WORLD BANK LATIN AMERICAN AND CARIBBEAN STUDIES fleupoin/ts Work in progress 19636 for public discussion July 1999 Beyond the Center tecertra lz o 9v4 Shahid Javed Burki, Guillermo Perry and William Dillinger with Charles Griffin Jeffrey Gutman Fernando Rojas Steven Webb - Donald Winkler WORLD BANK LATIN AMERICAN AND CARIBBEAN STUDIES ViewPpoilts BEYOND THE CENTER: DECENTRALIZING THE STATE ShahidJaved Burki Guillermo E. Perry William R. Dillinger THE WORLD BANK WASHINGTON, D.C. Copyright © 1999 The International Bank for Reconstruction and Development/THE WORLD BANK 1818 H Street, NW Washington, D-C. 20433, U.S.A. All rights reserved Manufactured in the United States of America First printing July 1999 The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) and should not be attributed in any manner to the World Bank, to its affiliated organizations, or to members of its Board of Executive Directors or the countries they repre- sent. The World Bank does not guarantee the accuracy of the data included in this publication and accepts no responsibility whatsoever for any consequence of their use. The boundaries, colors, denominations, and other information shown on any map in this volume 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 material in this publication is copyrighted. Requests for permission to reproduce portions of it should be sent to the Office of the Publisher at the address shown 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 class- room use is granted through the Copyright Clearance Center, Inc., Suite 910, 222 Rosewood Drive, Danvers, Massachusetts 01923, U.S.A. Shahid Javed Burki is vice president, Guillermo E. Perry is chief economist, and William Dillinger is principal urban economist in the World Bank's Latin America and the Caribbean Regional Office. Cover illustration by David McLimans Library of Congress Cataloging-in-Pablication Data Burki, Shadid Javed. Beyond the center: decentralizing the state / Shahid Javed Burki, Guillermo E. Perry, William Dillinger. p. cm. - (World Bank Latin American and Caribbean studies) Includes bibliographical references. ISBN 0-8213-4521-4 1. Decentralization in government. 2. Intergovernmental cooperation. 3. Political participation. 4. Intergovernmental fiscal relations. I. Perry, Guillermo. II. Dillinger, William. 1951- . III. Title. IV Series. JS113.B86 1999 352.2'83'098-dc2 1 Contents Acknowledgments ...........................................vii Acronyms and Abbreviations .......................................... ix Prologue: Matching Authority and Accountability .......................................... Chapter 1: Decentralization: Politics in Command .......................................... 9 The Roots of Centralism ............................................... 9 The Wind Changes ........................ ....... ... 10 Eight Cases . ................................................................................. 10 Federal Countries Reverting to Type .....................1....................... 1 Civilian Regimes Unleashing Local Electoral Competition ....................1........................ 1 Two-tier States ............................................ 13 Chapter 2: Getting the Rules Right: A Framework for Subnational Government ............................. 17 Dividing Up National Political Power .............................................................. 18 Assigning Functions, Structures, and Revenues ........................................................ 21 Functions ........................................................................ 21 Structure ........................................................................ 24 Revenues: Taxes and Transfers ................................................................... 26 Controlling Subnational Debt .................................................................... 29 Local Politics ......................................................................... 31 Strategies ........................................................................ 32 Synchronizing the Elements of Reform ........................... ................................ 33 Demonstrating Early Commitment to the New Rules of the Game ....................................... 33 C onclusion ....................................................................... 34 Chapter 3: Addressing the Macroeconomic Threat: The Quest for Hard Budget Constraints .................... 35 Evidence of Macroeconomic Problems: Econometric Results .............................................. 37 Outcomes and Institutions in Major Decentralized States of Latin America ................................... 39 Hard Budget Constraints for Subnational Governments ............................................... 39 Borrowing Constraints ....................................................................... 40 Autonomy of Subnational Governments to Stay Within Budget Constraints ................................ 40 Revenue Autonomy ....................................................................... 41 Political Ability to Enforce Hard Budget Constraints ........................................... ..... 41 Outcomes and Institutions in Four Latin American Countries ..................................... ..... 41 Summary of Cases ....................................................................... 47 Decentralization Institutions and Fiscal Management ................................................... 47 Hard Budget Constraints for Transfers and Service Responsibilities ....................................... 47 Borrowing Constraints ....................................................................... 49 111 BEYOND THE CENTER Subnational Autonomy for Fiscal Adjustment ...................................................... 50 Intergovernmental Political Relations . ............................................................ 51 Conclusion .................................................................. 51 Chapter 4: Empowering Municipalities or Schools? The Decentraization of Education ......................... 53 Rationale for Education Decentralization . ........................................................... 54 The Educational Context of Decentralization . ......................................................... 55 Typology .................................................................. 55 Level of Decentralization .................. ................................................ 55 Decisionmaking Powers .................................................................. 56 Structure and Content .................................................. ................ 56 Typology Applied to Recent Latin American Experience .................... ............................. 57 Level of Decentralization ......................................... 57 Decisionmaking Powers ......................................... 58 Structure and Content ......................................... 59 Evaluation Decentralization .......................................... 60 High-Performing Schools ......................................... 60 Characteristics of Effective Schools ......................................... 60 The Consequences of School Decentralization ......................................... 62 Leadership ......................................... 62 Teaching Excellence ......................................... 63 Focus on Learning ......................................... 63 Responsibility ......................................... 64 Empirical Findings ......................................... 64 Conclusion ......................................... 66 Annex ......................................... 69 Chapter 5: Empowering Mayors, Hospital Directors, or Patients? The Decentralization of Health Care .... ......... 75 Three Kinds of Health Services . .................................................................. 75 Many Modes of Service Delivery and Financing ....................................................... 76 What has been Decentralized in Latin America and How ................................................ 77 Argentina: From Central Monopoly to Provincial Monopoly ............................................ 77 Mexico: From Deconcentration to Decentralization .................................................. 78 Bolivia: Municipalizing the Physical but not the Human Assets ......................................... 79 Brazil: Public Financing with Private Production .................................................... 80 Chile: Municipalizing Production, Centralizing Financing ............................................. 81 Colombia: Conflicts Between Decentralization and Privatization ......................................... 81 Major Issues in Decentralizing Health .............................................................. 82 Labor Management .......................................................................... 82 Recurrent Financing ......................................................................... 82 Investment Decisions . ....................................................................... 83 The Political Economy of Reform ................................................................. 84 C onclusion ................................................................................ 85 Chapter 6: Decentralizing Roads: Matching Accountability, Resources, and Technical Expertise ................. 87 Decentralization in the Road Management Sector ...................................................... 87 In Search of a Model: The Latin American and Caribbean Experience ........ .............................. 89 What is Decentralization in the Road Sector? ................................................... 89 Road Functions ................................................... 89 Management Tasks ................................................... 90 Financing Mechanisms ..................... .............................. 91 Road Decentralization in Latin America ................................................... 92 Where Political and Fiscal Decentralization is Well Established ......... ............................... 93 Dangers of Rapid Decentralization ................ ................................... 94 Benefits to National Networks of Devolving Non-National Roads ............ .............. ............ 95 Contribution of Parallel Reforms ................................................... 95 iv CONTENTS Delegation of National Roads ....................... 97 Conclusion ....................... 97 References ........................ 99 Boxes Box P1 What is Decentralization. 3 Box 2.1 A Glossary of Political Terms .20 Box 2.2 How Far to Decentralize? .22 Box 2.3 Are Local Governments Incompetent? .24 Box 2.4 Revenue Assignment in Federal Countries .29 Box 2.5 The Impact of Transfers on Local Political Behavior .30 Box 3.1 Risks and Costs in the Source and Terms of Borrowing .50 Box 4.1 Chicago: An Initial Emphasis on Governance .66 Box 4.2 Memphis: Decentralization Focused on Improving Learning .67 Box 4.A. 1 What Has Been Learned from Effective Reform Strategies? .73 Box 5.1 Lack of Reliable Information .84 Box 6.1 The New Generation of Road Funds .92 Box 6.2 Key Features of the Contrato de Recuperaci6n Mantenimiento (CREMA) .96 Tables Table P.1 Summary of Decentralization Measures. 2 Table 1.1 Summary of Decentralization Measures in Selected Countries .14 Table 1.2 Indicators of Decentralization .15 Table 2.1 Electoral Rules, Selected Latin American Countries .19 Table 2.2 Electoral Rules, Selected OECD Countries .20 Table 2.3 Structure of Subnational Governments: Selected OECD and Latin American Countries .26 Table 2.4 Principal Revenue Sources of Subnational Governments .27 Table 2.5 Sources of Revenues: Percent Distribution, Selected Countries .28 Table 3.1 Basic Model: Cross Section; Overall Sample Estimates .37 Table 3.2 Panel Data, Fiscal Variables in First Differences, 1980-94 .38 Table 3.3 Institutional Arrangements to Set and Keep Hard Budget Constraints on States .39 Table 3.4 Suriinary of Conditions for Rcduced Danger of Excess Deficits .48 Table 3.5 Channels for Control of Subnational Borrowing .49 Table 4.1 Types of Decisions that may be Decentralized .57 Table 4.2 The Locus of Key Educational Decisions and Responsibilities, Selected Latin American Countries . 59 Table 4.3 Characteristics that can be Stimulated Through Decentralization .61 Table 4.4 Assessment of Education Decentralization .63 Table 4.A.1 Characteristics of "Decentralized" School System Argentina, Education Reforms 1976 1991 .69 Table 4.A.2 Characteristics of "Decentralized" School System Chicago, IL, Education Reforms 1988-1999 .69 Table 4.A.3 Characteristics of "Decentralized" School System Chile, Education Reforms 1981 .69 Table 4.A.4 Characteristics of "Decentralized" School System El Salvador, Community-Managed Schools (EDUCO). 70 Table 4.A.5 Characteristics of "Decentralized" School System Memphis, TN, Education Reforms 1995-1999 . 70 Table 4.A.6 Characteristics of "Decentralized" School System Mexico, Education Reforms 1992-1998 .70 Table 4.A.7 Characteristics of "Decentralized" School System Minas Gerais, Brazil .71 Table 4.A.8 Characteristics of "Decentralized" School System Nicaragua, Education Reforms 1993-1998 .71 Table 4.A.9 Summary of Characteristics of Decentralized Schools .72 Table 5.1 Suggested Financing and Provision for Different Types of Health Services .76 Table 5.2 Authority to Manage Personnel by Different Levels of Government .83 Table 5.3 Authority to Make Infrastructure Investments .......................................... 83 Tablc 6.1 Decentralization Options for the Road Sector .91 Figures Figure 4.1 Latin American Countries Implementing Education Reforms .54 Figure 4.2 Average Math Achievement Test Scores of Eighth Graders, Selected International Comparators . 55 Figure 4.3 Third-Grade Language Achievement, Latin American Countries .56 Figure 4.4 Level of Decisionmaking in Education Sector .58 v Acknowledgments Ti his report is the product of a multisectoral team. The research and writing of the report were conducted by a group coordinated by the Office of the Chief Economist for Latin America and the Caribbean (LAC), led by Guillermo Perry and William Dillinger. Main authors were William Dillinger (Chapters 1 and 2), Steven Webb and Guillermo Perry (Chapter 3), Donald Winkler (Chapter 4), Charles Griffin (Chapter 5), and Jeff Gutman (Chapter 6). Each team member, in turn, benefited from background papers and advice provided by specialists in their respective fields. In particular, we wish to acknowledge Fernando Rojas, whose survey of decentraliza- tion in Latin America provided the basis for Chapter 1; Jose Luis Irigoyen and Francisco Wulff, who assist- ed in drafting Chapter 6; and Tarcisio Castefieda and Girindre Beeharry, who assisted in drafting Chapter 5. The group benefited from initial discussions at a workshop on the topic, carried out with lead economists, lead specialists, and staff from the LAC regional office of the World Bank. Richard Bird, of the University of Toronto, and Ernesto Stein, of the Inter-American Development Bank, also provided valuable input into the initial design of the report. vii Acronyms and Abbreviations Acronyms and Abbreviations ACE Community Education Association (El Salvador) ARS Administradora del Regimen Subsidiado (Colombia) BANOBRAS Banco Nacional de Obras Publicas (Mexico) CEPAL Centro Estudios para America Latina CONAFE National Board for Educational Improvement (Consejo Nacional de Fomento Educativo) CREMA Contrato de Recuperaciucat Mantenimiento CVF Consejo de Vialidad Federal (Federal Road Council, Argentina) DNV Direcci de Vialidad Federal (Federal Road Council, Argentina) DRG Diagnostic Related Group EDUCO Educaci6n con la participaci6n de la Comunidad (Education with the Participation of the Community) FINDETER Financera de Desarrollo Territorial S.A. (Colombia) GDP Gross domestic product GNP Gross national product INEP Instituto Nacional de Estudos e Pesquisas Educacionais IMSS Social Security Institute (Mexico) LAC Latin America and the Caribbean NGO Non-governmental organization OECD Organization for Economic Cooperation and Development OPD Decentralized public organism (Mexico) PRI Partido Revolucionario Institucional PSF Programa de Saude de Familia (Brazil) SIMCE Sistema de Medicion de la Calidad de la Ensenanza SN Subnational SSA Ministry of Health (Mexico) SUS Single health system (Brazil) TVAAS Tennessee Value-Added Assessment System UNESCO United Nations Educational, Scientific and Cultural Organization VAT Value-added tax ix PROLOGUE Matching Authority and Accountability ecentralization is transforming the structure of governance in Latin America. Since 1983, all but one of the largest countries in the region have seen a transfer of power, resources, and responsibilities to subnational units of government. In much of the region this began as a shift from appointed to elected governors and mayors (see Table P.1). In some coun- tries this occurred in the course of a reversion from military to civilian rule (as in Argentina in 1983, Brazil in 1985, and Chile in 1992). In other countries it represented the intro- duction of local elections for the first time in decades (as in Colombia in 1986 [mayors] and 1991 tgover- nors}, and Venezuela in 1989 [governorsl). In Mexico the expansion of local democracy has been more sub- tle. After decades of one-party dominance in state and local elections throughout the country, opposition parties now constitute serious competition. Decentralization has also appeared in the form of a devolution of major functional responsibilities. In Colombia and Argentina primary education has been decentralized to the intermediate levels of government. In Chile it has been transferred, in part, to the municipal level. Parts of the highway network have been decentralized to state governments in Brazil. Major increases in fiscal transfers to subnational governments have also occurred, particularly in Brazil, Mexico, and Colombia. What is at Stake This dispersion of power is a global trend (see World Development Report {1999-20001). Political scientists have Political Stability and the Deepening of Democracy proposed various explanations for it. In broad terms, they The dominant force behind decentralization is, in the final have suggested that it is an outcome of the declining cred- analysis, political. It is part of a broader trend toward ibility of the centralized state. Groups that have historically democracy in the region. As shown in Table P1, of the 14 been denied power now demand it, and central govern- countries in the region with populations over 5 million, ments are increasingly reluctant to combat this demand only three were multiparty democracies in 1960. (Mexico with force. The state's declining credibility, in turn, has may be best classified as a single-party democracy from been attributed to economic failure (with the consequent 1929 to 1988, but can now be considered a multiparty alienation of important business and labor support), to the democracy.) As of 1999, all 14 countries fall into this cate- relative absence of war and civil unrest (with the conse- gory. At the national level, this has meant a dispersion of quent decline in the acceptance of strong authoritarian gov- formal political power to elected presidents and congress- ernment), and to the emergence of educated urban middle men. At the subnational level, it has meant a dispersion of classes (with the consequent decline of traditional patron- power to governors, mayors, and subnational legislatures. client relationships between the government and the gov- I BEYOND THE CENTLR: DECENTRALIZING THE STATE TABLE PI Summary of Decentralization Measures FIRST YEAR OF ELECTIONS ADDED LOCAL DEVOLVED DEVOLVED BY LEVEL OF GOVERNMENT* ELECTIONS' FUNCTIONS REVENUE COUNTRIES WITH POPULATION POPULATION 2 5 MILLION (IN MILLIONS) NATIONAL PROVINCIAL MUNICIPAL Brazil 164 1985 1982 1985 X X Mexico 95 1917 1917 1917 X X X Colombia 38 1958 1992 1986 X X X Argentina 36 1983 1983 1983 X X Peru 25 1980 -- 1981 Venezuela 23 1961 1989 19892 X X X Chile 15 1990 1 1992 X X X Ecuador3 12 1978 1978 1978 X Guatemala 11 1985 1985 X Bolivia 8 1985 1987 X X X Dominican Republic 8 1966 -- 1966 X Honduras 6 1982 -- 1982 X Nicaragua 5 1986-90 -- 1990 X Paraguay 5 1991 1994 1991 X -Following military rule, where applicable. -- = Countries with no provincial level of government. 'Victories by opposirion parties at the municipal level began in 1982 and at the stare level in 1989. ' Open list replaces closed, blocked party list. 3 Provincia! executives and councils have traditionally been elecrcldexcept during military regimes-but have limited powers. Parallel provincial governments, headed by a centrally appointed governor, are responsible for the field administration of central government iunctions. 'Although Chile, Guatemala, and Bolivia do nor have elected governments at the regional level, these countries have assigned major planning and investment responsibilities to regional governments, with significant revenues and expenditure autonomy. s Compares current regime to previous civilian regime. (Local elections are generally absent during periods of militart rule.) Ssewce: Willis, Garman, and Haggard 1997. erned) (O'Donnell, Schmitter, and Whitehead 1986; Potter was in part motivated by the desire of central party leaders 1993). In a fundamental sense, democratization, and with to gain grass roots support in areas under rebel control. But it, decentralization, can be seen as a strategy to maintain it also reflected a concession to long-standing demands political stability-to provide an institutional mechanism from the established parties for regional autonomy. The for bringing opposition groups into a formal, ritualized bar- decentralization of education and health in Colombia also gaining process. As such, it constitutes an alternative to reflected concerns with efficiency and quality in those sec- civil war or other forms of violent opposition. tors. In Mexico, the decentralization had its origins in a Specific cases, of course, do not lend themselves to such long-running series of political reforms that began with the global generalizations. In Latin America the motivations for decision to allocate 25 percent of the seats in congress to decentralization are complex. In Brazil, for example, decen- opposition parties. A subsequent financial crisis caused the tralization accompanied the transition from military to government party to lose control of congress and several civilian rule. By shifting political and fiscal resources to the key governorships and mayoralties. Opposition parties have municipal level, the negotiators of Brazil's transition were therefore tended to favor decentralization, and their able to accommodate popular dissatisfaction with military increasing influence-particularly in congress-has per- centralism and allow the military to withdraw in good order mitted them to demand further measures as a price for their (Hagopian and Mainwaring 1987). In Guatemala, decen- support on key government legislation. tralization was one of the key tenets of the peace accords (1996) and raised the expectation of greater self-govern- More Efficient, Responsive Public Services ment for indigenous communities. In Colombia, the insti- Political stability and democratic government are worth- tution of elected mayors, and later provincial governors, while ends in themselves. But there is more at stake. On the 2 PROLOG UE positive side, it is argued that decentralization can increase trators to local elites may worsen the quality of services, at the efficiency and responsiveness of government (Oates least for the majority of constituents. Questions have also 1972). According to this argument, devolving resource been raised about the technical capabilities of local govern- allocation decisions to locally elected leaders can improve ment staff the match between the mix of services produced by the While the evidence to date does not point definitively in public sector and the preferences of the local population. either direction, it is clear that there has been an increase in Because local officials have better knowledge of local condi- the variance of public service performance. Centralized tions and are more accessible to their constituents, they ministries were capable of delivering a fairly standardized have the means and the incentive to be responsive. Decen- level of services nationwide. Decentralization has improved tralization, according to this argument, may also improve services in some jurisdictions and worsened it in others. the management of public services since, through sheer Concern with these risks has prompted some Latin proximity, local officials can be held more accountable for American and Caribbean countries to favor slow, incremen- their performance (Ostrom, Schroeder, and Wynne 1993). tal, or partial decentralization. This has taken the form of Where the population is mobile and citizens can "vote with micromonitored earmarking (as for example, in the Mexi- their feet," decentralization may also result in local govern- can approach to sector decentralization or the Colombian ments competing with each other to better satisfy the wishes government's requirement that provincial and local gov- of citizens (Tiebout 1956; Inman and Rubenfeld 1997). ernments be "certified" before assuming responsibility for There are downside risks to decentralization. First, of education or health). Programs geared to strengthening course, is the risk that service delivery could decline. Grant- subnational government technical capacity have been ing political autonomy to local governments does not guar- implemented in practically every country in the region, antee an improvement in public services. There is, to start although this has been more successful when organized as with, a risk of capture by local political elites. Transferring information sharing among peers than as top-down techni- decisionmaking power from central government adminis- cal assistance directed at recalcitrant mayors. BOX P. i What is Decentralization? Decentralization, as the term is used here, refers to the regional governments into independent political entities. process of devolving political, fiscal, and administrative Another is privatization, through the sale of assets, the grant- powers to subnational units of government. Although ing of concessions, and through public-private alliances. there are many entry points and strategies for decentral- This report focuses on elected subnational government ization, for purposes of this report a country is not consicd- for two reasons. First, it is the most ambitious-and per- ered to have decentralized unless it has a locally elected haps most risky-of the three forms of structural reform subnational government. Decentralization may consist of going on in the region. Second, it is unique in its behav- bringing such governments into existence, restoring them ioral implications. Decentralization shifts the structure of after a period of authoritarian rule, or expanding the local accountabilty from central government to local con- resources or responsibilities of existing elected subnational stiruents. Deconcentration, in contrast, preserves the hier- goveriments. archical relationship between central government and field There are, of course, other ways to disperse the power staff. Privatization introduces the profit motive as an influ- of central government ministries. One is deconcentration, ence on behavior. Although in practice the three can be in which the central government increases the autonomy employed simultaneously, their political, fiscal, and admin- of its regional offices. This has been widely used in Latin istrative implications are quite different. Decentralization America, often as a precursor to the transformation of must be considered unique. 3 BEYOND THE CENTER: DECENTRALIZING THE STATE Widening Disparities ernments may reduce deficits by cutting spending in the In theory, decentralization may also widen regional dispari- sectors that remain under their responsibility. This can dis- ties in the provision of public services. Such disparities, per tort the sectoral allocation of spending in the public sector se, are not undesirable. As with private goods, effective as a whole.) demand for local public services is likely to vary with Decentralization can also hamper a government's ability income. But disparities matter when they have distribu- to respond to economic shocks. Fiscal decentralization- tional implications. Thus disparities in spending on primary whether in the form of tax reassignment or mandatory rev- education or primary health may be cause for concern. If enue sharing-reduces central control over aggregate pub- the financing of education or health is decentralized, spend- lic sector revenues and expenditures. Governments, such as ing levels will reflect variations in local tax bases, rather Brazil's and Colombia's, which must share nearly half their than the allocations of centralized government ministries. tax revenues with subnational governments, may find it dif- ficult to raise tax rates sufficiently to compensate for eco- Macro Risks nomic downturns (Tanzi 1996). Decentralization may also pose macroeconomic risks in the But it is not merely the technical issues-transfers, form of recurring central government deficits, an overex- expenditure assignments, and tax levels-that are a source panded public sector, or the inability to use fiscal policy to of macroeconomic risks. There is a more subtle risk that adjust to economic shocks. newly empowered local governments may use their politi- Central government deficits may arise if governments cal muscle to undermine the national interest for the bene- are unable to reduce expenditures or increase taxes to fit of their individual constituencies. match the increasing cost of intergovernmental transfers. Mayors or governors, for example, may borrow exces- Brazil's 1988 constitution, for example, required a major sively in the expectation of central government debt relief. increase in fedcral tax sharing without providing for any While central governments are generally under no obliga- devolution of federal spending responsibilities. This threat- tion to provide such relief, decentralization can change pol- ened to provoke recurrent deficits at the federal level (see itics at the national level. The emergence of powerful Chapter 3.) In other cases, central governments have decen- regional voting blocs can weaken the government's resolve. tralized both revenues and expenditures, but have been This was the case, for example, in Brazil, where the Federal unable to reduce existing levels of central government Senate permitted states to accumulate unsustainable levels spending after the functions have been decentralized. of bond debt. This ultimately required a federal assumption Colombia, for instance, maintained and sometimes of the state debt, at massive cost to the government (see increased central government spending on education and Chapter 3). health several years after these functions were transferred to subnational government. Mexico substantially increased transfers to municipalities in 1998 and 1999 without a re- ciprocal reduction in spending at the federal level. As a In this report we examine these claims in the light of empir- result, both countries have been forced-at least temporar- ical evidence from both Latin America and the long-stand- ily-to maintain existing levels of spending while funding ing multitier democracies of Europe and North America. an expanding volume of intergovernmental transfers. In This report finds that the impact of decentralization-its some cases this has resulted in significant central govern- effect on the efficiency of public services, on equity, and on ment deficits. macroeconomic stability-depends very much on the Attempts to eliminate such deficits by raising taxes can specifics of the case. One has to look beyond the question of raise another macroeconomic problem: an overexpansion of whether a country has elections at the local level, whether a the public sector. Brazil's mismatch between revenues and nominal transfer of functional responsibilities has taken expenditures, for example, was ultimately resolved not by place, or whether intergovernmental transfers have grown, reducing federal government expenditures, but by increas- to the whole range of factors that affect the behavior of the ing federal taxes. (Other ways of reducing deficits can have people who are involved in public policymaking and service equally adverse microeconomic consequences. Central gov- provision. Who are these people? They are the newly 4 PROLOGUE elected mayors and governors, of course. But not only makes it easier for central governments to maintain a pol- them. They are also the teachers, doctors, and road engi- icy against bailouts. neers who are involved in delivering public services. We A similar pattern emerges in the chapters that examine find that the microeconomic impacts of decentralization- the impact of decentralization on the quality of public ser- the impacts on the efficiency and responsiveness of ser- vices. To examine this issue, the report looks at three spe- vices-depend not merely on decisions made in provincial cific sectors: education, health, and roads. Each is a major or municipal governments, but also in individual schools, expenditure item of government. Each has been the object health clinics, and road departments. of decentralization in Latin America. Here again we find Mayors and governors, moreover, are not the only politi- that success lies in (a) clearly establishing responsibility for cians who matter. It is the politicians at the national level- performance (accountability) while (b) allocating sufficient the congressmen, the president, the national party leader- authority to deliver results. In the case of roads, for exam- ship-who write the rules and decide which ones will be ple, some success has been achieved by allocating responsi- enforced. Rules decided at the national level have an over- bility for different parts of the road network to the level of riding impact on the behavior of politicians at lower tiers of government that best represents the roads' users. Thus government. To understand how these rules are made and rural roads are assigned to rural local governments, urban enforced, one has to understand how decentralization streets to urban local governments, secondary highways to affects national as well as local politics. provincial governments, and the major intercity highways If there is one common theme that runs through the six to the national government. But such functional decentral- chapters that follow, it is the importance of accountability: ization has failed where it has not been accompanied by suf- the need to clearly demarcate who is responsible for what. ficient authority. As Chapter 6 details, Peru's attempt at But accountability is not enough. Those who are account- road decentralization ran aground when the central gov- able must also have the authority to deliver results. This ernment failed to establish a financing mechanism for the means not merely the legal authority to make decisions, but roads it had decentralized. also the financial and human resources to carry them out. The education and health chapters tell a similar story, This lesson emerges, first, in the chapter on macroeco- but at a more micro level. Decentralization in these sectors nomic stability. Chapter 3 describes how four countries have can improve performance but only if it extends to the level responded to pressure for excessive borrowing by subna- of the individual schools and health units. Where the school tional governments. Two factors appear to determine suc- director is made accountable to the parents (through a cess. The first is a hard budget constraint imposed by rhe school board) and has the authority to make key manage- central government. Where a central government main- ment decisions (including who gets hired and how the cur- tains a credible policy against bailouts, it forces subnational riculum is adapted to meet the particular needs of the stu- governments and their lenders to live with the conse- dent body), improvements can be expected. By the same quences of their decisions. They become accountable for token, in the health sector, the benefits of decentralization their borrowing and lending. Under these circumstances, do not appear until responsibility and authority are lenders tend to refrain from lending to uncreditworthy sub- extended beyond the level of the municipality. In fact, one national governments. Where responsibility for debt is of the most effective forms of health care decentralization is ambiguous-as it is when there is reason to expect national to shift power to the patient by changing the focus of gov- bailouts of failing subnational entities-subnational gov- ernment funding from inputs (salaries of health workers in ernments and their lenders have a greater incentive to take government hospitals) to outputs (reimbursements to risks at the expense of national taxpayers. providers, whether public or private) thus allowing patients The second factor is subnational management auton- to choose where they will go for treatment. omy, that is, the ability to adjust to macroeconomic shocks Does this mean that subnational governments should be in a way that forestalls default. Where subnational govern- organized as a collection of sectoral autarkies, with each sec- ments have the authority to cut personnel costs or raise tax tor organized as an independent unit of government rates, they can respond to adverse economic circumstances responsible to its particular clientele? No. As Chapter 2 without reneging on their debt obligations. This in turn describes, there are important constraints to this approach. 5 BEYOND THE CENTER: DECENTRALIZING THE STATE First, the beneficiaries of a service are not easily defined. Rules must be compatible with incentives. Politicians at While there may be a general consensus that the benefits of both the local and national levels have their political careers national defense are national and the benefits of local street to protect. Interest groups have their axes to grind. Ordi- lighting are local, this simple prescription is of little use in nary citizens have their lives to worry about. A system can- allocating responsibility for the "big ticket" items that are not depend on altruistic politicians or a constantly alert cit- now the object of decentralization. Parents may be the pri- izenry poised to intervene whenever its interests are mary beneficiaries of public education, but there are threatened. national interests in primary education as well. (Education, This report, therefore, looks at the broader set of rules for example, may be a key element in a poverty reduction that affect political behavior, focusing particularly on elec- strategy.) toral systems and political parties. What we observe is that What we find as a solution in both Organization for weaknesses in the formal rules can be offset by strengths in Economic Cooperation and Development (OECD) and the political system. In examining the case of subnational some Latin American countries is a structure of government debt, for example, Chapter 3 finds that there are several dif- that consists of a complex set of principal-agent relation- ferent institutional setups that can produce a hard budget ships, in which subnational governments act both as agents constraint on subnational borrowing. One is a credible pol- of higher levels of government and as principals (or more icy against central government bailouts, demonstrated over precisely as agents of their constituents) in the delivery of the course of many years and many economic circum- local services. Services, such as education, health, or roads, stances. This tends to discourage lenders from increasing are "unbundled," with central and subnational units divid- their exposure to shaky subnational governments. But ing up aspects of service delivery according to interest and another institutional setup is a hegemonic and internally comparative advantage. This arrangement often finds cen- disciplined political party with the power to suppress any tral governments playing a financial and regulatory role- defiant behavior on the part of subnational politicians. As providing financing to maintain at least a minimal level of long as the national party continues to pursue a conserva- social services in all jurisdictions, for example, but leaving tive fiscal policy, such a system can force mayors and gover- detailed management decisions to lower tiers of government. nors to refrain from excessive borrowing. What is important is that the role of each tier is fixed- But hegemonic political parties are becoming rare in that the division of roles among tiers of government is Latin America. This report therefore examines ways in clear-and that subnational units of government have suf- which electoral rules can be revised at the national level to ficient authority to perform the roles assigned to them. The enable governments to maintain a firm policy against most striking problems in the decentralizing countries in bailouts-an examination that calls into question the wide- Latin American and the Caribbean occur where these con- spread use of proportional representation in the region. The ditions are not met. Chapter 4 recounts the problems in report also attempts to address the issue of voter apathy and education decentralization in Colombia, for example, where the risk that subnational politics will be captured by local responsibility was nominally transferred to the provincial elites. This again involves changes in electoral rules and (department) level, but where key controls over teacher internal party practices. salaries and working conditions were retained by the Min- Decentralizariors in Latin America is clearly driven by istry of Education. Chapter 5 critiques the case of health national political imperatives. Progress on the rest of the care in Bolivia, where physical facilities are the responsibil- agenda-improving public service delivery, forestalling ity of local governments, personnel is the responsibility of macroeconomic risks-depends very much on astute polit- departments (deconcentrated units of central government), ical timing. What is right for a given country will depend and medicine is the responsibility of the central Ministry on its particular historical conditions and the other impera- of Health. tives it faces. This report also addresses the issue of politics. We have In most of the region, national and federal government learned that even the best set of intergovernmental rules policies and institutions have been the focus of previous has little effect if it is not consistent with political culture. reform and modernization efforts. Though by no means Successful decentralization requires more than good rules. completed at that level, the reform agenda must increas- 6 PRU LO(, U I ingly focus on subnational and intergovernmental institu- tional level is becoming essential for the quality of democ- tions and policies. The strong drive toward decentralization racy in the region. We hope that this report will help in the that is taking place throughout the region, and that we required refocusing of the reforms and modernization agen- expect to continue, means that already today some of the das toward subnational government and intergovernmental public services that matter most for citizens and the overall institutions and policies, and that it will provide some use- economy and society-such as basic education and health, ful guidance to countries that-willingly or not-are in the subnational roads, water supply and sanitation-are in the throes of the process. hands of semiautonomous subnational governments. It also means that adequate rules are essential to avoid fiscal and Shahid Javed Burki macroeconomic threats that this process may entail. Finally, Guillermo E. Perry it means that progressively good governance at the subna- William R. Dillinger 7 CHAPTER 1 Decentralization: Politics in Command atin America has a long tradition of centralized government. To understand the causes of decentralization, it is necessary to understand the causes of the centralism that preceded it. Some observers have traced centralization back to the colonial era. Spain's legacy to Latin America was a tradition of extreme centralization in governmental decisionmaking. This auto- cratic control reflected the crown's long-standing solutions to conflicts with local elites in Spain (similar measures were also adopted into law for Latin America) and the crown's fear of losing control to the antecedentes-the early explorers who established their own fiefdoms in Latin America (Nickson 1995). The Roots of Centralism Historians also see cultural and economic roots of centralism in (a) the acceptance of authority fostered by the particular form of Catholicism practiced in the region (authoritarian and allied with conservative groups), (b) the deep inequality in social relations, (c) the high concentration of land ownership (which in turn fixed the economic boundaries between owner and peon), (d) the low prevailing levels of education, and (e) the marginalization of indigenous people from national politics (Dominguez 1994). Centralism was also a reaction to the civil wars that Industrialization and urbanization, beginning in the occurred throughout Spanish-speaking Latin America in early twentieth century, changed the political context some- the nineteenth century (Germani and Silvert 1994). After what. Theoretically, these trends generate demand for the Latin Americans gained their independence from Spain, decentralization of power. Because of its inherent require- they established fledgling "democracies" on the U.S. model, ments, a modern economy distributes political resources but these failed quickly. Political competition degenerated and political skills to a wide variety of individuals, groups, into open civil war between competing caudillos (war lords) and organizations. Business groups, labor unions, and other in most of the countries of the region. Peace was achieved economic interests all demand a voice in government (Hag- only through the victory of centralizing dictators (Reyes in gard and others 1995; Przeworski 1977). Peasants, once Colombia, Gomez in Venezuela, Porfirio Diaz in Mexico). In under the control of landlords, move to the cities and become this respect, centralization was a key condition for consoli- an urban underclass, better organized and better positioned dating the nation-state. (Brazil, which remained a monar- to pressure the existing power structure (Lipset 1983). chy throughout the nineteenth century, is the only conti- Interludes of democratic government did in fact follow nental Latin American country that survived the century in in some Latin American countries in the early twentieth relative peace.) century. Political participation-which was initially confined 9 BEYOND THE CENTER: DECENTRALIZING THE STATE to traditional elites-was broadened in an attempt to cope expectation that once the crisis had subsided, they would be with the rise of the middle class and industrial workers. In restored to office. When it became clear that the military some countries elites attempted to co-opt the middle class had other plans, they withdrew their covert support, help- while leaving industrial workers disenfranchised. In other ing to isolate the regime (Little 1997). countries political leaders recognized the potential political The legitimacy of the military regimes also depended on base of organized workers and co-opted them (Skidmore successful economic performance. Military failings in this and Smith 1997). But these governments were unstable. arena further eroded their standing. Even in those cases Because the government itself was a source of economic where there had been some initial success (Brazil 1968-73, power, political groups fought among themselves to get Argentina 1977-79, and Chile 1977-82), the military control of it (Kling 1994). Military interventions during found that sustained improvement was beyond them. Not this period were common, with the military entering as an only did this prevent the military from reaching out to civil agent of civilian forces to topple the existing president and society to "buy" support, but it also led to disenchantment replace him with someone more to their liking. with their rule by their natural allies among the property- The 1960s saw the emergence of a new kind of military owning classes. Although in aggregate terms Latin Amer- intervention. Justifying their actions as a response to leftist ica did develop during the period in which military regimes threats, military governments took power with no fixed were common, the perception of politically influential deadline for return to civilian rule. By 1980, 9 of the 14 classes was that authoritarian economic policies were no largest countries in the region had military governments. more successful than democratic policies (Little 1997). Mexico, although civilian, was firmly under the control of One final element that seems to have been widely pre- the Partido Revolucionario Institucional (PRI). In Colombia sent in the move toward democratization was the percep- and Venezuela, multiparty democracies were politically cen- tion in military circles that holding power inevitably meant tralized: In Colombia, all political offices below the national the politicization of military institutions as different fac- government were appointive rather than elective; in tions split over policy issues. Since military power and Venezuela, state governors were appointed and mayors were autonomy rested above all on institutional unity, the longer elected from closed, blocked party lists. (See Box 2.1 for a the military stayed in power, the greater became institu- glossary of political terms.) tional pressures to withdraw. This era also saw an increase in the role of the central Crises of legitimacy have occurred not only in the mili- government. Rapid urbanization laid bare the inability of tary-ruled countries, but also in the nominally civilian local governments to meet growing demands for urban democracies. In Colombia, the government's decentraliza- infrastructure. Virtually every government in the region tion efforts since 1983 have been linked to the challenge to reacted not by strengthening municipal government, but political stability posed by the growing guerrilla move- by creating new central government bodies that stripped ment. Although previous administrations had responded municipalities of their formal role as service providers at the with force, the new (Betancur) administration initiated a local level. As a result, in the 1950s and 1960s a widespread peace process based on a democratic opening of the politi- process of demunicipalization took place throughout the cal system. The new administration argued that a strong subcontinent. Responsibility for the delivery of such impor- connection existed between political and functional decen- tant services as the urban water supply, transportation, tralization: Without increased autonomy and resources at public housing, primary health care, and education were the local level, political autonomy would be meaningless transferred to the central government (Nickson 1995). (Garman and Haggard 1998). The Wind Changes Eight Cases As the 1970s wore on, it became apparent that the state The specific form in which power has been dispersed from faced a growing crisis of legitimacy in many countries of the central governments varies considerably in Latin America. region. In part this was due to the declining threat from the In some countries, the most striking change has been polit- left. In Argentina and Chile, political groups on the center ical: the first-time introduction of elected governors in and right had tacitly condoned military intervention in the Venezuela, for example, or the emergence of genuinely 10 DECENTRALIZATION POLITICS IN COMMIAND competitive multiparty elections for subnational offices in and the majority of secondary schools). The Federal District Mexico. In other cases, it is functional decentralization that was granted the status of a state, with an elected governor is more striking, as in the decentralization of education to and legislature. At the same time, the federal government the departments in Colombia and to the municipalities in increased its control over the provinces-particularly the Chile. Radical changes in the distribution of revenues have smaller ones-by encouraging the closure of provincial also occurred, as in the doubling of revenue sharing to banks and federalizing provincial pension funds. Changes in municipalities in Brazil and Colombia. electoral laws also eliminated the power of provincial legis- Countries also vary in the amount of change they have latures to elect their representatives in the federal senate. experienced. Argentina is arguably one of the most decen- (Senators are now directly elected.) Below the provincial tralized countries in the region but has essentially the same level, Argentina remains highly centralized. While munici- political and fiscal structure it had before the military inter- pal mayors and councils are elected, their functional respon- vened in 1976. In contrast, Colombia has radically sibilities are limited and their independent revenue-raising increased the powers and responsibilities of subnational powers virtually nonexistent. units of government. Brazil is another case of a long-standing federal country Any understanding of decentralization in the region reverting to type after a period of military rule. Since the requires some examination of individual cases. The eight founding of the republic (in 1891) Brazil has been constitu- Latin American countries discussed below appear to be in tionally organized as a federation, with elected state gover- the forefront of decentralization, but they represent eight nors and legislatures. Centralized government has existed distinct decentralization stories. only during the Vargas dictatorship (1937-50) and the recent military period (1964-85). With the departure of Federal Countries Reverting to Type the military, Brazil reverted to its former federal structure. For descriptive purposes, the eight countries can be divided Brazil's postmilitary constitution (1988) is far more into two groups: the federal (or three-tier) countries and the decentralist than the former civilian constitution, however. unitary (or two-tier) countries. The three-tier countries can This is due to changes in its treatment of the municipal tier be further divided between those in which political decen- of government. The 1988 constitution recognizes munici- tralization largely consisted of a reversion to type after a palities as a third, independent tier of government (rather period of authoritarian rule and those that have opened sub- than as creatures of their respective states, as formerly.) It national offices to electoral competition for the first time. radically increased municipal revenues by doubling the pro- Argentina is the least radical of the group. Expressed in portion of federal taxes that must be shared with the terms of shares of total public expenditures, it appears to be municipalities and by requiring the states to transfer an one of the most decentralized countries in Latin America: increasing share of state-collected taxes to the municipali- Provincial spending accounts for nearly half the total. But ties within their jurisdictions.1 The new constitution made in arriving at this condition, Argentina has traveled only a no attempt to redefine the distribution of functions between small distance during the last 30 years. tiers of government, despite its reallocation of revenues. Argentina's federal structure dates from 1853. From the After an aborted effort to legislate a systematic devolution outset, provinces have had elected governors and legisla- of functional responsibilities, the federal government fell tures. Despite a history of coups beginning in the 1930s back on case-by-case negotiations, which are still in process. (and extended periods of dictatorship during 1946-55, 1966-73, and 1976-83), Argentina has functioned as a Civilian Regimes Unleashing Local Electoral multiparty federal democracy for much of its history. Competition With the departure of the military in 1983, Argentina Colombia, Venezuela, and Mexico have relatively long- reverted to its former federal structure. Decentralization standing democratic regimes. In political terms, decentral- efforts under the subsequent civilian regimes have been rel- ization in Colombia and Venezuela meant the replacement atively limited. Under the Menem administration, control of appointed governors by elected ones. In Mexico, it meant of federal hospitals and secondary schools was transferred to the emergence of genuine political competition at the the provinces (which already operated all primary schools subnational level. 