay levels are low, they may be assigned to one position but be paid for another. Many may be ghost workers who appear on the payroll but not at work; some may hold two or more positions, and those who show up on the job may put in less than a day's work because the official salary scale is a lot less than a day's reasonable pay. Thus there are two coexisting civil service systems-one based on formal rules, the other on actual practices. To say that there is an informal system is not to conclude that the rules always are ignored or that corruption always flourishes, although these pathologies may occur. Rather, it is to argue that the informality contributes to public order; in the case of the civil service, it enables the government to recruit and retain skilled persons. Informality also reigns in the budget arena. The government has two budgets: the public one that is presented to the parliament and the real one that determines which bills are paid and how much is actually spent. The formal budget promises spending that exceeds the government's fiscal capacity; the informal budget facilitates macro- economic stability by not making some of the expenditures approved by the parlia- ment. The formal budget is known in advance, the informal one after the spending occurs. Because there are two budgets, the temptation is for the formal document to be unrealistic and unachieveable. The process thus feeds on itself. Inasmuch as the official budget will not be implemented, why not cram into it spending that will not be made? This behavior leads to cash flow budgeting, in which the amounts actually spent are determined more by cash payments than by the amounts authorized by law. It also breeds repetitive budgeting, in which the government "rebudgets" several times during the year to align disbursements and resources. Informality is a mixed blessing. On the one hand, it cuts through red tape, unre- sponsive bureaucracies, and bad policies; on the other hand, it opens the door to (and sometimes institutionalizes) corruption and inefficiency. The positive side of informality in public management includes the maintenance of fiscal discipline de- spite unrealistic budgets and the provision of public services despite rigid rules and controls. But the costs are high; they include widespread evasion of civil service rules and other controls, the time and resources spent in beating the system, distrust of government, routinized corruption, and inattention to the outputs and results of public programs and the performance of government agencies and officials. It would not be surprising if some of the most esteemed and productive civil servants in devel- oping countries are those who use their entrepreneurial and managerial skills to out- wit the formal controls. But when bureaucrats are valued for their verve in operating informally, it is easy for them and others trapped in the system to lose sight of the 128 The World Bank Research Observer, vol. 13, no. I (February 1998) public purposes they are serving and the outputs they are supposed to produce. It is only a short step from disabling the controls to bending the rules for dishonorable purposes. It is a much longer step for them to adopt New Zealand-style reforms, and a much riskier one. No country should move directly from an informal public sector to one in which managers are accorded enormous discretion to hire and spend as they see fit. New Zealand did not make this leap, and neither should other countries. Before reform New Zealand operated under budgets that controlled spending and corresponded to actual transactions; it also had a civil service system that governed how public employees were hired and paid. In other words, it had a formal public sector. This is an essential precondition for adopting elements of the New Zealand model. The Logic of Development If contract-based public management is beyond reach and informality is an unsatis- factory state of affairs, what can developing countries do to improve government operations? In my view, significant progress can be made through a logical sequence of steps that diminish the scope of informality while building managerial capacity, confidence, and experience. This concluding section outlines some of the key steps. First, progress in the public sector requires parallel advances in the market sector. As long as the economy operates according to informal norms and property rights are defined more by practice than by contract, the government is not likely to make much headway in installing rule-based public management. There may be special situations (in colonial regimes, for example) that enable a developing country to establish a skilled civil service system, modern financial management, and other trap- pings of formal public management even though the market sector is lagging be- hind. But the much more typical situation is one in which market development precedes or coincides with the development of robust public institutions. Singapore and Chile are countries in which economic development and modernization of pub- lic management have proceeded in tandem. Formalizing the market sector does not ensure reciprocal changes in public insti- tutions, however. Informality is as much a matter of culture as of practice; it defines social roles, relationships, and legitimate and expected behavior, and it persists even when the underlying conditions that gave rise to it vanish. There are quite a few countries in which the development of the public sector has not kept pace with economic changes. These countries typically have a competitive sector that is open, formal, and lightly regulated, as well as a heavily regulated sector that depends on informal contracts, embedded traditions, and government protection. The two cul- tures can operate independently of one another for an extended period, but sooner Allen Schick 129 or later they will be driven by scandal, financial mismanagement, or citizen pressure to modernize the public sector. Second, modernizing the public sector means establishing reliable external controls, as described above. As old-fashioned as external controls may seem to be, they are building blocks for a formal, rule-based, honest public sector. Operating in an exter- nally controlled environment is an essential phase in the development process. It gives managers the skills to manage on their own, builds trust between central controllers and line managers and confidence between citizens and government, and encourages managers to internalize a public ethic of proper behavior. As these basic conditions of formal management take root, it should be possible for central controllers to ease the regulations by giving line managers broader discretion in operating their programs. This process, however, can bear fruit only if the controls are exercised in a fair and realistic manner. In the case of civil service rules, this means that pay levels rise as the economy develops, the number of ghost positions declines, and public employees are given opportunities to acquire new skills and advance professionally. If these condi- tions are absent, learning will take place, but it will be pathological: how to beat the system, how to outmaneuver the controllers, how to get paid without really working, and so on. Realism must also pervade budgeting, another arena that often is infected by pervasive informality. The budget presented to the parliament must be one that can be implemented, not a political wish list that promises more than the govern- ment intends to spend. Moreover, agencies must inculcate the habits and ethic of spending according to the plans laid out in the budget. In other words, the budget must be treated as an implicit contract. Only then does it make sense to convert the budget into an explicit contract. Third, politicians and officials must concentrate on the basic process of public management. They must be able to control inputs before they are called upon to control outputs; they must be able to account for cash before they are asked to ac- count for cost; they must abide by uniform rules before they are authorized to make their own rules; they must operate in integrated, centralized departments before be- ing authorized to go it alone in autonomous agencies. Once the basics have been mastered, the public sector should be organized ac- cording to the principles of internal control. External control and New Zealand- type managerial discretion are not the only options for organizing governmental operations. Internal control is a third possibility. In a formal sense, internal control refers to the systems and procedures used by agencies to assure compliance with rules and to safeguard public assets. In a behavioral sense, internal control means that the rules are accepted as fair, workable, and legitimate. Without this normative under- pinning, no system of internal control can be effective. In practice, internal control gives managers broader discretion; it shifts the focus from ex ante control to ex post audit, from control of individual actions to control within a broad band, from reviewing specific actions to reviewing systems. It means, 130 The World Bank Research Observer, vol. 13, no. I (February 1998) for example, that civil service rules dictate the total number of positions or the total within broad employment categories and that operating departments make their own hiring decisions subject to oversight by central agencies. In the financial sphere it means that if funds are available, agencies can make purchases, authorize travel, and take other spending actions without obtaining prior approval. Singapore illustrates the progression from external to internal controls and thence to New Zealand-type arrangements. On gaining independence and for many years afterward, Singapore had a line-item budget that specified the positions to be filled and the items to be purchased. During the 1980s block budgets were adopted that shifted the government from external to internal control, and in the mid-1990s a "budgeting for results" system was adopted that implements several elements of the New Zealand model. Singapore provides another lesson for countries seeking to advance to the first rank of developing countries. The process of development does not have to stretch out over generations. If development proceeds in a logical order, progress can be rapid, especially if modernization of public institutions advances apace with mod- ernization of the market sector. To many developing countries, New Zealand is at the cutting edge in public management, but they will not get there by taking short- cuts that turn into dead ends. Notes Allen Schick is a consultant in the Public Sector Group, Poverty Reduction and Economic Manage- ment, of the World Bank. He wrote this article while on assignment to the Economic Development Institute. 1. The term is taken from Davis, Sullivan, and Yeatman (1997). See especially Matheson (1997) in that volume. References The word "processed" describes informally reproduced works that may not be commonly available through library systems. Davis, Glyn, Barbara Sullivan, and Anna Yeatman, eds. 1997. The New Contractualism? South Melbourne: Macmillan Education Australia PTY Ltd. de Soto, Hernando. 1989. The Other Path: The Invisible Revolution in the Third World. New York: Harper and Row. Matheson, Alex. 1997. "The Impact of Contracts on Public Management in New Zealand." In Davis, Sullivan, and Yeatman, eds. (1997), pp. 164-79. Schick, Allen. 1996. "The Spirit of Reform: Managing the New Zealand State Sector in a Time of Change." Report prepared for the State Services Commission, Wellington, New Zealand. Pro- cessed. Allen Schick 131 Deforestation and Forest Land Use: A Comment Jeffrey R. Vincent * Malcolm Gillis Hyde, Amacher, and Magrath (1996) imply that deforestation and timber rents (logging revenue minus logging costs other than timber fees) are not subjects that justify policymakers' attention, arguing that market responses limit the scope of de- forestation and that rents are usually small. But they fail to recognize that land markets will not develop efficiently, nor will efficient levels offorestry investments occur, when policy distortions and other factors obstruct the conversion of open-access forests to private or communal ownership. For these reasons rates of deforestation can be far above optimal levels. Contrary to the authors' claims, timber rents often (al- though not always) are large in developing countries. Moreover, the allocation of rents between loggers and the government owners ofpublic forests can indeed affect the profitability offorestry (and thus deforestation), the intensity of timber harvest- ing, and national welfare. "Deforestation and Forest Land Use" by Hyde, Amacher, and Magrath (1996) ad- dresses one of the most prominent environmental issues of the last two decades: global deforestation. We agree with many of the article's main points, in particular that "market responses . . . create limits to potential deforestation" (p. 242). As the authors note, this is a modest argument. Many economists have written that, in theory at least, one would expect forest areas first to decline and then to stabilize in response to changes in the prices of land, labor, capital, and forest products (see, among others, Clawson 1979; Berck 1979; Sedjo and Lyon 1990; Vincent 1992; and Vincent and Binkley 1992). The historical pattern of land use in the mature market economies of North America, Europe, and East Asia illustrates this process nicely, as does the emerging pattern of land use in some rapidly industrializing coun- tries such as Malaysia (Vincent and Hadi 1993). But because the story is an impor- tant one, it is worth retelling, especially with the fresh information contained in the section "The Empirical Evidence." The World Bank Research Observer, vol. 13, no. 1 (February 1998), pp. 133-40 C) 1998 The International Bank for Reconstruction and Development / THE WORLD BANK 133 The authors push their argument too far, however, in three instances: first, in downplaying the potential inefficiencies associated with deforestation; second, in stating that "forest rent is seldom large" (p. 240); and third, in minimizing the eco- nomic importance of rent allocation between public owners and private users. In so doing they may have led readers to conclude that policymakers in developing coun- tries and their advisers need not be concerned about deforestation or timber rents. In our view, such a conclusion is far off the mark. Deforestation The authors make several provocative assertions about deforestation. In particular, they claim that capitalized forest values net of "the costs of establishing and protect- ing permanent rights to the land" are "generally" negative (p. 226), and that the capitalized values of "most remaining tropical forests" are negative even if one ig- nores the costs of establishing property rights (p. 227). They offer no empirical evi- dence to support these assertions. This is a critical omission. If the assertions are true, then it follows that the degradation and conversion of open-access tropical forests to other uses do not generate significant efficiency losses. In effect, the authors assert away the essential economic concern about deforestation. They also implicitly assume that individuals, households, communities, and firms face few obstacles in converting land that does have a positive value in forestry uses from de jure or de facto open-access to private or communal ownership (pp. 225-7, 229). If that is true, then property rights and land markets might indeed emerge in an efficient manner over time, and deforestation might not be associated with sig- nificant inefficiencies. We agree that rural areas are much more marketized than is popularly supposed. But the efficiency of markets depends critically on the existence of well-defined, enforced property rights. In the case of plantation forestry invest- ments in Chile, cited more than once by the authors to illustrate the presumed insig- nificance of policies relative to market adjustments, markets did not work their magic until the