11 BEYOND THE CENTER DECENTRALIZING THE STATE Except for a military intervention in the early 1950s and urban roads, in practice these functions were performed by a 16-year period of joint administration by the two major agencies of the federal government. parties (1958-74), Colombia has functioned as a multi- Venezuelan decentralization began in 1989 with legisla- party democracy since 1886. The country has, nevertheless, tion authorizing the direct election of state governors. been politically centralized. Prior to the recent reforms, the Municipal legislation enacted in the same year replaced the president appointed the provincial governors, who in turn closed, blocked, party list system with an open, unblocked appointed municipal mayors. In functional terms, Colom- system in which voters could express their preferences for bia was also highly centralized. Education and health were individual candidates. New legislation also established a either directly provided by the central government or process for decentralizing federal government responsibili- deconcentrated to the provincial level. Except in large ties to the states. This specified a list of functions-includ- cities, the water supply, sanitation, and roads were also ing education, health, and "public works of state inter- responsibilities of the central government. est"-that could be transferred to the states at the initiative Colombia's decentralization began in 1983 with the of individual state governors. The legislation thus permit- decision to strengthen subnational sources of revenue and ted each state to choose which functions it would assume to grant subnational governments more discretionary and which it would leave in the hands of the federal gov- authority on tax rates and overall tax administration. This ernment. Implementation has, however, been thwarted by path was reinforced in 1986 with the decision to permit the a lack of agreement on how the decentralized functions direct election of mayors and the transfer of significant rev- would be financed. Since the states themselves have virtu- enues and responsibilities to municipalities. Responsibilities ally no taxing powers, they are entirely dependent on fed- transferred to the municipal level included oversight of edu- eral transfers. Although the legislation established general cation, health, water supply, and local road construction guidelines on how states would be compensated, the and maintenance. Although the electoral and revenue process of federal-state negotiations was cumbersome and transfer reforms were implemented, municipalities were left wide discretionary margins to federal budget authori- slow in assuming-and the central government in effec- ties. As a result, although many functional transfers are tively transferring-the new responsibilities. A new consti- under study, few functions have been effectively transferred tution in 1991 and the 1993 law on the distribution of to any states. The only functions that have in fact been resources and responsibilities called for a phased transfer of transferred have been those that generate their own rev- education and health, first to departments, then to munic- enues-toll roads and bridges, for example. ipalities (through a certification process). In addition, the Although nominally a multiparty democracy with a constitution of 1991 authorized the direct election of three-tier federal structure, Mexico has historically had nei- provincial governors. The new constitution also raised the ther of these characteristics. Mexican politics have been 1986 level of transfers to subnational governments to dominated by a single party-the PRI-since the party's almost 50 percent of current revenues and made them pre- founding in 1929. The PRI has in turn functioned as a dominantly formula based. The increase was particularly highly disciplined party under the direct control of the pres- large for municipalities, though their responsibilities were ident, who is "designated" by his predecessor and serves a not increased and remain ill defined. All three measures single six-year term. As head of the PRI the president has have since been implemenred. customarily designated the candidates for state governor- Venezuela, like Colombia, has no recent history of ships. At the municipal level the prcsident has also exer- authoritarian government (since 1958) but has a long tra- cised final control over the choice of PRI candidates for dition of political centralism. Until 1989, state governors political office.2 PRI gubernatorial and mayoral candidates were appointed by the president. Mayors were elected, but have then been "elected" in political contests with only on a closed, blocked, party list system, which effectively token political opposition. In his capacity as chief executive, gave national party leaders control over local politics. States the president has had the power to remove sitting gover- functioned as administrative agencies of the central govern- nors and replace them with provisional governors (with sen- ment. Although municipalities were constitutionally ate approval). In functional terms, Mexico has traditionally assigned responsibility for water supply, sanitation, and been highly centralized, with the federal government 12 DECENTRALIZATION POLITI(CS IN COMMANI) responsible for education and health and the most infra- president. Smaller municipalities elected their own mayors, structure investment. Municipalities, in particular, have but these operated under the close prefectoral supervision of exercised a minor and ill-defined role, and have been largely presidentially appointed officers. Municipal responsibilities dependent on ad hoc grants from their respective state were extremely limited, with the central government governments. directly responsible for education, health, welfare, and most Mexico's decentralization began in 1977 with the pas- infrastructure services. sage of legislation setting aside one-quarter of all seats in Although the military regime abolished all traces of local the lower house of the legislature for members of opposition democracy, it greatly increased the functional responsibili- parties. While tentative, this marked the beginning of mul- ties of local government. Primary education was transferred tiparty politics. Further reforms (and increasing disenchant- to the municipal level (where the newly municipalized ment with the PRI) led to opposition victories in five schools were required to compete with private schools for municipal capitals in 1982 and in four states during central government education funding). Municipalities also 1989-95. assumed responsibility for the administration of the Changes were also made in the division of functions and national income support program and for primary health resources between levels of government. In 1984, the con- care. With the departure of the military in 1990, the stitution was amended to regularize revenue sharing to incoming civilian administration kept intact the functional municipalities, devolve the property tax from states to assignments imposed by the military. Political autonomy municipalities, and specifically define the functions of was not only restored but expanded. Mayors and councils municipal government. These include urban water supply, were elected in even the largest cities. Thus, through the street paving, and public security. While the fiscal reforms intervention of an authoritarian regime, Chile became one have been implemented, smaller municipalities have been of the continent's more functionally decentralized nations. unable to assume all the infrastructure responsibilities Bolivia has traditionally been highly centralized. Boli- assigned to them, and continue to rely on state agencies. In vian municipalities have historically not been geographical 1992 the government embarked on a more radical transfer subdivisions of the national territory. Instead, their bound- of functions from the federal government to the states, aries have been limited to urban areas. Much of Bolivia's effective in 1993. The federal government transferred territory thus had no local government, but instead fell responsibility for all primary, lower secondary, and teacher directly under the control of the central government. training institutes to the states. A new intergovernmental Municipal political autonomy has been a sometime thing. transfer system was created to finance the corresponding From 1878 to 1942, mayors were appointed directly by the costs. However, human resource policies and negotiations president (from candidates nominated by directly elected with the powerful federal teachers' union remain at the fed- councilors). In 1942 a new municipal code authorized direct eral level. Government health services were also transferred elections, but it was in force only five years. From 1949 to to the states, and a corresponding financing mechanism set 1987, no local government elections took place, municipal up. A growing number of sector-matching grant programs councils were abolished altogether, and mayors were once now complements earmarked transfers and general revenue again appointed by the central government. sharing for the states. Bolivian decentralization began with municipal elections held-for the first time in decades-in 1987. A compre- Two-tier States hensive package of decentralization legislation was then The three remaining cases are all two-tier countries. Two passed in 1994. This had three major components. First, it have reverted to local democracy after a period of military redefined the boundaries of municipalities to cover both rule. The third has introduced municipal elections for the rural and urban areas. In effect, this extended the jurisdic- first time in decades. tion of municipalities to the entire national territory. Sec- Prior to the recent military regime (1970-90), Chile had ond, it increased municipal revenues by doubling the scale a two-tier structure of government-central and munici- of revenue sharing (from 10 to 20 percent of central pal-but with a high degree of central control. The mayors national sharable taxes) and by giving municipalities exclu- of thc four largest cities were directly appointed by the sive (rather than concurrent) authority to impose vehicle 13 BEYOND THE CENTER: DECENTRALIZING THE STATE and property taxes. Third, it transferred a wide range of during the Peace Accords of 1996 when a 1 percentage physical assets-along with responsibility for maintenance point surcharge on the value-added tax (VAT) was trans- and new construction-to the municipal level. Notable ferred to municipalities. Eighty percent of the constitu- among these assets were primary and secondary schools, tional transfer and 90 percent of the VAT are now broadly health posts, hospitals of the second and third level, and earmarked for capital investment. These formula-driven, neighborhood roads. (Note that staff employed in these revenue-sharing transfers are complemented by a variety of facilities-including teachers and health workers-contin- planning (Consejos de Desarrollo), grant (social investment ued to be paid by the central government.) funds), and credit mechanisms-some of them similar to More than decentralization in any other country in Latin matching grants-that partially preserve the influence of America and the Caribbean, Bolivia's reform efforts put the the central government on allocation of resources at the emphasis on the organization of civil society for participa- local level. Although Guatemalan municipalities have a tion in allocation decisions and control actions at the local wide range of responsibilities of their own, local revenues level. As a result, participatory organizations grew dramat- are extremely limited. Indeed, the national level manages ically during the initial years of implementation of the Law most local tax and fee policies and the central government of Popular Participation (1994), and citizens have demon- administers the property tax for most municipalities. strated an active interest in local investment planning and Table 1. 1, below, summarizes the decentralization expe- monitoring of municipal governments. rience of the eight countries, distinguishing decentraliza- Municipal elections were reestablished in Guatemala uon among its political, functional, and fiscal dimensions. soon after the transition from military to civilian rule. The Table 1.2 attempts to rank these eight countries accord- 1985 constitution also established a transfer of 8 percent ing to their current degree of decentralization as of 1999. (now 10 percent) of national current revenues to munici- The four variables in the table are intended to measure the palities. Transfers to municipalities were again strengthened extent to which power is held by "autonomous elected sub- TABLE 1 I Summary of Decentralization Measures in Selected Countries COUNTRY POLITICAL FUNCTIONAL FISCAL Federal (or multitier) countries Argentina Federal country reverts to type Secondary education and health transferred Minor adjustments in revenue sharing to provinces Brazil Federal country reverrs to type, with new No explicit reallocation of functions Increase in revenue sharing to municipalities constitutional guarantees for municipalities Colombia Introduces election of mayors and governors Transfers primary education, health to Increases earmarked transfers for social provinces service expenditures in provinces, general revenue sharing to municipalities Venezucla Introduces elcction of governors, Introduces optional decentralization of social Increase in funding is to follow direct election of mayors services, infrastructure to states decentralization of functions Mexico Concedes opposition victories in states, Transfers education and health to states; 'water Increases earmarked transfers for social municipalities supply, paving, public security to municipalities services in states Unitary or single-tier countries Chile Reverts to elected mayors in small Transfers primary education, health to Increases earmarked transfers for social municipalities, introduces election of municipalities services to municipalities mayors in cities Bolivia Introduces elec-ion of mayors Transfers primary schools, clinics (physical Increases general revcnue sharing to assets only) to municipalities municipalities Guatemala Reverts to elected mayors Increases general revenue sharing 14 DECENTRALIZATION POLITICS IN ( OMMAND TABLE 1.2 Indicators of Decentralization INDEPENDENCE OF SUBNATIONAL SHARE LLECTORAL SUBNATIONAL PARTY Of TOTAL SPENDING' FUNCTIONAL RESPONSIBILITIES AUTONOMY ORGANIZATION '.1, IN MAJOR SECTORS Brazil Both levels elected Strong 3 X Education, health, police, most highways Argentina Both levels elected Weak 45 Education, health, some highways Colombia Both levels elected Strong 47 Education, health, somc highways Venezuela Both levels elected Weak 19 Mexico Both levels elected Weak 30 Education, health Chile Local level elected Strong 15 Edducation, health Bolivia Local level elected n.a. 10 Guatemala Local level elected n.a. 11 1 Excludes social security from cenrad government expendrrares. 2 Excludes expenditures financed through cntral government transfers to man,cipal schools and health posts. n.a.- nou available S,o,re,: GFS, country data. national governments capable of taking binding decisions ers. A large subnational share of total public spending does in at least some policy areas."3 The first variable, electoral not imply that these expenditures are made independently autonomy, measures the extent to which subnational offices of central government control. Even nominal functional are filled through elections, rather than by appointment. responsibility does not imply management control, as illus- The second, independence of subnational party organiza- trated by the case of Mexico. Despite many similarities in tions, conveys the degree to which central political party the decentralization experience of Latin America, precise organizations control the choice of candidates for subna- comparisons continue to be elusive. tional office and can influence their subsequent careers. The third, subnational share of total spending, indicates the Notes degree of expenditures accounted for by subnational gov- 1. The new conlstitution also mandated ani inicrease in federal tax ernments. The fourth variable, functional responsibilities in sharing to states. As a group, the states nevertheless made little net major sectors, shows the degree of management autonomy gain from the transfer reforms, because the increase in federal tax held by subnational governments over major sectors. sharing was offset by the federally mandated increases in state tax These are obviously crude indicators. The existence of sharing with municipalities. 2. As with other appointments to the highest state position, these local elections does not imply complete political autonomy. deiin weemd.ocretyb h rsdn,hg-akn decisions were made concurrently by the president, high-ranking (Nor does the absence of local elections imply that central party officials, and the state governors. government decisions are immune from local influence.) 3. This is the classic definition of decentralization, used by Smith Party control may be strong in some areas and weak in oth- (1996). 15 CHAPTER 2 Getting the Rules Right: A Framework for Subnational Government iven the magnitude of changes in these Latin American countries, it would be useful to have a normative framework for decentralization-a set of guidelines that would help countries that are under pressure to decentralize to not only maintain political stability, but also improve the efficiency and responsiveness of public service delivery while preserving macroeconomic stability and equity in the delivery of social services. The academic disciplines of public finance, political science, and organization theory have all attempted to produce such guidelines. Public finance sees this as an assignment issue-a question of dividing functional responsibilities among levels of government and allocating corresponding revenue sources. Local politics, in this model, serve to "clear the market" in local public goods and services: Citizens are confronted with a menu of alternative combinations of services and tax levels and can express their pref- erences through the ballot box. But this model assumes that all citizens have equal access to the political process and that politicians act simply as the agents of voters. Political science takes a more jaundiced view of the polit- None of these conceptual approaches alone is adequate. ical process. It sees politics as a device for mediating conflict This report takes the view that the most useful way to see among interest groups, and it takes a more complex view of decentralization is as a nmanagement reorganization of the the motivations of politicians. It explains the behavior of public sector, in which the rules of hierarchical bureaucracy politicians as the pursuit of political power and emphasizes are replaced by a much more limited set of constraints on the role of key constituencies and party leadership in deter- the behavior of subnational political actors. The difference mining the rate of advancement in a political career. can be understood by contrasting a stylized authoritarian Organization theory calls attention to the problem of system with one that is politically decentralized. In an incentives and the difficulties that arise when one group- authoritarian system, policy is made at the top and exe- say, local taxpayers-must rely on another-say, local cuted through a hierarchy. Each level is accountable to the politicians-to act on their behalf Conflicts can arise one above. Detailed instructions emanate from central gov- because the interests of the taxpayers, the "principals," dif- ernment ministries. Local administrators are motivated by a fer from those of their "agents," the politicians. When career path that leads upward through the ministerial hier- agents have privileged information and are difficult to mon- archy. Advancement depends on the favorable opinion of itor, the interests of the principals can be thwarted. Orga- immediate superiors.' With decentralization, this hierarchi- nization theory attempts to overcome this problem, by cal relationship is cut-or more accarately, attenuated. proposing ways in which the interests of principals and Because local politicians now owe their office to the local agents can be made compatible. electorate, they become increasingly mocivated by a political 17 BEYONI) THE CENTER: DECENTRALIZING THE STATE career that requires the favorable opinion of local interest of political decisionmaking, rather than to an ideal world of groups and party leaders. As a result, subnational politi- theory. cians may find it in their interest to act counter to the wishes of the central government. But not entirely. In a Dividing Up National Political Power country (as opposed to a voluntary association of states) It is the national government that makes the rules under subnational units are not sovereign. As long as a country which subnational governments operate. The power of sub- remains intact, rules will constrain the behavior of subna- national interests in the national government, therefore, has tional politicians. Decentralization thus implies not subna- a key bearing on how the intergovernmental rules are tional sovereignty, but a new set of rules that define the revised and enforced. For Latin America the risks involved relationship between national and subnational governments are exemplified by Brazil's recent state debt crisis. There, and replace the rules of hierarchical bureaucracy. (In prac- strong regional interests in the national legislature acqui- tice, this contrast between authoritarian and decentralized esced to the accumulation of massive state debts, ultimatelv regimes is somewhat exaggerated. Few authoritarian requiring a federal bailout of equally massive proportions. regimes in Latin America have managed to govern without The obverse case is Argentina, where a more centralized conceding some important autonomy and influence to sub- system of governance-both in the formal structure of gov- national interest groups.) ernment and in the working rules of the political parties- The rest of this chapter deals with these rules. Such rules permitted the national government to force spendthrift come in a variety of forms, not all of them legislation. Party provinces to adjust. nominating practices, for example, can have a strong influ- Democracies have long struggled with the question of ence on whether central government interests prevail over how to represent regional interests in the national govern- local ones. For example, the traditional Mexican practice in ment. Over three decades ago, Piker (1964) noted that in which the president designates party candidates for guber- any federal system (defined as a system with multiple tiers natorial office reinforces the authority of the national gov- of elected government), the constituent units have an incen- ernment. In Brazil, these decisions are made at the state tive to undermine federalism by trying to get a free ride- level, thus weakening the authority of the national govern- that is, trying to get benefits without paying for them. ment. Rules also take the form of customs and expectations. Are there ways to guard against this? Political science No law prohibits the U.S. government from bailing out suggests several. One strategy is to strengthen the office of defaulting states; no Brazilian law requires the federal gov- the presidency in relation to the legislature.2 Constitutions ernment to do so. Yet in both cases these are well-estab- vary in terms of the powers granted to presidents. 'T'hese lished practices that profoundly influence the expectations include the power to rule by decree (that is, to promulgate of banks and politicians. laws without legislative approval), and the size of the Explicit legislation is nevertheless a key element of suc- majority required to overturn a presidential veto. In Brazil, cessful decentralization. Explicit rules reduce uncertainty for example, a veto can be overridden with a mere 50 per- and provide a common set of ground rules for the political cent majority plus 1. In Argentina, it requires two-thirds. process. They provide a focal point to coordinate subna- In principle, presidents with broad powers to rule by decree tional governments' reaction against an abusive central and an unassailable veto are better positioned to defend the government, while restricting the subnational govern- national interest against the parochial interests of individual ment's scope for bargaining. "Superlaws" can be used to legislators. strengthen the credibility of the most important rules. Electoral rules that discourage party fragmentation can These superlaws can take the form of constitutional provi- also help. Party fragmentation makes policymaking, in sions or other forms of legislation that can be altered only general, more difficult and weakens the position of the pres- by exceptional majorities or complicated amendment pro- ident, in particular. In a country with numerous parties, a cedures (Linz and Stepan 1997; Elster 1995). It is impor- presidential candidate typically has to run as the head of a tant to note that even superlaws are made and enforced by coalition. He is thus forced to promise key offices or key self-interested human beings. The discussion that follows is policies to other parties. Once victorious, he is constantly therefore intended to be relevant to the second-best world faced with the prospect of abandonment by coalition 18 GETTING THE RULES RIGHT. A FRAMEWORK FOR SUBNATIONAI. GOVERNMENT members (Carey 1997). Coalition partners may have age party consolidation, since a party has to win a majority enough power to block change, but not enough leverage to (or plurality) in order to gain any representation at all. In effect positive change on their own (Roubini and Sachs principle this would argue for a shift to the single-seat 1989; Alesina and Perotti 1998). approach. Mexico and Guatemala have, in fact, opted for a Many Latin American democracies are subject to this mixed system, in which half the seats in the lower house are difficulty. As shown in Table 2.1, all the listed countries, elected from single-seat districts (60 percent in Mexico), except Argentinia, Colombia, and Mexico, have at least five with the rest elected on a proportional basis. The Brazilian parties represented in the lower house of the legislature.3 Congress is currently considering a similar plan. The rule is One way to reduce party fragmentation is to change not infallible, however. As shown in Table 2.2, France, electoral rules. The proliferation of parties in Latin America which uses the single-seat approach, has nine parties repre- is traceable in part to the widespread use of proportional sented in the national assembly. representation as a means of electing candidates (see Box Rules that affect party discipline also can strengthen 2.1). This encourages the proliferation of parties, because it central control over subnational governments. Consider the allows any party to scour an entire electoral district in procedure used to nominate candidates. The Mexican prac- search of enough votes to win a single seat. (If a district has tice, in which the president ultimately approves the selec- 10 seats, a party can win representation with just 10 per- tion of all Partido Revolucionario Institucional candidates cent of the vote.) Single-seat districts, in contrast, encour- for public office, fosters considerably stronger party disci- TABLE 2.1 Electoral Rules, Selected Latin American Countries NO. OF EXECUTIVE PARTIES* LOWER HOUSE UPPER HOUSE Argentina President directly elected 3 Directly elected by multisear provincewide Directly elected by province, equal seats per province electoral districts Brazil President directly elected 7 Directly elected by multiseat statewide Directly elected by state, equal seats per state electoral districts Bolivia President directly elected 7 50% by single-seat electoral districts: Directly elected by department, equal seats per 50% by multiseat districts department Chile President directly elected 6 Directly elected by electoral districts (not Directly elected by region with equal seats, 7 regions specified in Constitution whether single seat) having 2 seats each, the remaining 6 being divided into 2 electoral districts, each with 2 seats Colomnbia Presidenit directly elecCed 2 Directly elected by multiseat departmentwide Directly elected by party list from nationwide electoral electoral districts district Guatemala President directly elected 5 80% elected by multiseat departmental None districts; 20% by single-seat districts Mexico President directly elected 3 40% elected by large multiseat districts; Directly elected by state, equal seats per state (with one 60% by single-seat districts seat in each state reserved for the runner-up party) Nicaragua President directly elected 6** Directly elected by multiseat "regional" None districts Peru President directly elected 6 Directly elected by single national multiseat None district Venezuela President directly elected 5 Directly elected by multiseat statewide Directly elected, equal seats per state, plus ex oficio electoral districts seats for ex-presidents *Number of paries represented m lo-er hoase of congress, excluding partmes ith less than 3 percent cif seats. *tlncludlmg four parties in government coalition. 19 BEYOND THE CENTER: DECENTRALIZING THE STATE A GlosIy of Poitical Terms~ Proportional~ ~~~ rr'snaini yfrtemostcommonsys~ set o yte at.Prylsts May also be 'bloke,i tern for choosn congresional candidates~ in atn" 'Amr whiccs voters are not allowed to cast votes~ for candi- icR. As shown in Table 2.2, it is also used in Italy and dates of different parties. Unblocked systems permit vot-~~~~~~~~~~~~~~~~~~~~~~~~~~0 Al If iien SainE Underaprportional rpresentatio system,the eriodvd hi oe mn difentpris county isdividd ino reatiel larg districts-usall The altenative is agk .tnl.s lcoa itrict ~systemn in correpondig to he bondaris of xistig sAtates or whic te national territory sdvdditvml lcoa provinces. Each districtis assigned severalseats. Cani. districts, each of whfich is assigned onily one se T.inde dats or parties~compete atlarge6 truhout th dstrict, this sysem a stingle winne ielcdeither bypualIty EThenAumbe of seats Vwonbye.ach partyin tedistrict is or-throug ~h6aunffby anasoue aority.The sin bedonr itspe"rcetge of teoalve.if the patylit is~ Rgleeat sy$Tf t tvpr stem isused in the U.S. the U.K. and Fran ce. closed, th party ledership cooses whic of its cndidates o Latin A erican country reis excluivelyo it. Meico actuall ~takes office.~ If the party lis is open, voters can And Gatemdala hiow"ever, use a mixed'system,rn which choose amonag indivdual~ candidateswithinl a party list, soesasaeeetdfocigesat districts and others with th c ~andidates' w1inn inthe most votes occupying he nlby pootoa ersnain pline than its U.S. counterpart, in which this task is per- her choice. Voters vote for a single candidate by name. formed by local branches of the national parties. By the Although the number of seats is awarded to each party on same token, the closed voting system used in Venezuela fos- the basis of proportional representation, parties are obliged ters greater party control than the open system used in to assign seats to the candidates on the party's list who win Brazil. Brazil, in fact, has a uniquely open system of candi- the largest number of votes. As a result, parties may find date selection. 4 Anyone meeting age and residence qualifi- themselves represented in congress by members who have cations can run for office under the party banner of his or no links to the party leadership. TABLE 2.2 Electoral Rules, Selected OECD Countries NO. OF EXECUTIVE PARTIES LOWER HOUSE UPPER HOUSE France President directly elected; prime 9 Directly elected by small (577) single-seat Indirectly elected bv electoral col'ege, no minister elected by lower hoose electoral districts specific regional representation Germany Prime minister elected by lower house 7 1/2 elected from nationwide multiseat Ex oficio 3-5 seats per stare (figurehead president) party list; 1/2 elected by single-seat subprovincial districts Italy Prime minister elected by lower bouse 6 Directly elected by muttiseat subprovincial Directly elected by regions; nomber of seats districts (32) proportional to populatioo Spain Prime minister elected from lower house 4 Directly elected by multiseat provincial By province equal seats per province, 81% (constitutional monarch) districts by proportional representation di rectly elected; 19% appointed by provincial legislatures U.K. Prime minister elected by lower house 3 Directly elected from small single-seat Hereditary and appointed (figurehead mioniarchi) eleCLoral districts U.S. President directly elected (pro forma 2 Directly elected from small (435) single-seat Directlv elected by stare; equal seats per state electoral college) electoral districts 20 GETTING THE RULES RIGHT A FRAMEWORK l'OR SUBNATIONAL GOVERNMFENT Assigning Functions, Structures, and Revenues and in Mexico and Argentina. In Brazil it became a mixed In principle, the rules governing the structure, functions, responsibility of the states and the municipalities. and revenues of subnational government should have an What should be the functional role of subnational gov- important bearing on the impact of decentralization. The ernments? The literature on fiscal federalism provides a use- "right" design should produce a more efficient and respon- ful framework for thinking about this question. Fiscal fed- sive public sector, while guarding against macroeconomic eralism assigns three roles to the public sector as a whole: instability or growing disparities in the provision of social macroeconomic stabilization, income redistribution, and services. Unfortunately, the empirical basis for recommen- resource allocation (in the case of market failure) (Oates dations in this area is weak. While rules have been derived 1972; Inman and Rubenfeld 1997). The model assigns the from first principles from such fields as organization theory first two roles to the central government. The central gov- and public finance (as discussed below), the empirical evi- ernment is assigned responsibility for stabilization on the dence to support any particular approach is limited. The grounds that local economies have no access to an indepen- Organization for Economic Cooperation and Development dent monetary policy and are too open for effective coun- (OECD) models are of some use. The OECD countries- tercyclical measures. The income redistribution function is with relatively long histories of multitier democratic gov- also assigned to the central government, on the grounds ernment-might be expected to have converged upon a that local attempts to address income disparities are likely single-best-practice approach. But in fact the differences to provoke higher-income groups to move to low-tax areas among them are as pronounced as the similarities. The and low-income groups to move to high-benefit areas. combination of first principles, evidence from OECD coun- Subnational governments enter the picture only with tries, and the ongoing experience of Latin American coun- respect to the third function: resource allocation. The the- tries in the process of decentralization nevertheless provide ory argues that if the benefits of particular services are some basis for normative statements. largely confined to local jurisdictions, welfare gains can be achieved by permitting the level and mix of such services to Functions vary according to local preferences. Local consumers, con- Traditionally, the Latin American municipio has performed a fronted with the costs of alternative levels of service relatively minor role in the production of public goods and (according to this viewpoint), will reveal their preferences services. The nineteenth century municipality provided ser- by voting for rival political candidates or moving to other vices fitted to the small scale of urbanization at the time- jurisdictions. In this respect, local politics can approximate the regulation of public nuisances, street repair, public the efficiencies of a market in the allocation of local public lighting, and garbage collection. In the large cities that had services by "pricing" municipal services and relying on the emerged by the end of the century, private (often foreign- local political process and household mobility to clear the owned) concessionaires normally provided public utilities- market. water supply, electric power, and mass transit (streetcars). There are a number of limitations to this approach, how- In response to the large-scale urbanization of the 1950s ever.5 As a practical matter, it is often difficult to define the to 1970s, national governments reacted not by strengthen- scope of benefits of a specific service and match it to a spe- ing the role of municipal governments, but by expanding cific jurisdiction. Although the benefits of defense may be the role of the central government. In Brazil, the private unambiguously national and the benefits of public lighting water and power companies were nationalized and consoli- unambiguously local, the largest sectoral expenditures of dated into state-level companies under the control of feder- government-education, health, and transportation-fall ally owned holding companies. Similarly, in Chile, Mexico, somewhere in between. Education, for example, can be con- Peru, and Venezuela, national water companies assumed the sidered to have extremely localized benefits, reflecting the responsibility for the water supply. The expansion of educa- impact of education on the future income of pupils. But it tion and health services was also usually a central govern- also has national benefits as a vehicle for poverty alleviation ment initiative. Thus primary education, once it passed and political acculturation. Spending on education, for from the domain of the church, became a central govern- example, is often the largest form of income transfer to low- ment responsibility in all the unitary states in the region, income populations. A common curriculum is often one of 21 BEYOND THE C:ENTER: DECENTRALIZING THE STATE the few effective means of weaving diverse ethnic interests local governments to act on their behalf Even programs into a functioning societv. It is, therefore, difficult to imag- that are unambiguously central have to be administered on ine that a central government would be content to Fermit the ground. A program of direct income transfers to the a subnational government to abolish all public education poor, for example, requires field offices to determine who is within its jurisdiction. But neither is it necessar-y for central eligible. Creating a separate unit of field administration in government to manage every detail of public education to each municipality would result in unnecessarily high ensure that its interests are addressed, administrative costs. There are also administrative constraints to the theoret- What in fact characterizes the division of functional ical model. Public services are subject to economies of scale, responsibilities-in the long-standing multitier democra- particularly in the use of technical staff. Not every village cies and many countries in Latin America and the Carribean can use a full-time homicide detective, even if the benefits -is a complicated set of principal-agent relationships, in of his services are entirely local. Nor can a town with 200 which subnational governments act both as agents of students afford to offer 12 years of grade-differentiated edu- higher levels of government and as principals (or more pre- cation. Local governments may, therefore, find it advanta- cisely as agents of their constituents) in the delivery of local geous to contract with larger, higher-level authorities to services. This typically involves at least two tiers of subna- produce specialized services. By the same token, central tional government. It may also involve the private sector in governments may find it advantageous to contract with various roles as concessionaires or operators of public utilities. BOX 2:2 Howi Far to DeWhtn1ra z? One of the vaunted betiefits 6 d6et ntalization4is its abil- Mefxico-is therefifre unliklya to impro-ve quality Oni the ity to improve the efficiency and resp3onsi-veness of govern- other,hand, effoirts in ElSalvdor,Nicaragua,and the state ment. According to this argument, local officia hv bet ofMiasN Geas ~razil, to sh M maa et, power to ter knowledge of local conditions than central governiment COmmubnity grusadsho ietr r ieyt how officials and are thuis better positioned torspnd to4 loca results'. tastes and prefrences. Anid b~cas the a ~re mre~ ace scs- i In additin decnraato mynobe thems ble to their constituentis, they can~ More eail bhed imuportant stucualrfiormgvrmet a tke. For accountable for res~ults. Buit the evidence inChpters 4, 5, exm ple, in te hathTstor~ (Catr )he mostimie por- and 6 of this report rais~es some interestinlg questios:~ tant strutra efr ha noti6been'the decentrlizatio of H-ow much decentraliz: tion is enlough?.~ Anid i it always publi hospitals to 1oca gover mens lnsead it ha been necessary? a oefUdmnalsiti h ublic sector rle as a Cha~pter 4 st Is' that transferingmanagement whol from tatio prdueofhat servcestothto resp~onsibI11i4 for i ima . education to local goenments a insrrohelhcaepoie yth rvt etr may no bc e sffci mt. It is onywhen mtanagement s', TheNC dscussio ofroa decentralization (Chapter 6)~ sug- decentralized to cohimnunity school boards "and ~school Aesft ~thiat piV~ari ZAtion-AriuarlyI In road mainte- directors that the positive impacts of decetrlzain nancer h ha fa orte makdImpact oni road perfor- appear. As descrfibe in Chapter 4, improvement in school mance t&hani thffo66rts to decentralize parts of the hi'ghway quality is poisitivel~ cot reated with (a commn ity influ- network'i to sbnaionall governments.i While decentfraliza.- ence over the hirin anId dismissal of school directors, and tion from onetiero overnkmendtitto another ma imnprove (b) the influe~nc oef scho ol directors ovepr the recruitment, efficewny anid responsiveness in~ some sectors, it may~ not be promotion, and d ismnissail of teachers. Decetaidzing edo- sufficient in irseWo ee the~ ~ighesr priorl for sitruc- cation to state gov~ernm ents per se--as In iColomibia or tuaPrfom,K in all cases. 22 (ETI IN6 THk RULES RIGHT. A FRAMEIWORK iOR SU3NA'IIONAL GOVbRNMEN I Individual services are often unbundled, with each level each tier are well defined. Revenue assignment is fixed. performing a role that reflects its interest or comparative Intergovernmental transfers are allocated on the basis of advantage. In Northern Europe and North America, for formulas, rather than through negotiations. example, primary education is de facto a concurrent respon- The absence of these characteristics has complicated sibility of central or state governments and local govern- government operations in countries that have decentralized ments. But the role of the higher level of government is in Latin America. In Brazil, for example, there is no explicit largely confined to financing, rather than day-to-day oper- division of functional responsibilities between state and ations. Because education is seen, in part, as a tool to alle- municipal governments. Under the constitution, states are viate poverty, central governments seek to ensure a mini- authorized to legislate about "any subject not otherwise mum level of education funding in all jurisdictions, assigned to the federal government." Municipal govern- regardless of their local tax bases. Local governments or ments are authorized to "legislate over any matter of local school boards, however, are assigned responsibility for day- interest." As a result, crises tend to be met with conflict and to-day management decisions. Thus schools boards are finger pointing among politicians at different levels, rather often given responsibility for personnel management (deci- than by the concerted action of one tier of government. sions on the hiring and dismissal of staff, salary schedules, Another problem is that central governments appear to conditions of work, and career structure) and the allocation delegate management responsibility but then fail to relin- of the budget (for teaching and administrative personnel, quish enough management control for the recipient gov- teaching materials and supplies, maintenance of school ernment to exercise it. In Brazil, for example, state and buildings, and other programs). The central government municipal governments were (until recently) constitution- role may also consist of establishing the rules under which ally prohibited from dismissing staff or reducing nominal local decisions are made. In England (United Kingdom), for salaries.6 This hamstrung subnational governments in their example, primary education is nominally the responsibility efforts to adjust to falling revenues and abrupt increases in of counties and metropolitan districts and boroughs. The nonpersonnel costs, resulting in a massive accumulation of 1988 education reform act, however, requires local govern- debt. Similarly in Colombia, the central government- ments to adopt a nationally standardized curriculum and a while nominally decentralizing education to the mixed public-private governing board for each school dis- provinces-maintained its authority to determine teacher trict, and requires central approval of the formula to be salaries. The government then granted a major salary used to distribute funding among individual schools. increase to the teachers. As the provinces lacked the In some sectors, local governments may also operate resources to pay the increase, the central government was strictly as agents of the national government. In Germany, ultimately required to create a separate transfer program to for example, Lander (states) act explicitly as agents of the enable them to meet their payrolls. federal government in the construction and maintenance of In other cases, the central government takes responsibil- federal highways, but act as principals in the provision of ity for financing but then fails to follow through. Chile's police services. In France, the departments are paymasters system of education financing, for example, went through a in the administration of national social assistance, but act period of crisis in the 1980s when the central government on their own in providing local roads and regulating urban dramatically reduced per-pupil capitation grants in an transport. The principal-agent relationship can work in attempt to reduce its deficit. Peru, similarly, attempted to other ways. Small gemneinden (municipalities) in Germany decentralize much of its road network to provincial and contract with county governments to provide primary and municipal governments but then failed to provide financ- secondary schools. In France, small communes contract ing. (This resulted in widespread deterioration of the net- with private international water companies, and contiguous work and, ultimately, a recentralization {see Chapter 61.) municipalities often create a metropolitan association to The form and content of the rules governing intergov- manage services such as transit or waste disposal. ernmental relationships involve a delicate balance between While these relationships are complex, they are charac- the central government's interest in subnational perfor- terized by a high degree of stability and clarity in the link mance, on the one hand, and the demands of local auton- between each tier of government. The responsibilities of omy on the other. 