government assured potential investors that forests would not be national- ized (Vincent and Binkley 1992). In many other developing countries, including the world's two leading tropical timber producers, Indonesia and Malaysia, the owner- ship of forests by parties other than the state is (or was until recently) prohibited. Without secure legal rights to forest land, economic agents are understandably reluctant to make the investments in forest management or forest plantations that are necessary for dynamically efficient forest-sector development. The authors dis- miss such problems as "delays," stating that "there is no empirical evidence of their economic importance" (p. 236). But they provide no empirical evidence to demon- strate that institutional obstacles and the "delays" they generate are not important. Indeed, on pp. 238-39 they provide several examples of policy distortions that prob- 134 The World Bank Research Observer, vol. 13, no. 1 (February 1998) ably do generate significant inefficiencies. Others are outlined in one of the articles they cite (Gillis 1988a). Surely governments ought to be concerned about removing obstacles to the devel- opment of efficient land markets (see Repetto and Gillis 1988, Vincent and Binkley 1992). The observation that deforestation might not "attain its physical limit" (p. 223) is not very reassuring from a social welfare standpoint: the cumulative amount and rate of deforestation might still be far above socially efficient levels. The Magnitude of Timber Rents Timber rent is the return over and above the factor costs of logging: the difference between total revenue from timber sales and total costs of harvesting and delivery (including a normal profit margin, but excluding the fees charged by the govern- ment in the form of taxes, licenses, and so forth). In light of evidence from the literature and from the field, the authors' claim that "forest rent is seldom large" (p. 240) is astonishing. Many studies conducted since the 1 970s have documented the existence of large rents in the virgin or semi-virgin (logged over, but with residual virgin trees) forests that have been the locus of most commercial logging in develop- ing countries (Page, Pearson, and Leland 1976; Ruzicka 1979; Gillis 1980; Repetto and Gillis 1988; Vincent 1990; Sizer and Rice 1995; Sizer 1996). The authors cite almost none of this literature. In fact, they appear convinced that such rents simply cannot exist, commenting that "Surely all private operators would have had suffi- cient incentive to harvest each year all the way to the geographic point where their access and harvest costs depleted the entire value of the standing resource" (p. 240). There are several reasons why large rents can, and indeed do, exist in many tropi- cal forests. The first comes from basic theory. An old-growth timber stock is like a deposit of a nonrenewable resource. When a nonrenewable resource is depleted at an optimal rate, its marginal rent (the rent on the last unit extracted in a given period) rises over time at a rate equal to the discount rate (the return on an alternative invest- ment). This reflects the rising cost of depletion as the resource becomes more scarce, and it provides the fundamental economic reason why a rational user concerned with future as well as present returns will not harvest "all the way to the geographic point ...." The only difference in the case of forestry is that the rate of increase in the marginal rent (which forest economists call stumpage values) diminishes as log- ging proceeds and more vigorously growing, second-growth forests, which offer a biological rate of return because of growth, replace old-growth forests, which have no net timber growth. Contrary to the authors' claims, long-run data on stumpage values do exist (for instance, see the sources in Binkley and Vincent 1988 and Nordhaus 1992), and they are broadly consistent with this theoretical story. For example, softwood stumpage Jeffrey R. Vincent and Malcolm Gillis 135 values in the U.S. South rose in real terms by 4.6 percent a year from 191 0 to 1943 and 3.1 percent a year from 1943 to 1985; they are predicted to rise at around 2 percent a year into the early 2000s (Binkley and Vincent 1988). Should one be surprised that rents for tropical timber, whose products compete with ones made from temperate timber, are high today after a century of steadily rising values for temperate timber? This theoretical explanation for large tropical timber rents is not entirely convinc- ing, however. The absence of secure property rights in many developing countries implies that loggers do not make socially efficient decisions: in comparing the re- turns to current and future harvests, they place too little weight on the latter. In the extreme, if loggers completely ignore depletion costs, then harvesting will indeed proceed "all the way to the geographic point. . ." where marginal rent equals zero. If that is the case, where do large rents come from? The first potential source is from timber within the margin: timber in forests with easy terrain and near roads, mills, and ports. Indeed, this "inframarginal" rent was the major source of the large timber rents in Malaysia estimated by Vincent (1990). The possibility of large rents within the margin coexisting with rents equaling zero at the margin has been known since the time of classical economists such as Ricardo and von Thunen. The second potential source of large rents is from government regulation of log- ging activity. Logging in Southeast Asia, the world's major source of tropical hard- wood logs and products, is not completely uncontrolled. State and federal govern- ments in Malaysia and Indonesia control access and utilization of the forest, although not perfectly, by issuing harvesting licenses, and by setting cutting limits at some minimum diameter. Such regulations prevent loggers from harvesting "all the way to the geographic point . . ." and result in positive rents at the harvesting margin. Third, the expansion of infrastructure (logging roads, ports, and so forth), the development of lower-cost logging technologies, and increases in the number of species and sizes of roundwood accepted by markets have combined to increase rents over time. Such changes, which are typically not fully anticipated by timber markets, make rents available for loggers, concessionaires, and their political pa- trons. The authors downplay these changes as "specialized cases" of a return to "a relaxation of a restriction on forestry activity" (p. 240). Such changes do indeed relax restrictions on forestry activity, but they are far from occasional curiosities. Indeed, they have been the driving forces of the expansion of the tropical timber industry since the end of World War II.L Today, they are causing a historic shift of the industry from Southeast Asia to Latin America. It is unlikely that timber bar- ons in Kuching and Jakarta will abandon their pursuit of concessions in Surinam, Guyana, and the Amazon after reading in Hyde, Amacher, and Magrath that tropical timber rents are no more than "relaxations of restrictions." Paraphrasing Gertrude Stein, a rent is a rent is a rent. Nor will Malaysian logging companies stop bidding 136 The World Bank Research Observer, vol. 13, no. I (February 1998) as much as 40,000 ringgit a hectare ($16,000) for harvesting rights in peninsular Malaysia (Awang Noor 1996). In sum, there is no mystery why large rents are both theoretically possible and do in fact exist in many tropical timber-producing countries. Some of the evidence the authors present to deny this situation simply does not withstand scrutiny. For ex- ample, they imply that rents were low in the Philippines in the 1980s by juxtaposing the example of the Philippines Bureau of Forestry Development, which lost money on its timber sales in the early 1980s (revenues from timber fees were less than the bureau's expenses), with that of below-cost timber sales in the United States. But timber fees were low in the Philippines not because rents were low, but because the