23 BEYOND THE CENTER DECENTRALIZING THE STATE BOX 2~3 Onie of the classic ojcintodcenrlztosta rcived ascnrlgoverhnment employees. But this is a local govenrnments ar noptn.Ctn ttsiso tw-de wr.While tt maAkes it easier to decentralize illiterare m-ayors, crdHconigsses n ie staf itcn aeit diffi6cult for local government to adapt spread nepotis,citc rgeta loa goene ts are wgsadenfts to local conditions or to introdue incapabeoraig onexadeaunton,manaeetadpersonnel reform.~ Thsargumentinoascmelnastmy first Whl rnfern tff canI address the immiieditate appear. As a practia matr he ao sriei issue raise bydcnrlzto,teoealmngmnt decentralized existn fedsafaenraldeeta-wknsofoclgoemnteaisaauefor concern. ized WIth it. Thus whnpiayeuato a eetal- Lo sari,lwpetg, an th ihtrnover tat ized to the departmnsadsae nClmi n e-reslsfo xesv oiia nefrneiesne ico respctivly),exising entrl goernmnt tachrs, deiinIaaeit diffcul to attac and retain compe- wer deetaie tth aetm.Te bcm ols ettf,priua ly i ery smiall jurisdicios (or oe optn hnte a enwe hywr eomhspoe difficult The managemrent of pb Technia coptnehseegda rbeuhn poliia poesThcalngfrlclgvrnments is to tralized. nPeufrexml,mncetagormntand poiial n ueurtial esbe(es1991) highway engineers choset eierte hnacp ua eoremngmn sas challenge. Local employment in loca goenet 1oa s taff prvdic-gvmet blt oitoueperonlrfrs pabe of asuig h ak nisown (a proble excr Anldn efrncceevaluation mechanims, training, bated by the abseneo eta oenetfnnin o a ikdt rdciiy and incentive to attract and the newly decentraie rod)UhseetuVyldt h reItancmentskld personnel-4is often constrained coll1apse of the decnrlzto for,wihwsfloed b oeflpbi emlyee uin.Tee problems are by rcenralzaton. ovenmets an aciltat th trns- not niqe t loal overnmients, however. They are the fer of centaf gvrnet tf b eqiindo gogoen aefcosta fet6tecompetenrce of staff in central entst fe hm th aewae nbnft hey goermet Detailed regulation is more appropriate in some situa- Structure tions than in others.Where subnational governments act as The structure of subnational governments-the number of agents uf the central government, central regulation is tiers, the number of units in each tier, and any legal dis- required to ensure that they in fact do so. In even the most tinctions among units within a given tier-can affect the decentralized country, the central government will require performance of subnational governments. A structure that centrally financed welfare payments to be distributed fails to distinguish between major metropolitan areas and according to federal criteria. Regulation is required to small villages makes it difficult to clearly define the func- ensure the validity of the local electoral process and to tional responsibilities of local government. A structure with address conflicts between units of subnational government, too many tiers or too many units of government at each The spread of political autonomy at the local level, never- level can exhaust its resources on administrative costs. Latin theless, should allow for less reliance on central regulation America suffers from both problems. and mote reliance on interest groups to monitor the perfor- Almost all the major Latin American countries have a mance of local government, common structure of subnational government. Subnational 24 GETTING THE RULES RIGHT A FRAMEWORK FOR SUBNATIONAL GOVERNMENT governments are territorial subdivisions of the nation. In laws permit not only distinctions among different classes of countries with two tiers of national government, the municipal government, but also the formation of single- national territory is first divided into states (or provinces or purpose local units to provide such services as education, departments). Each state is then divided into municipios. fire protection, water supply, and mosquito abatement. (Countries with a single tier of subnational government Some degree of distinction among local governments forego the first step.) In rural parts of the country, a munici- (though perhaps not to the degree practiced in the United pio may be largely rural, with only a few villages within it. States) has much to recommend it. Such distinctions allow In urban areas, a municipio may be an entire city, or part of functional assignments to be customized to the capacity one. No legal distinction is made between rural or urban and requirements of different jurisdictions, and they reduce municipios or among municipios of different size. Instead, the need for bargaining and negotiation among tiers of gov- within each tier, subnational governments have the same ernment. Ironically, the trend in Latin America is in the enabling legislation. Thus the legislation for municipios in opposite direction. Until 1994 Bolivia was unique in having Brazil applies equally to the municipio of Sao Paulo (with a municipalities whose jurisdiction was confined to urban population over 11.2 million) and the municipio of Pirapora areas. The law was later changed to place Bolivia in confor- de Bom Jesus, with only 4,585. mance with its neighbors. This degree of uniformity can make it difficult to define Are there too many municipal governments in Latin the roles of subnational governments with any degree of America? Proliferation can be a problem for two reasons. specificity. Under such a system, the only roles that can be First, there are overhead costs to government, distinct from specifically assigned to the biggest unit are those that can the costs of delivering services. Increasing the iiumber of be performed by the smallest. In a situation where the municipalities presumably increases these costs in the smallest municipio may have a population of only a few aggregate. Second, proliferation implies a reduction in the hundred, this is extremely limiting. In practice, the use of average size of local governments. This can lead to losses in uniform structure for municipal government results in an economies of scale (as, for example, a case where a munici- ad hoc division of labor between tiers of government, in pality does not have enough work to support a full-time which large municipalities carve out a large role for them- homicide detective). selves, while smaller jurisdictions remain dependent on cen- By the standards of the OECD countries, Latin America tral (or state) government intervention. Under these cir- would not appear to have too many local governments. As cumstances, responsibility is ambiguous. shown in Table 2.3, the average population of a municipal- The alternative is a system that recognizes the differ- ity in the larger Latin American countries is in the tens of ences among local jurisdictions and assigns different thousands, exceeding 20,000 in all cited cases except Peru. responsibilities according to needs and capabilities. British In continental Europe, on the other hand, average popula- local government legislation, for example, has traditionally tions are an order of magnitude smaller, ranging from 1,600 distinguished between services that must be provided to 7,000. Large average sizes are found only in Japan and the throughout the national territory (such as education) and United Kingdom. The statistically large average populations those that are strictly urban (such as solid waste collection in Latin America are deceptive, however. In every Latin and public housing). Counties have traditionally been American country (except Costa Rica) they mask enormous assigned the former, while urban districts have been disparities in population size. The vast majority of munici- assigned the latter. The German structure of subnational palities in the region have fewer than 15,000 inhabitants. government distinguishes among governments of different Sixty percent of the municipios in both Chile and sizes, granting broader responsibilities to three "city-states" Guatemala have fewer than 20,000 inhabitants. Population (Berlin, Bremen, and Hamburg) and allowing other large size can be as small as Rosario Tasna in Bolivia (population municipalities to assume the responsibilities of counties. 185) and Santiago Tepetlapa in the state of Oaxaca, Mexico Similarly, Japan grants provincial status to Tokyo, Kyoto, (population 149), (Nickson 1995). Moreover, Europe is not and Osaka; Russia grants Moscow and St. Petersburg the necessarily a model in all respects. The large number of status of oblasts (provinces).The United States illustrates local governments in Europe reflects the same political local government specialization at its most extreme. State pressures for local control that exist in Latin America. 25 BEYOND THE CENTER: DECENTRALIZING THE STATE Where OECD governments have attempted reform, the to allocate the transfers of resources to subnational govern- direction of reform has been toward reducing the number of ments in many Latin American countries favor the further municipal governments. This is usually justified on grounds subdivision of municipalities. of efficiency and cost. There was a massive decrease in the number of school districts in the United States in the 1950s Revenues: Taxes and Transfers as jurisdictions tried to combine enough students in a sin- A third important element of the relationship between cen- gle jurisdiction to run grade-differentiated primary schools. tral and subnational governments is the assignment of and Germany has reduced the number of gemeinden by half. The control over sources of revenues. Money is power. The United Kingdom has eliminated a tier of subnational gov- degree to which subnational governments have indepen- ernment in Scotland, Wales, and metropolitan areas of Eng- dent control over broad tax bases influences their ability to land. In contrast, the tendency in some Latin American act contrary to the wishes of the central government. The countries has been to increase the number of municipalities. fiscal relationships between tiers of government are fraught The number of municipios in Brazil increased from 3,000 to with subtleties. Central governments may assign broad- nearly 5,000 in the 15 years following the return of democ- based taxes to subnational governments but limit their rev- racy. In Venezuela, the number of municipios increased enues through controls over rates. A central government may from 202 in 1985 to 330 a decade later. The criteria used ostensibly refrain from guaranteeing subnational borrowing, TABLE 2.3 Structure of Subnational Governments: Selected OECD and Latin American Countries AVERAGE OECD INTERMEDIATE LOCAL POPULATION Canada 10 provinces, 2 territories 4,507 municipalities 6,700 France 22 regions, 96 departments 36,772 communes 1,600 Germany 13 states, 3 city-states 329 counties, 115 county-free cities; 14,915 5,500 municipalities Italy 22 regions, 93 provinces 8,100 municipalities 7,000 Japan 47 prefectures 655 cities, 2,586 towns 38,800 Spain 17 autonomous communities, 50 provinces 8,097 municipalities 4,800 United Kingdom* 49 counties 540 rural districts, metropolitan districts, and 92,600 London boroughs United States 50 states 39,000 counties and municipalities** LATIN AMERICA Argentina 23 provinces 1,617 municipios 21,600 Bolivia 296 municipios 27,000 Brazil 27 states, 1 Federal Discrict 4,974 municipios 32,400 Colombia 32 departments, I Federal District 1,068 municipios 34,600 Chile 325 municipios 43,000 Ecuador 20 provinces 176 municipios 68,200 Mexico 31 states, I Federal District 2,412 municipios 38,600 Peru 194 provincial councils 1,624 municipios 14,778 Venezuela 23 states, I Federal District 282 municipios 78,000 -England only. ** Excludes special districts. Suroes: U.S. CIA 1997; OECD 1997; country-specifc sources. 26 GETTING THE RULES RIGHT. A FRAMEWORK FOR SUBNATIONAL GOVERNMENT TABLE 2.4 Principal Revenue Sources of Subnational Governments THREE-TIER COUNTRIES PROVINCIAL GOVERNMENT MUNICIPAL GOVERNMENT Argentina Fixed shares of total central government taxes; taxes on Varying shares of provincial taxes and transfer revenues, fees gross receipts, property, motor vehicles Brazil Otigin-based VAT, fixed shares of central government Fixed shares of state VAT and central income and excise taxes, property income and excise taxes and services taxes Colombia Fixed shares of total central government taxes; minor Fixed shares of total central government taxes; property tax, industry taxes on alcohol and tobacco and commerce tax, gasoline surcharges Mexico Fixed shares of total federal taxes; earmarked transfers Fixed shares of federal taxes (passed through states), property tax, for education and health, minor payroll and motor vehicle taxes business licenses Peru Fixed shares of national sales tax, taxes on vehicle purchases Shares of central government revenues. property tax and business licenses Venezue!a Fixed shares of total central government taxes Fixed shares of state revenue-sharing receipts, taxes on property, vehicle tax TWO-TIER COUNTRIES MUNICIPAL GOVERNMENT Chile Earmarked capitation grants for education, health, property tax Ecuador Fixed shares of central government oil revenue; taxes on property, business assets, vehicles, business registration Guatemala Fixed shares of total central government revenue, miscellaneous local taxes Bolivia Fixed shares of total central government revenues; taxes on vehicles and property but find itself forced to come to the relief of overindebted that user charges allow individuals to express their demand borrowers to prevent tumult in financial markets. for services whose benefits are largely private, such taxes In the abstract, the principles of revenue assignment are enable taxpayers to express their demand for local services straightforward. Finance follows function. The appropriate that are consumed collectively. structure of subnational finance-the mix of user charges, What should be the role of transfers? In the traditional taxes, transfers-depends on the functions that have been fiscal federalism formulation, intergovernmental transfers assigned to each tier of government. This is because differ- have a very limited role. They exist to "correct" the decision- ent sources of revenue have different effects on behavior and making process of local governments. Left to their own different patterns of incidence. User charges, for example, devices, local governments would be expected to base their impose costs directly on individual consumers and can budget decisions only on the benefits captured by their ration consumption by price. Therefore, there is an argu- constituents. This would cause them to "underprovide" ment for assigning user charges to finance services whose services that confer benefits on neighboring jurisdictions. benefits are confined largely to individual consumers, such Transfers, in this view of the world, can be used to improve as water supply and bus transportation. Certain forms of allocative efficiency by inducing local governments to take direct taxation are appropriate for financing services whose these wider benefits into account. To perform in this role, benefits cannot be confined to individual consumers but do transfers have to be earmarked-to ensure they are spent not extend beyond the boundaries of individual subnational on the service in question-and must be subject to jurisdictions. A property tax, for example, can be an appro- matching requirements to ensure that local governments priate way to finance local police services. In the same way do not reduce their own expenditures on the service by a 27 BEYOND THE CENTER: DECENTRALIZING THE STATE corresponding amount. Earmarked grants for sewage treat- be borne locally. To be cheaply administered, a tax must be ment plants or cost-sharing arrangements for national imposed on a large volume of taxable activity flowing highways that serve urban areas are often cited as examples through a small number of collection points. Few taxes can of such transfers. meet both criteria simultaneously. Taxes that meet the test But transfers also have a distributional role to play. This of localized incidence (such as property or business taxes) role is becoming more important as social services-partic- tend to involve large numbers of small-scale taxpayers. ularly education and health are decentralized to subna- Taxes that meet the ease of administration test (such as tional governments. Transfers permit central governments taxes on manufacturing and imports and origin-based to use local governments as agents of income distribution value-added taxes [VATs)) tend to involve large-scale inter- policies. As noted earlier, local governments are badly posi- jurisdictional incidence shifting.7 Local taxes also tend to be tioned to pursue such objectives on their own. Local subject to arbitrary assessment and incidence patterns. attempts to redistribute income from the rich to the poor Property taxes are based on the gross (rather than net) value are likely to provoke inefficient migration. Nevertheless, of assets held in the particular form of real property. They many of the services that have been decentralized to local are thus a poor reflection of benefits received or ability to governments have important distributional implications. pay. Local business taxes also rend to be arbitrarily assessed Primary education, for example, is often the largest redis- (in the absence of good record keeping among small busi- tributive expenditure that the public sector undertakes. Pri- nesses). For all these reasons, transfers can be justified as a mary health care, similarly, tends to benefit the poor dis- substitute for local taxes. proportionately. In some countries, Chile, for example, local In practice, transfers pursue a mix of objectives. Not all governments have been recruited to act as administrators of of them are confined to strictly technical objectives. Politi- the national income support program. Where such redis- cally motivated transfers, for example, remain a key part of tributive services have been decentralized, transfers are an the intergovernmental relationship. Attempts to eliminate appropriate means to finance them. them have simply demonstrated their resilience (Rojas Transfers have also been justified on the grounds that 1998). In 1991 Colombia passed a constitutional reform they reduce the economic distortions or high administrative that prohibited the traditional form of politically oriented costs that would arise if subnational governmentcs relied discretionary transfers, known as auxilios regiocraloe. These only on their own tax bases. The characteristics that make were central government grants to private entities-often for a good local tax also make for high costs of administra- non-governmental organizations (NGOs) controlled by tion. To function as a local tax, the incidence of a tax must parliamentarians-and were rife with corruption. Soon the TABLE 2.5 Sources of Revenues: Percent Distribution, Selected Countries PROVINCES MUNICIPALITIES TAXES % TRANSFERS % NON-TAX REVENUE % TAXES & TRANSFERS 51 NON-TAX REVENUE % Argentina 36 57 7 Australia 36 40 24 55 17 28 Brazil 67 17 9 19 67 14 Canada 67 20 14 41 44 15 France 48 34 18 Germanv 68 19 13 32 32 36 Mexico 81 12 18 64 18 Peru 3 94 3 15 57 28 Spaini 19 78 3 52 36 12 U.K. 16 72 12 U.S. 50 22 28 41 37 22 N.ote: Taxes collected centrally for the benefit of saboari-nol governments ace classified as sabnoti-nal raoes. Data an French regions and departments i inclnded with commanes ander manicipalitirs; dare on German Lander exclrde inter-L-nde- equalizarion payments. Scare: GFS 1997. 28 GFTTING THE RtlES RIGHT A FRAMEWORK FOR SUBNATIONAL GOVFRNMENT BOX 2.4 Revenue Assignment in Federal Countries Public finance theory is not particularly enlightening when reduce regional revenue disparities. Australia has the it comes to financing large intermediate tiers of govern- most centralized form of state financing. Virtually all ment. The principle of using taxes as prices does not hold taxes are administered by the central government. States up wel when the jurisdiction in question has a population are financed from fixed shares of federal revenues, dis- of several millions (as in Brazil or Mexico). In practice, tributed on the basis on fiscal capacity and fiscal needs. countries use two approaches: They either assign a broad- Over the course of the twentieth century, Latin Amer- based tax to intermediate governments or they finance ican federal countries have moved away from the U.S. intermediate levels through fixed shares of national taxes. model toward one that more closely resembles Australia's. Most federal OECD countries do some of both. This has generally coincided with attempts to reform the In the United States, states rely largely on state- tax structure as a whole by replacing arbitrarily assessed administered taxes-income and retail sales-with fed- provincial taxes on business with more modern VATs. eral transfers confined to federally sponsored programs. Argentina replaced its system of provincial business taxes Canada also relies heavily on state-administered taxes- with a national tax-sharing program in the 1930s. Simi- a VAT and an income tax-but makes large federal larly, Mexico abolished provincial taxes and-after an transfers directly to poorer provinces. Germany relies on aborted attempt at a state-level VAT-replaced them with a system of constitutionally defined tax sharing in which revenue sharing. Brazil remains the exception. When the the principal taxes-personal income, corporate income, VAT was introduced in the 1960s, it was assigned to the and VAT-are divided, roughly equally, between the fed- states. And although the federal government transfers eral government and the states. The income tax is dis- fixed shares of its income and excise taxes to poorer states tributed among individual states on the basis of origin, of the federation, the state VAT remains by far the largest The VAT, however, is redistributed among states so as to source of aggregate state revenue in the country. aixilios reappeared in the form of parliamentary influence jurisdictions with weak tax bases, for example, there is a on the allocation of the rural, urban, and social special case for distributing the pool on the basis of tax capacity investment funds. Parliamentarians used their power over and earmarking it for education. Such methods are used in the budget and other legislation to pressure the executive to OECD countries (although earmarking is far more common divert these funds in favor of their particular constituencies. in the United States than in Europe or Japan). The ability (See Box 2.5 for a discussion of the impact of transfers on of Latin American countries to follow suit depends on local political behavior.) whether they possess a similar quality of information on Perhaps the most important remedial measure develop- local finances and a similar capacity to monitor spending. ing countries can take is to reduce the uncertainty and bar- gaining that now accompanies intergovernmental fiscal Controlling Subnational Debt relations. Fixed, formula-driven transfers have much to rec- Debt can play an important role in subnational financing, ommend them. The effectiveness of transfers in achieving particularly in financing long-lived capital works in cities their various objectives can, in principle, be improved that are growing rapidly Long-term borrowing enables the through changes in design: the method used to determine cost of such infrastructure to be spread out to match the the total amount to be transferred, the criteria used to duration of its benefits. It also permits a flow of savings divide this amount among individual jurisdictions, and the from areas of surplus savings to areas where subnational conditions attached to the use of the transfer. Where the infrastructure investments have potentially high rates of intent is to support expenditure on primary education in return. Access to credit is a particularly thorny aspect of 29 BEYONID THE CENTER: DECENTRALIZING THE STATE Oppoing hesetechicalarguent in favor ~oftasesi transferrepovdd Negotiabl transfiers-those that the iew hattranfershav advrsebehavirlmptso ae provieinrtnfo political support to central gov- thei recpiens;hat ocalpolticins rgar trasfer aserminet pltcasmydistr local spending decisions. "oter eope'smony" nd illbe oreincine tomis Trnsfrs esevedforsubatinal governments in fiscal use t thn thy wold txesraised loaly Ter1s itl crises ma enoraelca oermns to take unw ar- evidnce ithe wa on his ueston.Compaios ih ratdfinaca iss uhtransfers were widespre~ad in OECD cuntris nevrtheles sugest tat ex ensv Latin: Ameiai te16sand 1970s. Snce te hr relince n trnsfrs i noripsofaco assoiae wt wd- a been a eea0sitt formula-based tr~ansfers. As spreadcorrupion an mismaagemen. As sown inTable shown inTble 2., he major transeh rgas inthe 2.5, te relance f muncipaliies o tranfers n the region no ak h form of fixed share ofseiidcn- county fbrwhichinterationlly cmparale daa aregovenmns nthe bais oprdtmined fokrmulas. availale. Prvincia govenment elianc on trnsfer is Wilethes hae lim-inatdhemsobiu inicentives hige inSanta ti nAgnia rBai.t evrelclfiscal beairdh cl of inergvern- What would ore loial pert aeamact ~mental trnfesi Latin Aeiacnnusto ailarmi some intergovernmental relationships, however. Subnational bor- been limited in scale and largely confined to short-term rowing can destabilize national financial systems or lead to cash flow management. central government deficits. While in principle, borrowing In Latin America, it has been central governments that by subnational governments can be a private transaction have mobilized much of the long-term savings that went betweeni subuaational borrower anid private len-der, the national into inifrastructure during the 1960s and 1970s. Where government's role in ensuring the stability of the financial subnational governments were responsible for these func- system often draws it reluctantly into the transaction. tions, the funds were often mobilized and allocated through In principle, subnational governments can borrow from centrally owned institutions. Much of the state-level infra- a variety of sources. Arrears to suppliers and personnel are structure housing and water supply in Brazil in the 1960s probably the most widespread form of subnational debt. and 1970s, for example, was financed from resources gen- But this source has its limits. Suppliers and personnel wvill erated by a federally sponsored unemployment insurance eventually withdraw their services if they go unpaid too plan and allocated through the national housing bank. Fed- long. Another source is the domestic private capital market, eral taxes on electricity, telephones, and fuels financed but it too has historically played only a limited role in sub- investment in power, telecommunications, and highways. national finance. Shallow domestic financial markets have Mexico's federal development bank, the Banco Nacional de limited the aggregate supply of private capital. Doubts Obras Publicas (BANOBRAS), has long served as an inter- about the credit worthiness of subnational governments mediary for donor financing for subnational governments, have prompted private lenders to confine their government while federal tax revenues have been used to finance large lending to the national level. Long experience with macro- programs of federal capital grants to states and municipali- economic instability has discouraged private investors ties. Colombia's Financera de Desarrollo Territorial S.A. from making long-term commitments of any kind. Pri- (FINDETER) has performed a similar role. As a means of vate lending to subnational governments has therefore mobilizing savings, this practice has often been successful. 30 GETTING THE RULLS EIGHT: A FRANIEWORK FOR SUIENATIONAL GOVERNMENT But the recovery record of centrally sponsored financial willing to lend to them. This provides central governments intermediation is mixed. The allocation of loans and with an important source of leverage. If lenders are con- enforcement of debt service have tended to become politi- vinced that central governments will not bail them out, cized, with taxpayers ultimately bearing the financial con- lenders themselves will act as a source of restraint. Con- sequences of bad loans. vincing the lenders of this requires more than a statement With domestic financial deepening and the increasing of intent. It requires a central government to demonstrate financial credibility of some subnational governments, the its commitment in practice by allowing a subnational bor- market for private financing of subnational debt has grown. rower to go into default and leaving the lender and the bor- Private banks (as well as pension and mutual funds) have rower to work out a settlement between themselves. Once entered some markets on a significant scale. In Brazil the private lenders are persuaded that lending to subnational volume of state bonds in private hands at one time governments carries real risks, they are likely to restrain exceeded US$15 billion. Private lending to provincial banks their lending despite the supplications of subnational in Argentina now exceeds US$4.4 billion. politicians. In principle, this presents a way of avoiding politicized As will be discussed in Chapter 3, neither a purely regu- lending. With their own capital at risk, private lenders latory nor a purely laissez-faire approach is Likely to be would appear to have an incentive to make prudent loans entirely successful. Some degree of regulation-particularly and to insist on repayment. In practice, this is not entirely over foreign borrowing-combined with a credible policy the case. When the consequences have been a pending against subnational bailouts-is likely to be necessary. strike by public employees over unpaid salaries or the fail- ure of a major bank, central governments have often Local Politics stepped in to assume or restructure the debt of subnational So far this has been a story about the relationship between governments that have borrowed too much. This has set up central government politicians and local government politi- a cycle of perverse expectations. It has encouraged subna- cians. But what about the people? If decentralization does tional governments to overborrow and domestic banks to not increase local influence over the public sector, then the overlend, secure in the knowledge that the national gov- principal case for decentralization-other than political sta- ernment witl come to their relief. bility-vanishes. The presumption of some decentralization One common reaction to this risk is to increase regula- advocates is that decentralization automatically increases tion. The Brazilian government, for example, now prohibits the influence of all strata of society. But there is an equal the issuance of new bonds or any expansion in aggregate possibility that decentralization simply transfers power domestic bank lending to subnational governments. Other from national to local elites, and that improved access of central governments prohibit subnational governments local elites to public resources simply increases opportuni- from undertaking certain forms of debt, as in Mexico, ties for corruption. where subnational governments are prohibited from issuing One of the key determinants of the influence of local external debt. National governments also regulate subna- interest groups on local officials is, of course, the formal elec- tional borrowing through the supply side. In Colombia the toral process: how governors, mayors, and members of the national banking regulation commission has required pri- subnational legislatures are elected. But the existence of sub- vate banks to increase their reserves for loans to subnational national elections, while important, is not sufficient. Incen- governments. tives and mechanisms for participation need to be in place. Recent evidence suggests that regulation alone is not In principle, decentralization itself would be expected to sufficient. Brazil's recent state debt crisis, for example, arose increase local participation. Writers on institutional eco- despite a blanket ceilinig on subnational borrowing and a nomics have long observed that people's willingness to par- web of restrictions and controls on various forms of debt. ticipate varies according to their perception of how much These regulations failed to withstand the pressures of impact such participation will have (Hirschman 1970; regional politics. North 1990; Ostrom and others 1993). Individuals who are Fortunately, in matters of debt, it takes two to tango. expected to invest resources (including their own time and Subnational governments cannot borrow unless someone is labor) must believe that the benefits they receive will 31 BEYOND THE CENTER: DECENTRALIZING THE STATE exceed the costs of doing so. The more interest groups per- Members of the public become actively involved only ceive that they can influence political decisions, the more when their vital interests are at stake (Hirschman 1970). likely they are to mobilize (Smith 1996). Local democracy is not, therefore, a forum for mass deci- The empirical evidence is ambiguous. In countries sionmaking on all issues of public policy Rather, it provides undergoing a transition from military to civilian regimes, a mechanism for interest groups to reach political decisions the initial push to throw out the ancien regime has sometimes without resorting to open conflict. mobilized mass participation. O'Donnell and Schmitter Several measures can be taken, nevertheless, to encour- describe how trade unions, grass-roots movements, reli- age more widespread participation in local government. gious groups, intellectuals, and artists supported one The first is to adopt ward- or neighborhood-based electoral another's efforts to hasten the exit of military governments districts. The election of council members by ward or in the 1980s (O'Donnell and Schmitter 1986). The transi- neighborhood gives geographically defined interest groups tion to democracy in Latin America has never been revolu- an assured seat on the council and thus some prospect that tionary, however. (In fact, no stable democracy has ever their involvement in the political process will produce a emerged from mass actors gaining even temporary ascen- tangible result. It also reduces the costs of running for dancy over dominant elites in the region [Haggard and office, because a candidate needs to run only in a single others 19951.) Instead, most Latin American cases were ward, not throughout the entire city or province. Interna- transitions from above, in which traditional elites remained tional evidence confirms this: Turnout is higher when seats in control and successfully used some combination of com- on the assembly are elected by ward, rather than at large promise and force to retain at least some of their power (Galeotti 1991). Ward-based elections are the exception (Karl 1996). What followed the democratic revolutions was rather than the rule in Latin America, however. Councilors often a dissipation of civic energy. The broad fronts of reli- typically run at large (Nickson 1995). gious, professional, student, labor, and other associations A second measure is to adopt open, unblocked electoral broke up once their common goal of bringing down the systems for local elections. As noted earlier, the closed and regime had been achieved. blocked balloting system increases national party control This is not surprising. Political participation is limited over elected officials. While useful in fostering party disci- because it requires time and effort. Thus day-to-day politics pline among members of the national legislature, it may be tends to be the work of a small coterie of professional politi- less appropriate for local elections. cians and party workers for whom politics is a full-time job. A third measure is to change the timing of subnational To paraphrase O'Donnell, Schmitter, and Whitehead elections. In many Latin American countries, subnational (1986): elections coincide with national elections. This linkage means that local government elections are overshadowed in The (traditional) theory of liberal democracy was importance by the national elections held at the same time. based on the presumption that active citizens would This distracts voter attention from local questions and elect and hold accountable, individual representatives toward a consideration of national issues. It also effectively who would, in turn, produce substantively superior encourages clientelist considerations in the selection of decisions. Contemporary theories of democracy place party candidates for municipal office. Selection is often the burden of consent on party elites and professional determined by the electoral support that potential munici- politicians (sporadically subject to legitimization at pal officeholders can mobilize for the party's national politi- the polls). These agree among themselves that they cians. In exchange for this support, potential officeholders are will compete in such a way that those who win will 'rewarded" with a municipal office position (Nickson 1995). exercise their temporary political superiority in such a While changes in the electoral rules will help, political way as not to impede those who lose from competing parties will continue to provide the critical connection for office in the future, and those who lose agree, in between the electorate and the political system. Political exchange for being allowed to take office and make parties are an essential instrument for representing polirical decisions later on, to respect the winner's authority to interests, aggregating demands, recruiting and socializing make binding decisions in the interim. new candidates for office, organizing the electoral comocti- 32 GETTING THE RULES RIGHT A FRA-MEWORK FOR SUBNATIONAI. GOVERNMIENT tion for power, and forming effective governments (Dia- fiscal burden on the federal government was alleviated only mond 1997). By organizing class and other interests, par- by an increase in national taxes and a protracted series of ties are one of the instruments by which the poor and mar- case-by-case asset transfers (Dillinger and Webb 1999b). In ginal groups can gain voice in the formal political system Colombia, the central government committed itself in 1999 (Boeninger 1991; Przeworski 1995). Party members and to transfer 22 percent of central government revenues by leaders have an incentive to get out the vote, create a pres- 2002, without defining the responsibilities that would be ence in the community, seek out voters, and respond to vot- correspondingly devolved. The opposite has also occurred. ers' contacts. In Argentina, the federal government transferred its remaining secondary school and health programs to the Strategies provinces. While the federal government guaranteed a Decentralization often takes place amid political and compensatory payment, the payment was financed not by economic turmoil. The euphoria at the fall of a military the federal treasury, but from the provinces' own share of regime; the economic crisis that precipitates a regime's col- participaciones. lapse; the jockeying for power by newly emergent interest National governments have crippled local governments' groups-all these conditions create an environment in ability to perform newly decentralized functions by failing which a careful, rational, and orderly process of decentral- to decentralize key management controls. In Brazil, as ization is highly unlikely. Even where decentralization hap- noted earlier, the postmilitary constitution imposed rigid pens in a less dramatic context, questions of strategy and controls on state and municipal personnel policies, forbid- timing still arise. Experimenting, testing, adjusting, and ding the dismissal of staff or any reduction in nominal replicating are emerging as the prevailing methods of salaries, and requiring expensive pension benefits. These decentralization in the region (Rojas 1998). rigidities played a large part in Brazil's subsequent state There is clearly no blueprint for decentralization. Much debt crisis. Similarly, in Colombia the central government depends on the initial conditions in the country and the continued to set the salaries of public school teachers after particular political interests that support or oppose decen- the management of primary and secondary schools was tralization. The recent experience of countries in the ostensibly decentralized to the provinces. The central gov- throes of decentralization nevertheless leads to two general ernment's subsequent decision to grant a major increase in conclusions. salaries prompted a fiscal deficit in some provinces, which was only resolved through the creation of a special com- Synchronizing the Elements of Reform pensation fund. The most consistent lesson of recent decentralization expe- The recent decentralization of education in Mexico has rience is the need to synchronize the elements of reform. followed a more balanced approach. Along with the trans- The political impetus behind decentralization prompts cen- fer of assets and teaching staff, corresponding financing was tral governments to make political concessions hastily. But provided in an amount equal to federal spending on these granting local elections is a step that can be taken rapidly. functions in the previous year.' Similarly, in Chile the What is slow and difficult is the working through of new decentralization of education and health to municipal (and regulatory relationships between central and subnational private) providers was accompanied by a loosening of man- governments, the transfer of central government assets (and agement controls. staff), and the conversion of what had been annual budgetary transfers between units of a centralized administration to a Demonstrating Early Commitment to the New Rules system of tax assignment and intergovernmental transfers. of the Game The reccnt history of decentralization in Latin America A second lcsson that emerges from recent decentralization has been plagued by a lack of synchronization. Countries experience is the need to demonstrate commitment to the have failed to synchronize the decentralization of functions new rules of the intergovernmental relationship early in the with the decentralization of revenues. In Brazil, for exam- game. Precedents matter. They affect expectations. The ple, federal transfers to municipios doubled, while their benefits of a stable set of rules can only be achieved if the expenditure responsibilities remain undefined. The added rules are credible. Credibility requires demonstration. 33 BEYOND THE CENTER: DECENTRALIZING THE STATE One of the most important precedents a central govern- and controls. Ward-based local politics are more likely to ment can set for newly autonomous subnational govern- elicit participation than elections at large. ments is the hardness of the central budget constraint. Evi- Finally, we know that a strategy aimed at stopping dence suggests that even the possibility of a central decentralization is unlikely to succeed. The pressures for government bailout will prompt excess spending and decentralization are beyond the control of governments. deficit financing by subnational governments. This is The emergence of modern economies and an urban, literate shown in recent experience with government efforts to middle class has created nearly insurmountable pressures restrain the borrowing of recently redemocratized provin- for a broader distribution of political power. The ongoing cial governments in Brazil and Argentina. As elaborated in decentralization efforts in Latin America are a response to Chapter 3, Brazil began its current democratic era with a these pressures. Rather than attempting to resist them, mass rescheduling of debts owed by state governments to governments need to accommodate them in a way that foreign creditors. Three years later came a mass reschedul- maintains political stability while improving public sector ing of debt owed by the states to federal financial interme- performance. diaries. Together, these established the presumption that the federal government stood ready to provide debt relief to Notes any state that required it. The following five years wit- I1This is not to say that each member of the bureaucracy is an nessed an explosion of state borrowing, largely through automaton. Individuals within an organization will not necessarily find the interests of the organization to be congruent with their own. bonds and debt to state-owned banks. Ultimately, both had to bas ebt fdagvn nAgThey can nevertheless be expected to make self-interested responses to be assumed by the federal government. Argentina, in to the working rules that allocate rewards and costs within the contrast, has succeeded in enforcing a hard budget con- bureaucracy (Ostrom and others 1993). straint on its provinces. From the outset the current admin- 2. Note that all Latin American countries have presidential, as istration has refused to provide any significant debt relief to opposed to parliamentary, systems. the provincial governments. Moreover, it has taken steps to 3. The importance of the minor parties varies among countries. In minimize its potential exposure by prohibiting direct lend- recent elections for the chamber of deputies in Argentina, the two largest parties garnered 73 percent of the seats. In Brazil the two ing from the fcdcral trcasury to the provinces and by pto- ags ate anrdony3 ecn ftesas largest parties garnered only 39 percent of the seats. hibiting the central bank from discounting loans made by 4. Among countries using proportional representation, the only provincial banks to their governments. As a result, when other country with this system is Finland. provincial governments ran aground in the Tequila crisis of 5. These limitations are in addition to its unwarranted assump- late 1994, they w ere forced to adjust rather than rely on tions concerning the efficacy of tocal democracy as a market-clearing mechanism. federal relief. 6. In 1998 the Brazilian congress approved a constitutional amendment that would allow states to dismiss staff, provided their Conclusion personnel spending exceeded a threshold percentage of state rev- Decentralization in Latin America is very much a work in enues. The congress was also considering amendments to pension progress. There are many experiments under way and little legislation. If approved, these amendments would go a long way data on final outcomes. But we do know a few things. A toward providing states with the means to respond to fiscal pressures system that is based on rules is likely to produce better without resorting to default. 7. Income and payroll taxes would appear to meet both tests to the extent they can be imposed through withholding by employers. telism. Differentiated structures of subnational govern- But the coverage of such taxes is limited in most Latin American ments would help governments to define the functions of countries. A local income tax in a municipio with a population of a different units more explicitly. A hands-off attitude toward few thousand would probably have no more than a dozen contributors. subnational loan default is as critical to controlling subna- 8. Thereafter, funding has been based on a formula that gradually shifts the distribution from its historic pattern to one that provides an uional debt as the most comprehensive set of regularsons euaamntpruilcosalsae. equal amount per pupil across all states. 34 CHAPTER 3 Addressing the Macroeconomic Threat: The Quest for Hard Budget Constraints ome decentralized democracies like Germany, Switzerland, and the United States are known for their macroeconomic stability, and some scholars even give federalism some of the credit for the stability (McKinnon 1997; Qian and Weingast 1997). Recent events in Brazil, however, show how decentralization can contribute to macroeconomic instability. Other cases in Latin America, such as Argentina, show how decentralization can complicate achieving stability, as it did in the 1980s, but also how it can be made compatible with stability, as has happened in the 1990s. Colombia is also a good example of how rapid decentralization can put substantial stress on macrofiscal sta- bility (Dillinger and Webb 1999a, 1999b). Many other countries in the region, such as Ecuador, Guatemala, Mexico, Nicaragua, and Venezuela, have begun decentralization more recently, and it is crucial that they make institutional arrangements to prevent undesirable macroeconomic consequences. This paper considers two questions: sons for decentralization in Latin America and elsewhere have been political, not economic (Rojas 1998; Willis and ageHow t adoe dheentrization affecst macroeconomic man- others 1997). Decentralization has been an integral part of agement and the size of the state? programs to restore or deepen democracy in many coun- *What institutional arrangements and policies account tries. Even more, as we show below, understanding the for the good and the bad macroeconomic effects of political dimensions of decentralized states is crucial for d understanding their macroeconomic outcomes. The standard economic dimensions for evaluating public In the macroeconomic dimension, decentralization has finance policy are macroeconomic stability, equity, and effi- raised both hopes and fears. The first hope is that decen- ciency (Musgrave and Musgrave 1959). In analyzing tralization will improve the overall macroeconomic perfor- macroeconomic management in decentralized democracies, mance of the public sector by involving the taxpayer more we focus mostly on the first dimension, although we also closely in spending decisions. Some people think that citi- discuss some broad issues of efficiency, related to the size of zens are more tightfisted with the taxes they pay to local the public sector and the relative supply of local and government than with the taxes they pay to central gov- national public goods. ernment, with the result that when spending is decentral- We recognize that decentralization is not to be judged ized, aggregate public sector spending will be lower. It is by only its economic dimensions. Indeed, the primary rea- also believed that creditors will be more conservative in 35 BEYOND THE CENTER: DECENTRALIZING THtE STATE lending to local governments than they are to central not induce subnational governments to participate in the governments. required budget cuts (Tanzi 1996; Prud'homme 1995). Central governments, in contrast, seem more likely to Alternatively, fiscal adjustment may be achieved, but overspend and overborrow, both because taxpayers are through undesirable means. Central government expendi- more remote from the beneficiaries of services, employ- tures may have to be cut drastically, or rates on taxes not ment, and contracts, and because the banking sector is subject to revenue-sharing may have to be increased often a captive market for national debt. excessively. The macroeconomic fears about decentralization arise Finally, the transition from a centralized to a decentral- precisely because the conditions outlined above are fre- ized system may also be a source of macroeconomic prob- quently not met. First, decentralization often increases the lems. Central government deficits may grow if the govern- separation between spending and tax decisions, rather than ment fails to reduce spending as it increases transfers or bringing them closer. This happens whenever subnational gives up tax bases to subnational governments. Such an expenditures are financed mainly through transfers from outcome may develop because of a serious mismatch the central government, as is the case in most of Latin between the allocation of responsibilities and resources. It America. Subnational governments may then overspend, if can also arise because the subnational governments fail to they expect to get more resources from the common pool of do the job with the monies transferred, leaving the center national resources, either through additional discretionary with the double cost of continued service provision plus the transfers or bailouts. transfer (or forgone revenue from the devolved tax base). Second, creditors wiltl be conservative only if they Finally, it may happen because the central government does believe the central government will not bail out failing local not reduce its size, even after subnational governments take governments. But, in fact, central governments often do so.1 over its previous tasks, because it refuses to relinquish the Federal systems also present the danger that states' rep- powers of patronage or is constrained by inflexible labor resentatives to the national government will collude to codes or union power. extract more resources from the common pool through leg- Thus, together with the hopes, there are legitimate con- islation or negotiated agreements (Alesina and Perotti cerns that decentralization will lead to excessive decentral- 1998; Sauguinetti and Tommasi 1997). Such decentralized ized expenditures and to problems with macroeconomic or concerted attacks on the "commons" might lead to dif- management. As indicated above, compensating actions to ferent outcomes, depending on rhe central government's avoid macroeconomic instability-excessive reductions in reactions. If the central government responds by cutting its federal spending or excessive overall tax levels-can lead to own expenditures to maintain fiscal equilibrium, the result efficiency and equity problems.2 Decentralization can then will be an inefficient composition of public expenditures: result either in insufficient provision of federal public goods, larger than optimal spending on local public goods, lower in larger overall public expenditures and taxes, or in macro- than optimal spending on national-federal types of public economic instability. goods. If the central government responds by increasing In attempting to answer the two questions presented taxes so as to keep fiscal equilibrium, the result will be a above, this paper draws mainly on research conducted by bloated state: larger than optimal total public expenditures the Chief Economist Office for Latin America and the and taxes. If the central government cannot or does not do Caribbean at the World Bank (Dillinger and Webb 1999a, any of the above, the result will be excess fiscal deficits and 1999b; Fornasari, Webb, and Zou 1999). This chapter pre- macroeconomic instability. sents evidence on the effect of decentralization on public In addition, the central government's ability to carry out finances in a worldwide sample of countries; analyzes the stabilization policy in response to shocks may be more ham- effects of institutions on the macroeconomic outcomes of pered in decentralized economies. One reason is that the decentralization experiences in four Latin American coun- central government may have to share with subnational tries (Argentina, Brazil, Colombia, and Mexico); compares governments the more efficient tax bases or give them up these experiences schematically to see which combinations completely. Then, a necessary and desired fiscal contraction of institutional arrangements for decentralized fiscal man- may not be achieved because the central government can- agement suffice to avert severe macroeconomic problems, 36 ADDRESSING THE MACROECONOMIC THREAT THE QUEST FOR HARD BUDGET CONSTRAINTS and which do not; and summarizes the main conclusions of inflation, urbanization, central bank independence, and these analyses. major political transitions, were also used. When the data were averaged over time to make cross- Evidence of Macroeconomic Problems: section comparisons, only two coefficients on subnational Econometric Results fiscal variables turned out to have t-statistics indicating a To answer the question of whether decentralization con- significant difference from zero: In the equation with tributes to excessive public expenditures or unsustainable national spending on the left side, the subnational spending fiscal deficits, an econometric study analyzed the relation- variable had a positive coefficient and the subnational tax ships between subnational and national government fiscal variable had a negative coefficient (see Table 3. 1). The coef- indicators, using data from 32 industrial and developing ficient on subnational spending was close to 1.0, which countries during 1980-94 (Fornasari, Webb, and Zou implies that central government spending tends to grow in 1999). The dependent variables were central government a one-to-one ratio with subnational spending. On the other primary spending (including transfers and primary hand, the coefficient for subnational taxes is below -1.0 deficits).3 The main independent variables were the subna- (significantly). These results suggest that (a) decentraliza- tional fiscal variables-either subnational total spending tion of spending by transfers increases the size of total gov- and taxes, as a pair, or lagged subnational overall deficits. ernment, and (b) to the extent that subnational govern- Some standard control variables, such as GDP growth, ments finance themselves with their own taxes, the national TABLE 3.1 Basic Model: Cross Section; Overall Sample Estimates (All dependent variables are net of interest payments) DEPENDENT VARIABLE CG DEFICIT TO GDP CG EXP TO GDP CG DEFICIT TO GDP CG EXP TO GDP Constant 5.404** 0.619 5.867* -5.506 2.250 0.056 2.211 -0.566 SN Tot. Expenditure to GDP 0.041 1.009** -0.384 2.556 SN Tax Revenue to GDP 0.033 -2.110** 0.180 -3.132 Lag Overall SN Deficit to GDP 0.758* -0.199 1.884 -0.108 Lag CPI Inflation -0.073** -0.206 -0.068** -0.122 -2.691 -1.649 -2.167 -1.060 Percentage Urban Population -0.029 0.160 -0.026 0.128 -1.162 1.398 -0.959 1.268 1981) GDP Per Capita in US$ -0.272 2.157 -0.248 2.836** -0.799 1.380 -0.661 2.065 Real GDP Growth -0.539** 0.586 -0.570* 0.320 -2.096 (0.496 2.0]2 7. 09 Adj. R-Squared 1).25 0.36 0.13 0.52 No. Time Observations 1 1 1 1 No. Cross Section 32 32 32 32 CG - central government SN = subnat,otal. vIndictres a signifiancc level of 10 percent **Indicates a significance level of 5 percent N,tc The nu.nbers in italics represent the r-starisric associated with each -oefficient. 37 BEYOND THE CENTER: DECENTRALIZING THE STATE TABLE 3.2 Panel Data, Fiscal Variables in First Differences, 1980-94 (All dependent variables are net of interest payments) DEPENDENT VARIABLE CG DEFICIT TO GDP CG EXP TO GDP CG DEFICIT TO GDP CG EXP TO GDP Constant -0.045 0.459* -0.413 -o.(69 -0.092 1.838 -0.932 -0.312 SN Tot. Expenditure to GDP 0.322** 0.538** 4.007 8.121 SN Tax Revenue to GDP -0.176 -(.(73 -7.004 -0.571 Lag Overall SN Deficit to GDP 0.234** 0.261** 2.805 7.902 Lag CPI Inflation -0.054** -0.013 -0.057*' -0.028 2.806 -0.592 -3.368 -1.509 Percentage Urban Population -0.007** -0.0:3** -0.010** -0.012* 2.019 -4.657 -2.693 -4.959 1980 GDP Per Capita in lJS$ 0(.07( 0.089** 0.123* 0.128e* 0.985 2.113 1.878 3.393 Real GDP Growth -0.150** -0.194** -0.143** -0. 142** -8.141 -13.545 -8.896 -12.706 Adj. R-Squared 0.10 0.23 0.14 0.33 Durbin-'Watson Statistic 2.12 1.77 2.13 1.82 No. Time Observations 13 13 13 13 No. Cross Section 32 32 32 32 CG = cenral government. SN = sub.arona,. -Indicates a significance level of 10 percent. s5lndicares a significance level of S percent. Note: The only variables that are in first difference are the dependent variable and the fiscal explanatory variables; rhar is. total expendirure, tax revecue, and deficir. The numbers in italics represent the t-statistic associated with each coefficient. government tends to be smaller by about as much as the ever (see Table 3.2).4 They show that increases in subna- subnational spending, leaving the overall size of the public tional spending and deficits lead to higher spending and sector about the same. deficits at the national level. The relationships are strong The subnational deficits did not, on average, affect the economically, and statistically significant 5 The results in national spending or the national-level deficit in the cross- columns 1 and 2 have the clearest meaning-an increase in section regressions, nor did the subnational spending have subnational deficits is associated with an almost 1-to-I a statistically significant relation to the national govern- increase in central government spending and deficits in the ment deficits. This implies that when countries are decen- subsequent period. This is consistent with a pattern of the tralized in a long-run steady state-which is the interpreta- central government bailing out states and cities when they tion of these cross-country regressions on averages per have increased borrowing too much. country-they do not have higher national deficits on aver- In interpreting these results, it is important to recognize age than the less-decentralized countries. They have pre- that the use of first differences means that the results are sumably developed institutions-and raised taxes-at least determined mainly by the countries and periods when the to avert the macroeconomic fears concerning deficits. levels of spending and deficits at the national and subna- The panel regressions with changes in the national and tional levels are changing rapidly; that is, by those times subnational fiscal variables got very different results, how- and places where increased decentralization is taking place. 38 ADDRESSING THE MACROECONOMIC THREAT: THE QUEST FOR HARI) BUDGET CONSTRAINTS In these times and places, we can reject the hypothesis that *Increases in subnational deficits lead to increases in the transfers between central and subnational governments national-level expenditures and deficits and to macro- are usually determined exogenously by the center. The economic instability. process of fiscal decentralization tends to cause problems. Fears not realized: These results are powerful arguments against rapid decen- *Decentralization based on local taxes does not increase tralization without adequate safeguards. the size of the state in the steady state. The results in the panel data held up even with the *Decentralization is not associated with more or less inclusion of political-institutional variables-major national fiscal deficits in the steady state. political transitions and central bank independence; and two variables that pertained directly to decentralization-elec- These findings suggest that one would expect to see tion of subnational officials and a unitary-federal constitu- more problems with macroeconomic management during tion. Whether local officials were elected had no effect on and shortly after an increase of fiscal decentralization; then the intergovernmental fiscal relations considered here. In the problem would subside as institutions were strength- other words, local democracy does not seem to worsen or ened. The case studies show that this can happen. improve macroeconomic fiscal management on average. Being a unitary state (rather than a federation), however, Outcomes and Institutions in Major significantly increased the extent to which national govern- Decentralized States of Latin America ment spending was related to subnational spending and This section analyzes the macroeconomic outcomes from deficits. This might reflect the fact that the national spend- decentralized fiscal management in Argentina, Brazil, ing figure includes some transfers to states, and one would Colombia, and Mexico.6 These cases give some indications expect budgeted transfers (rather than tax sharing or dele- of which institutional features of decentralized fiscal sys- gated taxes) to be more important in a unitary state. Being tems lead to excess deficits at national or subnational levels a unitary state does not significantly affect the transmission and which features help avert these problems. To guide the of subnational to national deficits. analysis we use the following list of 12 institutional factors, To summarize, whereas in the cross-section analysis, the both fiscal and political, which are expected to affect the steady state level of subnational borrowing (implicitly sus- quality of fiscal management in decentralized democracies tainable) is not associated with higher central government (see Table 3.3). spending and deficits, when subnational governments increase their borrowing (potentially unsustainable) the Hard Budget Constraints for Subnational central government seems to have to spend and borrow Governments more in the subsequent period. This implies that transitions First and foremost, a firm allocation of expenditure respon- to decentralization and fluctuations of borrowing by subna- sibilities is critical for establishing a hard budget constraint tionals typically cause problems for macroeconomic man- for subnational governments. If the central government can agement, but evidently many countries with long-standing effectively delegate functions to subnational governments decentralized public sectors have developed institutions to to go along with the delegation of revenue sources, central prevent these problems. Although many of these countries spending and deficits are more likely to be contained. If this are outside Latin America, the experiences in the region is not the case-because, for example, the constitution or show important positive as well as negative lessons for the law mandates resource transfers without allocating macroeconomic management in decentralized democracies. explicitly equivalent responsibilities-the central govern- The results presented thus support some of the hypothe- ment may find itself with a constitutional obligation and ses of hopes and fears discussed in section 1, but not others. political expectation to continuc providing somc services, even after revenues or tax bases have been turned over to Fears realized: subnational governments with the understanding that they *Decentralization based on transfers from the center, will do the rask common in Latin America, tends to increase the size of Second, basing transfers on clear rules is a necessary the total government sector. ingredient for a hard budget constraint. Wherever there is 39 BEYOND THE CENTER: DECENTRALIZING THE STATE TABLE 33 able deficits if both sides expect to gain, most likely Institutional Arrangements to Set and Keep Hard Budget through some sort of federal bailout. The bailout can take Constraints on States many forms, including allowing the financial system *Hard budget constraints from che central government to subnational gov- (implicitly insured by the government) to count a debt that ernments is not being serviced as an asset. Unsustainable deficits are *Rule-based transfers *Firrn allocation of spending responsibilities also less likely if the central government credibly commits -Constraints on borrowing not to have bailouts, prohibiting explicit bailouts and forc- *Ex ante constraints *Ex pste conseruences an eslin ncnivsing subnational governments to service their debts, and if +Ex post consequences and resulting incentives 'Enforcing payment by subnational governments regulators force creditors to accept the losses implied by any *Enforcing losses on banks with bad loans to uncreditworthy subna- failure to service debt by subnational governments. It is still tional governments-bank regulation *Central bank independenice an open question whether ex ante regulation or ex post * Autonomy of subnational governments to stay within hard budget con- enforcement of debt service is more effective in preventing strannu s ..a,r to excessive subnational government borrowing. Both can -Ability to control spending and costs *Ability tO increase revenues work together, but conflicts are also possible, as noted above. 'Political ability of central government to enforce hard budget constraints Often financing from the central bank loosens the bud- * Power of the president -Power of the governors get constraint for the subnational governments, either *Political party discipline directly by discounting subnational debt or indirectly by * Electorall ruins easing the national government's budget constraint or allowing commercial banks to roll over bad subnational debts. Unsustainable deficits are expected to be less likely recourse to significant discretionary transfers, including when the central bank (and the bank supervisory agency) is matching grants, subnational governments will have an more autonomous and has a strong anti-inflation mandate. incentive to overspend because they expect that they can As mentioned above, subnational governments may also get a larger transfer. accumulate excessive contingent liabilities. This situation is more likely to happen wherever subnational governments Borrowing Constraints are allowed to run their own pension regimes, when they Although tax and spending policies create fiscal pressures, own banks, and when they make concessions to the private whether they cause problems for macroeconomic manage- sector without adequate regulation from above. ment depends on whether the subnational governments face hard limits on their ability to borrow or to seek other Autonomy of Subnational Governments to Stay ways of increasing their resources by spending more. Within Budget Constraints Unsustainable deficits are less likely if the central govern- The second group of institutional factors relates to subna- ment effectively controls subnational borrowing ex ante. tional governments' having the capacity and autonomy to But how to enforce this in practice is not always clear when stay within the budget constraint. The first one in this cat- the subnational governments have considerable political egory is expenditure auitonomy. Unless subnational govern- autonomy. Pseudo-strict controls could make matters worse ments (SNGs) have autonomy over their expenditure, no if central government approval creates the impression, and fiscal decentralization and no macrofiscal problem are likely perhaps the self-fulfilling expectation, that the central gov- to result. Giving SNGs autonomy over spending is, of ernment has also extended a guarantee. course, the way in which decentralization can improve effi- To run deficits, subnational government must find a ciency in matching the needs and desires of a diverse popu- source of financing, which potentially includes contractual lation. But to live within a sound budget constraint, SNGs borrowing from private foreign or domestic banks (espe- must have authority to control their costs. Too often central cially banks owned by the subnational governments), issu- governments keep for themselves decisions (such as teach- ing of domestic or foreign bonds, and the running up of ers' and doctors' wages) that critically affect the level of arrears to suppliers and personnel. A creditor and a subna- costs of SNGs, and a liberal decision may throw them into rional government would only agree to finance unsustain- deficits. In particular, SNGs must have the authority to 40 ADDRESSING THE MACROECONOMIC THREAT THE QUEST FOR HARD BUDGET CONSTRAINTS spend less, particularly to cut personnel, salaries, and pen- decentralization are less likely if presidents are constitution- sion benefits, collectively the largest item of subnational ally strong at the national level, and if governors and may- expenditure, in order to be able to adjust to shocks or to ors have little constitutional autonomy. These conditions contribute to needed fiscal retrenchment. If central rules are not necessarily beneficial for all aspects of governance constrain this authority, it is more difficult to reduce and, of course, fiscal decentralization presumes that gover- deficits, and expectations of a central government bailout nors have enough political autonomy to be considered a are higher. Unsustainable deficits should be less likely if politically separate level of government rather than just a subnational governments have authority to cut their costs. field representative of the central government. Governors and the president indirectly contend for resources, espe- Revenue Autonomy cially through congress and the parties, whose effects on the With fiscal decentralization, SNGs usually obtain certain intergovernmental balance of power we must also consider. tax bases, but for reasons of politics, equity, and efficiency, The central government should be more able to maintain a these bases rarely cover all the expenses of SNGs. It is com- harder budget constraint if electoral rules orient congress monly believed that SNGs will have smaller deficits if they toward national, not local, issues, and if party discipline is rely more on their own tax bases (and have the power to strong. The balance between these forces depends on the change tax rates on the margin), because they have the abil- interaction of the constitution, electoral procedures, and ity to adjust to shocks by increasing revenue. In addition, as party discipline. The constitution is a given at most times, discussed above, relying on one's own resources may as are electoral procedures. Therefore, party discipline is the strengthen the incentives co control spending. Therefore, dependent outcome. But constitutions and electoral proce- unsustainable overall public sector deficits are less likely if dures change with some frequency in Latin America, and subnational governments raise much of their own revenue they reflect the political balances and party characteristics and have enough flexibility to change rates or impose new at the time of each change. taxes. Outcomes and Institutions in Four Latin American Political Ability to Enforce Harcd Budget Constraints Countries Political institutions determine the capacity of central gov- With these 12 conditions as a framework, the next section ernments to enforce hard budget constraints. The rules for examines the outcomes and institutions in four countries. intergovernmental fiscal relations and the way in which they are implemented result from the political relations Argentina: The successful institutionalization of a hard between levels of government (Riker 1964; Stepan 1997; hudget constraint on provinces. Willis and others 1997). Why does the central government Fiscal deficits at the federal level were a major problem in say yes or no to an SNG's request for more resources? Why Argentina before 1991, leading to hyperinflation, which do the SNGs take on additional tasks or accept reduced reached over 5,000 percent in 1989. Provincial deficits and transfers in order to help the central government balance its indirect bailouts of provincial banks, which had access to budget? The answer depends in large part on the relation- central bank credit, contributed to the financial difficulties ship between the president, the national congress, and local of the period. Provinces accounted for at least half of the elected officials. Systems with strong presidencies should be public sector deficits that fueled the hyperinflation. The better able than systems with weak presidencies to fend off high inflation itself disrupted and substantially devalued pressures to cover the states' deficits. The presidency is the the tax and transfer system in real terms, making state only office with a national constituency and as such is more finances shrink in real terms. By the end of the decade, likely to take into account the interests of the overall econ- hyperinflation had also greatly eroded the real value of omy, while the members of congress represent regional con- domestic debt, leaving governments at all levels with little stituencies and their parties. domestic debt to repay, although external debts remained We can analyze the power relationships between the substantial. national and subnational levels in four steps, corresponding In addition to major improvements at the national to four conditions. Unsustainable deficits arising from fiscal level-committing legally to currency convertibility at a 41 BEYOND THE CENTER: DECENTRALIZING THE STATE fixed rate with the dollar (the Convertibility Law), cutting Provinces at first reacted by trying to borrow more. They the budget deficit, and privatizing major industries-the forced suppliers and personnel to accept bonds in lieu of steps to improve subnational finances in the 1990s were payments on overdue bills and wages. These bonds could be also important for the success of macroeconomic stabiliza- converted to cash at a discount at provincial banks and then tion. In the reconciliation of intergovernmental obligations used to pay taxes. This forced lending accounted for more inherited from before stabilization, it turned out fortuitously, than half of the debt incurred in 1994-95. Provinces with but not accidentally, that the debts between each of the debt servicing difficulties suddenly found that Banco de la states and the federal government netted out to about zero. Nacibonds could be converted to cash at a coparticipaciones From 1991 until 1994, the real value of revenues to subna- to pay to creditors. Eventually they had to cut investment rional governments grew rapidly, both from their own taxes programs, lay off nonpermanent staff, give regular staff and even more from the growth of taxes collected by the time off without pay, and implement emergency revenue federal government and shared with the provinces. The fed- measures. In some cases the federal treasury facilitated refi- eral government restrained its own spending rate to less nancing of provincial debt, but at market rates with no than the growth of revenue and negotiated a series of agree- bailouts. And neither the federal government nor its agen- ments with the provinces, getting them to accept additional cies took over any provincial debt; the law forbade it. Most responsibilities. To accomplish this politically, the govern- of the smaller states, along with Cordoba and the city of ment used party discipline and fiscal transfers (limited in Buenos Aires (both run by the opposition party) had to the aggrcgate) to small states that had a lot of congressional make strong structural adjustments. They have kcpt their votes per capita. In addition, the governors understood the debt in check since then. need for overall fiscal adjustment to sustain the economic The one channel for aid to the provinces in present-value revival and growth of their revenues, and the federal gov- terms was the federal takeover of most of the state ernment guaranteed floors on the level of transfers. employee pension plans. In the early 1990s the federal gov- In the early 1990s, many of the provinces borrowed ernment offered to take these over; most states resisted at heavily, even though their revenues were growing strongly. first, for political reasons, but the economic crisis forced Borrowing from banks, including those owned by the them to accept. This measure will prevent future accumu- provinces, was the most important form of subnational lation of unfunded pension liabilities by the provinces. indebtedness in the early 1990s. The provinces pledged The strong anti-inflation commitment after 1991 and their coparticipaciones (revenue sharing payments) as a guar- tight limits on central bank credit to the public sector in antee to private creditors. The federal government was a Argentina limited subnational spending and deficits in two party to this arrangement in that Banco de la Nacion would ways. First, it allowed the federal government to reject deduct the debt service from federal tax receipts and trans- provincial pleas for more resources after the Tequila shock, fer only the remainder to the provinces. Aggregate provin- with the rationale that it could not increase transfers with- cial debt grew more slowly than gross doemstic product, out endangering the stabilization gains and the survival of however, because the most important province, Buenos the Currency Board system. Second, it constrained the abil- Aires, and another large one of the next rank, Santa Fe, ran ity of the provinces to borrow from their own banks by balanced budgets or only small deficits. They exercised this tightening bank regulations and eliminating local govern- discipline, in step wirh the adjustment program at the ment access to the central bank rediscount facility After the national level, because the governors in these provinces 1994-95 economic shock, most provinces had to recapital- belonged to the president's party and supported the overall ize or privatize their banks-borrowing from them was not economic program. an option. Eighteen of the provincial banks were privatized The economic crisis of 1994-95, ignited by the currency during 1994-96 and more have gone through the process and debt crisis in Mexico, tested the hard budget constraint since then (World Bank 1998). put in place by the Convertibility Law, the new central bank The timing had been good. Making changes before the charter, and the debt-servicing arrangements. The crisis not crisis had forced the provinces to adjust turned out to be of only raised the interest costs of provincial debt but also critical importance for the institutionalization of the hard seduced revenues from local taxes and revenue sharing. budget constraint in subnational finances in Argentina. 42 ADDRESSING THE MACROECON(MI( THREAT THE QUEST FOR HARD BUDGET (ONSTRAINTS When the crisis came, the deficit hawks in the central gov- trary to initial expectations, and state and local govern- ernment needed to have just enough political leverage to ments expanded strongly, resulting in a larger public sector veto a change in the status quo, not to impose a new one, overall. which would have been a much harder task. Political tim- The main macroeconomic problem with decentraliza- ing in getting the budgets balanced in the largest states (of tion, however, arose from excessive state deficits and then the president's party) when revenues were growing was also mismanagement of the debt. In Brazil, as in Argentina, right. This made it politically easier during the economic subnational deficits contributed to the fiscal problems in crisis for the president to enforce a hard budget constraint the high-inflation periods prior to the stabilization that on the main opposition states. started in 1994. Dealing with the deficits and debt of the states has been one of the main challenges to sustaining the Brazil: Repeated reschedutling. No hard budget constraint stabilization since then. There have been four rounds of for the states. crises and rescheduling of state debt-in 1988, 1993, A state debt crisis was not the main macroeconomic prob- 1994-95, and 1998-99. lem that observers expected from decentralization in Brazil. In the earlier debt crises, the debt agreements estab- They feared that the large increase in tax sharing mandated lished three precedents that influenced subsequent agree- by the 1988 constitution would provoke federal deficits, ments. First, the federal government actually put the state because the federal government would not cut its ordinary debt on its books and then provided relief in the form of (non-transfer) expenditures or raise federal taxes by an rescheduling, rather than forgiveness. Second, through the equivalent amount. At the outset, this deficit situation combination of grace periods, rescheduling, and debt ser- seemed likely to happen, because the congress rejected a vice caps, the agreements reduced the debt service burden proposal for expenditure decentralization intended to of sitting administrations, leaving the fiscal consequences to match the new division of revenues. their successors. This reinforced the perception that the fed- Nevertheless, fiscal adjustment ultimately occurred at eral government was prepared to provide debt relief to any the federal level. Part of the adjustment took place on the state requiring or requesting it. The repeated cycle of the expenditure side in the early 1990s. Having failed to decen- federal government refinancing state debt, coupled with tralize federal functions formally, the federal government caps on debt service, had the perverse incentive effects that succeeded in giving the responsibility for certain expendi- one would expect. By the time some consensus for action tures-hospitals, highways, and urban transit systems-to had been reached, the number of bankrupt states was too subnational governments on an ad hoc basis. Service from large to allow them all to fail, and their debt had grown too the federal level declined to the point that it sometimes large for any solution without substantial debt relief to became politically attractive for a municipality or state to work. The lack of discipline in the political parties and the take on the responsibility. The largest part of fiscal adjust- strong autonomy of the state governors, in contrast to ment at the center, however, occurred through adjustments Argentina, put the federal government in a weak position in revenue. On the revenue side, at first the federal govern- to negotiate with the states, so the terms of agreement were ment used the guise of economic stabilization measures to often relatively generous. claw back some of the increase in the constitutionally man- At the beginning of the 1990s, Brazilian subnational dated transfers, namely reducing by 20 percent the portion debt as a share of GDP was at a level similar to Argentina's, of income and industrial-product taxes subject to sharing. but by 1997 it had more than doubled. Unfortunately, the The federal government also increased the rates of taxes Brazilian stabilization program of 1994, the most success- that it was not required to share with subnational govern- ful to date, left unchanged many rules and institutions, ments, such as social security taxes and a new tax on checks. which motivated the states to let their debt grow. Most of Because agreements required federal and state governments this debt was owed to the central governm1ent or to state to share the most efficient taxes, they created an incentive banks, and until the debt-rescheduling agreements in 1998, to increase inefficient taxes, as Argentina did in the 1980s. much of it was not being serviced by the states. Interest was Consequently, the federal deficit and spending remained being capitalized. As a result, state debt and deficits were a roughly constant as a share of GDP over this period, con- direct fiscal problem for the central government and for the 43 BEYOND THE CENTER: DECENTRALIZING THE STATE overall public sector. State debt accounted for one-third of Total government spending grew at the same pace as the increase in domestic public debt from 1994 to 1997. GDP in the late 1980s and then grew much more rapidly The situation has evolved in several ways since the mid- during the 1990s, increasing from 11 percent of GDP in 1990s, some of them positive. First, the Ministry of Finance 1991 to 18 percent in 1997. Revenues, in contrast, and the central bank have set some ex ante limits on state increased from only 10 percent of GDP during 1987-90 to borrowing. The limits are not air tight, and they reinforce 13.6 percent of GDP in 1997. This increase was just par- the unfortunate impression of a federal guarantee for exist- tially effective because current revenues are shared up to 40 ing state debt. Still, they tend to limit state deficits to percent with subnational governments. The national gov- unpaid service on outstanding debt. Second, the agree- ernment's primary spending (net of interest and of transfers ments of 1998 are written in a way that requires the federal to the territorial governments and entities) has roughly government to withhold debt service from transfers, as in doubled its share of GDP since the late 1980s, from 5 to 10 Argentina. Even when the much-publicized refusal of percent. The national government increased its own expen- iMinas Gerais to pay debt service resulted in the federal gov- ditures for the military and judges-as a response to ernment's covering the state's debt service in order to pro- increasing violence from drug traffickers and guerrilla war- tect the national credit rating, the state was effectively fare-and for social security, as a consequence of the social forced to pay, because the federal government dedocted the security reform of 1993. It did nor reduce central adminis- state's payments from transfers. (The most recent crisis in tration outlays, however, which should have occurred with 1998-99 has not yet been fully resolved.) Third, Sao the transfer of other functions to the subnational levels. Paulo-the largest state-has become one of the reform During the 1990s the national government deficit has states. This is likely to shift the balance of political power in grown from less than 1 percent of GDP to almost 5 per- the federal congress in favor of the reformers, making it eas- cent. The transfers to provinces (departments) and munici- ier to impose budget constraints on the remaining spend- palities increased by only about 1 percentage point of GDP thrift states. If this strategy works, the recent crisis may end each, corresponding to only about one-third of the increase up being as important for Brazil as the 1994-95 crisis was of central government expenditures (Clavijo 1998). for Argentina-as it tests the system and sets precedents. At the departmental level, spending and deficits increased by only 0.1 or 0.2 of a percentage point of GDP Colombia: Attempting to establish hard budget in the 1990s. Municipal spending, however, has increased constraints through ex ante regulations. by almost 2.5 percentage points of GDP The aggregate Unlike Argentina and Brazil, Colombia has had a strong municipal deficit rose, nevertheless, by only 0.2 of a per- centralist tradition for a century, since the 1886 constitu- centage point of GDP, because transfers from the national tion. This centralism was partly a reaction to geographic level and tax revenues had also risen by about I percentage fragmentation, civil war, and losses of territory in the nine- point of GDP each. The increase in transfers to depart- teenth and early twentieth centuries. Political decentraliza- ments can be attributed mostly to the central government's tion came recently in the form of elected mayors in 1986, objective of increasing the coverage of social services, espe- and elected governors in 1991. In contrast to Argentina cially basic rural health and education, and to salary and Brazil, Colombian political decentralization was imple- increases negotiated by the central government. A special mented without the additional trauma of a transition from fund had to be set up to assist departments in paying the military to civilian rule and without the complication of nationally negotiated salary increases. While initially small, hyperinflarion. Indeed, Colombia has enjoyed a tradition of this fund grew to about one-fourth of the total transfer to and reputation for sound macroeconomic management, departments by 1998-a major non-formula transfer. which was associated in some minds with central fiscal con- Scheduled increases in the transfers to municipalities have trol. In the 1990s, however, the expanding political and fis- increased and will continue to increase the fiscal burden on cal decentralization has coincided with the expansion of fiscal the central government. Some transfers came through deficits and a failure to share in the general increase of macro- matching-grant investment programs, which at least until economic stability in most other parts of Latin America. 1997 had considerable political discretion in the allocation 44 ADDRESSING THE MACROECONOMI1C THREAT THE QUEST FOR HARD BUDGET (:ONSTRAINT.r of projects. Until the mid-1990s the transfers increased in account of this inflexibility in their ex ante evaluations of importance, reaching 40 percent of total transfers to the ability to pay. In the case of municipalities, the debt municipalities. Since then they have declined slightly and in crises were related to runaway expenditures financed with 1997 were made into soft loans from the Colombian devel- the pledge of increasing transfers. The inclination of private opment bank Financera de Desarrollo Territorial S.A. banks to lend to subnational governments that had little (FINDETER), with perhaps more transparency (Ahmad debt to the national government meant that the national and Baer 1997). government was not heavily exposed to the threat of terri- Revenue sharing has complicated stabilization efforts. torial default, in contrast to Brazil. In Colombia, the sub- At present, more than 40 percent of all current revenues national borrowing has not been large enough potentially are automatically transferred to departments and munici- to cause a general banking crisis that would require palities, and the constitution has mandated that the share national government intervention. Nonetheless, subna- increase further. Unlike the other major decentralized tional governments in fiscal straits have often received ad economies of Latin America, the Colombian national gov- hoc bailouts from the center, most notably Bogota in 1991 ernment has virtually no unshared tax bases, not even inef- and Medellin in the late 1990s for its debt from the con- ficient ones (Perry and Rodriguez 1991). Successive tax struction of a metro. They have also restructured their debts reforms, attempting to mitigate the growth of the national with private lenders several times, and recently this has been deficit, have thus automatically increased subnational without financial help from the national government. expenlditures. The government has presented legislation to Witnessing the high rates of growth of subnational debt congress to earmark part of the transfers for pension to domestic banks in 1993 and 1994 (more than 60 percent reserves, a large unfunded liability of the states which per year in real terms, according to the Superintendency of looms large on the horizon in Colombia, as in Brazil and Banks) and the debt crisis of several subnationals in 1995, elsewhere. the national government attempted to exert some control In Colombia the controls on subnational borrowing have over it. On the supply side, in 1995 the Superintendency of varied over time. Up to and including the 1980s, all sub- Banks tightened banking regulations, which slowed the national borrowing had to have approval from the Ministry growth of subnational debt in real terms for a while. A law of Finance. It was rarely granted. This was natural, since enacted in 1996 aimed to prevent indebtness in excess of the subnational entities were appointed representations of the ability to pay, through a system of warning signals the central government and had no political or fiscal auton- that would prompt direct control from higher levels of omy. In the late 1980s and the 1990s, the ad hoc approval government. process gradually allowed more freedom for domestic bor- A rating system for subnational borrowers (established rowing, as subnational political and fiscal autonomy by the so-called Traffic Light Law) scheduled for full effec- increased. From 1991 to 1998 the total debt stock of tiveness in 1999 has not been able on its own to prevent departments and municipalities actually declined as a share unwise borrowing, but has worked as a complement to of GDP, but the debt nonetheless reached a crisis point for hank regulation. Now banks have to provision against loans several entities twice during this period (in 1995 and 1998). to departments and cities that do not score well enough on The decrease was due to amortization of external debt, part the rating system, which may make the banks more cau- of which was taken over by the national government in tious in lending. Whether these measures will be enough to 1992.7 Domestic debt of the subnational governments to prevent excess lending will depend on whether the grew in the 1990s, especially to the banking sector in which national government refuses to bail out overindebted states debt rose from 2.6 percent of GDP in 1991 to 4.6 percent and municipalities and their creditors in the present crisis. in 1997. In 1998 the consortium of creditors to the department of The departments' debt in Colombia has been problem- Valle de Cauca (Cali), under pressure of the combination of atic mainly because they have little discretion over their bank regulation and the Traffic Light Law, renegotiated the receipts or spending, most of which has to go for salaries. debt and the fiscal program. If this arrangement with other Neither the departments nor the creditors took sufficient departments requires that they and their creditors bear the 4 5 BEYOND TH- CENTER: DECENTRALIZING THE STATE primary responsibility and costs of fiscal adjustment and and fiscal management. State governments and a few debt renegotiation, it will provide an incentive to all credi- municipalities have borrowed for some time from the tors to make responsible borrowing decisions. As experience national development banks; recently they have been bor- in the other countries shows, creditors will not lend to bad rowing from commercial banks as well. In the aggregate credit risks where they have no hope for a bailout from above, the state debt levels are low-about 2 percent of GDP- although the debt has been a fiscal problem for some states. Mexico: Ex ante controls andpolitical grip as budget Thc low ovcrall indebtcdness rcsults mainly from cx antc constraints. What does the future hold? controls on borrowing. First, the constitution forbids any Constitutionally, Mexico is a federal structure, and since the subnational borrowing from abroad or in foreign currency. 