bureau captured less than 10 percent of the available rents. Low rent capture, not low rent, was in fact the main point of the article by Boada (1988) that they cite. The authors draw the forestry rent gradient in their figure 1 so that it "shows that" (p. 226) the value of forest land is low, but they could equally well have drawn it higher. The articles they cite lead us to speculate that the developing countries they had in mind in drawing this figure are ones whose commercial forests are a small proportion of total forests, such as China, India, Malawi, Nepal, Pakistan, and the Philippines. It is not surprising that average forest rents are low in such countries, which, because of ecological factors (low rainfall, say) or historical reasons (when considerable depletion has already occurred), do not have heavily stocked forests. But one should be wary of extrapolating from experience in such countries to the entire developing world, just as one should be wary of extrapolating from deserts to the global value of commercial agricultural land. Allocation of Timber Rents By contrasting two of our papers (Gillis 1988a; Vincent 1990) against two others (Paris and Ruzicka 1991; Hyde and Sedjo 1992), Hyde, Amacher, and Magrath imply that we have argued that timber rents in public forests in developing countries are always large and always belong to the government, and that the assignment of these rents always has an efficiency impact. This is not an accurate representation of our work. Regarding the first claim, in Gillis (1988a) and Vincent (1990), we em- phasized that rent can vary substantially from stand to stand. Gillis (1988a; see also Gillis 1980) dealt explicitly with variations in rent within and between stands. As noted earlier, the model in Vincent (1990) was based on the existence of rent varia- tion within the margin of utilization. Nor have we claimed that large rents in virgin forests imply that rents in second-growth forests, which are what matter for long-run land-use decisions, are also large. Vincent (1990) examined in great detail how the present value of net returns from second-growth harvests varies according to differ- ences in prices, costs, rotation lengths, growth rates, and discount rates. Jeffrey R. Vincent and Malcolm Gillis 137 As for the second and third claims, we have never stated that the complete capture of resource rent by government agencies is necessary for efficient land-use allocation or efficient levels of forest management investments. Vincent (1993) cited quota- tions from our work emphasizing that many aspects of rent capture involve primarily distributional issues. We, and others before us (Page, Pearson, and Leland 1976; Schmithuisen 1976; Ruzicka 1979), have argued that the fees that forestry agencies levy to capture timber rents can affect marginal harvesting decisions and result in inefficient harvest levels if they are not set appropriately. This follows from basic concepts of resource taxation. Hyde and Sedjo (1992) accepted the validity of this point, although as noted in Vincent (1993), they misinterpreted much of the analysis in Gillis (1980, 1988a) and Vincent (1990). Hyde, Amacher, and Magrath ignore this and several other potential links be- tween government rent capture and efficiency within the forest sector. For example, Page, Pearson, and Leland (1976) pointed out that a government's failure to set timber fees at levels comparable to competitive stumpage values can destroy rents by artificially maintaining the viability of inefficient logging firms. (Gillis 1988a showed how industrial policies-high levels of protection to Indonesian plywood mills- destroyed rents by promoting inefficient local processing.) That timber fees can af- fect land-use decisions is evident from figure 1 in Hyde, Amacher, and Magrath: timber fees reduce the private return to forestry, shift the forest rent gradient down- ward, and thus reduce equilibrium forest area. Whether this effect is good or bad depends on the relative social returns to different land uses. The allocation of rents from logging also has important implications beyond sectoral efficiency. One would expect national welfare to be enhanced if rents are retained within the country instead of being repatriated by transnational logging companies. Concerns about retained value triggered early studies on rent capture in the tropical timber sector, such as those by Page, Pearson, and Leland (1976), Ruzicka (1979), and Gillis (1980). The authors cite none of those works. We seriously doubt that Papua New Guinea and the Latin American countries that are the new frontiers for tropical logging will be better off if they grant harvesting rights to Asian logging companies for a song. The authors overlook a second link to national welfare: the potential efficiency gains that result from the state's ability to capture a higher proportion of timber rents and to reduce existing taxes that are more distortionary. If the authors doubt the significance of such measures, they should consider how low rates of rent capture in natural resource sectors have undermined economic recovery in Russia by forcing the state to impose crushing taxes on income, commercial transactions, and interna- tional trade. Indonesia and Malaysia, at the other extreme, have been able to keep general tax rates low because they generate revenue from petroleum (though not timber) rents. 138 The World Bank Research Observer, vol. 13, no. I (February 1998) The authors claim (note 4) that analyses of the impacts of general economic poli- cies upon the forest sector are scant. In fact, several such analyses have long been available. The book Public Policies and the Misuse of Forest Resources (Repetto and Gillis 1988) not only highlighted failures in forestry policies, but it also discussed the effects of macroeconomic, trade, monetary, and agricultural policies on the forest sector. Indeed, Gillis (1988a), which Hyde, Amacher, and Magrath cite, stressed these effects. Another chapter in the same volume depicted these effects in some detail for Ghana and several other African countries (Gillis 1988b). Xie, Vincent, and Panayotou (1997) cited nearly 20 computable general equilibrium modeling studies showing the impacts of general economic policies on the forest sector. In conclusion, it will be unfortunate if readers of the Research Observer conclude from Hyde, Amacher, and Magrath that deforestation and timber rents are not wor- thy of policymakers' attention. Both are associated with policy issues whose resolu- tion could have important impacts on the efficiency of land use, forest manage- ments, and economic growth in developing countries. Notes Jeffrey R. Vincent is a fellow at the Harvard Institute for International Development. Malcolm Gillis is president of Rice University. 1. And such changes, not timber growth (p. 225), are the main causes of pulse-harvesting in the tropics. References The word "processed' describes informally reproduced works that may not be commonly available through library systems. Awang Noor Abd Ghani. Faculty of Forestry, Universiti Pertanian Malaysia. Personal communica- tion, December 17, 1996. Berck, Peter. 1979. "The Economics of Timber: A Renewable Resource in the Long Run." Bell Journal of Economics 10(2):447-62. Binkley, Clark S., and Jeffrey R. Vincent. 1988. "Timber Prices in the U.S. South: Past Trends and Outlook for the Future." Southern Journal ofApplied Forestry 12(1):15-18. Boada, Eufresina L. 1988. "Incentive Policies and Forest Use in the Philippines." In Repetto and Gillis (1988). Clawson, Marian. 1979. "Forestry in the Long Sweep of American History." Science 204:1168-74. Gillis, Malcolm. 1980. "Fiscal and Financial Issues in Tropical Hardwood Concessions." Develop- ment Discussion Paper 110. Harvard Institute for International Development, Cambridge, Mass. Processed. . 1988a. "Indonesia: Public Policies, Resource Management, and the Tropical Forest." In Repetto and Gillis (1988). Jeffrey R. Vincent and Malcolm Gillis 139 . 