1980s considerable spending has been handled by the The rule has been observed, except for one or two state states. One party, the Partido Revolucionario Institutional bond issues that were effectively indexed to the U.S. dollar. (PRI), has dominated governments at all levels for decades, Second, each state congress annually sets borrowing ceilings however. So the country operated almost like a unitary for the state and its municipalities. (For the Federal District, state, with the president and the head of the PRI selecting where local elections were first held only in the 1990s, the the party candidates who would win the elections for gov- federal congress still approves the annual budget and bor- ernors and mayors of large cities. Two important changes rowing ceilings.) Third, creditors can get and almost always took place in the 1990s. First, the political system opened want a guarantee in the form of access to federal-state up, allowing other parties to compete with the PPI. To transfers. This requires a review by the Secretario de date, these new parties have won almost one-third of the Hacienda y Credito Publico and proper registration. The state governor seats and many municipal presidencies, and Ministry of Finance can deny an application or request a in 1997 the PRI lost its majority in the national congress. reduced amount.' The development bank, Banco Nacional This activity has stimulated a federalism that had lain dor- de Obras Publicas (BANOBRAS), also reviews the fiscal mant in the constitution. Governorships have become picture of borrower states. While some say that these important offices in the careers of politicians. The second reviews are perfunctory, they and the borrowing restrictions important change has been the rapid increase in the trans- seem to have prevented extremes of excess borrowing. It fer of resources to states and municipalities. At the same remains to be seen how this will work with the loosening time, the authority of the states to control this spending has political grip of the PRI, greater political autonomy of sub- been reduced, while the authority of municipalities has national governments, and a growing private financial sector. been increased. Defaults on state debt have arisen in three contexts in The transfer of nominal responsibility for health and Mexico. The largest occurred in the context of economic basic education to the states in the early 1990s required and fiscal contraction after the economic crisis of 1994-95. unification of the state and federal components, which had Many states could not service their debt, so all got some previously operated as parallel systems. The national min- degree of debt relief (and bailout for creditors) from the fed- istries remained in charge of establishing curriculums and eral government in exchange for fiscal adjustment, standards and negotiating wages and terms of employment although the terms for individual states varied. A second with the national union. In other areas of local public ser- context was large infrastructure projects, like metros and vices, the congress in 1997 and 1998 voted for more decen- toll roads that did not yield the return expected and neces- tralization transfers, but mostly to the municipalities and sary to service the bonds that financed them. Here again with reduced leeway for any control by the states over these the federal government provided debt bailouts along with a funds, although the constitution requires that they appear takeover of the infrastructure assets. on the books of the states. It remains to be seen whether the In addition to these discrete episodes, every year until fiscal trend toward municipal federalism will continue, or 1998 most states received some extraordinary transfers whether it will somehow be reversed, perhaps as a result of from the federal government, usually related to covering the growing political prominence of the governors. service on old debt or paying teacher salaries. The amounts These developments in overall intergovernmental rela- varied from year to year and across states, even adjusting for tions help explain what has happened in the areas of debt population. Without the extraordinary transfers, the states 46 ADDRESSING THE MACROECONOMIC THREAT THE QUJEST FOR HARD BUDGET CONSTRAINTS as a group would have been running primary deficits dur- strongly centrist systems, political decentralization has ing 1995-97, rather than primary surpluses. These bailouts advanced through a combination of two processes. First, were not important on a national macroeconomic scale, but the traditional elites (in Mexico the PRI and in Colombia they may have created an environment where creditors did the two traditional parties, particularly the Liberals) sought not worry much about the creditworthiness of the states, to increase their legitimacy with those who were defecting and states faced no harsh penalty for overborrowing. The from traditional parties and sometimes joining insurrection part of the budget allocated for these extraordinary trans- movements, especially in Colombia. Decentralization was a fers was cut back in 1998 and largely eliminated in 1999. way to appease these groups; perhaps it was less threaten- Thus, the federal posture toward state debt has been ing to traditional power relations than if the concessions changing rapidly in the last few years. The new system is had all come at the national level. Second, as the non-tradi- still taking shape-it remains to be seen how it will work tional parties gained footholds on the periphery of the polit- in practice. ical system-governors, mayors, members of congress and (in Colombia) the constitutional assembly-they voted and Summary of Cases lobbied for decentralization of expenditure to the lower lev- The first two cases-Argentina and Brazil-illustrate the els of government where they had a larger share of the problems of stabilizing high-inflation economies at the power. same time that finances are being decentralized and democ- racy is being restored. In Argentina, the federal executive Decentralization Institutions and Fiscal was able to use the political dynamics of the federal system Management and the fear of hyperinflation (as well as the joy for its end, What institutional arrangements reduce the likelihood of and the economic and fiscal rebound) to pressure the excess debt by national or subnational governments as a provinces to bring thcir fiscal policy to a stance consistcnt result of decentralization? with the national stabilization program. Today the country Table 3.4 summarizes the 12 economic and political con- seems to have institutions in place that impose an effective ditions identified at the start of the previous section as con- hard budget constraint on subnationals. In Brazil, the sta- tributing to containing deficits in a decentralized fiscal sys- bilization started later, and many states have not yet made tem, and rates the four countries discussed above in a the fiscal adjustment necessary to support the program. But comparative way. Some conclusions emerge from this com- in Argentina in the early 1990s (the first years after the sta- parison. bilization) there were also many states that did not adjust their fiscal stance. The situation in Brazil seems more severe Hard Budget Constraint for Transfers and Service in that none of the largest states adjusted early on and the Responsibilities federal government continued to give bailouts-debt The rules and institutions for raising revenue and for spend- rescheduling-that established perverse incentives. The ing are most directly important for determining the effi- state debt and exchange rate crisis of 1998-99 could turn ciency and equity in fiscal decisions. For overall macroeco- out to be a turning point for Brazil, similar to the Tequila nomic management they are indirectly important, shock for Argentina. That will depend on how the states facilitating or hindering decisions and adjustments that and federal government react. come out of the borrowing process and the political process. In the second two cases, Colombia and Mexico, the chal- Certainly, fixing the amount of the transfers in the lenge has been to sustain fiscal stability and moderate infla- aggregate is important. The absence of a firm fix on the tion at the same time that substantially more resources are total was a major problem for Argentina in the 1980s and, being transferred to subnational governments. Although counting the debt relief, has been a serious drawback for these countries did not have governmentwide political tran- Brazil in the 1990s. For Colombia and Mexico, having spe- sitions from military government to electoral democracy, as cial transfers that are decided on a case-by-case basis, with- in Brazil and Argentina, they increased democracy at sub- out any hard ceiling on the overall program, has been a national levels and widened the range of parties that could problem, although not yet an overwhelming one, thanks to effectively participate at the national level. Starting from debt constraints discussed below. Having some political dis- 47 BEYO\D THE CENTER: DECENTRALIZING THE STATE TABLE 3.4 Summary of Conditions for Reduced Danger of Excess Deficits VARIABLE ARGENTINA, 1991-98 BRAZIL, 1994-98 COLOMBIA, 1991-98 MEXICO, 1994-98 HARD BUDGET CONSTRAINT FOR TRANSFERS AND SERVICE RESPONSIBILITIES I. Transfers are specified by legal 'fes, set by formula and No, grants from the budget No, formulas exist but No, not at margin. formulas, not ad hoc. limited in aggregate, but and tax sharing set by 'ormula, with important exceptions. important for some small but large debt relief sometimes. provinces. 2. Central government effectively Yes No Yes, for departments, Yes for states; delegates functions to subnational No, for municipalities. No, for municipalities. governments tc go along with the delegation of revenue sources. BORROWING CONSTRAINTS 3. Central government strictly No, federal government Yes, in principle; very Yes Yes controls subnational borrowing has some authority but sophisticated rules but also cx ante. mainly requires reporring. sophisticated loopholes. 4. Central government credibly Yes No, cap on debt service allows No No commits not to have bailouts, interest capitalization. prohibiting explicit bailouts and fo-cing subnatioual governosents to service their debt. 5. Regulators force creditors to Yes, but rare because debt No Yes No accept the losses implied by any service is deducted from failure to service debt. transfers. 6. The central bank (and bank Strong autonomy and Limited autonomy; Yes No regulators) are more autonomous commitment to fixed discretionary monetary policy. and have a strong anti-inflation exchange rate. mandate. SUBNATIONAL AUTONOMY FOR FISCAL ADJUSTMfENT 7. Subnational governments raise No Yes, for large states; no, for No No much ofii their own revenue. small states. 8. Stubnational governments have Yes No, legally difficult to cut No for departments; No for states; aulthority to cut their costs. labor costs. Yes for municipalities. Yes for municipalities. INIERGOVERNMENTAL POLITICAL RELATIONS 9. Presidents are constitutionally Yes, very stron,g through Yes Yes Yes strong at the national level, decree posvers. 11:. Governors have little Partial, federal intervention No, no federal intervention. Yes Yes, but governors' constitutional autonoemy i(fr common, at least until autonomy growing. example, no reelection; central 199 4, but reelection allowed. government cani intervene). 11. Electoral rules orient congress Yes .No No, national at-large Yes rowvard riational, nor ltoal, elections for senate may iiecrests. change this. 12. Parry discipline is strong. Yes, closed lists. No, open lists. No Yes, very strong. 48 ADDRESSING THE MACROECONOMIC THREAT THE QUEST FOR HARD BUDGET CONSTRAINTS TABLE A.5 Channels for Control of Subnational Borrowing FOR BORROWERS FOR LENDERS Ex ante controls *Central government review of fiscal capacity *Credit rationing to states to carry debt *Prohibition of international borrowing * Prohibition of international borrowing Incentives: ex post consequences 'No bailouts *Government does not hold subnational -Regulations require provisioning againsr debr government debt from fiscally weak subnational governments *Debt service withheld from transfers *Strong supervision of banks cretion on the distribution of special transfers within a firm These may be circumvented, although local voters may overall limit, as in Argentina, does not seem to pose major subsequently punish elected officials. Central controls difficulties. Indeed, it may be a way to allow some leeway (national controls on states or state controls on municipali- for unavoidable patronage games while making it clear to ties) are sometimes firmer, but they may be hard to enforce all players that the game is zero-sum. if there is substantial political decentralization. In addition, A clear assignment of functions among levels of govern- getting adequate information is difficult; if the borrower ment is obviously important for a rational and efficient and lender both want to make a deal, they have incentives public sector. The aspect that seems most important for to collude in giving biased information to the regulator. macrofiscal management is clarifying what subnational lev- Most important, if the central approval carries, or is els will take over along with the transfers, and what the believed to carry, an implicit central government guarantee national level is not expected to take over in case of a sub- of the debt, it will increase the likelihood that the subna- national fiscal crisis. In Argentina, part of the national-level tional government and its creditor will want to agree on a adjustment was accomplished through agreements with loan even if they both know it is risky. They will conclude provinces that they would take on additional functions, that if something goes wrong, someone else will pay-the since transfers were increasing during the stabilizationl. In federal taxpayer. All four countries described above provide Mexico, on the other hand, a significant loophole in the examples of this problem, although it has been more fre- budget constraint of states has been the willingness of the quent and of larger magnitude in Brazil. Brazil has the federal government to take over failing state projects in most elaborate ex ante controls, mostly put in place after transport and electricity. In Colombia, the lack of an the debts were already oversized, but due to lack of ade- explicit assignment of exclusive responsibilities to munici- quate enforcement they actually helped reinforce the moral palities contributed to the outcome of high and rising cen- hazard of expected federal bailouts. In contrast, Mexico tral government expenditures at the same time that munic- seems to have had more success with enforcement. The rea- ipal expenditures were increasing significantly. sons for different outcomes with the same instrument seem to lie in differences in political institutions, as discussed Borrowing Constraints below. Controlling borrowing by subnational governments has Usually ex ante controls are imposed on the borrowers, two main dimensions: their type or timing-ex ante con- but they might also be imposed on the creditors, as with the trols or ex post incentives-and whether they act on bor- central bank moratorium on lending to states in Brazil or rowers or creditors. Together these make a matrix with four the Superintendency of Banks regulations in Colombia. In cells, as in Table 3.5. Mexico the constitutional provision against foreign borrow- One response to problem subnational borrowing, actual ing by (and lending to) states and municipalities is an ex or anticipated, is to have ex ante controls or even prohibi- ante control on both sides. tions (Ter-Minassian and Craig 1997). If they are well The second row of Table 3.5 represents complementary implemented, they can help eliminate debt problems or approaches that ensure the consequences of default provide make them more manageable. Sometimes the controls are either the borrowers or the lenders, or to both, with incen- self-imposed, for example, by a state in its constitution. tives to discipline themselves. Of course, the worst incen- 49 BEYOND THE CENTER: DECENTRALIZING THE STATE tives arise if the central government (national to states or it may seem no accident that in some top-performing states to municipalities) gives bailouts when subnational decentralized countries, like Switzerland and the United governments overspend and creditors finance the debt. States, the subnational governments (especially the cantons Bailouts can take several forms-special transfers, takeover or states) raise most of their own resources, while in devel- of costly functions, takeover of debt or contingent liabili- oping countries with many macroeconomic problems aris- ties, or subsidized refinancing of debt. All of these reduce or ing from the decentralization process, the states get much eliminate the cost to the original decisionmakers, allowing of their revenue as transfers from the center. But this over- them to make mistakes with other people's money. The simplifies the problem; the relatively high share of local rev- essence of improving incentives is to ensure that those who enue in the largest Brazilian states did not prevent them make mistakes pay a substantial share of the cost. This can from causing the biggest macroeconomic headaches. And put the burden on borrowers by forcing repayment, for some countries, such as the United Kingdom and Australia, example, through deductions from their share of centrally rely on transfers to finance subnational governments, with collected tax revenue, as in Argentina (but without com- no adverse macroeconomic consequences. pensating special transfers as in Mexico). It can also operate For short-term adjustment to macrofiscaL necessities, the by making creditors take losses for any subnational debt opportunity for subnational governments to cut costs, espe- that is not being serviced, as in Colombia and Argentina. cially by reducing the number of employees and their (See Box 3.1 for a discussion of the risks and costs in the salaries, seems critical. It helped in Argentina when source and terms of borrowing.) provinces were given the opportunity to reduce wages. The lack of spending flexibility in subnational governments has Subnational Autonomy for Fiscal Adjustment been a problem in Brazil, Colombia, and Mexico. In Colom- Reliance on transfers, rather than local taxes, does not bia and Mexico this lack has led to repeated special trans- appear to induce local governments to run deficits. At first, fers, cven with modcst debt-to-GDP and total rcvcnue BSOX 3.1 RWisk and Cost in te Source attd Terns of Boirrowig Subnational borrowers face impora tradeoffs in choos- may have broad external eects, iakig it mre likely ing the source and terms of their borrowing. Typicalyfor f Athat government would succumb to pressure 0eah choice, there Tis a lower 0risk or alower cost. Whichtheand bal Wout Sthe su:bnaonal- For thesefreasons, most coun- borrower prefers depends onthe source of revenues and tries place stronger ex ante controls on foreign than on the tmac>roeconlomic con text 000 t, 0000000000 0;000000000000000 j00domestic borrowinwg bysubnarionals.000000 0t0 0; 0000000000 In choosing 0betweent domestic0 and fotreign-currency000-0 Loa 0 0ns with floating 0interest rates ared similarly tcheaper 0 0 based loants,l which; arej typically linked to tethe glocationl00 of than fiMg-xe d-rare loans,0 but0 carryf the rinsk forthe borrower 0 the redtor,t foreignj curec oan sb .gare uual cheae, ht the rateswil ie hti ris worth takin nyfor rency 0devaluation.t But? theyf are only a better option for theV 000ready capcity toj increase i0nterest0payments ilf needed. A0 su bnaional borrowertif it fis well positioned 0to handlethe 000short 0term 0to mturiaty for aX loa alouulylwrh risk.: If it has a rev,enesource,suchas touis, taxe;Ws,0 or; rae n sago elfrabroe ht execstob 0 is a fgood opionpr for the biorrower. tIf the bSorroer £0does niot 0 gal seut re m larket access an expects thetrate ;tog b3e comilngf haearvnu ore fit 7:i,1s takna largeand cosrly drisk ;down. :Otherwise, ti otlygamle Aomnfaur handle mre 0cheaply. What0 isS more,0 foreign debt 0dealt: ing0 tly short-term debt. 0 $ if 0 0 000 0 00 0 00 0 0 t 0 00 00 0 50 ADDRESSING THE MACROECONOMIC THREAT: THE QUEST FOR HARD BUDGET CONSTRAINTS ratios, because the terms of the transfers from the national worked effectively in Mexico (at least until recently), rela- level commit almost all resources of states (or departments) tively well in Colombia, and in a rather perverse way in to paying salaries in education and health. Brazilian states Brazil. have somewhat more discretion over spending because the transfers are not explicitly linked to paying salaries, but Conclusion until very recently the constitution forbade layoffs or cuts in Econometric results reported here-on a worldwide sam- nominal wages. (The latter was not a constraint until after ple-suggest that fast decentralization normally leads to the stabilization of 1994.) In addition, national law (or the higher overall public expenditures and serious problems in constitution) mandates early and generous retirement for macroeconomic management. Indeed, the results show public employees at all levels. Such provisions encumbered strong evidence that increases of subnational expenditures the adjustment of states facing debt crises and contributed and deficits are associated with subsequent increases in to the expectation that the federal government had the national government spending and deficits. responsibility to solve the problem. The results also show, however, that in the long term, deeper decentralization is not associated on average with Intergovernmental Political Relations higher deficits-or surpluses-although it is associated These fiscal policy institutions do not stand in isolation, with larger overall expenditures if states do not collect their but derive their backing and commitment from the politi- own revenues. Therefore, decentralized countries seem to cal system, although once in place the institutions can set be able to eventually develop institutions and raise taxes, to some limits to political decisions. The power of the presi- avert the macroeconomic fears concerning deficits, though dent compared with that of governors has been important they have not escaped having large public sectors. for maintaining fiscal constraints in Argentina, Colombia, The four country cases examined show that it is not nec- and Mexico, and party discipline in Argentina. In Brazil, essary to fulfill all of the proposed 12 conditions to achieve meanwhile, the power of the governors, in their own states an effective budget constraint on subnational governments, and in influencing congress, and the lack of party discipline and hence sound macroeconomic management in a decen- all tended to make excessive state borrowing more likely. tralized context. However, the cases also show that if some The relations of political institutions to fiscal rules derive institutional conditions are missing, the others must be from three types of situations. In the baseline situation, the stronger. national executive (by itself or with the support of the In the area of revenues, whether states raise most of their attorney general and the courts) has enough power to own revenue (as in Brazil but not in the others) seems less enforce the spirit and letter of laws passed by the legisla- important than whether transfers to states (including debt ture. Colombia, Argentina, and Mexico, since the mid- bailouts) are non-discretionary. Clear formulas to set the 1990s, are in this situation. In the second type of situation, overall level of transfers in the aggregate and to each state the national executive has power to impose fiscal discipline are the first step to prudent fiscal management with decen- that goes beyond the regular laws, either through party dis- tralization. Without that, the states have little incentive to cipline (as in Mexico traditionally and Argentina in the take anything else seriously. 1990s) or through emergency interventions, as in In the area of spending, clarity in allocating spending, Argentina and Brazil during the military periods and in though important, does not seem as important as having Colombia at times before the constitution of 1991. In the the authority to cut or control costs (especially those related third type of situation, the states have so much power, rel- to personnel) in the areas where subnational governments ative to the national executive, that they can block the do take responsibility. national executive from carrying out even the rules agreed In the area of borrowing controls, ex ante restraints have to in the national legislature. Since the restoration of had favorable effects at times in Colombia, when they were democracy, Brazil has often been in this situation. complemented by ways to make creditors bear some of the These differences in the relations of political institutions cost of states' overborrowing or to force the states to pay In help to explain why in some situations the same instru- Brazil the ex ante constraints stand more in isolation and ment-for example, ex ante controls on state borrowing- therefore may have had the perverse effect of increasing the 51 BEYOND THE CENTER: DECENTRALIZING THE STATE impression that the federal government guarantees the 2. In addition to these problems, there are other efficiency and equity concerns with which we do not deal here, including those aris- states' debt. In Argentina, by contrast, the mechanisms to ing from the following: lack of capacity at the subnational tevel, ensure that provinces pay their debts in full seem to have igfo h olwn:lc fcpct ttesbainllvl eventual capture of government by local elites, and inequality of ser- worked well in the later 1990s (after a painful test of the vice provision because of unequal tax bases or poor allocation of system in the Tequila crisis), despite the absence of strong transfers, etc., which are examined elsewhere. ex ante borrowing controls. In Colombia and especially 3. Primary spending is net of interest payments and budgeted Argentina (after 1991), the debt constraints for states were transfers to subnational governments. Primary deficit is the usual def- reinforced by the autonomy of the central banks, which inition of overall deficit minus the interest on debt (see also Treisman helped keep a relatively hard budget constraint on govern- 1998). 4. First differences were used because with the regression in lev- ment at all levels. In the end, the name of the game is effec- es h iesre eennsainr n h tosi h ets ' ~~~~~~~~~~els, the time series were non-stationary and the errors in the regres- tive hard budget constraints on subnationals, and these can sions were serially correlated. be achieved-and undone-in several ways. 5. These results show up when we look at changes between five- The political factors are largely given from the point of year periods, as well as annual changes. view of the policymaker, but they are important. Political 6. The Argentina and Brazil accounts draw directly on Dillinger centralization, as in Mexico and Colombia before 1990, and Webb 1999a, and the Colombia account on Dillinger and Webb makes the other fiscal/institutional concerns almost redun- 1999b. 7. Much of this debt was to multinationals for electrification pro- dant. But as the political system delegates more and more jccs in the 198Os, before the subnational governments had political true autonomy, the fiscal rules become more and more or fiscal autonomy. important for providing the proper incentives for decen- 8. Until 1997, the federal government essentially had an obliga- tralized democratic fiscal management. tion to give these guarantees if the lending was within approved lim- its. In 1997 the federal government tried to absolve itself of this func- Notes tion by making the states responsible for themselves. But the 1. Bailouts of subnational governments occur not only as a conse- commercial banks and states pressured the federal government into quence of excessive debt. Subnational governeni.icts miauy end up an equivalent arrangement where states could legally mandate this being insolvent due to unfunded contingent liabilities (financially function back to the federal government. The arrangement was tem- unbalanced pension systems for state employees; excessive guarantees porary and the finance ministry would like to terminate it in 1999. to private investors in infrastructure, losses of subnational public banks or public enterprises). Such cases are theoretically equivalent to excessive debt issuance. 52 CHAPTER 4 Empowering Municipalities or Schools? The Decentralization of Education s we have seen, over the past decade, decentralization of government has become common throughout Latin America. The education sector is no exception; in the 1990s the number of countries implementing significant decentralization reforms has increased rapidly (see Figure 4.1). At the same time, there has been a worldwide trend to give schools greater decisionmaking autonomy, in the interest of improving school performance and account- ability. School systems as diverse as those in Victoria, Australia; Memphis, Tennessee; and Minas Gerais, Brazil, have given authority to school heads, and then through a variety of mechanisms held them responsible for school performance. The two types of education decentralization-to lower levels of government and to individual schools- have very different origins and objectives. The decentralization of education to lower levels of government has almost without exception been undertaken in the context of a more general decentralization of govern- ment, the causes of which vary widely. In contrast, the decentralization of education to individual schools typically has been motivated by concerns about poor school performance. Both types of education decen- tralization are well represented in Latin America, and this chapter reviews the evidence to date of their var- ious impacts on schooling. The literature on education decentralization is growing merit mention. The World Bank recently completed several rapidly, but is still primarily descriptive in nature. Attempts studies on education decentralization worldwide (Fiske to assess the impacts of decentralization have suffered from 1996; Gaynor 1998); the Inter-American Development weak baseline data and poor research design, mainly result- Bank sponsored research on the effectsO of different organi- ing from inadequate data. Weak evaluations are not limited zational arrangements in education in Brazil, Chile, and to Latin America or developing countries. For example, Venezuela (Savedoff 1998); and the Centro Estudios para Summers and Johnson (1996) reviewed over 600 evalua- America Latina (CEPAL) worked with researchers in five tions of school-based management in the United States and countries (Bolivia, Brazil, Colombia, Mexico, and found only two with an adequate research design. Nicaragua) to assess education decentralization strategies Several recent studies and evaluations of primary and (di Gropello 1999). In addition, this chapter draws on sev- secondary education, both in Latin America and in other eral country-specific evaluations from Latin America and regions, provide the basis for this chapter, of which three selected evaluations from outside the region. 53 BEYOND THE CENTER: DECENTRALIZING THE STATE FIGURE 4.1 Latin American Countries Implementing Education Reforms 14 1998 NIEX, PAR, NIC, PAN 12 - HON, BOL, 10 CHI 1997 1994 VEN GUA, BOL 8~~~~~~~~~~18 1976 CHI~~~~~~~~~~~~~~~~~~~19 -~~~~~~~~~~~~~~~~~~~~~~~~E G ,/OL. N IC 4 1988 /19 BRZ ES, BRZ;, 1981 2 / ARG, COL 2 ARG 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 Rationale for Education Decentralization ing coming from centralized sources. Ensuring equality of The economic rationale for decentralizing education is to educational opportunity, measured at a minimum by equal- improve social welfare and technical efficiency (Winkler ity in educational spending, is a further argument for a high 1993). Decentralized decisionmaking, it is argued, will give degree of centralized financing in countries where income local voter-consumers greater voice in the service mix they inequality is high. receive and, hence, raise their welfare. Presumably, the Improved technical efficiency is the other rationale for more local the decision the greater the voter-consumer education decentralization. Here the argument has several voice will be-that is, greater at the school level than at the elements. First, to the extent that prices and production municipal level-and greater in single-purpose govern- processes vary across localities, there are obvious efficiencies ment (for example, a school district) than in general-pur- resulting from letting local decisionmakers allocate budgets pose government. If the financing and supply of education across inputs. Second, in situations where the capacity of are determined locally, the improvement in social welfare central government ministries to monitor and supervise will be still greater, because the median voter-consumer will local schools has been weak, devolving these responsibilities tax himself or herself only up to that point where the mar- to local voter-consumers may increase the accountability of ginal tax costs and marginal educational benefits are the school for its performance. The interest of local voter- equal. consumers may be higher if they are also contributing However, these arguments presume a world in which resources financial or non-financial--to the school. democracy works well, and in which all externalities are A final argument for decentralization is that having captured locally. If there is the risk that local elites will cap- many suppliers rather than just one supplier is likely to lead ture local decisionmaking, social welfare may not improve; to a wider variety of experiences and innovations. If there this risk may be higher in societies with little experience in are adequate means for communicating and exchanging participative democracy at the local level. If the externali- information on these experiences, a decentralized system ties alleged to result from education, especially basic educa- may lead to more rapid innovation and change than a cen- tion, are distributed beyond the confines of the locality, tralized one. There is some evidence for this argument in there is a strong argument for a high percentage of financ- the case of Brazil (Xavier, Sobrinho, and Marra 1994). 54 EMPOWERING MUNICIPALITIES OR SCHOOLS THE DECENTRALIZATION OF EDUCATION The Educational Context of Decentralization tests as Colombia (see Figure 4.3). The low quality of basic The problem of access to basic schooling has been solved for education constrains the quality of higher levels of educa- most children in Latin America. Now, there is a growing tion and puts LAC at risk in its capacity to compete eco- consensus that it is the quality of education that must be nomically with the rest of the world. In addition, while chil- improved, especially in the public schools and especially for dren from all income groups now have access to basic poor children (Summit of the Americas II 1998). Low qual- schooling, there remain large inequalities in educational ity is reflected in high rates of repetition and dropout, and opportunity as measured by quality of schooling. Compared low performance on standardized tests of scholastic achieve- with children from economically advantaged homes, chil- ment. Colombia, in an international test of science and dren from poor households are likely to receive lower mathematics, scored well below East Asian countries and schooling investments from both the home and the school. only slightly above African countries (see Figure 4.2). While the rationale for decentralization is at least as In addition, the evidence coming from a United Nations much political as it is educational, the proponents of decen- Educational, Scientific and Cultural Organization tralization expect one impact to be improved quality. Other (UNESCO) test of educational achievement administered possible impacts are changes in efficiency and equity. in 11 Latin American and Caribbean (LAC) countries shows Because of the importance of raising quality and the limited that, excluding Cuba, the performance of most countries in information available on efficiency and equity, this paper LAC does not differ greatly, suggesting that most LAC focuses on the impact of decentralization on educational countries would fare as poorly on international achievement quality in LAC. Typology FIGURE 4.2 Decentralization takes many forms. It varies by the level of Average Math Achievement Test Scores of Eighth Graders, government to which decisions are devolved, the kinds of Selected International Comparators decisions moved to other levels of government, and the ori- entation of the decentralization-whether it emphasizes governance or pedagogic changes. 600- Singapore -Level of Decentralization 600- Korea i 607 1 The level to which educational decisions are decentralized ranges from regional and local government to the commu- 500 - Ireland 527 nity and the school. In many federal countries-Brazil, 0Ireland 527 Canada, Germany, India-the states or provinces that make up the federation have had a constitutional responsi- 400 - Portugal 454 bility for education. In other countries-Argentina, Mex- Colombia a 385 ico, Venezuela-education responsibilities have historically - 354 been located in the central government, but have been 300 S. Afrca largely devolved to states or provinces over the past decade. Local governments quite often have educational respon- sibilities, especially for primary and secondary schooling. In 2(0 -: the United States, most state governments have devolved educational management to single-purpose local govern- ments, or school districts. In other countries-Brazil, Chile, I00 Colombia-municipalities have been given increased edu- cational responsibilities over the past decade. Finally, some countries have given school councils and o schools significant autonomy in managing (but rarely S-- IEA, 1996 financing) education. The Netherlands is perhaps the best 55 BEYOND THE CENTER: DECENTRALIZING THE STATE FIGURE 4.3 Third-Grade Language Achievement, Latin American Countnies (Median 25%, 75%) Argentina Bolivia Brazil -- Chile Colombia - Cuba Honduras - Mexico - Paraguay - Dom. Rep. Venezula -- 150 200 250 300 350 400 Sqa-: UNESCO data in PREAL/CIDE 1999 example of a country that has empowered parents to create the purposes of this paper, we adapted these definitions to their own schools with financing and other support from be consistent with Latin American experience and available the central government. Recently, in cities like Chicago and information. The content of each group is given in Table 4.1. Memphis in the United States, it is the school district that has given the school significant management autonomy. Structure and Content Just as the composition of educational functions that are Decisionmaking Powers decentralized varies across countries, so too does the goal Some educational functions are decentralized, even within and orientation of the decentralization reforms. In some centralized systems, and others are centralized, even within reforms, local control is the goal, either for political reasons decentralized systems. An Organization for Economic or to strengthen accountability by the schools to their Cooperation and Development (OECD) survey of its mem- clients. The focus of these reforms is on structure, that is, bers, for example, shows that schools make most of the transferring decisionmaking powers and responsibilities to decisions about the organization of instruction, even in cen- lower levels of government or to school councils. Implicit in tralized systems. These decisions include choice of teaching these reforms is the expectation that local control and methods, textbooks, criteria for grouping students within accountability will improve efficiency, both in the uses of schools, and day-to-day methods of student assessment. On resources and in the match between client demand and the the other hand, in most European countries, most person- supply of school services. nel management decisions are made at a central level. In other reforms, the goal is improved learning, and the The OECD methodology for measuring the degree of transfer of decisionmaking powers is simply a vehicle for education decentralization divides educational functions attaining that goal. These reforms put more emphasis on into four groups: the organization of instruction, personnel the content of education reform than on the structure itself. management, planning and structures, and resources. For Parental participation is valued by these reforms because it 56 EMPOWERING MUNICIPALITIES OR SCHOOLS' THE DECENTRALIZATION OF EDUCATION TABLE 4.1 Level of Decentralization Types of Decisions that may be Decentralized The level of education decentralization varies widely within Organization School attended by student. Latin America. In Argentina, primary and secondary edu- of Instruction Instruction time. cation and the normal schools were transferred from the Choice of textbooks. central government to the provincial governments (in 1976 Curriculum content. Teaching methods. and 1991, respectively), and today most decisionmaking authority remains concentrated in the provincial education Personnel Hire and fire school director. Management Recruit and hire teachers. ministries. In this respect-the concentration of decision- Set or augment teacher pay scale. making authority at the regional level-Argentina presents Assign teaching responsibilities. a unique model in Latin America, although Mexico appears Determine provision of in-service training. to be quickly evolving in a similar fashion (see Annex Planning and Create or close a school. Table 4.A. 1). Structures Selection of programs offered in a school. Definition ofcourse content. Brazil has a long traditon of decentralized education, Set examinations to monitor school performance. with most authority concentrated at the state government level. The state's preeminent role in secondary education Resources Develop school improvement plan. Allocate personnel budget. was confirmed by the 1988 constitution, and municipalities Allocate non-personnel budget. were given the preeminent role in financing and delivering Allocate resources for in-service teacher training, preschool and primary education. In addition, during the 1990s, some states (for example, Minas Gerais) have trans- is viewed as contributing to the success of education, and ferred significant decisionmaking authority to the level of not because it improves accountability. Matching client the school. demand with what the schools offer is important only to Chile's education decentralization effort is long and the extent that client demand is consistent with raising complicated. It began in 1981 with the transfer of deci- quality. sionmaking authority to the municipalities, on the one Although it is tempting to contrast structural reforms hand, and to nonprofit schools, on the other. It continued with reforms that emphasize content, this typology is in in the 1990s with the central government's exercising fact a continuum with most decentralization reforms stronger pedagogic leadership and working directly with encompassing elements of each, the schools to bring about school-level improvements. El Salvador's decentralization effort was not universal Typology Applied to Recent Latin American but, instead, targeted rural areas where central government Experience schools failed to function during the civil war. Hence, while Education decentralization has taken many forms in Latin for traditional public schools, educational decisionmaking America and the rest of the world. It always includes the remained concentrated at the level of the central govern- transfer of authority and responsibility from higher to lower ment, the new rural Educaci6n con la participaci6n de la levels of government, but it varies considerably in terms of Comunidad (Education with the Participation of the Com- which decisionmaking powers are decentralized and who munity [EDUCO]) schools were given significant decision- receives those new powers. Figure 4.4 illustrates the varia- making authority and autonomy. The success in imple- tion in the location of important educational decisions in menting the EDUCO model has led to current efforts to Latin American and OECD countries. In addition, since decentralize traditional schools as well. education decentralization is often part of a broader educa- Mexico's education decentralization is a combination of tion reform effort, there is considerable variation in practice the Argentinian and El Salvadorian models. The 1993 Ley in terms of accompanying school improvement measures. General de la Educaci6n transferred most educational In the discussion that follows, the typology will be decisionmaking authority for primary and secondary applied to the experiences of Argentina, Brazil (with a focus schools to the state governments, but the central govern- on the state of Minas Gerais), Chile, El Salvador, Mexico, ment's important role in financing education through nego- and Nicaragua. tiated transfers to the states resulted in continued de facto 57 BEYOND THE CENTER: DECENTRALIZING THE STATE FIGURE 4.4 Level of Decisionmaking in Education Sector 100 - 90 80- 70- 60 Centra] 50 El Regional 40 - Local 30 - School 20- 10- 0 Neth. N.Z. Spain Chile Parag. Arg. U.S. France Soxnce: OECD 1998. centralization. Real decentralization to the states occurred Region, only Costa Rica, Ecuador, Panama, and Uruguay only in 1998 when educational transfers became automatic. have chosen to retain centralized educational systems. In addition, the central government continues to directly operate a system of rural (National Board for Educational Decisionmaking Powers Improvement [Consejo Nacionalde Fomento Educativo- What does it mean that education has been decentralized to CONAFEI) schools to ensure learning opportunities for a particular level? As noted earlier for OECD countries, sev- remote rural, and especially indigenous, children. While eral educational decisions, such as choosing textbooks, not nearly as autonomous as El Salvador's EDUCO selecting teaching methods, and taking responsibility for schools, the CONAFE schools give parents a considerably implementing school improvement plans, tend to be more important role than is found in the traditional pub- located at the school level irrespective of the level of decen- lic schools. tralization. Others, such as setting the core curriculum or Finally, Nicaragua's education decentralization has administering and reporting results on achievement exami- evolved from an emphasis in the early 1990s on muncipal- nations, tend to be located at the national level irrespective ization to a clear policy in the late 1990s to transfer most of the level of decentralization. Table 4.2 illustrates the important educational management and finance decisions focus of key educational decisions in several countries in to the level of the school. Latin America. Several other countries in the region not included in Fig- Decentralization is mainly characterized by the locus of ure 4.4 have also adopted education decentralization poli- decisions on personnel and budgets. The greatest consis- cies during the 1990s. Colombia decentralized primary and tency is found in teacher and school director recruitment secondary education to departments (regional govern- and hiring decisions and the budgeting of non-personnel ments) and municipalities, and Bolivia is slowly imple- expenditures. In Argentina and Mexico these decisions are menting a similar policy. Guatemala and Honduras have located at the regional (provincial) level, in Chile at the followed the model of El Salvador's EDUCO schools with local (municipal) level, and in El Salvador and Nicaragua at their own programs of autonomous rural schools. In the the school level. Teacher pay decisions are sometimes 58 EMPOWERING MUNICIPALITIES OR SCHOOLS' THE DECENTRALIZATION OF EDUCATION retained at higher levels of government (as in Minas Gerais, with content and viewed as a vehicle to raise quality? The El Salvador, and Mexico), and in most cases are heavily answer, of course, is not simple. influenced by national policy setting minimum pay condi- The education decentralization experiences of Argentina, tions (for example, Chile) or national decisions about edu- Chile in the 1980s, El Salvador, and Mexico can be viewed cation finance (for example, Minas Gerais). as mainly structural in nature, but for very different rea- Of course, simple descriptions of decentralization fail to sons. In Argentina, primary and secondary education were capture important nuances. A case in point is the school devolved to provincial governments for mainly fiscal rea- improvement plan. Almost every country in LAC now sons. Hence, the goal of the reform was simply to move requires that schools or local jurisdictions develop improve- expenditure responsibilities to the provincial governments. ment plans, but as a recent assessment of the Chilean expe- There was little concern as to whether this would lower or rience illustrates, such plans are often carried out as a raise quality. bureaucratic exercise and fail to meet minimum standards In Chile, the Pinochet government simultaneously intro- of quality and community participation. When schools do duced a modified voucher scheme and municipalized public develop plans, they often lack the authority to implement education to increase competition among schools for stu- them, as in Colombia. And even when they have the author- dents, and thereby raise the accountability of schools to par- ity to implement, they may have no source of financing. ents. In El Salvador, the EDUCO model has put the Another case in point is the allocation of the personnel emphasis on the creation of school councils to receive and budget. The multiple constraints of national or regional pay manage government funds for the purpose of providing scales; collective bargaining agreements on working condi- schooling. While the main objective of EDUCO has been tions, including class size; and national curriculum require- educational-to improve access in rural areas-its focus has ments may translate into little real discretion at the decen- not been on interventions to alter the content and raise the tralized level. quality of schooling. In Mexico, education decentralization has been an integral part of a broader decentralization of Structure and Content powers to state governments in keeping with the political Have decentralization reforms in LAC been mainly struc- liberalization of the country. Finally, Nicaragua's policy of tural in nature-focused on increasing local control and school autonomy as the principle focus has been giving raising accountability-or have they been more concerned voice to parents and civil society on educational issues and, TABLE 4.2 The Locus of Key Educational Decisions and Responsibilities, Selected Latin American Countries GROUP DECISIONS ARG MIN GER CHILE EL SAL MEX NIC Level of decentralization R S L S R S ORGANIZATION Choice of textbooks S S S S N S Teaching methods S S S S S S PERSONNEL Hire-fire school director R s L S R S Recruit/hire reachers R R L S R S Set or augment teacher pay R R L N N S PLANNING Set performance exams N R N N N N Implement school improvement plan S S S S RESOURCES Determine expenditures R R N, L N R N, S Allocate personnel budget R R L N R S Allocate non-personnel budget R S L S R S N = national, R = regional, L = local, S = school. 59 BEYOND THE CENTER: DECENTRALIZING THE STATE in this way, increasing operational efficiency (Arcias and Given the difficulty of isolating the effects of decentral- Belli 1998). ization on learning and educational attainment, our In contrast to the cases described above, Minas Gerais approach is to look at how decentralization changes factors and Chile (since 1990) have focused on changing the con- known to be related to learning. First, we ask what is the tent of education and raising its quality through decentral- received wisdom on what characteristics define effective or ization. Minas Gerais granted a significant degree of auton- high-performing schools. Second, we ask how these charac- omy to the public schools financed by the state government teristics are reflected in the school environment. Third, we to define their goals, develop a school pedagogical project, ask how decentralization directly or indirectly affects any of and manage financial resources with the overall goal of these factors. improving education. Chile has attempted to balance the structural reforms of the 1980s with content reforms in the High-Performning Schools 1990s to raise educational quality, especially for the poor. There is a growing qualitative and quantitative research lit- While the recent reforms have been top-down in their erature on the characteristics of high-performing or effec- design and the goals they pursue, they have attempted to tive schools (Mohrman and Wohlstetter 1994; Creemers deepen the decentralization process and move pedagogic 1994; Darling-Hammond 1997) that mirrors the much decisionmaking to the level of the school. For example, larger literature on successful organizations (Barzelay 1992; beginning in 1992, teachers have been encouraged to work Lawler 1992). This literature concludes that high-perform- together to develop school improvement projects, which ing schools are characterized by strong leadership, highly the education ministry funds on a competitive basis. The qualified and committed staff, a focus on learning, and Teachers' Statute was revised (in 1995) to allow school responsibility for results. Another set of literature reviews directors to directly manage funds and to provide school- the evidence for the process by which schools improve, and based financial incentives for performance. Further, begin- yields coniclusions that are consistent with the effective ning in 1997, a competition to fund the best education schools' research. For example, in an evaluation of school improvement projects proposed by secondary schools both improvements on three continents, Dalin (with others provides financial incentives for performance and gives 1994) concludes that essential ingredients in successful school directors full management responsibility for imple- reforms are a sustained commitment to quality improve- menting the projects. ment, local empowerment to adapt programs to local con- ditions, strong emphasis on school and classroom practice, Evaluation Decentralization and strong support linkage between education authorities While the reasons for educational decentralization in Latin and the school "via information, assistance, pressure and America are often political or fiscal in nature, from an edu- rewards" (see Annex Box 4.A. 1). In the discussion that fol- cational perspective there is the expectation that decentral- lows, we group the variables associated with high-perform- ization will improve schooling outcomes. Schooling out- ing schools into the four characteristics used in the research: comes can be defined in a variety of ways, but at a leadership, excellent teachers, focus on learning, and minimum involve measures of the level and distribution of responsibility for results. learning and years of schooling attained by schoolchildren. For three reasons it is difficult to use these measures to Characteristics of Effective Schools evaluate educational decentralization. First, time series of Strong leaders have the capacity to effectively develop and these measures are seldom available. Second, these school communicate a schoolwide and communitywide commit- outcomes usually change slowly in response to any kind of ment to a common mission and vision for the school, and to educational intervention, including decentralization. Third, manage the implementation of the school's improvement it is very difficult to control for external shocks-ranging plan. The common mission and vision fosters teamwork from natural disasters and fiscal crises to teacher strikes and inside and outside the school and, most importantly, the changes in national education leadership-which may also process of developing them makes teachers and parents the influence school outcomes. 'owners" of efforts to improve learning. Leadership is espe- 60 EMPOWERING MUNICIPALITIES OR SCHOOLS) THE DECENTRALIZATION OF EDUCATION TABLE 4.3 and student learning, and do their work in a supportive Characteristics that can be Stimuilated Through Decentralization work environment. Teacher commitment is essential to CHARACTERISTICS DECENTRALIZATION VARIABLES THAT developing the teamwork required for schools to continu- OF EFFECTIVE CAN CONTRIBUTE TO SPECIFIC CHARACTERISTICS ally diagnose their own problems and devise their own solu- SCHOOLS OF EFFECTIVE SCHOOLS tions. Teamwork is also essential to permit the sharing of LEADERSHIP School directors are selected by the communitv using teaching experience required to continually improve teach- transparent criteria. School improvement plans are developed locally. ing practices. Effective evaluations of teaching performance Resources are transferred tO schools for the are critical to giving teachers information on what and how imnplemnenitation of school plans, they need to improve. The time required to participate in SKILLED AND Schools are given the authority to make the management of the school and the improvement of COMMITTED coririculum and pedagogic changes COMMITTD currculum nd pedgogic hangesteaching is unlikely to materialize in a work environment TEACHERS Teachers havc significant responsibility for developing r i school improvement plans. where teachers are not given time for these activities within Directors are given the authority to provide a their normal work schedule. In many LAC countries, substanrtive evaluation of teachers' performance. Schools are given the authority (and resources) to where double and triple shifts are common, it may be make their own decisions about the type of training logistically challenging to find che space and time for to be given teachers. teacher participation. FOCUS ON The school improvement plan emphasizes goals of Decentralization can contribute to excellent teaching in LEARNING improving learning (and associated results, such as a variety of ways. When decisions on significant pedagogic RESULTS reducing dropout and repetition). matters are transferred to schools, teachers are empowered Information on learning at the level of the school is transparent. and motivated to work collectively to improve the services delivered to students. When school directors are given the RESPONSIBILITY Directors have fixed-term appointments that may not authority to carry out meaningful evaluations of teaching FOR RESULTS bc renewed if improved learning goals are not met. staff, teachers can focus their training on what they need to improve. When resources for training and training deci- sions are given to the school, teachers and directors can pur- cially important in a service industry like education, where chase the training they need (demand-driven) rather than the the contribution of individual teachers is difficult to mea- supply-driven training provided by the education ministry. sure, and thus difficult to directly reward. In the absence of Excellent teaching focuses on student learning. A school strong individual incentives, leaders must motivate teachers system that is focused on learning provides a pedagogy, a to improve. These characteristics can be stimulated through curriculum, and resources appropriate to student needs. In decentralization. Table 4.3 summarizes our findings, most cases, it is the local school and its teachers who are Decentralization cannot, of course, convert school direc- best placed to diagnose and find pedagogic solutions to tors who are used to passively following ministerial orders individual student and collective school learning problems. into dynamic leaders overnight, but it can and often does Different kinds of students-rural, indigenous, poor, urban provide a transparent, competitive selection process for youth, and so forth-are also likely to have different learn- school directors, which selects in part for leaders. A good ing needs, with implications for the distribution of financial example of this is the Minas Gerais decentralization which resources to schools by higher levels of government. Rural (a) established a procedure for certifying qualified candi- children may require smaller class sizes, reasonable commut- dates to compete for school director positions, (b) required ing distances, or bus transportation. Indigenous children may candidates to present their proposals for school improve- require more costly bilingual instruction. Poor children may ments as part of the competition, and (c) empowered school require school lunches and subsidized textbooks. councils to make the final selection of the school director. Decentralization can facilitate and reinforce a focus on Excellent teachers commit to the high goals and stan- student learning by providing the information required to dards of the school, have the strong teaching skills required assess learning problems, devolving appropriate pedagogic to meet those goals, continually work to improve teaching decisionmaking to the school, and allocating additional 61 BEYOND THE CENTER: DECENTRALIZING THE STATE resources to schools with special needs. The visible product school system replaced tenure for school directors with four- of this process is a solid school improvement plan, con- year contracts and required each director to sign an annual structed with the active participation of teachers and the performance contract with the system specifying measur- community, and with real possibilities of being imple- able goals for the year. Schools that consistently fail to meet mented. Good information on student learning and on the goals may see their director dismissed and teaching staff value added by the school is essential to the diagnosis of reassigned (see Annex Table 4.A.2). learning problems that is a fundamental part of the school improvement plan. Good information is also essential to The Consequences of School Decentralization monitoring progress toward attaining learning goals. The In this section, we attempt to evaluate each of the educa- devolution of appropriate pedagogic decisions is critical to tion decentralization cases discussed in this paper in terms the local design of solutions to local learning problems. of its potential to raise learning, especially among children Finally, financing is important, both as a means of imple- from poor households. In some cases, such as Argentina, menting school improvement plans and making possible decentralization was just one component of a larger educa- the adoption of pedagogy that meets special needs. In par- tion reform. In other cases, such as Chile, education reform ticular, in the absence of additional resources, children and changes in decentralized responsibilities have evolved from educationally disadvantaged homes are unlikely to over more than a decade. Given the complexities of evalu- meet the educational goals required for them to escape ating reforms, we do not attempt to separate out the their parents' poverty. "decentralization" component for evaluation, nor do we try Establishing responsibility for results provides the incen- to evaluate the initial reform. Rather, we try to make an tives necessary for sustained educational improvement. A assessment of the reform as it looks today. school system with responsibility for results requires a set of The criteria for this evaluation are the characteristics of measurable learning goals, up-to-date information on decentralization that the research literature and professional school performance toward meeting those goals, rewards opinion attribute to high-performing schools (mentioned in for meeting goals and sanctions for not meeting them, and the section on "High-Performing Schools," above). Below active monitoring of progress. The participant held account- we give a summary assessment for each country reviewed in able is typically the school director or the staff of the school. this paper (Table 4.4); more complete information on each The one holding the school accountable may be the educa- country's education decentralization is given in Annex tion ministry, a school council, or both. In Latin America, Tables 4.A. 1 through 4.A.8. the failure by ministries to hold schools accountable is often cited as the rationale for the creation of elected school coun- Leadership cils, which have local knowledge of the school but often lack The decentralization experiences reviewed here vary greatly sophistication in systematically evaluating performance. in terms of the extent to which they have created the con- There can be no accountability at the local or school ditions that may give rise to strong local leadership. Neither level in the absence of devolution of authority to make ped- Argentina nor Mexico have given school directors any sig- agogic and resource allocation decisions at the local level. nificant authority and responsibility. Chile has recently Decentralization can contribute to accountability at the granted more authority to directors of municipal schools, local level by (a) devolving decisionmaking; (b) establishing and, of course, the directors of the private, subsidized performance contracts between schools and financing bod- schools have long had a high degree of authority. The ies (including central government ministries and parent-led EDUCO schools of El Salvador are mostly small and often elected school councils) that specify learning goals; (c) without school directors, and school autonomy is only inventing information systems, including standardized tests slowly being granted to the traditional public schools. of students' knowledge, to permit contract enforcement; Minas Gerais and Nicaragua are the two examples where and (d) creating performance-related rewards and sanc- school directors have significant authority, and in Minas tions, including the dismissal of school directors. For exam- Gerais, in particular, the open selection process implicitly ple, the decentralization reform in the Chicago, Illinois, values the leadership qualities of candidates. 62 EMPOWERING MUNICIPALITIES OR SCHOOLS? THE DECENTRALIZATION OF ED)LCATION TABLE 4.4 Assessment of Education Decentralization CHARACTERISTICS OF EFFECTIVE DECENTRALIZATION VARIABLES SCHOOLS RELATED TO EFFECTIVE SCHOOLS ARG MIN GER CHILE EL SAL MEX NIC LEADERSHIP Community selects director n.a. School improvement plans V ' Transfer funds to school N SKILLED AND School curriculum authority V '1 COMMITTED Teachers develop improvement plans V N TEACHERS Directors evaluate teachers N / n.a. Schools decide training a 'I FOCUS ON Learning goals specified 4 LEARNINIG. . Transparent information V RESPONSIBILITY Fixed-term appointments for directors N n.a. V FiOR RESULTS.. Competition for students n.a. Parents have effective voice na. = not available Teaching Excellence schools as assessed using school performance indexes, and Strengthening the teaching capacity of teachers has been a (c) provision of staff time to participate in professional high educational priority for most countries in Latin Amer- development groups, with financial support from the edu- ica in recent years. Argentina has embarked on a major cation ministry. upgrading of its normal schools. Minas Gerais has empha- sized the use of distance education to upgrade teacher skills. Focus on Learning Chile has provided competitive grants to universities to The emphasis on improving quality and raising student improve their teacher training programs and has sent large achievement is clear in the Argentine education reform, the numbers of teachers abroad to strengthen their teaching Minas Gerais decentralization reform, the evolving Chilean skills. Mexico has introduced a program, the Carrera Mag- reform of the 1990s, and some of the policies and programs isterial, to strengthen teacher evaluation and performance carried out in Mexico. It is less clear in El Salvador, where incentives. the emphasis has been more on raising access; and in However, few of the region's efforts to upgrade teaching Nicaragua, where the focus has been more on parental par- capacity have been accompanied by in-depth evaluation of ticipation than on scholastic achievement. However, even in teachers, additional compensated time to participate in those countries where national education reforms and poli- school activities and prepare lessons, and incentives for cies are focused on student learning, the conditions are not teachers to work and learn in teams. Among the countries always present for effectively creating a school-based focus reviewed here, Chile has the policies best aligned with on learning. changing teacher behavior and training. Teamwork among Argentina has adopted an ambitious reform to train a school's teachers in Chile is encouraged through (a) com- teachers, supply sophisticated feedback on individual stu- petitive funding of teacher-designed and implemented dent performance (at the secondary level), and provide school improvement plans, (b) bonuses (equal on average to additional financing for children with special needs. How- one month's salary) to the 25 percent highest-performing ever, schools, teachers, and local communities have almost 63 BEYOND THE CENTER: DECENTRALIZING THE STATE no authority to diagnose their own needs and design their Responsibility own interventions. Minas Gerais, in contrast, encourages It is in the realm of responsibility for results that Latin schools to diagnose, monitor, and evaluate; and expects American decentralization reforms are found to be most schools to produce improvement plans. The state govern- wanting. In Argentina, Chile, and Mexico at least one crit- ment provides funding for these plans and provides feed- ical element is missing for there to be real accountability. In back on student achievement. However, the focus of all this Argentina, performance goals are not specified, systems to effort is not necessarily specific learning goals, and teachers systematically evaluate performance are still under develop- and community members are not always active participants ment, and no one is at risk of losing their job or suffering in the process. lower pay due to the low performance of the school in As in Argentina, the Mexican educational reform has which they work. Performance goals are not specific in been guided and driven at the national level. While decen- Chile either, and there are few risks to schools that do tralization efforts have not been focused on improving poorly. The same is true for Mexico. Furthermore, in all learning, other components of the reform-including three countries school councils are largely nonexistent, so changes in teacher evaluation and pay, and additional schools are accountable to neither parents nor higher levels resources for poor and indigenous rural children-are of government. focused on learning. However, excluding the CONAFE In contrast, school councils are active in Minas Gerais, El schools, teachers and parents are not yet actively engaged in Salvador, and Nicaragua; school staff can lose their jobs for bringing about learning improvements at the level of the poor performance in El Salvador and Nicaragua; and school school (Gershberg 1998a). directors are at risk of losing their jobs in all three countries. Chilean reform efforts since 1990 have been focused on On the other hand, learning goals are rarely specified with student learning, especially for poor children. Teachers have any precision, and the systems for monitoring and measur- been actively involved in diagnosing their own needs and ing school performance with respect to specific goals need developing their own school improvement projects. The considerable strengthening. Catholic University (1998) evaluated the school improve- ment projects carried out during 1992-95 and concluded Empirical Findings that the largest change was increased innovation in teach- While rigorous evaluations of education decentralization ing practices, especially increased use of interactive learning are difficult to find, a few do exist. We review the findings processes, and increased teamwork among teachers. The to date of evaluations carried out in El Salvador and evaluation also found that, on average, schools that imple- Nicaragua with the assistance of the World Bank, and we mented improvement projects experienced increased stu- complement these findings with evaluations of decentral- dent achievement. However, only 60 percent of all schools ization in Brazil and Chile, and in two large U.S. cities, experienced achievement gains, reflecting the fact that not Chicago, Illinois, and Memphis, Tennessee. all improvement projects were focused on improving learn- The evaluation of El Salvador's EDUCO program by ing, and some projects attempted to simultaneously accom- Jimenez and Sawada (1998) compares teacher absenteeism plish too many objectives. and student achievement in EDUCO schools with that of In addition to funding school improvement projects, the traditional schools, controlling for student characteristics Chilean education ministry has provided additional funding and selection bias (since the EDUCO schools were not ran- for special needs, such as with the P-900 programn, which domly selected). Two results merit attention. EDUCO provided extra resources for the 900 poorest schools in the schools, with their close community monitoring of the country. Average student test scores are annually published school and the potential sanction that teachers would not for each school in the country, and the schools making the be rehired, had fewer days of teacher absenteeism than tra- most progress over time are eligible for financial rewards. ditional schools, but student achievement in EDUCO While the education ministry could improve the monitor- schools was no different from that of traditional schools. ing and evaluation of specific learning standards, Chile has These findings on student achievement complement an ear- most of the conditions in place to bring about significant lier evaluation that compared mean differences in variables learning improvements. between EDUCO and traditional schools (Ministry of Edu- 64 EMPOWERING MUNICIPALITIES OR SCHOOLS? THE DECENTRALIZATION OF EDUCATION cation 1997). Surprisingly, that study found no difference achievement as measured by the Brazilian national educa- between EDUCO and traditional schools in terms of the tional test. They found statistically significant but mixed number of decisions made at the level of the school, which results. The establishment of school councils and the direct suggests that the EDUCO model may not be fully imple- transfer of resources are associated with increased atten- menred. On the other hand, EDUCO parents are three dance and reduced age-grade lags, but have no statistically times more likely to engage in day-to-day classroom activ- significant relationship to student achievement. The local ities than parents in traditional schools, teachers in election of the school director, on the other hand, is posi- EDUCO schools spend considerably more time meeting tively associated with student achievement gains, but nor with parents, and EDUCO teachers are much more likely with the other measures of schooling outcomes. to visit the family to inquire why a student has been absent As noted earlier, Chile has passed through two reform from school. phases. The first, begun in 1981, emphasized changing the In contrast to the El Salvador findings, an evaluation by structure or organization of education through municipal- King and Ozler (1998) of Nicaragua's autonomous schools ization and the introduction of competition and choice. A found that autonomous schools make significantly more simple comparison of student achievement scores across the schooling decisions than do traditional schools, especially 1980s shows a decline in learning, but during this period on personnel matters and in determining the school plan real per-student education expenditures also declined, mak- and budget. However, even the autonomous schools seldom ing it difficult to isolate the reform effect. However, a 1998 make teacher-training decisions. Another key finding of the study by McEwan and Carnoy assembled school-level panel evaluation is that the degree of decisionmaking actually data to examine how the degree of competition and choice exercised by autonomous schools varies greatly, and there is across municipalities and over time affects public school a positive and statistically significant relationship between quality, as measured by changes in student achievement the degree of decisionmaking exercised and student rest scores. They conclude that this aspect of Chilean edu- achievement. Furthermore, the strongest positive relation- cation reform has had no effect on public school quality. ship to learning was found for variables measuring deci- This finding confirms the qualitative evaluations made by sionmaking on teacher staffing and monitoring of teacher other scholars that municipalization did not lead to any activities. Nicaragua also illustrates the potential role of the substantive changes in behavior and achievement in the central government within the context of decentralization: public schools (Espinola 1997). A recent qualitative assessment of Nicaragua's school The second phase of the Chilean reform began in 1990 autonomy discovered that educators strongly welcome the and, as noted earlier, simultaneously deepened decentraliza- active intervention of the central government in promoting tion and set clear goals of raising quality and equity. In con- a pedagogy of active learning (Fuller and Rivarola 1998). trast to the 1980s, student achievement on Chile's stan- The Minas Gerais reform has not been systematically dardized exam, Sistema de Medicion de la Calidad de la evaluated, but the results of the Brazilian national educa- Ensenanza (SIMCE), increased significantly, both in Ian- tion test put Minas Gerais at or near the top of student guage and mathematics (Cox and Lemaitre 1998). Nation- achievement in every grade and subject matter (Instituto ally, the number of correct answers increased by about 18 Nacional de Estudos e Pesquisas Educacionais [INEP percent. However, here, too, it is difficult to separate the 19981) The reforms undertaken by the state of Minas effects of decentralization reforms, such as introduction of Gerais in Brazil have been replicated in part by several school improvement projects, from other reforms (for other states. In particular, several states have now adopted example, in teacher training), and from significantly (a) the establishment of school councils, (b) the direct trans- increased spending over the decade. fer of resources to schools, and (c) the local election of The findings for El Salvador, Nicaragua, and Chile are school directors. Using state-level, pooled time-series, complemented by two careful evaluations carried out in two cross-sectional data, Paes de Barros and Silva Pinto de Men- large U.S. cities having large populations of poor and donca (1998) have analyzed the relationship between these minority students-Chicago and Memphis. As discussed in reforms and a number of schooling outcomes: gross enroll- Box 4.1, Chicago introduced largely structural reforms in ment rates, repetition rates, age-grade lags, and student the 1990s and followed up with a much stronger content- 65 BEYOND THE CENTER: DECENTRALIZING THE STATE based reform in 1995. A consortium of academic institu- 1998). Finally, an evaluation of the Memphis decentraliza- tions led by the University of Chicago has carefully moni- tion confirmed that school director leadership and teacher tored and evaluated the Chicago reform from day one. The support for reforms were critical to its implementation. most recent evaluation report concludes that year-to-year Taken together, the El Salvador, Nicaragua, Chile, gains in student learning have risen significantly (for exam- Chicago, and Memphis evaluations provide strong evidence ple, a 19 percent gain in achievement for fifth graders that educational decentralization can improve learning. between 1992 and 1996) since the beginning of the reform, What is notable is that those cases demonstrating the despite the fact that the socioeconomic level of students has largest positive gains have emphasized school autonomy been gradually decreasing (Bryk and others 1998). Earlier with pedagogic reform, especially true in Chicago since evaluations demonstrated, also, that school reform efforts 1995, in Memphis, and in Chile since 1990. resulting from autonomy are as likely to be initiated in poor as in rich neighborhoods. Conclusion In contrast to Chicago, the Memphis reform has been Education decentralization is a worldwide phenomenon, heavily content-based from the beginning (see Box 4.2). and Latin America is no exception. Although there are eco- The evaluation of the Memphis school reform confirmed nomic and education arguments for decentralization, the the Chicago results of sustained improvements over time. particular forms of decentralization in most Latin American Prior to implementation of the reform, the experimental countries have been driven mostly by politics. Given the schools (those subsequently undertaking school-based magnitude of education decentralization efforts in the reforms) had smaller student gains in learning than a group region over the past decade and the forms they have taken, of control schools. After one year of implementation, the it is time to assess their effects. gains of the experimental and control schools were the The evaluation of decentralization reforms is difficult same, and after two years of implementation, student due to lack of baseline data, incomplete implementation of achievement gains in the experimental schools were signif- many reform elements, lags between implementation, and icantly higher than in the control schools (Ross and others changes in such factors as behavior and resource allocation, ch00icago w ;h.asaoted t;@wo f dctiorefors Th0e .fir i- itook4 control an naed aetra dstrct jscho bo'X rdf and tia+ted in198 0focse on$' gvernance, whlethe second,ta cororate-syle ml5anagement tea. Thme boardwas g iven effetie beginning in 1995, S dcetraie ome poers the ;right to0 imos sntion on pt;f ;oorl peror iin and pt th focs onimprvinglearirig The1988refom schols,incldingdisbndin theuchol lconci d andvaal creaed leced, aret-ld scoolcoucil wit th poer urin an disissng rinipal (i cojuncionwittth to hre ad fie th schol drectr. Te cunci wors wih concil). Oe ofitsfirs actons as t put109 ef th 55 the iretor o pepar an montora shooldevlopmnt ubli scoolsin hicgo o prbatin ebcaus ofpoo plan Tenre fr diectos wa repacedby fur-yar cn- aademc peformnce.The 995 efor als estbliseda tracts.Directorse were given increasedpowers to 0hire central body responls0ible forthereview and evaluation of teachers, iKncre&ased retion in allocaig the budget, a t r eac so with recommendations for increased;contl o c eased the By 195 thre wa thewide pra pecpint d-bdeayatnoyo ahsho,icuiggvn each cario.al mproemens intiate in 988 ere ot ocur- irecor te fredom to usource a ievaitf school ring rapidly enough in Chicago. As a result, the mo s 66 EMIPOWERING MUNICIPALITIES OR SCHOOLS' THI. DECENTRALIZATION OF EDUCATION BOX 4.2 Memphis: Decentralization Focused on Improving Learning The schools of Memphis, Tennessee, serve a largely poor school development plan; performance contracts with spe- and educationally disadvantaged population. Frustrated cific, quantifiable targets between the school director and with the persistently poor academic performance by stu- the central administration; extensive teacher development dents, in 1995 the city decided to grant limited autonomy activities at the school level; teamwork within the school; to individual schools with the objective of stimulating and constant monitoring of progress, including the use of school-level educational reforms. Each school formed an standardized examinations. advisory school council consisting of the director, teachers, The central Memphis education office continued to play parents, and community members. The principal func- a strong role in setting high standards (for example, today rions of each council were technical-diagnosing needs, all students in grades 3 through 8 must pass set exams in agreeing on reforms, and monitoring progress in student mathematics and science in order to be promoted. The edu- learning. But although the council was Legally advisory in cation office was also committed to (a) mandating minimum nature, its opinions were taken seriously. standards and core curriculums; (b) facilitating teacher devel- Each school in the Memphis district was required to opment by offering a broad menu of training options and adopt a school-based reform from a menu of eight differ- opportunities; (c) providing additional financing to cover the ent school-restructuring models. While the pedagogic ori- costs of implementing school development plans (with entation of the models differ, they share several character- larger amounts for schools serving the poor); and (d) estab- istics: increased school autonomy (especially, on pedagogic lishing monitoring and evaluation systems to provide con- matters); a common vision of school goals reflected in the stant feedback to individual schools on their performance. which affect learning. The difficulty in evaluating reforms is difficult for anyone outside the school to monitor the argues for caution in interpreting results. The lack of much school's performance and hold the school accountable. After rigorous evaluation of Latin American experiences has led all, the outputs of the school are several, and almost all are us to rely to some extent on well-supported evaluations of difficult to measure. Experience has shown it is especially decentralization efforts outside the region for our overall difficult to measure the value-added of the school in pro- conclusions. ducing scholastic achievement (Ladd 1996). Other charac- The fact that few evaluations exist of the impact of teristics impede reform. Teachers often work in isolation decentralization on learning outcomes has also led us to an and may shirk their responsibilities, with little risk of neg- alternative approach to infer impacts by looking at the ative consequences. Finally, strong labor unions and regula- extent to which characteristics of decentralization reforms tory protection (often embodied in teacher statutes in Latin are consistent with the characteristics associated with high- America) make it difficult to penalize poor-performing performing schools. The fact that two well-evaluated and teachers even when they can be identified. successful U.S. school reforms-in Chicago and Mem- To economists, these agency problems argue for a num- phis-have shared the decentralization characteristics that ber of solutions. First, intense efforts should be made to professional educators associate with public schools lends provide good information on the performance of schools credence to this approach. Interestingly, many of the rec- and teachers, taking into account the complexity of the ommendations made by educators for creating effective educational production process. This may require establish- schools are consistent with the prescriptions economists ing an independent agency to carry out external audits of might make. schools that go beyond merely identifying outputs, but also Designing decentralization reforms to improve learning provide diagnoses of problems and propose solutions as well. is complicated by the nature of education. For example, it Second, school directors should be given a large degree of 67 BEYOND THE CENTER: DECENTRALIZING THE STATE authority, presuming they have considerably better capacity dence for Chile is positive. Two other Latin American to monitor school and teacher behavior than do local polit- reforms-more limited in scope than Chile and Minas ical agencies, including school councils. Third, teaching Gerais-have been evaluated in terms of impact, with should be organized in a way that minimizes shirking and somewhat contradictory results. El Salvador's EDUCO pro- provides peer rewards and sanctions for performance. This gram has not yet demonstrated positive effects on learning, requires that teachers share experiences and work together while Nicaragua's charter school program has. Nicaragua's as much as possible. Fourth, given the high risk of shirking, reform granted substantial authority to school directors, teachers must themselves become the proponents of efforts which Brazilian research has found to be associated with to improve teaching, including deciding on their own train- learning gains. ing. Externally imposed (that is, top-down) solutions to In sum, there is growing evidence that at least some of educational problems are likely to fail in the absence of an the characteristics of education decentralization reforms effective campaign to enlist the support of teachers. that focus on school autonomy, as opposed to municipal or Of the Latin American reforms reviewed here, two- regional autonomy, contribute to higher-performing those in Chile and Minas Gerais, Brazil-include a large schools. Decentralization to subregional governments may number of the elements that arguably give rise to the char- also yield some educational benefits by allowing greater acteristics of effective schools. Neither reform has yet been innovation and greater flexibility to adapt to local condi- subjected to rigorous evaluation, although the available evi- tions, but they have not yet been demonstrated. 68 Annex TABLE 4.A. 1 Finance Schools cannot raise their own funds, though revenues Characteristics of "Decentralized" School System Argentina, do follow students within the system. Education Reforms 1976-1991 Pedagogy School directors, in conjunction with the council, can influence the non-core curriculum and have some power of pedagogical approach. Resource All power is at level of provincial government. Management Information Public test scores and other school-level evaluation measures are readily available, though the public mea- Personnel All power is at level of provincial government. sures do not assess value added. The evaluation is done Management independently of the school; in addition, the central district has created hoth an inspector general to investi- Governance All power is at level of provincial government. School gatricias matementh an actountality councils have no deciionmaking authority.gate financial management and an accountahitity councils have no decisionmaking authority, council to review school performance and determine Choice No significant choice, which schools are failing. Incentives Incentives within the system appear to be strong. Direc- Finance Provincial government is the financial source of all rs lost tenure and can nnw he fired hy either the regular operations of the schools. Central government council or the central district for failing to improve is source of finance for compensatory education outcomes. Teachers are largely still protected. programs. Pedagogy Nor enough information Information Test scores are publicly available at the level of the TABLE 4.A.3 schools. Characteristics of "Decentralized" School System Chile, Educa- Incentives Very weak incentives for teacher or d&rector tion Reforms 1981 performance. COMMENT/SUMMARY JUDGMENT Resource Most resource decisions are made at municipal manage- ment level, including development of a municipal education plan. School directors have newly granted authority to manage funds. Most schools have developed TABLE 4.A.2 school improvement plans that have been financed by Characteristics of "Decentralized" School System Chicago, IL, the central government. Education Reforms 1988-1999 Personnel Municipalities have the authority to recruit and hire COMMENT/SUMMARY JUDGMENT Management teachers. Teacher pay is set at the municipal level, but it Resource Schools must develop school improvement plans. is constrained hy the national minimum pay scale. Ar Management They have gained greater control over allocation of the the school level teachers can define their own training non-personnel budget and now receive most of federal needs. Title I funds as discretionary revenue. Governance No school councils. Weak parental participation in Personnel School councils can hire and fire the school municipal schools. Management director, and directors have increased authority over new Choice High degree of choice of school for parents. hires. The process for firing incompetent teachers already at schools has been streamlined somewhat. Directors can Finance 90 percent of revenues come from central government, assign teaching responsibilities. which allocates according to a set formula. Revenues follow students. Governance Councils have a voting majority of parents and can select the director. Pedagogy Strong central government role. Choice Little or no traditional choice plans. Parents can select high schools but schools cannot generally have admis- sions requirements. (Table continues on next page) 69 BEYOND THE CENTER: DECENTRALIZING THE STATE (Table 4.A3 conrinaed) TABLE 4.A.5 Characteristics of "Decentralized" School System Memphis, TN, Information Published test score data are available at the school level. Very strong financial auditing tradition but no perfor- Education Reforms 1995-1999 mance audit is conducted. COMMENT/SC MMARYJUDGMENT Incentives There is competition for students and competitive grants Resource Schools must develop a school improvement plan that for school improvement and teacher training which Management includes the strategy for allocating resources to support requires teamZwork. Salary bonuses to teachers in highest the chosen pedagogical restructuring model. This model quartile of performing schools. comes complete with textbooks. Non-personnel funds were essentially decentralized to the directors through school-based management reforms that took place before the pedagogically driven reform in 1995. TABLE 4.A.4 Personnel The school councils cannot hire and fire the principal, Characteristics of "Decentralized" School System El Salvador, Management who in turn has relatively limited power over teaching Community-Managed Schools (EDUCO) personnel: They cannot really fire teachers, and they have some limited extra control over hiring new teachers. COMMENT/SUMMARY JUDGMENT They have some limited power over assigning teaching Resource Community Education Associations (ACEs) devise an responsibilities but must remain within union guidelines. Management annual plan for the school and are responsible for admin- Pay scales are set centrally Professional development is istering funds according to its assessment of the educa- design-specific depending o1 cIse pedagogical miiodel tional needs of the school. They are also in charge of the chosen and takes place at both the school and at sites maintenance and equipment of schools. The selection of determined by the organizations that provide a particu- textbooks falls under the responsibility of the Ministry of lar design. Education. Governance School councils are advisory in nature and have a Personnel ACEs have the authority to hire and fire teachers majority of school staff They do not select the director Management according to the Ministry's criteria for selecting teachers, or any other school staff and supervise their attendance and performance. The Choice Parents can choose from among public schools, although Ministry is in charge of teacher training and setting a this choice is limited by availability, and school neighbor- pay scale. hood residents are given priorities. Only magnet schools Governance Creation of ACEs. These are bodies of elected members can have admission requirements. drawn from the community and are usually parents of Finance Schools cannot raise their own revenues, but funds do the children attending EDUCO schools. ACEs do nor follow students. select the school director. In most cases the principal Peda works closely vwith the councils; hovwever, some view the Pdgogy The entire pedagogical approach is determined by the model selected by the school. Schools must choose one of ACE as a threat to their authority. the models approved by the central school district, but Choice EDUCO schools are located in very remote areas and the range of choices is very wide. tend to be the only means of delivering educational Information The Tennessee Value-Added Assessment System (TVAAS) services there. As a result, parents do not have a choice. one of the most well-respected evaluatn systems Finance ACEs enter into a one-year renewable contract with the nationally. It specifically provides value-added achievement Ministry of Education, through which they receive results, which are available to the public on a school- earmarked funds on a monthly basis to cover teacher by-school basis. salaries and operating costs, including school materials. Incentives Directors are evaluated heavily based on the TV7AAS, Another source of funds for the school is the "bonus," and poor performers can be removed by the central which is a small discretionary amounr transferred to the district. Teachers, in contrast, have more substantial job school. Both of these funds are based on a formula that security. considers size of school and number of teachers. ACEs i,iay raise additional finids by negotiating with other government agencies and international donors and by TABLE 4.A.6 mobilizing local support. Characteristics of "Decentralized" School System Mexico, Edu- Pedagogy The EDUCO Coordinating Unit within the Ministry of cation Reforms 1992-1998 Education is responsible for aspects of the program designed to increase classroom effectiveness, including COMMENT/SUMMARY JUDGMENT curriculum development. Resource Because schools have little power over resources, they Management have little ability to influence a school improvement Information The Ministry conducts audits of the ACEs wvhen regional plan. The Ministry of Education pays only for its text- supervisors report the existence of serious problems. books. Incentives Teacher job security is affected by student performance. Personnel Little or no school-level power over personnel. As a result, the main reasons teachers leave EDUCO and Management go to the Ministry's regular system are job stability Governance Officially councils exist but they do not function and and related benefits. have few powers. 70 EMPOWERING MUNICIPALITIES OR SCHOOLS' THE DECENTRALIZATION OF EDUCATION Choice No significant choice. Information The SEE-MG has established a mechanism of external assessment of the school unit called the Assessment Pro- Finance Schools cannot raise significant revenues and it is not gram of the Public Schools of the State of Minas Gerais. clear that funds followv students. It is a periodic testing designed to measure students' Pedagogy Schools have little choice over curriculum and pedagogy. achievement on basic knowledge and skills, and the scores are made public. Information School-level test scores available for most schools. There are no measures of value added. Incentives The election process has given the community the power to replace principals who are not performing according Incentives Very low level of incentives for either teachers or directors. to the community's interests or expectations. TABLE 4.A.7 TABLE 4.A.8 Characteristics of "Decentralized" School System Minas Gerais, Characteristics of "Decentralized" School System Nicaragua, Brazil Education Reforms 1993-1998 COMMENT/SUMMARY JUDGMENT COMMENT/SUMMARY JUDGMENT Resource Schools are in charge of developing a school improve- Management ment plan, an aspect of wvhich is to prepare a pedagogi- Resource School-level resource management increased significantly cal project defining the aspects to be improved in teach- Management in some autonomous schools. School directors receive ing-learning process. Schools also have the auhority tco monthly capitation grants to pay teachers and all other select their own textbooks and allocate non-personnl tmajor expenses, including utilities. Schools also charge sludget. fees (or, rather, voluntary contributions) and any rev- enues generated are allocated by the school council. Personnel The State Education Secretariat/Minas Gerais (SEE-MG) Schools can select their own textbooks, but the central Management is responsible for establishing the academic requirements government pays only for the ones that it chooses. and qualifications for teachers and promoting and pro- vidqaing them swit tra. Schoos an reing car pof Personnel School councils can officially hire and fire the school viding, determi whih teainincherls will'reicive t rain Management director and, with greater difficulty, teachers. However, ing, determining which teachers will receive training, in practice most councils have nor challenged centrally evaluating teacher performance, handing down discipli- appointed directors. There are cases of councils firing nary actions, and firing teachers. Salary levels and bene- directors, There are over of hiring fits are set at the state level. The schools can assign directors, but their direct control over the hiring process teaching responsibilities. Principals may not be dismissed is more suspect. Still, the incentive for the director to without an SEE-MG investigation. Professional develop- please the council members seems credible. If fee rev- mnent is demand driven because the school board (com- enue is substantial, councils tend to use it to augment posed in part of teachers and other school staff) has the reacher salaries. responsibility of developing training plans for teachers. Governance Councils have a voting majority of parents and include Governaiice School councils, called colegiado escolar, are elected by the director, teacherse and vtarous obectilve criteria, the community from teachers, other school staff, parents, but inisome sctey are ocrical elected ile and students over age 16. Candidates for principals are but in some schools they are democratically elected while first required to take a written exam. Those with the in others they are selected by the director or the Min- highest scores are asked to present a work program to the community, which then elects the principal by a Choice No significant choice. secret vote. Finance Schools can raise revenues through parental contribu- Choice No significant choice. tions and the annual school budget is determined by number of students enrolled. Finance Bulk of revenue transferred from SEE-MG, however schools may raise additional funds through public or pri- Pedagogy Directors, i conjunction with the councils, have somne vate donations. School revenues are formula-driven. Dis- discretionary power over the non-core curriculum, hot it cretionary funds (those spent according to priorities is limited and rarely exercised. Primary pedagogical established by schools) are determined using a formula approach, the methodologia actova, is developed and dis- that takes into account the number of students enrolled, seminated by the central ministry. the socioeconomic level of the clientele attending the Information There are no public, school-level measures of value school, and the location of the school. Earmarked funds added or other test score results. The central ministry is (materials and services approved by the SEE/MG) are developing a national achievement evaluation system in based on a per capita amount multiplied by the number conjunction with the World Bank. of students, plus a certain percentage according to type Incentives There are incentives for directors and teachers to raise of school (full-time, special education, and so forth). revenue, and for teacher attendance in those cases where Pedagogy Taking into account basic educational patterns and legal school revenues (which in turn depend to some extent on requirements regarding the annual number of school parent satisfaction) are sufficient to provide bonus pay. days, each school has the authority to design its own cal- Teachers can also be fired by the director, perhaps in endar, curriculum, and pedagogical approach, including consultation with the school council. Overall, the incen- methods for evaluation and organization of students in tives for all staff to perform appear higher in classes. autonomous schools. 71 BEYOND THE CENTER: DECENTRALIZING THE STATE IABIL z, A.' Summary of Characteristics of Decentralized Schools GOVER;N ANC('E School councils elected.v iNj j Councils select school directo)r. vA iA CHOICE Parents selecr school for c:hild. ln. a. ii School sets admission requirements. n.a. n.a. FINANCE Raise ownn revenues. i i Reventies follow students. n.a. v n. a. - -, S n. a. PEDAGOGY Set th,e non-core culrricullum-. - S n.4a. Choose pediagogical approach. ri. i na. RESOURCE MANA(EMENT Develop school imprl venNenN plan. ji Allocace non-personnel budiget.v lv Selection of textbo dks. rt N N N N ' 1 PERSONNEL MANAGEMENT Hire and fire school directhr. N n N N N Hire and fire teachers. r e na n Se-. or augment teacher pay scale. Ni Assign teaching responsibilities. n.a Demnd-driven professit nal development. - v n. a. n.a. INFOR2vMATION Independent public audit or inspection.v Public test scores no-t n.ca. i n.a. n Public measures of value added. N' N INCENTS VES Director job security ried to implementation of school improvement plan. N N N N Teachier job securitv affected by student performance.rsonnl h RESULTS Increased tesc scores. v n.Na. N' N Itproved teacher absenteeism Improved stug ent attendeace. \i Higher parental satisfaction. Public~~~~~~ tes scrs N a Nn. n.a =nor applicable 72 EMPOWERING MVNICJPALITIES OR SCHOOLS; THE DECENTRALIZATION OIE EDiCATIOON BOX4.A.1 What Has Been Learmed from Effective Reform Strategies? 