1988b. "West Africa: Resource Management Policies and the Tropical Forest." In Repetto and Gillis (1988). Hyde, William F., Gregory S. Amacher, and William Magrath. 1996. "Deforestation and Forest Land Use: Theory, Evidence, and Policy Implications." The World Bank Research Observer 11(2):223-48. Hyde, William F., and Roger A. Sedjo. 1992. "Managing Tropical Forests: Reflections on the Rent Distribution Discussion." Land Economics 68(3):343-50. Nordhaus, William D. 1992. "Lethal Model 2: The Limits to Growth Revisited." Brookings Papers on Economic Activity 2(1): 1-43. Page, John M. Jr., Scott R. Pearson, and H. E. Leland. 1976. "Capturing Economic Rent from Ghanaian Timber." Food Research Institute Studies 15(1):25-51. Paris, Remy, and Ivan Ruzicka. 1991. "Barking Up the Wrong Tree: The Role of Rent Appropria- tion in Sustainable Tropical Forest Management." Environment Office Occasional Paper 1. Asian Development Bank, Manila. Processed. Repetto, Robert, and Malcolm Gillis, eds. 1988. Public Policy and the Misuse of Forest Resources. Cambridge, U.K.: Cambridge University Press. Ruzicka, Ivan. 1979. "Rent Appropriation in Indonesian Logging: East Kalimantan, 1972/73-19761 77." Bulletin of lndonesian Economic Studies 15(2):45-74. Schmithiusen, Franz. 1976. "Forest Utilization Contracts on Public Land in the Tropics." Unasylva 28(1):52-73. Sedjo, Roger A., and Kenneth S. Lyon. 1990. The Long-Term Adequacy of World Timber Supply. Washington, D.C.: Resources for the Future. Sizer, Nigel. 1996. Profit without Plunder: Reaping Revenue from Guyana's Tropical Forests without Destroying Them. Washington, D.C.: World Resources Institute. Sizer, Nigel, and Richard Rice. 1995. Backs to the Wall in Suriname: Forest Policy in a Country in Crisis. Washington, D.C.: World Resources Institute. Vincent, Jeffrey R. 1990. "Rent Capture and the Feasibility of Tropical Forest Management." Land Economics 66(2):212-23. .1992. "The Tropical Timber Trade and Sustainable Development." Science256(5064): 1651- 55. - . 1993. "Managing Tropical Forests: Comment." Land Economics 69(3):313-18. Vincent, Jeffrey R., and Clark S. Binkley. 1992. "Forest-Based Industrialization: A Dynamic Per- spective." In Narendra P. Sharma, ed., Managing the World's Forests. Dubuque, Iowa: Kendall/ Hunt. Vincent, Jeffrey R, and Yusuf Hadi. 1993. "Malaysia." In National Research Council, Sustainable Agriculture and the Environment in the Humid Tropics. Washington, D.C.: National Academy Press. Xie, Jian, Jeffrey R. Vincent, and Theodore Panayotou. 1997. "Computable General Equilibrium Models and the Analysis of Policy Spillovers in the Forest Sector." Environment Discussion Paper 19. Harvard Institute for International Development, Cambridge, Mass. Processed. 140 The World Bank Research Observer, vol. 13, no. 1 (February 1998) Deforestation and Forest Land Use: A Reply William F. Hyde In our recent article (Hyde, Amacher, and Magrath 1996), we argued that the "topic of deforestation is seldom examined from the perspective of prices and responses to resource scarcity" (p. 223). Our point was not that the market is perfect but rather that market responses limit the extent of deforestation-even in regions of attenu- ated property rights or open-access forest resources. The market sets this limit be- cause as deforestation proceeds, the increasing scarcity of the resource induces rising prices, which in turn induce private investments to satisfy both commercial and subsistence household uses of forest resources. As a result, deforestation proceeds to the geographic points where the harvest and access costs associated with removing the remaining natural stock equal the costs of reforestation on land that has estab- lished and enforceable property rights. This means that a considerable area of the world's natural forest is not endangered. Therefore, we argued that concern for the total forest resource and its protection is misguided. It would be better to concentrate on the price increases induced by deforestation and the costs of private reforestation with an eye to identifying the level of prices and costs that will alter the boundaries of several important classes of forest land, and on concentrations of nonmarket, forest-based environmental ser- vices and public policy failures that distort market incentives. Vincent and Gillis, in their article in this issue, feel that we err on three counts: deforestation, the magnitude of timber rent, and rent allocation. I will address these in order. Deforestation Vincent-Gillis suggest that we diminish the inefficiencies associated with deforesta- tion. The source of our difference is their focus on our "Simple Model" of market The World Bank Research Observer, vol. 13, no. I (February 1998), pp. 141-45 ( 1998 The International Bank for Reconscruction and Development / THE WORLD BANK 1 41 forest values (pp. 225-30). We qualified the model with a more extended discussion of nonmarket factors (pp. 234-36) and policy failures (pp. 236-42), Vincent-Gillis are concerned with our argument that net market forest values are generally negative. They feel that we deny the essential economic concern with de- forestation. But surely some forest lies beyond the reach of any active harvest activ- ity. Some even lies beyond the reach of open-access activity. Surely, this is land with a negative net value in the local market. A recent report from the World Resources Institute (Bryant, Nielson, and Tangley 1997) suggests that 40 percent of the world's forests lie beyond the reach of market activity. Table 1 (p. 230) reports estimates that are at least this large. In addition, there is a region of the world's forest where previ- ous harvests of the natural stock have left standing trees that are either too immature or too sparse to have positive net value. In most time periods, this land also has a negative net market value. Vincent himself was the first to my knowledge to use the term "pulse" harvests (Vincent and Binkley 1992) to identify this region. These observations, however, do not deny the essential economic concern about deforestation. My own impression is that this concern has to do with nonmarket values (global change, biodiversity, erosion, or watershed protection), especially those attached to the remaining stock of natural forest. In examining these values, we argued that global concern for their protection should be concentrated in those higher market-value forests where nonmarket values are at risk. We find this approach preferable to wasting environmental policy on regions that are not accessible in any case and where the nonmarket environmental values are protected without policy intervention. Vincent-Gillis state that we "implicitly assume that individuals and firms face few obstacles in converting (open-access) land" to private property (p. 134). But we ex- plicitly discussed the problems of establishing secure tenure and cited some of the recent literature on pp. 238-39. Vincent-Gillis's arguments and ours are similar. Contrary to Vincent-Gillis's claim, we did not dismiss the problems of establish- ing secure tenure as a market delay. The "delay" they speak of (p. 236) specifically referred to our simple model and the response by investors to rising market prices. We observed that there is no empirical evidence on the magnitude of this particular market lag. This does not mean that the lag does not exist, or that it is the only identifiable market failure, but only that the lag and its effects have not been mea- sured successfully. As for institutional obstacles to forest investment, we discussed macroeconomic policy (p. 237), legal institutions (p. 238), spillovers from agricul- tural policy (p. 238), and public forestry agencies (pp. 239-40). In each case, we illustrated the problems these pose for private forestry activities, and either the in- centives created for deforestation or the obstacles for reforestation. We also discussed publicly funded research, technology transfers, and research to enhance forest pro- duction and to identify improved institutional arrangements (pp. 241-42). Con- trary to Vincent-Gillis's claim, we never presume