1. Educational reform is a local process. The school, run; only institution building based on sustained commit- not the ministry or the district administration, is the cen- ment works. ter of change. Schools determine the degree of success; 6. Focus on classroom practice. The clue is to focus they can block implementation, enfeeble it, or bring it to on the dynamics of the classroom and the individual effective life. For schools to improve the quality of their school, since this dynamic to a large extent determines programs effectively, they need to play an active and cre- implementation success. It is essential that the supporting ative role. materials are of good quality, whether nationally devel- 2. Central support is vital. The issue for the central oped and locally adapted, or locally built from the start. ministry is learning to support local schools in their efforts, 7. See teachers as learners. Good materials and facil- in other words, how to make demands on support, encour- ities are a necessary but insufficient condition. Teacher age, empower, enable, and build a strong local school. mastery is crucial for impact on students, and that can best More responsibilities to the individual school presuppose a be developed through a systematic local learning process strong support structure from the system at large, one that that includes in-service training, supervision, and coaching must be built around the real needs of schools in develop- in a collegial atmosphere. ment. For the central level it implies that a system of 8. Commitment is essential at all levels. It is crucial reform and a division of labor is needed to effectively sup- at the central level for sustained effort and the mainte- port the local level. nance of needed support structures. It is also essential at 3. Effective system linkages are essential. The strat- the district and school level; however, it cannot be trans- egy in complex systems is to identify effective linkages, mitted directly to schools. Commitment at the school lcvcl nonbureaucratic in nature, between the national, district, results from empowered successful action, personal mas- and local levels. For communication within the system to tery that starts with good assistance and develops from be effective, local empowerment is needed, usually as a practice. In effect, local empowerment builds emotional as consequence of more decentralization. A clear administra- welfl as administrative and problem-solving capacity. tive role that combines pressure and support and secures 9, Both local and central initiatives work. An inno- the delivery of needed resources is also required. vative idea that starts locally (Colombia), nationally 4. The reform process is a learning process. The (Ethiopia), or with external donors (Bangladesh) can suc- process is evolutionary and developmental in nature. Ir ceed, if programs meet the criteria of national commit- cannot be blueprinted ahead of time. The key to success is ment, local capacity building -and linkage, in a configura- to get good data from all parts of the system on a contin- tion that makes sense for the particular country. uous basis, studied and worked on at the school district 10. Parent and community participation contribute level, and subsequently at the central level. This implies a to success. Parent and community participation lead to competent supervision and monitoring system. comnmitment and contribute to outcomes, and are essential 5. Think systemic and big. A vision of reform that for the development and maintenance of primary schools affects school life substantially will have more effect than a in rural areas. Effective participation includes a real role for cautious, incremental approach. Any ma;or reforms in parents in school decisionmaking. complex systems need to build structures and capabilities at all levels. Ad hoc solutions will not work in the long Fr.: Dalin and others 1994' 73 CHAPTER 5 Empowering Mayors, Hospital Directors, or Patients? The Decentralization of Health Care H ' T ealth sector reforms are becoming common throughout Latin America. This is not sur- prising as most health systems suffer from rising costs, low coverage, inefficient operation, and patient dissatisfaction. Reform has occurred in a variety of ways. Some countries have attempted to decentralize the provision of public health services, transferring responsibil- ity for primary health care and lower tiers of acute care (hospitalization) to subnational units of government. Others have introduced competition into health services, by forcing government owned hospitals to compete with private hospitals for government health care funds. This chapter reviews how decentralization reforms in the health sector have been designed and imple- mented. It is based on the experience of six countries: Argentina, Bolivia, Brazil, Colombia, Chile, and Mex- ico. It begins, however, with a clarification of the nature of health care itself. Three Kinds of Health Services ties. Central governments must therefore be concerned that Unlike education programs, health care programs provide finances are adequate and a regulatory structure is in place some goods and services that are almost "pure" public to ensure that local government units or private firms rec- goods. That is, they are both "nonrival" in consumption ognize these externalities across subnational political borders. (consumption by one person does not reduce the amount From the consumer-citizen standpoint, individuals may available for others) and "nonexclusive" (benefits cannot be benefit from the pure public health goods and services restricted to only those people willing to pay for them). without even knowing that they exist (for example, vector Examples are provided in the first row of Table 5.1. These control), although they are generally well aware when the goods and services (or activities) require central government services are not adequately supplied. Management of the financing if they are to have any chance of being provided activity may not include interactions between suppliers and in quantities that would be efficient from the standpoint of beneficiaries, so the direct client for services may be a gov- the economy as a whole. Some of these activities have ernment functionary rather than the end user, creating largely localized benefits, that is, benefits that, while not principal-agent issues but also leaving open to cost and confined to individuals, are confined to local jurisdictions. quality considerations whether the relevant governmental Vector control and disease surveillance are examples. While unit prefers to own-mrianage the activity or contract it out. there may be a case for some local financing of such activi- A second set of services (Table 5.1, row 2) in health is ties, they also have important cross-jurisdictional externali- rival in consumption but nonexclusive in benefits. One 75 BEYOND THE CENTER: DECENTRALIZING THE STATE TABLE 5.1 Suggested Financing and Provision for DifFerent Types of Health Services CHARACTERISTICS OF GOODS AND SERVICES EXAMIPLES REGULATION FINANCING PRODUCTION PUBLIC OR PRIVATE? CENTRAL OR LOCAL? 1 Pure public goods Hcalth information *Central *Public: otherwise 'Decentralization for *Public in health (nonrival, 'General health underfinanced localized public goods, 'Contract out (public nonexclusive) education as long as national bureaucracy is the client) *Disease surveillance standards are maintained *Environmental health *Vector control *Regulation (market failures in information and insurance) 2. Household health Immunizarions -Cenrral -Public and some *Decentralization for *Privare preferred (reduces inputs with strong 'Sanitation private goods with strong local principal/agent problem externalities (rival, *Safe water characteristics, as long of public provision) nonexclusive) *Prevention of as national standards communicable disease are complied with 3a. Household health *Acute care -Central Private 'Central Private inputs that are * Public (in case of fundamentally private insurance market goods (rival, exclusive) failures) 3b. "Equity goods" in *Acute care *Central *Public (to finance *Central *Private health indigent) example is immunization. A large part of the population them if they are to be effective as medical interventions for must use this service if it is to be effective as a medical inter- the population. It is important that the service function in vention for the population as a whole. Immunizations may a manner that pays attention to the client. Thus the man- provide widespread immunity, however, at a level of cover- agement of the service at the point of delivery is important. age substantially lower than 100 percent. Thus they pro- Because these two groups of services involve inherent vide benefits to people who choose not to be immunized. market failures, they require some form of government Although, there is a willingness to pay for the service, there intervention in determining how much, and what, will be is also an incentive to take advantage of the free ride; hence provided. The case for government intervention in the third a strong rationale exists for, at least, partial government group of services is less clear-cut. These services-consisting financing or compulsion to force private parties to purchase largely of acute (or curative, as opposed to preventive) adequate quantities of the service. In practice, these services forms of care-are essentially private goods. The govern- tend to receive heavy if not exclusive public financing, ment may find itself involved in the financing of these as a although some jurisdictions do handle them primarily substitute for private insurance systems that fail to develop through regulations rather than through financing (for (Table 5.1, row 3a). Governments may also be involved to example, requiring children to have immunizations before ensure that those who cannot afford insurance have at least they enroll in school, and requiring coverage in basic insur- a minimum level of access to acute care (Table 5.1, row 3b). ance benefit packages). From the consumer's standpoint, these services require Many Modes of Service Delivery and Financing direct interaction between medical personnel and patients No government is required to become directly involved in (or for water and sanitation, between the seller and the con- the delivery of the services. Even in the realm of public sumer). A large share of the population actually must use goods, it may choose to contract for such activities as disease 76 EMPOWERING MAYORS, HOSPITAL DIRECTORS, OR PATIENTS' THE DECENTRALIZATION OF HEALTH CARE surveillance. Its choice should probably depend on the rel- allocated as line items in the budget: personnel, equipment ative quality and efficiency of contracting for the service or and supplies, and so forth. This gives health unit adminis- providing it directly. Table 5.1 demonstrates, however, that trators little incentive to reduce costs. Under an output- as benefits of health services become increasingly confined financing system, health units are paid according to the vol- to individual consumers, the case for government participa- ume and type of services they provide. This not only tion in production diminishes. While the public sector's role encourages efficiency, but can also boost quality because in regulating and financing acute care remains strong, its health units are forced to compete for clients. comparative advantage in provision is less so. The second element is management autonomy. Without ln the history of health care provision, this distinction autonomy, hospital administrators cannot respond to the between public and private goods, on the one hand, and incentives inherent in output-based financing. Manage- impersonal versus personal services, on the other hand, has ment autonomy can be achieved in different ways. In some been confused . As a result, governments and social security countries (for example, Chile), hospitals and clinics remain institutes have found themselves financing and delivering in the public sector but are organized as corporations to all kinds of services even though many could be delivered as ensure their management autonomy. In other countries (for well or better in a competitive market. Central govern- example, Brazil), the majority of health care providers ments, having entered the business of medicine, then con- belong to the private sector, and the government's role in sider decentralization to be a matter of decentralizing the health is largely limited to finance and regulation. management of facilities and personnel, when in fact lower levels of government may not have any advantage in skills, What has been Decentralized in Latin America incentives, motives, or negotiating power to improve the and How management of such services. Decentralizing health care has meant different things in dif- This framework is perhaps too simple to capture the true ferent places. This section summarizes the experience of six role of the public sector in health or to fully understand the countries. decentralization efforts that have taken place in Latin America (see, for example, Musgrove 1999; and Preker, Argentina: From Central Monopoly to Provincial Harding, and Girishankar 1999). However, it emphasizes Monopoly two simple issues that should be addressed in decentraliza- Judging by the proportion of the health budget that is exe- tion policies in health. cuted at the subnational level and by the decision powers First, should responsibilities and financing of the high- that have been transferred to the provinces, decentraliza- externality functions in health be distributed across politi- tion of health services in Argentina both in the territorial- cal entities, recognizing that many of these goods would political and the economic sense is quite advanced. Only benefit from local customization while also meeting mini- about 14 percent of health spending in the public sector is mum national standards for results (for example, manage- done at the national level; 70 percent is done at the provin- ment of water-, air- and vector-borne diseases)? Second, cial level, and another 16 percent is done at the municipal how can services be improved in historically large, expen- level. Less than 1 percent of inpatient facilities are adminis- sive, and poorly functioning government-owned-and-oper- tered by the national health authorities, some 70 percent ated delivery systems? While movement along the political are administered at the provincial level, and 20 percent are decentralization axis may affect service delivery to some administered at the municipal level (Gonzalez-Prieto and degree, this second issue is fundamentally an incentive Alvarez 1999). problem that can be addressed through a variety of actions. The public sector provides health coverage to about 46 One of the most promising involves splitting the financing percent of the population and accounts for about 23 per- function from the provision function. cent of total health spending in the country. The social This reform-termed the purchaser-provider split- security system (based on the so-called "Obras Sociales") involves two elements. The first is a shift in the mode of provides coverage to 47 percent of the population and financing, from inputs to outputs. Under a traditional accounts for 3 5 percent of health spending. The private sec- input-based system, government health expenditures are tor covers 7 percent of the population and accounts for 42 77 BEYOND THE CENTER: DECENTRALIZING THE STATE percent of health spending (insurance and out-of-pocket). for the uninsured population (currently about 42 million The public sector owns about 37 percent of health facilities people) through its own extensive network of health facili- and 54 percent of hospital beds, while the private sector ties. In addition, the Social Security Institute (IMSS) pro- owns 61 percent of health facilities and 43 percent of hos- vides comprehensive health insurance to some 41.5 million pital beds (Gonzalez-Prieto and Alvarez 1999). people through its own provider network. Sixty-seven per- The first wave of health sector decentralization in cent of IMSS health insurance revenues come from payroll Argentina (1978) was aimed at the provincial-rather than taxes and 33 percent from general taxation. the municipal-level. The primary motive for decentraliza- Another insurance scheme, jointly funded by the SSA tion was the alleviation of the fiscal burden at the central and the IMSS, brings coverage to an additional 11 million level, rather than a quest for efficiency or equity. Responsi- people, mostly in rural and indigenous communities. Paral- bility for running health facilities and budgets was trans- lel social security schemes exist, such as those for public ferred to the provinces. The resources transferred from fed- employees and for the national oil company. The private eral to the provincial level were not earmarked for the sector, on both the provision and financing sides, is small health sector; provinces could therefore choose how much but growing. Currently some two million Mexicans have they wanted to allocate to the health sector from their own private health coverage. budgets and from the federal transfers. The early 1990s Decentralization efforts, initiated in 1983, constituted brought a second wave of decentralization with the trans- an effort to share political power with the state govern- fer of the last federal hospitals to the Municipality of ments, reduce the fiscal burden at the central level, ratio- Buenos Aires, and the transfer of some provincial health nalize the supply structure, and improve management. The responsibilities (especially primary health care) to the process was a gradual one involving only 14 of the 32 states municipalities. and limited in scope. It did not involve either health juris- The second wave also introduced the concept of the dictions or health facilities. Resourcc allocation autonomy autonomous public hospital. The scope of autonomy of at the state level was limited to revenues obtained locally these hospitals was to include the ability to bill the social Budget execution remained highly centralized: The share of security system and private health insurers for services pro- their budget executed at the stare level was the same for vided to their beneficiaries and to retain a part of their earn- decentralized states (24 percent in 1995) and for those ings. Adoption of the autonomous hospital model has been states that had not been involved in the process (21 percent uneven, however. Two provinces (Cordoba and Neuqu6n) in 1995) (Centro Estudios para America Latina tCEPAL have rejected the concept entirely, fearing that it would 19981). The period was characterized by rival efforts on the lead to duplication of facilities and equipment, and to dis- part of the IMSS to deconcentrate (more for self-preserva- crimination against the uninsured population (as hospitals tion than to promote devolution). become more and more dependent on reimbursements With the arrival of the Zedillo administration in 1994, from insurers). Other provinces have adopted the concept decentralization was back on the agenda. Although the but have implemented it slowly, fearing that it would government was primarily motivated by pressure to share increase the public deficit. They also feared strong opposi- political power with the state governments, the Zedillo tion from labor unions and reduced political power at the reforms also aimed to increase coverage and improve the qual- provincial level. With one exception-the province of ity of care for the uninsured population as well as increase the Salta-no province has given its autonomous hospitals con- efficiency of public administration. The second phase of trol over personnel management. decentralization was initiated in 1996. It differed from the first phase in that it involved all of the states and made the Mexico: From Deconcentration to Decentralization sharing of roles and responsibilities between levels more The public health sector in Mexico consists of several enti- explicit, but it resembled the first phase in that decentraliza- ties. Historically, the Ministry of Health (SSA) has been tion did not reach the level of health jurisdictions or facilities. responsible for the definition of health sector policies and The second phase of decentralization brought about the the regulation and supervision of, and strategic planning creation of "decentralized public organisms" (OPDs), semi- for, the health system, and for the provision of health care autonomous state agencies whose governing board includes 78 EMPOWERING MAYORS, HOSPITAL DIRECTORS, OR PATIENTS' THE DECENTRALIZATION OF HEALTH CARE the state governor, a representative of the federal Ministry commitments by the IMSS to sign agreements with spe- of Health, a trade union representative, and the state healrh cialty hospirals and medical zones. minister. Clarity has been achieved in a number of areas. Resources from the federal level to the states are allocated Bolivia: Municipalizing the Physical but not the according to well-established criteria; the transfcr of human Human Assets resources from the federal level to the states has been nego- The public sector provides health coverage directly to tiated with the national union; infrastructure, goods, and about 30 percent of Bolivians, and the social security sys- equipment have been transferred to the states; and munic- tem provides coverage to another 14 percent. The private ipalities have been given limited responsibilities in the areas sector provides coverage to some 30 percent of the popula- of planning and infrastructure. Monitoring and conflict res- tion. The remaining 26 percent either have no access to olution are done by the National Health Council. health services or use traditional medicine (Ruiz Mier and With the reform, the states became accountable for all Guissani 1997). health care services for the uninsured population, and Health care decentralization in Bolivia has been used as obtained control over the execution of its health budget. an instrument of democratization. After a number of failed The decentralized budget increased from 4.8 million pesos efforts in the 1980s, the process was precipitated by the so- in 1995 to 16.4 million pesos in 1999. The functions of the called "popular participation law" of 1994, which provided federal Ministry of Health were concomitantly redesigned for the recognition of 311 municipalities. This law trans- and its normative and planning role strengthened. Spend- ferred assets of health centers and medium- and high-com- ing at the federal ministry level decreased from 12.2 million plexity hospitals, including national referral hospitals, to pesos in 1995 to 9.5 million in 1999. the municipalities. The responsibilities transfered included The IMSS has been involved in parallel deconcentration infrastructure maintenance, equipment, and input supply. efforts, with the creation of seven regional directorates in The process was further advanced with the administrative 1995, and 139 medical zones in 1997 (each providing decentralization law of 1996 that transferred some person- health care to a population of between 100,000 and nel administration responsibility to the departmental level 200,000 people). These medical zones are expected to (departments are administrative subdivisions of the central evolve into budget-holding "medical areas of autonomous government). The central ministry, however, continued to management." So far deconcentration has fallen short of the be responsible for personnel recruitment and firing, and for planned full management autonomy in such areas as per- negotiating salary levels every year with health sector sonnel, procurement, equipment, infrastructure, and main- unions. Decisions on the number of staff, both administra- tenance. Plans for the next few years, however, hold tive and medical, allocated for public health centers is promises of a purchaser-provider split involving develop- determined centrally, and municipal governments do not ment of the purchasing function within the IMSS and intro- participate in the decisionmaking process. duction of reimbursement to providers through risk- Thus one of the particularities of the Bolivian experience adjusted capitation and Diagnostic Related Group (DRG) of decentralization has been the sharing of roles and respon- systems. sibilities between the municipal and the departmental lev- The political decentralization process within the SSA is els. Although many countries have delegated primary making slow progress, and autonomy and capacity at sub- health care to municipalities, and more complex health care national levels are gradually being strengthened. In con- to departments, the demarcation line in Bolivia has been trast, there scems to be little or no progress in separating the control of particular types of inputs. The municipalities financing of purchasing from the provision function. The are responsible for the maintenance and equipment of the decentralization process in the IMSS has been equally slow, health infrastructure, and the departments are responsible but there are more explicit plans to operate a purchaser- for personnel administration. In this model, municipalities provider split. The initial steps toward this split are evident and departments each have control over some of the factors in the recent agreements signed with the Ministry of of production, and close vertical coordination is needed to Finance and the Ministry of Auditing and Administrative ensure effective service provision. In addition, municipali- Development, which are aimed at financial autonomy and ties and departments have to coordinate upward with the 79 BEYOND THE CENTER: DECENTRALIZING THE STATE national Ministry of Health, which plays a regulatory role demand and provision. Health providers, for their part, are and is responsible for the administration of essential drugs, not given the instruments to behave in a more profiL-ori- and downward with the local communities to ensure that ented fashion; indeed, the last few years have seen a reduc- local concerns are considered. tion in the degree of autonomy of the health facilities to Local health directorates have been created in an retain payments for services provided, and to decide attempt to achieve coordination among all the participants resource allocation and medical inputs. in the health sector. These are formed by representatives of the municipal government, the departmental health direc- Brazil: Public Financing with Private Production torate, and the local community. The mandate of these Brazil's decentralization process in the health sector was ini- directorates includes (a) preparing and proposing to the tiated in conjunction with efforts to integrate the social municipal government annual operating budgets and security institute, which provided health care to urban for- investment plans for the health sector, (b) negotiating for mal sector employees and their dependents, with the Min- health personnel, and (c) proposing and negotiating health istry of Health, which provided health care to rural unin- service provision agreements with different participants. To sured workers and indigents. The integration was achieved date, these directorates have fallen short of these goals. In in 1988 with the establishment of the Unified Health Sys- fact, since all three levels of government have a stake in ser- rem (Sistema Unice de Saude [SUSD), which provides near- vice provision, accountability for quality and coverage has universal coverage. SUS operates as a subsidized insurance become increasingly diluted. scheme, in which the federal SUS administration reim- Decentralization efforts have increased spending at local burses providers for services rendered. The vast majority of levels and have allowed municipalities to invest more in the such reimbursements go to private hospitals and clinics. maintenance and operation of the health facilities. Only State and municipal facilities of course are also eligible. about 12 percent of health spending is done at the national While SUS is generally considered a success, it has suf- level; some 62 percent is done at the departmental level; fered from financial problems. The reimbursements to and the remaining 26 percent at the municipal level. The public and private providers have been substantially major source of funding for the health sector remains trans- lower than the costs of services (especially for preventive fers from the central level. This transfer-based financing care). Private participation in the SUS is therefore system limits the actual degree of autonomy and control decreasing, and private health care insurance is becoming over inputs at subnational levels. The new model has also more widespread. actually diminished financial autonomy at the facility level: Publicly owned health care facilities are generally oper- Facilities have lost their autonomy to municipalities that ated by municipalities. The process of enabling municipali- now manage the user fees collected by the facilities. Finan- ties to manage their own health systems has been very cial autonomy at the facility level has also been reduced gradual and has involved a formal qualification process. By with the implementation of Maternal and Child Insurance December 1996, 137 municipalities, accounting for 16 per- (1998). Public health providers, which are required to pro- cent of the Brazilian population, administered their own vide free care to pregnant mothers and children under 5 health systems. These mostly urban municipalities admin- years of age for diarrhea and respiratory diseases, are typi- istered some 20 percent of hospital expenditures of the cally inadequately compensated by the municipalities for SUS. Approximately 2,300 more (or 42 percent of the total) these services. municipalities had gained incipient autonomy, which Although there has been a definite transfer of responsi- allowed them to participate in planning activities and bilities to the lower administrative levels in the health sec- licensing private providers in their territories. Their ambu- tor, little progress has been made in separating the financ- latory and hospital care budgets were, however, still pre- ing from the purchasing functions. Apart from those who pared and approved by the federal government. For their require Maternal and Child Health Insurance and Old Age part, states were responsible for reviewing policy imple- Medical Insurance, health purchasers are not clearly identi- mentation, monitoring and evaluating systems, and provid- fied. The type and scope of services to be provided are not ing technical and financial assistance to municipalities in clearly determined, leading to a disconnect between their jurisdictions. 80 EMPOWERING MAYORS. HOSPITAL DIRECTORS, OR PATIENTS? THE DECENTRALIZATION OF HEALTH CARE Because SUS reimbursements fall short of the cost of municipalities, the simplicity of the outputs to be delivered, health services, municipalities are often forced to subsidize the clarity and simplicity of payment mechanisms, and the their hospitals from general municipal revenues. improvements in municipal administrative capacity. Notwithstanding the numerous shortcomings of the sys- Decentralization was not the government's only reform tem, Brazil's health system has definite virtues: The pur- in the health sector. In pursuing its market-oriented chasing side is integrated; the purchasing and provision approach, the government also encouraged the use of pri- functions are separated; the provision of private health vate health insurance. Private health funds were allowed to goods is largely done by the private sector (private operate and collect compulsory payroll contributions to providers account for over 70 percent of publicly funded provide private health care. As a result, the share of people hospital admissions); and the financing of public goods, covered by private insurance plans increased substantially goods with externalities, and "equity goods" (basic health (to about 27 percent of the population in 1997 from near measures, nutrition, epidemiological and sanitary surveil- zero in the early 1980s) and the proportion of hospital beds lance, and so forth) is central, although their delivery tends in the public sector decreased from 90 to 75 percent.2 to be through federal agencies rather than municipalities. Colombia: Conflicts Between Decentralization and Chile: Municipalizing Production, Centralizing Privatization Financing Over the last 30 years, health sector reform in Colombia has The health care decentralization process in Chile initiated in evolved from focusing on deconcentration (1960s) to the the early 1980s was part of a wider reform process that decentralization-devolution and privarization (1990s) of aimed at introducing market mechanisms in the health sec- service provision. Since none of these reforms was ever fully tor. Primary health care facilities and personnel were trans- implemented, the current situation inherits a little from ferred to the municipalities. This had two important effects. each of the earlier phases. Devolution started in 1990 when First, the transfer of personnel to the municipalities led to the Ministry of Health mandated the transfer, over a five- their loss of civil service status. Henceforth, primary health year period, of primary health care facilities, personnel, and care workers would negotiate their work conditions directly budgets to municipalities, and of secondary and tertiary with the municipalities. In principle, this gave local gov- hospital care to provincial governments (departamentos). ernments considerably more autonomy in setting wages The transfer ran into implementation difficulties (burden- and benefits (although in recent years there has been some certification requirements and insufficient resources increasing pressure from the labor unions to recentralize the to pay for the cost of decentralization) and only a handful of personnel hiring and management functions.) municipalities assumed their responsibilities. Second, it led to a change in the mechanism for financ- Progress was made in the decentralization process with ing public health providers, from one based on inputs (line Law 60 of 1993 (Ley 60). Transfers to departments for item budget allocations) to one based on outputs, or the health and education increased significantly (from 15 per- quantity and quality of services provided. This was cent of government revenues in 1993 to 24.5 percent from expected to encourage municipalities to expand coverage 1996 on). These transfers are administered autonomously and make more efficient use of staff and other rcsources.' in those departments that have been certified. Transfers to While municipalities in Chile were given relatively more municipalities for health, education, and other infrastruc- control over inputs than in other Latin American countries, ture expenditures increased from 15 percent of the central their autonomy is constrained by the fact that to a large extent government's revenues in 1994 to 17 percent in 1997, and the Ministry of Health determines their primary health care are expected to increase to 22 percent by 2001. The main spending. The quality of health care they provide is there- problem with these transfers is that they bear little relation fore heavily dependent on the level of government financing. to local health needs or to the cost of services. The municipalization of primary health care delivery in Chile In the midst of implementing decentralization reform, a has nonetheless been considered a success that owes much more radical reform was enacted at the end of 1993 (Ley to the continued commitment by the Ministry of Health to 100). The reform strategy included separating financing primary health care, the clear definition of responsibilities of from the provision of services, promoting competition 81 BEYOND THE CENTER: DECENTRALIZING THE STATE among insurance suppliers and service providers, providing Labor Management direct targeted subsidies for the poor, and aiming at pro- Even this much decentralization has been rather tentative. viding health services to the population irrespective of abil- Labor management, for example, continues to be central- ity to pay. Two insurance plans were created to implement ized. In most of the case study countries, the price of labor the new approach: a contributory plan designed to cover and the conditions of employment are controlled not by salaried and independent workers, and a subsidized plan for subnational governments or administrators of health facili- the poor (a direct subsidy to pay for a basic package of ties, but by the central government, which enters into health care services). national negotiations with health sector unions every year. Implementation of the Ley 100 reform is to be highly This has stymied efforts to develop management autonomy decentralized. The subsidized regime is to be administered to the level where it is most likely to have an impact-to the and financed through municipalities, which will be respon- individual health unit. Directors of government-owned health sible for (a) identifying poor beneficiaries, (b) affiliating the facilities have generally little power over their labor input. beneficiaries with a private or public health insurance com- Table 5.2 summarizes where the six countries stand in pany (Administradora del Regimen Subsidiado [ARS}), and terms of personnel management. It indicates that in most (c) authorizing payment of corresponding premiums to the countries, subnational governments and health care facili- ARS. Other decentralized features of the system include the ties have only limited personnel management responsibili- devolution to departments and municipalities of a number ties and are constrained by strict national labor codes of financial, organizational, and personnel responsibilities, regarding health care workers. In some cases, state health with departments achieving a greater degree of autonomy secretaries have the power to hire, fire, and transfer hospi- than municipalities. The area of slowest progress has been tal directors and staff (Argentina, Bolivia, and Mexico). personnel management; the margin of maneuver for rene- Chile is perhaps the sole case where primary health care gotiating salary levels and employment conditions has been workers were effectively transferred to municipalities and almost nonexistent. At the departmental level, government lost their status as federal civil servants in the process. hospitals are being transformed into autonomous enter- prises. The process of hospital autonomization has been Recurrent Financing hampered by lack of clarity as to who will pay the pension Most health care providers (hospitals and primary care net- debts and within what time frame, and the conditions, works) continue to be financed by direct historically based including the state of the infrastructure, under which hos- budget allocations either through deconcentrated Ministry pitals will be transformed. of Health offices in states or through earmarked transfers to states and municipalities (in Bolivia, Chile, and Mexico, and Major Issues in Decentralizing Health in Colombia to departments where health has not yet been As these six case studies suggest, health decentralization in decentralized). In Argentina, central transfers to provinces Latin America has been a varied experience. Although five are not earmarked for health; transfers from provinces and of the six countries are experimenting with various forms of municipalities to health care facilities are historically based. purchaser-provider splits, health care continues to be dom- Transfer formulas to political sublevels are generally com- inated by the public sector in both its financing and pro- plex and have a large population-based weighting. These duction. Decentralization has largely been limited to the transfers then become the basis for local governments to transfer of parts of the public sector system from the central finance services, which is typically done through an input- government to subnational governments. Central govern- based, historical budgeting approach that ignores output ments have tended to retain responsibility for policymak- and quality. Correspondingly, most countries charge no fees ing, overall financing, and operation of highly specialized (or only nominal fees) for health care in central or decen- medical care centers (cancer, for example), while the states tralized facilities, and highly subsidized fees, if any, for hos- or provinces have often been given responsibility for provi- pital services and inpatient or outpatient diagnostic-related sion of secondary and tertiary hospital care, and municipal- activities. ities responsibility for primary health care (Colombia, This continued reliance on input-based financing pro- Brazil, Chile, and to a lesser extent Bolivia).3 vides no incentive for improvement in efficiency or cover- 82 EMPOWERING MAYORS, HOSPITAL DIRECTORS, OR PATILENTS' THE DECENTRALIZATION OF HEALTH CARE TABLE 5.2 Authority to Manage Personnel by Different Levels of Government GOVERNMENT LEVEL SALARY DETERMINATIONS (PRICE) HIRING AND FIRING IQUANTITY AND INC:ENTIVES) Central Bolivia, Chile, and Mexico Bolivia State/Department/Province Argentina, Brazil, Colombia Argentina, Brazil, Chile (Health Regions), Colombia (municipalized*) (if non-municipalized) (If non-municipalized), and Mexico Municipal Colombia (if municipalized*) Brazil (for service providers); Chile for PHC workers: Colombia (if municipalized) Facilities where purchaser-provider split Brazil (limited autonomy in public hospitals); operates Colombia (corporatized hospitals) * subject to central government guodelines age. Moreover, the introduction of subnational govern- ticipation Law are a fixed percentage of government rev- ments into the health care system has raised new financing enues and represent additional resources for social services, problems. Health decentralization has generally been pred- including health. Budget allocations to pay subnational icated on continued central government financial support, governments for increased salaries of health staff and accu- but this has not always been forthcoming. This has been a mulated pension and other benefits are another source of problem particularly where there is central control over increased net expenditures on health. Overall, increased salaries. In Colombia, while labor costs are set by the expenditures in health have represented over 2 percent of national government, the funds transferred to subnational GDP in some countries (Bolivia and Colombia), although governments to finance health services do not fully reflect not all the increase can be attributed to decentralization. the costs, resulting in constant shortfalls, disputes among government levels, and financial pressures on local govern- Investment Decisions ments. One result in Colombia and other countries is little Unlike most other health activities, investment decisions interest on the part of subnational governments to accept have been fully decentralized in many Latin American health service responsibilities in their territories because the countries (see Table 5.3). While this has quickened respon- responsibility is transferred without adequate resources to siveness to local needs, it has also created problems. First, carry it out or authority to manage the costs. since all three government levels are allowed to invest in At the same time, decentralization reforms have resulted infrastructure and have the money (often earmarked trans- in a heavier fiscal burden for central governments. To a fers) to do so, a tendency exists to overinvest in infrastruc- large extent this has resulted because formula-based trans- ture and equipment. Typically, there is no explicit charge for fers to municipalities, and in some cases to states, have been rental of capital and equipment or for depreciation, so there made, in part for political reasons, with new monies rather is no countervailing incentive to use capital productively. than with reallocations from existing health expenditures. Second, wherever states and municipalities have not also In Bolivia, municipal transfers created by the Popular Par- been responsible for financing personnel, they have tended TABLE 5.3 Authority to Make Infrastructure Investments GOVERNMENT LEVEL PRIMARY HEALTH CARE HOSPITALS Central Bolivia, Chile State Mexico Argentina, Colombia, Mexico Municipal All except Mexico Brazil (plus some state cofinancing) Purchaser-provider split operated Colombia (from user fees) 83 BEYOND THE CENTER: DECENTRALIZfNG THE STATE to overlook the recurrent cost implications of those invest- reforms cost additional money because, at least initially, ments. In those countries it is common to find empty health there is (a) some duplication of expenditure (central gov- care facilities because the central government has not filled ernment employees are rarely transferred to other locations, the new staff positions or vacancies because of central bud- for example), (b) a need to rehabilitate hospital and primary get constraints. The central government has even been health care networks, and (c) a necessity to train personnel reluctant to allow municipalities and community organiza- to handle newly decentralized responsibilities. Where more tions to fill those vacancies for fear that the new positions radical decentralization reforms have been planned, the cost will eventually have to be absorbed by central government of accumulated pension benefits for health workers has payrolls as a result of pressure from labor unions. been one of the main obstacles. Under planned reforms in Colombia and Venezuela, provinces were to take charge of The Political Economy of Reform all personnel and payroll (under national civil service codes) The partial and haphazard way that health care decentral- and pension benefits. The central government promised to ization has been implemented reflects to some degree the pay all accumulated pension benefits up to the moment of powerful forces arrayed against it. While the potential ben- the transfer or when the decentralization law was enacted. eficiaries-the citizen-patients-are many and diffuse, the This has proven difficult. opposition groups are concentrated and focused. One such National ministries of health have also dragged their feet opposition group has been health sector workers, who have on decentralization, fearing a loss of their prerogatives. seen decentralization as a threat to their union power and as Arguing that subnational governments are incapable of a way for the government to divide them. In most cases, receiving and providing health services, they have imposed health sector workers have agreed to decentralization- cumbersome and often unnecessary certification require- deconcentration only after receiving assurances (usually by ments before permitting service provision to be decentral- law or supreme decree) that labor code regimes and associ- ized. Furthermore, those certification requirements are ated benefits are maintained for themselves, and often for never clearly specified with respect to content, dates, and new staff as well. process, but are left to the interpretation of central govern- Central governments have also been less than enthusias- ment bureaucrats, leaving ample room for discretionary tic. In part this reflects financial concerns. Decentralization practices and negotiations, and for stopping the process lack of Rlal nomto i available,: tare oftenp unsrelal. eorts o infiini essw.00 X and ther stW srcue and revenuei psects, andte publicX In uhssesteeaefwtasctosivlig srrsn htte nomto ytm paei ten- money, thereislittle competition,thereisalmostnoincen- traied systemfallfarsh j;any :reform that; multples 0the lo it of deciion -atn-taiigepniiiy EMPOWERING MAYORS, HOSPITAL DIRECTORS, OR PATIENTS' THE DECENTRALIZATION OF HEALTH CARE with a change of minister or other policies. In Brazil, vision (as opposed to the financing) of health care, and a Colombia, and Venezuela, for example, certification shift in the focus of government financing from inputs to requirements include the following: a detailed yearly health outputs. operation and investment plan, the development of an Happily, there are signs of movement in this direction. information system, the incorporation of health personnel Brazil, as noted earlier, has already largely shifted to the into state or municipal payrolls, and a working referral net- private sector mode of production and output-based work. Most state and local plans are never implemented financing. (Government-owned hospitals in Brazil, how- because states and municipalities control few of the inputs, ever, continue to receive input subsidies from their respec- except in the few fully decentralized states and municipali- tive subnational governments.) Chile has also introduced ties in Colombia. output-based financing for primary health care (although it Finally, there has been a marked lack of interest by sub- recently shifted to a capitation-based reimbursement sys- national governments in taking on the new responsibilities. tem for primary care, while introducing payments for To a large extent, this lack of interest results from a fear DRGs in regional hospitals). In Colombia, public hospitals, that the central government will not fulfill its financial which have adopted the form of Social State Enterprises obligations and a fear of having to deal with powerful under Law 100 of 1993, have a hospital directorate and health sector unions. In Argentina these fears also apply to director who can manage labor contracts, time schedules, hospitals that are in the process of obtaining autonomy and contract services. With the recent introduction at the from the provinces. In Colombia, for example, because of provincial level of reforms giving more autonomy to hospi- their desire to retain control over hospitals and budgets, tal facilities, hospital managers are just beginning to have departmental politicians have successfully resisted or more power over personnel in Argentina (Abrantes and slowed elements of the national reform to improve account- Diaz Legaspe 1999). These trends are perhaps indications ability of services to clients (by splitting provision from that policymakers are coming to terms with the limitations financing). of simply transferring assets and staff as a means of improv- ing the delivery of health services. Conclusion Overall, it might be argued that decentralization has dam- Notes aged health care provision, at least in some Latin American 1, One result was a vast expansion in output, resulting in a countries. It is difficult to imagine that sharing responsibil- change of the financing system for municipal health services from fee- ity for factors of production between levels of government, for-service to capitation in 1994. Public hospitals-as opposed to pri- as is done in Bolivia, is conducive to improving the delivery mary clinics-were transferred to the 25 quasi-autonomous health of health services. Decentralization in Argentina and Mexico, service units, which have been given less autonomy over personnel. o Personnel are paid according to fixed payrolls, following pay scales which has consisted of transferring resources and responsi- and personnel administration rules that are established by the Min- bilities down one level of government, is equally unlikely to istry of Health. With the introduction of new payment systems (pay- improve service delivery or contribute to containing costs. ments by DRGs), it was expected that personnel services could be The perverse result of the partial decentralization that has included in the fee schedules. However, this has been difficult to occurred in many Latin American and Caribbean countries implement. 