the "insignificance of policies rela- 142 The World Bank Research Observer, vol. 13, no. I (February 1998) tive to market adjustments" (p. 134). I would say that we stressed the importance of these policies in modifying market adjustments and that our conclusion specifically encouraged the assessment of their impacts on the boundaries of the various classes of forest land. The Magnitude of Forest Rent Vincent-Gillis are concerned that we underestimate the importance of timber rents. Because our article was about deforestation and because the concern with deforesta- tion is comprehensive, our context is the entire global forest, while Vincent-Gillis's context is with the share of mature natural forests where commercial timber harvests are important. Bryant, Nielson, and Tangley (1997) suggest that the share of mature natural forest that is subject to commercial timber harvest and that may have a positive timber rent is 28 percent of the world's total forest. We agreed that concerns with timber rent are valid in this region (p. 238), citing the primary literature as well as examples from our own experience to illustrate that these rents can be large (p. 239). These examples raise questions about the prospective occurrence of large rents in other cases. We think that the important questions have to do with what caused the large rents to arise and whether these sources of rent are instructive for policy. The mature natural forests that are the sources of commercial harvests and the common sources of timber rents are generally at or near the margins of previous economic harvest activity-otherwise they would have been harvested earlier. Gradual long-term price increases do give rise to rents at these locations, but they are unlikely to induce large rents before harvesting extends into these regions. Large rents must be caused by sharp market changes or by something that constrained gradual harvest expansion. The former are uncommon. The two obvious candidates for the latter are a change in infrastructure and successful enforcement of some government logging restriction, as Vincent-Gillis point out. Most of our discussion of rent was devoted to the same two sources, particularly changes in access (for example, new roads) and the particular biological formulation of the "allowable cut" harvest policy that is common in one form or another to most government forestry agencies. These observations still "beg questions about the management activities or poli- cies that created the rent. Does the same policy exist elsewhere? Could more rents be captured by applying the same management activity or the same change in forest policy to other timberstands ... ?" (p. 240) And, in our context, what would be the impacts on deforestation and on incentives to reforest? Finally, and contrary to Vincent-Gillis, we did not deny the existence of long-run stumpage price series. Rather we noted that these data are "not reliable" for purposes of prediction (p. 23 1). They are unreliable because stumpage value is the residual left William F Hyde 143 after all harvest, transport, and processing costs are deducted. This residual varies greatly, even within a region and often within a harvest site, according to differences in wood quality and location, and on-site harvest difficulties. Therefore, stumpage values from one time or place are not good predictors of stumpage values at another time or place, and composite stumpage price series from one time are not good predictors of composite stumpage prices in another time. This is one reason why Barnett and Morse (1963) used lumber prices, not stumpage prices, in their classic analysis of intertemporal trends in resource prices in the United States.' Rent Allocation Vincent-Gillis believe that we minimize the importance of rent allocation, and that we imply they believe three things are always true. We would never say someone else "always" believes anything. To clarify any possible misunderstanding I would restate our sentence (p. 239) to read as follows: "Vincent-Gillis have emphasized cases of large timber rent, and they have been concerned with claiming a larger share of this rent for the government. Vincent has also argued that rent allocation has an effect on efficient output levels." I think the articles Vincent-Gillis cite in their comment sup- port this revised statement. Nevertheless, the restatement does not alter our point that all discussions of timber rent should lead to questions about the further occur- rences of these rents, the policies that give rise to them, and the application of similar policies in other timberstands and forests. As for the merit of rent capture by public agencies, the difference between Vincent- Gillis's position and ours is that we believe that allocation is a distributive matter of undetermined merit. That is, it is an empirical question that must be judged by the performance of the government agency or forest concessionaires receiving the rent in each particular case. We have a preference for the most socially productive use of the rent-rather than for a favored agent.2 Of course, policy can modify and even destroy rents, as Vincent-Gillis argue and we agree. Administrative behavior can also modify or destroy rent. We all know of government agencies as well as private operations that dissipate rents with poor deci- sions, just as we know of those that collect and reinvest rents in ways that improve the general welfare.3 This is what makes the preferred allocation a distributive issue whose resolution depends on the performance of the agents involved. In conclusion, our concern was with global deforestation, whether due to com- mercial forestry, agricultural land conversion, or subsistence household uses of the forest. Our primary argument was that broad-based forest policy interventions designed as general protection against world deforestation are not well advised because specific markets for the various products and services of the forest already exist. These specialized markets divide forests into at least three distinct land use 144 The World Bank Research Observer, vol. 13, no. I (February 1998) classes: sustainable private forests and forested commons; unsustainable open- access forests near the frontier; and an unused residual forest. Forest policy will affect each of these land categories differently. Therefore, it would be better to address the market and policy failures that shift the boundaries of one or another of these classes of land use, and to anticipate that most policies will selectively affect some, but not all, of these boundaries. I think our original article was clear on these points. Notes William F. Hyde is professor in the Forestry Department at Virginia Polytechnic and State Univer- sity, Blacksburg, Virginia. 1. Smith (1979) raised doubt about the statistical reliability of many primary resource prices for prediction. Jung, Krutilla, and Boyd (forthcoming) demonstrate the unreliability of composite stump- age price series. 2. Contrary to Vincent-Gillis, allocation does not alter the efficient level of production in our figure 1. Also contrary to Vincent-Gillis, rents are not low because of the way we drew the rent gradient. The forest rent gradient almost always starts below the agricultural land value gradient and extends to its right. Forest rents are low because forests are lower-valued uses of the land. A steeper gradient in our figure 1 would not alter these basic points. 3. Repetto and Gillis (1988), who are cited by Vincent-Gillis, give many examples of government actions that dissipated timber rents or government agencies that did not have the administrative ability to capture some rents. These examples would seem to urge caution on any view that govern- ment is a generally preferred collector of rents. References Barnett, Harold J., and Chandler Morse. 1963. Scarcity and Growth: The Economics of Natural Resource Availability. Baltimore, Md.: Johns Hopkins University Press. Bryant, Dirk, Daniel Nielson, and Laura Tangley. 1997. The Last Frontier Forests: Ecosystems and Economies on the Edge. Washington, D.C.: World Resources Institute. Hyde, William F., Gregory S. Amacher, and William Magrath. 1996. "Deforestation and Forest Land Use: Theory, Evidence, and Policy Implications." The World Bank Research Observer 11(2):223-48. Jung, Chulho, Kerry Krutilla, and Roy Boyd. Forthcoming. "Aggregation Bias in Natural Resource Price Composites: The Forestry Case." Energy and Resource Economics. Repetto, Robert, and Malcolm Gillis, eds. 1988. Public Policy and the Misuse of Forest Resources. Cambridge, U.K.: Cambridge University Press. Smith, V. Kerry. 1979. "Natural Resource Scarcity: A Statistical Analysis." Review of Economics and Statistics 61:423-27. Vincent, Jeffrey R., and Clark S. Binkley. 1992. "Forest-Based Industrialization: A Dynamic Perspec- tive." In Narendra P. Sharma, ed., Managing the World's Forests. Dubuque, Iowa: Kendall-Hunt. Vincent, Jeffrey R., and Malcolm Gillis. 1997. "Deforestation and Forest Land Use: Comment." The World Bank Research Observer 13(1):133-40. William F. Hyde 145 Coming in March . .. Global Development Finance 1998 .~~~ n::. S .Ei SS-.; In GlobalDevelopment Finance, you'll find infor- mation on external debt and financial flows data for 150 countries, including the 138 that report to the .~~~~ '~~~~ ~World Bank's Debtor Reporting System. Global .X..~ . '.........................,. Development Finance is a must for economists, ....0 | Z a and capital flows worldwide. The 1998 edition of GlobalDevelopment . .. .. . . 0 ; S: Q :Finance continues to be the World Bank's most authoritative and up-to-date review of developing countries' external debt and financial flows, as it includes a special focus on East Asia's current economic crises. The main issues raised are the causes and implica- tions of the crisis, its evolution, and its prospects. March 1998.Analysis and Summary Tables. Stock no. 14117 (ISBN 0-8213-4117-0) $40.00. Analysis and Summary Tables and Country Tables (Country Tables not sold separately). Stock no. 14118 (ISBN 0-8213-4118-9) $300.00 GlobalDevelopment Finance is also available on CD-ROM and diskette. View, select, and extract the data you need through the World Bank's Socioeconomic Time-series Access and Retrieval System (*STARS*). Data may be exported to other software programs, including Lotus 1-2-3 M, or to ASCII. CD-ROM individual version: Stock no. 14121 (ISBN 0-8213-4121-9) $300.00. Network version: Stock no. 14122 (ISBN 0-8213-4122-7) $825.00. Diskette individual version: Stock no. 14119 (ISBN 0-8213-4119-7) $200.00. Networkversion: Stock no. 14120 (ISBN 0-8213-4120-0) $545.00. World Bank Publications For US customers, contact The World Bank, P.O. Box 7247-8619, Philadelphia, PA 19170-8619. Phone: (703) 661-1580, Fax: (703) 661-1501. Shipping and handling: US$5.00. For airnail delivery outside the US, charges are US$13.00 for one item plus US$6.00 for each additional item. Payment by US$ check drawn on a US bank payable to the World Bank or by VISA, MasterCard, or American Express. Customers outside the US may also contact their local World Bank distributor. 1300 A...this may be the only source needed for international social and economic data." Choice lagazine World Development Indicators 1998 The second annual edition of the World Bank's flagship statistical reference-WorldDevelopment Indicators 1998. This award-winning publication provides an expanded view of the world econo- my for 148 countries-with chapters focusing on people, economy, environment, states and markets, world view, and global links as well as introductions highlighting recent research on major development issues. The 1998 edition includes some key indicators for 1997. An award-winning series! Included in Choice Magazine's 34th annual Outstanding Academic Book list! April 1998 370 pages Stock no. 14124 (ISBN 0-8213-4124-3) $60.00 Also Available on CD-ROM This comprehensive database contains underlying time-series data for the World Development W Indicators and World Bank Atlas, now covering 1965-1996 for most indicators with some extending to 1997. Powerful features allow you to generate maps and charts and download your results to other software programs. Requires Windows 3.1 ._ April 1998 Individual Version: Stock no. 14125 (ISBN 0-8213-4125-1) $275.00 Network Version: Stock no. 14126 (ISBN 0-8213-4126-X) $550.00 World Bank Atlas 1998 One of the Bank's most popular offerings, theAtlas is designed as a companion to the World Development Indicators. Tables, charts, and colorful maps address the development themes of people, economy, environment, states and markets, world view, and globaL links. This easy- to-use, book is an intemational standard in statistical compilations and an ideal reference for office or classroom. Text, maps, and references appear in English, French, and Spanish. April 1998 64 pages Stock no. 14127 (ISBN 0-8213-4127-8) $20.00 w World Bank Publications For US customers, contact The World Bank, P.o. Box 7247-8619, Philadelphia, PA 19170-8619. Phone: (703) 661-1580, Fax: (703) 661-1501. Shipping and handling: US$5.00. For airmail delivery outside the US, charges are US$13.00 for one item plus US$6.00 for each additional item. Payment by US$ check drawn on a US bank payable to the World Bank or by VISA, MasterCard, or American Express. Customers outside the US may also contact their local World Bank distributor. 1300 DISTRIBUTORS CHINA FINLAND IRAN China Financial & Economic Publishing Akateeminen Kirakauppa Ketab Sara Co. Publishers OF W ORLD House PO. Box 128 Khaled Eslamboli Ave., 8, Da Fo Si Dong Jie FIN-00101 Helsinki 6th Street BANK Beijing Tel: (358 0) 121-4418 Delafrooz Alley No.8 PUJBLIC..ATION5 S Tel: (8610) 6333-8257 Fax: (3580) 121-4435 PO. Box 15745-733 Fax: (86 10) 6401-7365 E-mail: akablaus@stockmann.ti Tehran 15117 Prices and credit terms vary URL: htip:llwww .akaleeminen.com/ Tel: (98 21) 8717819; 8716104 frotm country to country. China Book Import Centre Fax: (98 21) 8712479 Consult your local distributor Ms. Zhou Rui FRANCE E-mail: ketab-sara@neda.net.ir before placing an order. PO. Box 2825 Woded Bank Publications Beijing 66, avenue dI16na Kowkab Publishers ARGENTINA 75116 Pads PO. Box 19575-511 Oficina del Libro Intemacional COLOMBIA Tel: (331) 40-69-30-56/57 Tehran 15117 Av. Cordoba 1877 Intoenlace Ltda. Fax: (331) 40-69-30-68 Tel: (98 21) 258-3723 1120 Buenos Aires Carrera 6 No. 51-21 Fax: (98 21) 258-3723 Tel: (54 1) 815-8354 Apanado Aereo 34270 GERMANY Fax: (54 1) 815-8156 SanlatI6 de Bogol:6, D.C. UNO-Vedag IRELAND E-mail: olilibro@satlink.com Tel: (57 1) 285-2798 Poppelsdorler Allee 5 Government Supplies Agency Fax: (57 1) 285-2798 53115 Bonn Oifig an tSoldthair AUSTRALIA, FIJI, PAPUA NEW Tel: (49 228) 949020 4-5 Harcour Road GUINEA, SOLOMON ISLANDS, COTE D'IVOIRE Fax (49228)217492 Tel 2 VANUATU, AND WESTERN SAMOA Center d'Edition et de Diffusion URL. htp1/Wwwwuno-veriag.de Tel: (353 1) 661-3111 D.A. Informaion Services Aticalnes (CEDA) E-mail: unovedag@aol.com Fax:(3531(475-2670 648 Whitehorse Road 04 B.R 541 GHANA Mitcham 3132 Abidjan EppBoksSeHiesINAA Victoria Tel: (225) 246510;24 6511 EppooksServicesYozmotLiteratureLtd. Victoria ~ ~ ~ ~ ~ Fx (2) 5057Mr. Gibrine Adam Y0.zBox 56055ue td Tel: (61)39210 7777 FPO (225(250567 P. Box 44 3 Yoa S6055 Fax: (61) 39210 7788 P.Bx4 Fax: (61 ) 3 9210 7788 ~~~~~~~~~~~TUC 3 Yohanan Hasandlar Street E-mail: service@dadirect.com.au CYPRUS Tc Tel Aviv 61560 URL: http:/Awww .dadirect.com.au Center for Applied Research Accra Tel: (972 3) 5285-397 Cyprus College GREECE Fax: (972 3) 5285-397 AUSTRIA 6, Diogenes Street, Engomi, PO. Box Papasotinou SA. Gerold and Co. 2006 35, Stournara Str. R.O.Y Intemational Weihburggasse 26 Nicosia 106 82 Athens PO Box 13056 A-1011 Wien Tel: (3572)44-1730 Tel: (301)364-1826 TelAviv 61130 Tel: (431) 512-47-31-0 Fax: (357 2) 46-2051 Fax: (301) 364-8254 Tel: (972 3) 5461423 Fax: (43 1) 512-47-31-29 Fax: (972 3) 5461442 URL: httpiAww .gerold.oo/at.online CZECH REPUBLIC HAiTI E-mail: royil@netvision.net.il National Information Center Culture Diffusion BANGLADESH prodejna, Konviktska 5 5, Rue Capois, C.R 257 Palestinian Authorty/Middle East Micro Industries Development CS -113 57 Prague 1 Pon-au-Pnnce Index Information Services Assistance Society (MIDAS) Tel: (42 2) 2422-9433 Tel: (509) 23 9260 PO.B. 19502 Jerusalem House 5, Road 16 Fax: (422) 2422-1484 Fax: (509) 234858 Tel: (9722)6271219 Dhanmondi R/Area URL: hltpi/www .nis.cz/ Fax: (972 2) 6271634 Dhaka 1209 HONG KONG, MACAO Tel: (880 2) 326427 DENMARK Asia 2000 Ltd. ITALY Fax: (880 2) 811188 SamfundsLitteratur Sales & Circulation Department Licosa Commissionana Sansoni SPA RosenoemsAlle 11 Seabird House, unit 1101-02 Via Duca Di Calabria, 111 BELGIUM DK-1970 Frederiksberg C 22-28 Wyndham Street, Central Casella Postale 552 Jean De Lannoy Tel: (45 31) 351942 Hong Kong 50125 Firenze Av. du Roi 202 Fax: (45 31) 357822 Tel: (852)2530-1409 Tel: (55) 645-415 1060 Brussels URL: http:llwww.sl.cbs.dk Fax: (852) 2526-1107 Fax: (55) 641-257 Tel: (32 2) 538-5169 E-mail: sales@asia2000.com.hk E-mail: licosa@ftbcc.it Fax: (32 2) 538-0841 ECUADOR URL: httpJ/www .asia2000.com.hk URL: http:l/www .ftbcc.it/licosa Libr Mundi BRAZIL Libreria Intemacional HUNGARY JAMAICA Publicacoes Tecnicas Intemacionais RO. Box 17-01-3029 Euro Info Service Ian Randle Publishers Ltd. Ltda. Juan Leon Mera 851 Margitszgeti Europa Haz 206 Old Hope Road Rua Peixoto Gomide, 209 Quito H-1 138 Budapest Kingston 6 01409 Sao Paulo, SP Tel: (5932) 521-606; (5932) 544-185 Tel: (36 1)111 6061 Tel: 876-927-2085 Tel: (55 11)259-6644 Fax: (5932) 504-209 Fax: (361)3025035 Fax: 876-977-0243 Fax: (55 11) 258-6990 E-mail: librimul t libdmundi.com.ec E-mail: euroinfo@mail.matav.hu E-mail: irpl@colis.com E-mail: postmaster@pti.uol.br E-mail: librimu2@1ibrimundi.com.ec URL: hftp:llwww .uol.brINAJPN EGYPT, ARAB REPUBLIC OF Allied Publishers Ltd. Eastem Book Service CANADA Al Ahram Distribution Agency 751 Mount Road 3-13 Hongo 3-chome, Bunkyo-ku Renouf Publishing Co. Ltd. Al Galaa Street Madras - 600 002 Tokyo 113 5369 Canotek Road Cairo Tel: (91 44) 852-3938 Tel:(813)3818861 Ottawa Ontaro KIJ 9J3 Tel: (20 2) 578-6083 Fax: (91 44) 852-0649 Fax: (81 3) 3818-0864 Tel: (613) 745-2665 Fax: (20 2) 578-6833 INDONESIA E-mail: orders@svt-ebs.co.jp Fax: (613) 745-7660 PtI ndraLimited URL:StIp:ffw.bekkoame.orpl-svtebs E-mail: order.dept@renoufbooks.com The Middle East Observer Jalan Borobudur 20 URL: http./www .renoufbooks.com 41, Sherif Street PO. Box 181 Cairo PO o 8 Tel: (20 2) 393-9732 Jakarta 10320 Fax: (20 2) 393-9732 Tel: (62 21) 390-4290 Fax: (202) 393-9732 Fax: (62 21) 4214289 KENYA NORWAY SINGAPORE, TAIWAN, SWITZERLAND Afdca Book Service (E.A.) Ltd. NIC Info AIS MYANMAR, BRUNEI Libraide Payot Quaran House, Mfangano Street Book Department Asahgate Publishing Asia Service Institutionnel PO. Box 45245 Postboks 6512 Etterstad Pacific Pte. Ltd. CBtes-de-Montbenon 30 Nairobi N-0606 Oslo 41 Kallang Pudding Road #04-03 1002 Lausanne Tel: (254 2) 223 641 Tel: (47 22) 97-4500 Golden Wheel Building Tel: (41 21) 341-3229 Fax: (254 2) 330 272 Fax: (47 22) 97-4545 Singapore 349316 Fax: (41 21) 341-3235 Tel: (65) 741-5166 ADECO Van Diermen Editions KOREA, REPUBLIC OF PAKISTAN Fax: (65) 742-9356 Techniques Daejon Trading Co. Ltd. Mirza Book Agency E-mail: ashgate@asianconnect.com Ch. de Lacuez 41 RO. Box 34, Youida 65, Shahrah-e-Quaid-e-Azam CH1807 Blonay 706 Seoun Bldg Lahore 54000 SLOVENIA Tel: (41 21) 943 2673 44-6 Youido-Dong, Yeongchengpo-Ku Tel: (92 42) 735 3601 Gospodarski Vestnik Publishing Group Fax: (41 21) 943 3605 Seoul Fax: (92 42) 576 3714 Dunajska cesta 5 Tel: (822) 785-1631/4 1000 Ljubljana THAILAND Fax: (822) 784-0315 Oxford University Press Tel: (386 61)1338347; 132 12 30 Central Books Distnbutfon 5 Bangalore Town Fax: (386 61)133 80 30 306 Silom Road MALAYSIA Sharae Faisal E-mail: repansekj@gvestnik.si Bangkok 10500 University of Malaya Cooperative PO Box 13033 Tel: (66 2) 235-5400 Bookshop, Limited Karachi-75350 SOUTH AFxRICA, BOTSWANA Fax: (66 2) 237-8321 RO. Box 1127 Tel: (92 21) 446307 For single titles: Jalan Pantai Baru Fax: (92 21) 4547640 Oxford University Press TRINIDAD & TOBAGO, 59700 Kuala Lumpur E-mail: ouppak@TheOffice.net Southem Atfca AND THE CARRIBBEAN Tel: (60 3) 756-5000 Vasco Boulevard, Goodwood Systematics Studies Limited Fax: (60 3) 755-4424 Pak Book Corporation PO. Box 12119 St. Augustine Shopping Center E-mail: umkoop@tm.net.my Aziz Chambers 21 N1 City 7463 Eastem Main Road, St. Augustine Queen's Road Cape Town Tdnidad & Tobago, West Indies MEXICO Lahore Tel: (27 21) 595-4400 Tel: (868) 645-8466 INFOTEC Tel: (92 42) 636 3222; 636 0885 Fax: (27 21) 595-4430 Fax: (868) 645-8467 Av. San Femando No. 37 Fax: (92 42) 636 2328 E-mail: oxford@oup.co.za E-mail: tobe@tnnidad.net Col. Todello Guerra E-mail: pbc@brain.net.pk 14050 Mexico, D.E For subscnpfion orders: UGANDA Tel: (52 5) 624-2800 PERU Intemational Subscdption Service Gustro Ltd. Fax: (52 5) 624-2822 Editonal Desarrollo SA PO. Box 41095 PO Box 9997 E-mail: intotecrtn.net.mx Apartado 3824 Craighall Madhvani Building URL: http://rtn.net.mx Lima 1 Johannesburg 2024 Plot 1614 Jinja Rd. Tel: (51 14) 285380 Tel: (27 11) 880-1448 Kampala Mundi-Prensa Mexico S.A. de C.V. Fax: (51 14) 286628 Fax: (2711) 880-6248 Tel: (256 41) 251 467 c/Rio Panuco, 141-Colonia Cuauhtemoc E-mail: iss@is.co.za Fax: (256 41) 251 468 06500 Mexico D.F PHILIPPINES E-mail: gus@swiftuganda.com Tel: (52 5) 533-5658 Intemational Booksource Center Inc. SPAIN Fax: (52 5) 514-6799 1127-A Antipolo St. Mundi-Prensa Libros, S.A. UNITED KINGDOM Barangay, Venezuela Castello 37 Microinfo Ltd. NEPAL Makati City 28001 Maddd PO. Box 3 Everest Media Intemational Tel: (63 2) 896 6501; 6505; 6507 Tel: (341) 431-3399 Alton, Hampshire GU34 2PG Services (P) Ltd. Fax: (63 2) 896 1741 Fax: (34 1) 575-3998 England GPO Box 5443 E-mail: libreria@mundiprensa.es Tel: (441420) 86848 Kathmandu POLAND URL: http:/hvwww.mundiprensa.es/ Fax: (44 1420) 89889 Tel: (977 1) 472 152 Intemational Publishing Service E-mail: wbank@ukminfo.demon.co.uk Fax: (9771) 224 431 Ul. Piekna 31/37 Mundi-Prensa Barcelona URL: http:/A/ww .microinfo.co.uk 00-677 Warzawa Consell de Cent, 391 NETHERLANDS Tel: (48 2) 628-6089 08009 Barcelona VENEZUELA De LUndeboom/lnOr-Publikables Fax: (48 2) 621-7255 Tel: (34 3) 488-3492 Tecni-Ciencia Libros, S.A. P.O. Box 202 E-mail: books%ips@ikp.atm.com.pl Fax: (34 3) 487-7659 Centro Cuidad Comercial Tamanco 7480 AE Haaksbergen URL: httpJ/www.ipscg.wawp.ips/export/ E-mail: barcelona@mundiprensa.es Nivel C2 Tel: (31 53) 574-0004 Caracas Fax: (31 53) 572-9296 PORTUGAL SRI LANKA, THE MALDIVES Tel: (58 2) 959 5547; 5035; 0016 E-mail: lindeboo@woddonline.nl Livraria Portugal Lake House Bookshop Fax: (58 2) 959 5636 URL: httpAwvw .worldonline.nV-lindeboo Apartado 2681 100, Sir Chittampalam Gardiner Rua Do Carmo 70-74 Mawatha ZAMBIA NEW ZEALAND 1200 Lisbon Colombo 2 University Bookshop EBSCO NZ Ltd. Tel: (1) 347-4982 Tel: (94 1) 32105 University ot Zambia Pdvate Mail Bag 99914 Fax: (1) 347-0264 Fax: (941) 432104 Great East Road Campus New Market E-mail: LHL@sr.lanka.net P.O. Box 32379 Auckland ROMANIA Lusaka Tel: (64 9) 524-8119 Compani De Librarii Bucuresti S.A. SWEDEN Tel: (260 1) 252 576 Fax: (64 9) 524-8067 Str. Lipscani no. 26, sector 3 Wennergren-Williams AB Fax: (260 1) 253 952 Bucharest P 0. Box 1305 NIGERIA Tel: (40 1) 613 9645 S-171 25 Solna ZIMBABWE University Press Limited Fax: (401) 312 4000 Tel: (46 8) 705-97-50 Longman Zimbabwe (Pvt.)Ltd. Three Crowns Building Jedcho Fax: (46 8) 27-00-71 Toude Road, Ardbennie Pdvate Mail Bag 5095 RUSSIAN FEDERATION E-mail: mail@wwi.se P.O. Box ST125 ibadan lsdatelstvo Southerton Tel: (234 22) 41-1356 9a, Lolpachniy Pereulok Harare Fax: (234 22) 41-2056 Moscow 101831 Tel: (263 4) 6216617 Tel: (7 095) 917 87 49 Fax: (263 4) 621670 Fax: (7 095) 917 92 59 V7198 Want to know more about Research Observer? You can view the tables of contents and the abstracts of articles from current and past issues at: http://www.worldbank.org/html/extpb/ObserverContents.html We're interested in your comments. Please write to us at: Editor, The World Bank Research Observer The World Bank 1818 H Street, N.W. (N-7-039) Washington, DC 20433 HE W OR LD BAN K 780821 341131