2. This has led to some concern that Chile is developing a two- isthate although sevomved, e ntsbiliy o s m gement finan. track system of health care, with high-quality care for those able to have been devolved, accountability has been dispersed. afford private insurance and inferior care for those who must rely on These results are not the inevitable consequences of the publicly financed system. decentralization. They reflect the particular form in which 3. One practical problem with this division of responsibilities decentralization has been implemented. Evidence from the between government levels-with hospitals devolved to states and six case-study countries suggests that improving the qual- primary health centers to municipalities-is that it complicates the ity of health care will have to go beyond transferring assets workings of the referral system, however well it might have func- tioned previously. This occurs to a large extent because the financial and staff to municipal governments. It may require more transfers for states and municipalities are usually disconnected, with fundamental changes in the role of the public sector, both receiving direct transfers, often by formula, from the central level including an increased role for the private sector in the pro- to meet their specific service responsibilities. In addition, neighboring 85 BEYOND THE CENTER: DECENTRAEIZING THE STATE municipalities find it difficult to develop compensatory economic defined or have changed repeatedly over a short period, hampering arrangements to pay for services rendered to their residents in other the consolidation of institutional and other arrangements at subna- jurisdictions (Argentina, Bolivia, Brazil, and Colombia). In practice, tional levels. however, responsibilities among government levels have often been ill 86 CHAPTER 6 Decentralizing Roads: Matching Accountability, Resources, and Technical Expertise ver the past decade a wave of political and administrative decentralization has swept over Latin America. Today, local authorities are publicly elected in almost every country in the region, and local governments are responsible for the execution of a broad range of pub- lic functions previously handled by national governments. Decentralization has been pur- sued as a strategy to improve efficiency in the provision of public functions and reduce public sector spending, based on the principle that policymakers and public administra- tors make better decisions about how to use resources when they are closer to the level at which activities take place. However, acting in the opposite direction are the institutional capacity and human resource con- straints that result from the loss of economies of scale as responsibilities are moved from the national to the local level. It is easy and often misleading, therefore, to make simplistic statements about the desirability of decen- tralized sector management. The transport sector, with its "lumpy"" investments and combination of national, regional, and local interests, represents a particular challenge to decentralization strategists. Road, rail, port, airport, and mass transit systems all pose a series of complex issues that must be addressed in order to evaluate the costs and benefits of decentralization, and the preconditions and strategies for implementing it. In this regard, the road sector has been the focus of numerous decentralization efforts. This chapter will present the Latin American experience in road management decentralization, provide an analytical framework for the review of these experiences, and propose opment. As Latin America continues to move toward some directions for the furure of the road sector based on regional market integration and strives to compete in the the lessons from these experiences. global economy, the extent and quality of its road network will need to grow, providing access to remote regions and Decentralization in the Road Management increasing the mobility of people and goods within the Sector region. Currenr road densities are still low compared with Roads are an important part of any country s economy. those of industrial economies, suggesting that access prob- They are a productive asset used for the transportation of lems still exist and that future extension of the networks 60 to 80 percent of passenger and freight cargo, and they will be required. often provide the only form of access to rural communities. Roads also represent an important proportion of public As such, roads are at the heart of economic and social devel- expenditures in the region, and in many cases they repre- 87 BEYOND THE CENTER: DECENTRALIZING THE STATE sent an asset value of up to 40 percent of annual gross sector labor force, to carry out works. Accountability to the domestic product (GDP). At the same time, road use gen- road users is muted because of the indirect nature of the erates significant tax revenues through the application of financing mechanism (general budget allocations), which is road user charges, such as fuel taxes, road tolls, vehicle reg- unrelated both to what users pay in terms of taxes and to istration, and license fees. Road construction and mainte- increased vehicle operating costs of deteriorated roads. nance also contribute significantly to employment, especially A number of reforms have been applied in Latin Amer- in less-developed, labor-abundant countries. Overall, roads ica and elsewhere that are directed at improving the are big business, contributing as much as 6 percent to the accountability of road managers and improving the respon- total GDP siveness of road agencies to deteriorating road conditions. Although significant improvements have been achieved Decentralization of road agencies-devolving the responsi- over the past four decades in the expansion of road net- bility for planning, building, and maintaining "non- works in developing countries, the sector confronts a per- national" and in some cases "national" roads to state and sistent problem of inadequate maintenance of roads, which local governments-is one reform that has been pursued as results in greater need for rehabilitation and reconstruction a way of bringing road users and managers closer. The investments. Road agencies around the world have gener- objective is to facilitate a greater correlation between the ally shown significant ability to build or improve roads, but quality of roads that users are willing to support and those have repeatedly faced difficulties developing and sustaining that are provided. There are a limited number of analyses of adequate maintenance practices. A decade ago worldwide the impact of road decentralization. A World Bank review estimates indicated that an estimated 15 percent of the cap- of 42 developing countries found that where road mainte- ital invested in main roads in developing countries had nance was decentralized, backlogs were lower and road eroded. This represented a loss of roughly US$43 billion, or conditions better. In the aggregate, the percentage of about 2 percent of the countries' GNP at that time, a sharp paved roads in poor condition dropped from 22 to 12 per- contrast with the estimated US$12 billion in preventive cent with decentralization, and the percentage of unpaved maintenance that would have been required to keep the roads in poor condition dropped from 33 to 15 percent roads in proper condition (Harral and Faiz 1988). In Latin (World Development Report 1994; see also Humplick and America, early in the 1990s, it was estimated that it would Moini-Araghi 1996). cost about US$2.5 billion per year to deal with the mainte- Nonetheless, the available experience with decentraliza- nance backlog on the region's roads and to prevent further tion of road management in Latin America shows mixed deterioration. results. While countries such as Argentina, Brazil, and The persistence of maintenance problems cannot be Colombia have generally benefited from decentralized road explained solely as the result of a lack of technical capacity management (mostly of secondary and tertiary networks), to carry out the necessary work programs. The road agen- negative experiences have led to partial reversals of the cies that have been able to plan, design, and execute the process in Peru and Bolivia. This has created uncertainty in significant number of road investments carried out in the the sector about the merits of reform. There are many past are certainly technically capable of performing the debates over the reasons for the success or failure of differ- much simpler tasks of maintaining roads. A more broadly ent decentralization experiences, and a deeper look into the accepted explanation is found in the institutional frame- nature of road management decentralization is required. work within which roacls are managed, and the distorted Several difficulties emerge in the attempt to understand nature of the signals and incentives which these institutions the various factors involved. First, there is a need to be clear receive. Road administrations have traditionally favored about which roads and which road management tasks are to new construction over maintenance, a mentality often rein- be decentralized. Second, there has been little compilation forced by political pressure. Funding has followed suit. of the experience in road decentralization across countries, Moreover, even when resources have been sufficient, funds and little data are available to assess the impact that decen- are used inefficiently, especially in those administrations tralization may have had on the performance of the sector. that depend on a "force account," that is, their own public Finally, parallel to decentralization, other reforms have been 88 DECENTRALIZING ROADS: MATCHING ACCOUNTABILITY, RESOURCES, AND TECHNICAL EXPERFISE initiated including the practice of contracting road mainte- establish an appropriate set of policies for countries that are nance in place of using a force account, establishing road in the process of decentralization. In most instances, decen- boards and related road funds, concessioning, and long- tralization of roads is part of a broader political and fiscal term, performance-based maintenance contracting, all of decentralization, with limited attention paid to sectional which increase accountability and provide a "voice" for road conditions. The lack of an effective strategy for transition users, but whose benefits cannot easily be distinguished in the sector has been the major weakness in most of these from those of decentralization. initiatives. This chapter seeks to address these issues by proposing a conceptual framework that can be used to understand the What is Decentralization in the Road Sector? implications of road sector decentralization, and reviewing Decentralization in the road sector entails the assignment of Latin America's experiences in road decentralization, in full responsibility for specific roads to subnational levels of order to identify lessons and directions for the future of the government. Full responsibility in the road sector involves sector. It argues that while decentralization can contribute a combination of road functions, management tasks, and to effective road management by bringing road users and financing mechanisms that are closely interrelated to ensure managers closer together, there are a number of precondi- the integrity of the road network. The first step toward tions which must be in place or instituted as part of a longer- understanding decentralization in the road sector is to term strategy to ensure the realization of projected benefits. clearly define what it is that is being decentralized. This requires the separation, or "unbundling," of the different In Search of a Model: The Latin American and building blocks that comprise a road management system Caribbean Experience and which could be subject to decentralization. Thus, our In general, three broad organizational systems can be iden- conceptual framework for the analysis of road sector decen- tified for the management of road networks: tralization is based on the separate treatment of three build- Centralized systems, in which the network is managed ing blocks of effective road management: road functions, by a single, national road agency (as in the Dominican management tasks, and financing mechanisms. Republic); Road Functions Deconcentrated systems, in which regional agencies Roads are classified by different levels according to their reporting to the national agency are set up to manage function. The function of a road refers to the role it serves parts of the road network (as in Chile); and parts of teranewr(within the network and the predominant type of demand *Decenrralized systems, In which local or regional agen- Diestrarelaed undter,i thejuriictlona of subnagional age by its users. Is it a road that mainly serves a national pur- govesarnmnsomaae p artsdofder the juri onewor (nastiona pose (such as connecting major cities with key border points for trade), or is it a road that mainly serves local needs (such Argentina and Brazil). as getting locally produced goods to local markets)? This These organizational systems usually reflect the broader distinction helps to identify the constituency that cares the political organization of the country. Countries with a cen- most about the characteristics and conditions of that road. tralized system of government will normally exhibit a cen- Four different levels of networks are identified here: tralized road management system. Similarly, countries with a high level of political decentralization and strong subna- *National networks (primary or main trunk), made up tional governments will usually have decentralized road of roads that play a function of national interest that management organizations. A discussion of road manage- goes beyond the interests of the immediate communi- ment decentralization, therefore, must take into account ties near the road; the broader context of the country's political organization, *Regional networks (also known as secondary, depart- and considerations of country size, population distribution, mental, or state networks), consisting of interurban and the complexity of the road network. The major chal- and rural access roads that benefit mainly the broad lenge for Latin America is to determine a strategy and communities of a region, state, or province; 89 BEYOND THE CENTER: DECENTRALIZING THE STATE *Rural networks (including community roads, tracks, 'Formulation of policies and regulations, including reg- and trails), which provide access to outlying rural com- ulation of road characteristics to ensure network con- munities; and nectivity and safety standards, and ensuring compli- 'Urban networks (in turn divided into metropolitan and ance with social and environmental regulations. large urban, medium-size urban, and small urban net- works), which provide circulation within citics and Table 6.1 presents a framework for the consideration of towns. options for the decentralization of management tasks in the road sector. Again, the general principle is that managerial Conceptually, responsibility for roads should be assigned responsibilities for a road should be given to the level of to the level of government that represents the community government most associated with the predominant users of of users who benefit most from the road. In practice, how- that road, as a means of ensuring greater efficiency in the ever, gray areas exist regarding the designation of some establishment of priorities and the corresponding allocation roads as national or regional, and others as regional or rural, of resources. mainly due to the difficulty in distinguishing among the Concerns are often raised, however, about the lack of various users and their travel purposes. Many roads often technical and legal capacity to absorb these tasks and the serve both local and national traffic. Moreover, political loss of economies of scale. Indeed, smaller countries with considerations often play a significant role in the final des- more limited secondary networks are not decentralizing ignation. Local representatives will often lobby for inclusion road management. Experience has shown that transferring of much of their network within the national network as a responsibility for roads from one inefficient public agency to means of ensuring national funding. another will not provide the necessary benefits. A range of There are a number of considerations in defining each technical capabilities is required to manage roads, including type of road classification and the political, technical, and planning expertise and traffic engineering, investment financial issues involved. For the purposes of this chapter, analysis, engineering design and costing, and works plan- the focus is on the national and regional networks rather ning. Changes in the use of the private sector have begun than on the rural and urban networks. Generally it is to alter the skill mix required in public agencies, reducing accepted by sector professionals that rural and urban roads some of the in-house capacity previously considered neces- should be devolved to state or local governments, although sary. Many agencies had depended on their own staff (force there remain issues of financial sustainability of such account) to carry out maintenance works with the ineffi- arrangements. The key debate, however, is over the ciencies associated with excess staff and no fuel or deterio- national and regional networks. rating equipment. Over the last decade, governments have contracted out such works to private contractors, which has Management Tasks increased the need for contract management over works The second element of road decentralization is the transfer management skills. There have been similar contracting of managerial responsibilities for the roads placed under the arrangements for design and other services. This reduces jurisdiction of subnational governments. In the manage- some of the technical demands on decentralization. ment of road networks, four clearly distinct tasks can be Another related option has been to maintain ownership identified: at one level of government while delegating certain road management tasks to another level. Construction and *Planning of road investments, which includes the defi- maintenance activities can be delegated from the national nition of future network requirements and the estab- to the regional governments, in the case of the national lishment of priorities for investment; network, to take advantage of economies of proximity. 'Rehabilitation and construction of roads, which This is especially true in larger countries with the under- includes detailed engineering and the contracting and standing that sufficient funding is provided by the dele- supervision of works; gating level of government. Delegation can also take place *Operation and maintenance of roads, which includes from the municipal governments to a regional or multi- periodic and routine maintenance, signaling, and municipal entity to achieve economies of scale and enhance sometimes policing; and network coordination. The delegation of maintenance 90 DECENTRALIZING ROADS MATCHING ACCOUNTABILITY, RESOURCES, AND TECHNICAL EXPERTISE TABLE 6.1 Decentralization Options for the Road Sector PLANNING CONSTRUCTION MAINTENANCE REGULATION NATIONAL 'Central Central 'Central 'Central NETWORK * Concession * Concession *Rehabilitation and maintenance 'Service contract to private contract sector 'Delegation to regional agency 'Delegation to regional agency REGIONAL 'Regional -Regional * Rcgional 'Regional NETWORK -Concession 'Concession 'Rehabilitation and maintenance *Service contract to private contract sector RURAL 'Regional 'Regional Regional 'Regional NETWORK 'Municipal *Municipal 'Municipal 'Municipal -Delegation to regional agency -Delegation to regional agency 'Delegation to regional agency 'Delegation to regional agency URBAN 'Municipal 'Municipal 'Municipal 'Municipal NETWORK *Multimunicipal (metropolitan) 'Delegation to regional agency 'Delegation to regional agency 'Delegation to regional agency functions to state governments is the most prevalent of these national government to monitor the ability and commit- possibilities. ment of local governments to meet minimum standards. Full funding through transfers, however, raises concerns Financing Mechanisms about local commitment. The depth of decentralization can The third building block of road management is the financ- be measured by the percentage of expenditures financed ing arrangements for the construction and maintenance of locally. A mixture of national transfers combined with a sig- the various road networks. It is perhaps the most contro- nificant local contribution, assuming local ability to raise versial element and is essential for setting the right institu- such funds, should offer the greatest balance between local tional incentives and ensuring the sustainability of the ownership and efficient reliance on user charges. It also pro- decentralization process. Perhaps the most common com- vides a basis for the national government to monitor plaint of local authorities toward road decentralization is progress and adjust incentives and technical assistance as that responsibilities are being transferred to them without required. the transfer of sufficient resources to carry them out. With- Using road funds is another means of implementing the out careful consideration of what sources of funding to use user-pays principle, although indirectly. The funds are gen- and how to allocate funding for the different networks erally financed from revenues generated through user under decentralized systems, the objective of increased charges collected at the national level, mainly fuel taxes. accountability can be jeopardized. Established at the national level, their use is often restricted There are three major mechanisms for funding roads: to maintenance. Formulas are often agreed upon for the budgetary resources, road funds based on assigned user- allocation of the money among the different networks and charge revenues, and tolls. The dominant source of funding the corresponding levels of government. Transport sector for roads is the government budget. Where this method is professionals have promoted funds as a means of ensuring used, road users have an impact only insofar as there is an adequate and stable funding for maintaining existing road effective representative governance system through which networks. Experience, however, has shown such funds to be expenditure priorities are determined. When road manage- highly sensitive to macroeconomic shocks. Thus, rather ment is decentralized, budgetary financing can be provided than protecting road maintenance from budgetary fluctua- from either central or subnational revenues. Financing from tions, in practice, they have exacerbated it. Finance min- central government revenues-that is, through transfers- istries have also complained of a lack of clear accountability has some advantages. Earmarked transfers allow the in the use of such funds. More recently, a new generation of 91 BEYOND THE CENTER: DECENTRALIZING THE STATE BoxX 6.1 0000000000000 0000000$000-00000000000000t0 000:00 tt0X00 0 l^ 00 00400 000000000 00 tg commercializ0ome lvrodm ag-ement.00- nac activi-;0-0.i:l^ g0-00 0000000ties. 00000000000 2X00t0000000 On000 ly roa us0Eer00 chags should he pi int th ra *Th roa fun shud ful fiancethe expniue ';0000-The arangement 0mus:t n sutrc revenue from eioa an uicpa roa expnd ture shoudb other s00000etosrs (aXdditi-onalspending on rom.s 0mus3t.. f,iStnaned onl-yi-thog transparent cotshrn cf ; i:ome from extra payments by road users>.- ar0;X0igi i0; Si:0 raneents that strengtenfinancial discliSne.l fSUt0 *0 i The road fnd shouldbe loverseeln bya rprsetatx:[ 'ive -0 Susdir roa funds ma be se pt ac ra ;;:0tt-publcpivate :board: ithmembe;irs<4 g appointd for o rurlras j;: jt three- 0to four-year: tem. 'Rvne shul be 0 :; :00;tj:i;ij04tj: 0. : t0iS-S- .0 0 t. t:o 0s -) 0cected ;using ia tjarifconsisting0 3 'Th ro:ad fuind should be managedb astng:, ne- prmail of aul tax,..................... veile registraton fee,; a sup 0t00-0pend;enttsecretariat,t: seaae frm: i teti min road: plementary: heavy-vehicl e,itratoarni Sairnre000 *rkgk ad vickers 19 9&0mi- - road funds has been developed (see Box 6.1) which incor- Road Decentralization in Latin America porates a more robust accountability mechanism by creat- Over the past 10 years, the majority of countries in Latin ing road boards to manage the funds. The road boards America have decentralized road management responsibili- include representatives of the various affected road users ties for non-national roads. Today, approximately 80 per- and of the public sector, and offer a management frame- cent of the roads in the region are under the responsibility work in which the users have a voice. It is still to be seen of subnational governments. The exceptions are Chile, how these boards and related funds will evolve over time which has a deconcentrated road organization, and small and whether they will be able to maintain a representative countries with a centralized political organization (such as and dynamic exchange between the users and providers of the Dominican Republic, Honduras, and Uruguay). In this road infrastructure. section, we review some of the Latin American experiences The use of tolls offers a clear application of the user-pay with road management decentralization in order to extract principle by being location specific. Thus, toll roads have lessons and make recommendations for the sector. been the basis for concessions to the private sector, which is Testing the benefits of decentralization is difficult. As the purest form of commcrcialization of road management. indicated earlier, the main weakness in road management Indeed, users in Latin America and elsewhere have given has been the lack of maintenance. Thus, the key measure of clear signals of their dissatisfaction when faced with what success would appear to be the condition of the network they consider high tolls for a poor level of service. However, before and after decentralization, as indicated by a series of tolls are only viable on very high-traffic segments of the engineering indexes. While this measure is essential, it is network (generally above 5,000 vehicles per day). On low- complicated by the need to define what is an appropriate, traffic segments, which is the case for most state and local econoTnically viable condition. Roads with lower-traffic vol- roads, the cost of collection is too high (and therefore inef- umes do not warrant the same level of maintenance or con- ficient) compared with other tax options. dition as roads with higher-traffic volumes. One would also 92 DECENTRALIZING ROADS: MATCHING ACCOUNTABILITY, RESOURCES, AND TECHNICAL EXPERTISE want to test the cost and efficiency of maintenance. Unfor- network are provided from direct user charges (tolls) and tunately, not enough data are readily available to properly from the national government's general revenue compare the performance of road agencies in Latin Amer- Argentina's decentralized system of road management ica. This is particularly the case where road management has been successful in ensuring adequate maintenance of decentralization has taken place, since it makes it harder for the road network at the various levels. The division of national agencies to collect and organize data from the var- responsibilities between levels of government has been sta- ious regional agencies. Nonetheless, some clear observa- ble over the last 10 years, as has been the financing of the tions can be made based on a review of some of the experi- sector. At the national level, there has been a clear trend ences in the region. toward increased use of private sector contracting for road maintenance. Current discussions in the sector focus on the Where Political and Fiscal Decentralization is Well extent to which concessions can be used for the develop- Established ment of new roads and on the sustainability of contracts Road decentralization has worked best when it has been with the private sector in the event of macroeconomic insta- implemented in a country where political and fiscal decen- bility or external shocks. tralization is mature and well established, and where it is Brazil's experience also shows that when a clear separa- treated as part of a set of sector reforms that include (a) a tion of institutional responsibilities is made and adequate clear separation of institutional responsibilities between dif- financial mechanisms are put in place, the decentralization ferent levels of government and (b) the establishment of of road management results in improved road conditions. adequate mechanisms for financing road maintenance and Brazil is also a federal system, organized into 27 states with investment. This is illustrated by the Argentine experience. strong traditions of regional government. In 1988, a new Argentina has a federal system of government, with 24 constitution gave ample powers to the states to raise their provinces. The country's road network is classified into own taxes, including a fuel tax. Until then, most taxes had national, provincial, and municipal roads. The National been centrally collected and then partially redistributed to Road Directorate (Direcci6n Nacional de Vialidad [DNV}) the states (this included fuel and vehicle taxes, which were is responsible for management of the national network, partly earmarked for roads and redistributed to the states while each province has its own Provincial Road Directorate through a formula). The new constitution transferred some (Direcci6n Provincial de Vialidad) for managing the provin- taxes to the states and municipalities, including a con- cial networks. Municipal networks, including urban and sumption tax (similar to a value-added tax) and the fuel and rural roads, are the responsibility of municipal govern- vehicle taxes. Earmarked funds were terminated. Expendi- ments. Each level of government is responsible for financ- ture responsibilities, however, remained unchanged. The ing road maintenance and rehabilitation of its correspond- state governments, now funded through their own consump- ing road network, with resources coming from general tion taxes and transfers from federal tax revenues (income revenues and from the revenues generated by the applica- and industrial taxes), have increased their revenues. How- tion of road-user charges. Since revenue generated by a fuel ever, the combination of excessive staff, high debt, and the tax is collected at the national level, the federal government disappearance of the inflation tax (after the implementation transfers funds to provinces and municipalities according to of the Real Plan in 1994), left few states with sufficient a transparent formula agreed to by all and monitored by resources for investment. Nonetheless, although important the Federal Road Council (Consejo de Vialidad Federal regional differences in institutional and fiscal capacity exist, [CVFI). The CVF serves as a forum for the coordination of the states have assigned a high priority to road maintenance national and provincial agencies, mainly regarding the and rehabilitation, and most of them have been able to raise planning of new roads and the appropriation of earmarked adequate funds in spite of important fiscal constraints. funds from the fuel tax revenues for road investments at the Brazil's experience illustrates the complexities of decen- provincial and municipal levels. These transfers account for tralized road management. In Brazil, the granting of addi- about 35 percent of the resources used by subnational gov- tional fiscal powers to the states was unrelated to the allo- ernments for roads, evidencing the high level of local own- cation of responsibilities for road management, even ership over the local road networks. Funds for the national though one of the key new revenue sources (the fuel tax) 93 BEYOND THE CENTER: DECENTRALIZING THE STATE was directly associated with road use. The National Roads available evidence. The arguments in favor of a gradual Department (Departamento Nacional de Estrada de approach to road decentralization are strengthened by the Rodagem) remained responsible for over 67,000 kilometers experiences in Peru, Bolivia and Venezuela with rapid, of federal roads, even though it had lost the significant rev- wholesale decentralization. enues from the fuel tax. Between 1990 and 1997, however, In Peru all roads were transferred in 1984 to the the conditions of the federal network improved, reducing recently established regional governments, including por- the number of kilometers of paved roads in poor condition tions of the national network that had profitable tolls. The from 30 to 11 percent. These improvements have been recurrent use of toll revenues to finance activities outside financed from federal general revenue, which demonstrates the road sector led to inadequate road maintenance and the the priority of the sector relative to other sectors. decay of key portions of the national network. As a result, A recent review of the federal network based on road in 1991 the national authorities decided to recentralize the functions suggested that about 18,000 kilometers should national network. While 78 percent of the total road sys- be reclassified as state roads. This is the next challenge for tem remains in the hands of the regional and local govern- the road sector in Brazil. The states have shown interest in ments, regional agencies are poorly funded and institution- assuming responsibility for additional roads, but have ally weak. Today, they depend heavily on technical and expressed concern about the poor condition of the roads financial assistance (and at times direct involvement) from that are to be devolved. Agreements have been reached the Ministry of Transport for almost every important activ- between the federal and some state governments for the ity. Road conditions in key sections of the national network rehabilitation and subsequent devolution of the roads that have improved since recentralization, but the majority of are not of national interest. Federal roads have also been the roads are deteriorating beyond the capacity of the cen- delegated to some states for their subsequent concessioning tral ministry to respond. by the state governments. However, the sustainability of In 1995 Bolivia embarked on a broad process of admin- some of these agreements is still uncertain. These situations istrative decentralization. The road sector had been origi- emphasize how difficult it is to establish and maintain a nally organized in a system of nine deconcentrated districts clear classification of road functions when different levels of under the national roads agency (Servicio Nacional de government share network responsibilities. In the end, the Caminos), which was responsible for the primary and sec- challenge is to establish transparent financing mechanisms ondary networks and some municipal roads. With decen- for road management (maintenance, rehabilitation, and tralization, these organizations were placed under the juris- new investments) that correspond to the allocation of diction of the regional governments (prefecturas). In responsibilities across the different levels of governments addition, a number of regional development corporations (the principle that "finance follows function") and provide that had been partially responsible for the tertiary network the right incentives for the behavior of road agencies. were pulled out of the sector. The Ministry of Transport, reorganized as a secretariat under the Ministry of Economic Dangers of Rapid Decentralization Development, received an unclear mandate and lost con- Both Argentina and Brazil are countries with long-standing siderable power and support at the national political level. traditions of decentralized government and exhibit strong While the entire network was being transferred to the pre- regional governments with relatively high levels of institu- fecturas at the same time that other sectors were being tional capacity. Naturally, we expect decentralized road decentralized, no source of financing other than tolls was management to work best under such circumstances. Else- provided. Many of the experienced staff of the deconcen- where in the region, decentralization of road management trated districts retired or left the sector during the reforms, has taken place over a rather short period of time, usually leaving the departmental agencies with few resources for concomitant with a broader process of administrative managerial and road maintenance practices. As a result, decentralization. Under such circumstances the challenges road conditions deteriorated to such an extent that in are greater and so is the risk of failure. Therefore, prudence August 1998 the central government decided to step in and suggests a gradual approach, which is supported by the take back the management of the national network. 94 DECENTRALIZING ROADS MATCHING ACCOUNTABILITY, RESOURCES, AND TECHNI-CAL EXPERTISE The Venezuelan experience, a case of mixed results, also risk that public pressure would mount for the national deserves attention. The Decentralization Act of 1989 agency to retake the roads, resulting in the reversal of the brought broad political decentralization to the country and decentralization process. established the responsibilities of the municipal, state, and Almost 10 years later, the results of this arrangement national governments in a number of aspects of public life. have been mixed. Local governments have proven to be bet- With respect to roads, the Act permitted states to have ter able to manage the road network than was the Ministry "exclusive or shared" jurisdiction over all primary and sec- of Transportation, but most states face institutional and ondary roads, bridges, ports, and airports in their territo- financial challenges to various degrees, which may threaten ries. It also established that (a) any transfer of responsibili- the sustainability of the reforms. One of the main problems ties from the central government would take place through is that financing for the sector varies significantly from year negotiated agreements, and (b) the states would annually to year due to fluctuations in the value of oil exports (which prepare coordinated investment plans consistent with spe- are the source of over 50 percent of total public sector rev- cific amounts to be provided by the central government. enues) and the absence of significant road-user charges. In addition, significant differences exist in the institutional Benefits to National Networks of Devolving Non- and fiscal capacities of the 23 states. As a result, this National Roads demand-based decentralization process has produced Decentralization of secondary and municipal networks as a important discontinuities in the conditions of the national first step in the process of reform has contributed to the network, because many states have not been willing or able clarification of the roles and responsibilities of the various to fully assume their responsibilities Therefore, it is not yet levels of government, directing political attention from clear whether this process will be sustainable over the long each level of government to the corresponding network term. This experience highlights the importance of clearly level. In this respect, the first positive impact of decentral- separating the roles and responsibilities of the different lev- ization of regional and municipal roads is that it serves to els of government, setting up adequate financing mecha- keep local politics at the local level, thus protecting the nisms for the sector based on road-user charges, and tak- national network from political pressures that tend to ing measures to preserve the connectivity of the overall divert resources and managerial attention to issues that are road system. not of national importance. The Colombian experience illustrates this point. In Contribution of Parallel Reforms 1993, 32 percent of the road system under the national In addition to establishing a clear separation of responsibil- government was transferred to the regional departments ities between levels of government, road agencies in the under an arrangement that included agreed financing for region have made strategic use of the private sector for the operation and maintenance for three years. As a result of management of high-volume roads, awarding long-term this transfer, the national road agency was able to focus on concessions for their rehabilitation and maintenance. Pri- the improvement of the national network. Within five vate sector participation has been sought as a tool to (a) years, the national network has become a well-defined sys- secure adequate technical and managerial capacity for road tem of trunk roads and corridors, and rehabilitation and management (maintenance, rehabilitation, construction); maintenance practices have improved significantly. This is (b) establish clear accountability for performance; and (c) clearly a positive effect of decentralization, even if some- provide adequate financing for road management through what counterintuitive. In contrast, departmental road commercialization of road use. agencies showed weak administrative and technical capac- As discussed earlier, concessions are the purest form of ity, and conditions of the regional and local networks road commercialization. In the 1990s more than US$35 remained poor. The challenge under these circumstances billion has been infused into the road sector through private was to strengthen institutional capacity and to set up ade- sector concessions in 13 countries in Latin America. The quate and reliable financing mechanisms at the local level leaders in this area have been Mexico (53 concessions), in a relatively short period of time. Otherwise, there was a Argentina (19 concessions), Colombia (12 concessions), and 95 BEYOND THE CENTER: DECENTRALIZING THE STATE Chile (10 concessions). The broad experience in the region investments (often 20 years), concessions are highly vulner- with road concessions confirms that the commercialization able to macroeconomic fluctuations and require a clear and of roads does have a number of positive results: reliable regulatory framework to avoid uneconomic results. As an alternative to road concessions, long-term, perfor- *Projecrs ate completed on time. mance-based rehabilitation and maintenance contracts Modern tchnologes are pplied.(Contrato de Recuperaci6n y Manrenimiento {CREMA})) *Higher road standards are achieved and maintained. were introduced by Argentina's Direcci6n de Vialidad Fed- Adirvional services arenprovided emergencyatoterst(forn eral (DNV) in 1997. The CREMA is a long-term, perfor- servicsttionsl fiandil remurgen att madentvaion). e. mance-based contract with a private firm designed to *Addirional financial resources are made available. secure funding for road maintenance while creating incen- * Greater clarity is achieved in the analysis of the tisk associated wirh investments. tives for cost-effectiveness and promptness in the execution of works and routine maintenance activities. Since 1997, 61 However, the results have not always been positive. It is CREMA contracts have been awarded in the country, cov- clear that the applicability of this model is limited due to ering 12,000 kilometers, with an average cost of high-traffic volume required to be financially sustainable, US$11,000 per kilometer per year (including rehabilitation even with government guarantees or subsidies. The process and maintenance). Box 6.2 presents the key features of this of concessioning requires careful design and execution to new instrument. The CREMA model is also being imple- ensure that the expected benefits are realized. Furthermore, mented in Chile, Colombia, Guatemala, Paraguay, and due to the long periods of time required for the recovery of Uruguay. Box 6.2 Key Feature offte Con de Recuperaci6n y Maute mb (CRE-A) A ctivities include:0: C 0 :ontract icombiines in1 0 i 0 00-tia works 00 0000 0000000000000 activities, 0 signals, 00;drainage, st u-: .-'','-'g''.=of th ortr Ad- - ies inslbilx ff ;Mi+i#@i;f:Vf:or the aci&veinSenr of. predeter . contractor.E sW Award: Contract isawarded throug rai0 0nes: usually upIn tq 45 0percent0 00 competitive bidn to00 the0 t * 1 tlowe:;st the Forexeuino nitia wok dot- bdd 0 Wt?I Xi i [ (2\ : \equal edE motl paymets during Road, agnc rle: Deegte oprtoa reposbii 96 DECENTRALIZING ROADS IMATCHING ACCOUNTABILITY, RESOURCES, AND TECHNICAL EXPERTISE Delegation of National Roads deterioration through a combination of strategies that has As the capacity of local agencies improves and the national included institutional strengthening, concessions and ser- mechanisms for the collection of road-user charges and the vice contracts to the private sector, increased budgets and allocation of funds across networks become more reliable, dedicated trust funds, and decentralization and delegation national authorities have also considered the delegation of of responsibilities to subnational governments. The most management tasks to regional agencies under performance- striking result has been the visible improvement in several based contracts with well-defined sources of financing. For countries of the conditions of the national (primary) road example, Argentina's DNV was reorganized in 1992 and network. Although it may be difficult to attribute this given a mandate to gradually involve the provinces in the result to decentralization, it does appear that decentraliza- operation and maintenance of the national network. Since tion has played a role in the process by allowing central 1995, routine maintenance of about 7,000 kilometers (25 agencies to concentrate human and financial resources on a percent of the national network) has been delegated to the more limited network of national roads. Just being able to provinces through two-year, activity-based and unit-price- draw a political firewall between the national and local net- paid contracts. To a lesser extent, a similar experience is tak- works seems to provide benefits that are reinforced through ing place in Brazil, where the authorities are discussing the the institutional reforms of national road agencies and, in merits of completely delegating the management of the particular, through the move away from force account to federal network to the states. These experiences, however, private contracting. have come after the states (or provinces) developed a repu- The evidence of the impact of decentralization on state tation for effective management of their regional networks. and local roads, however, is less clear. It is probably prema- The experience of the region strongly suggests that consid- ture to draw any conclusions at this point, because the erations about thc decentralization or delegation of the reforms are less than a decade old, and the development of national network should really take place after the decen- local governance and technical and managerial capabilities tralized management of regional and municipal networks at the local level takes time. In those countries with more has been completed consolidated. developed forms of local governance, road sector decentral- ization has been more successful. The danger is that inade- Conclusion quate design of decentralization strategies in other coun- It is clear that although the concept of decentralization is tries will lead to frustrations and disincentives that could accepted as a means of increasing road users' control over jeopardize the whole process. road management decisions, in practice, the results have From the experience in Latin America and elsewhere in varied widely. Much of the variation can be attributed to the world, a number of preconditions can be established as the fact that the main impetus for decentralization has critical to the success of a road decentraliiation program: come from the broader macro and political democratic ini- tiatives that were not specifically designed for any one sec- "Decentralization of road management can take place tor in particular. As a result, the transport sector has had to only where there is an adequate level of local gover- adapt to the various systems that have evolved. There are nance in terms of legal, financial, and community par- countries that have begun to show positive results (for ticipation. example, Argentina, Brazil, and Colombia), and there are *When there is little tradition of decentralized gover- countries that have had to seriously reconsider and reverse nance, a gradual approach to road decentralization course. (Bolivia and Peru). These cases, however, offer a set should be followed, transferring first responsibilities of lessons on which to build future initiatives as decentral- for regional and urban municipal roads, while retain- ization spreads throughout the region. ing management of the national network at the There has been a marked change in the attitudes of most national level. national governments toward the importance of maintain- *When decentralizing road responsibilities, the desig- ing their road systems as well as the costs of neglecting nation/classification of road functions should be clearly them. In the past decade, Latin American countries have related to the assignment of political responsibility for pioneered a wide range of efforts to address road network the roads. Municipal roads should be placed under 97 BEYOND THE CENTER: DECENTRALIZING THE STATE municipal governments, state roads under state govern- reform within a time frame that takes into account the ments, and national roads under national governments. effort required to put these preconditions in place. Within *Similarly, the management responsibilities associ- countries, the capabilities of local governments, whether ated with the roads to be decentralized (planning, states, provinces, cities, or municipalities, will vary consid- construction, operation and maintenance, and regula- erably. Experience shows that it is possible to move more tory policy) must be clearly designated, avoiding over- aggressively in certain localities, while taking a longer-term laps among the responsibilities of different levels of strategy for those localities that require support in any of the government. key areas. *The mechanisms and framework for ensuring ade- The most difficult but essential elements, however, are quate financial resources should also be aligned with creating the financing mechanism and designing the the devolution of responsibilities. They should be national-to-local resource transfer mechanisms. As has been structured to reflect local commitment to road main- said earlier, the difficulty arises from the indirect nature of tenance and development, and to allow for cross- the user-charge instruments for raising road revenues, most regional redistribution of resources in order to main- of which accrue to the national government and are then tain national standards and network connectivity, transferred through various approaches to local govern- * Special attention should be given to the assessment of ments. Moreover, the design of the transfer mechanism technical capabilities of the receiving institutions, and must be consistent with the overall transfer of resources the time frame for devolution should recognize the under the decentralized regimes, many of which have been need for the transfer of technology and the dynamics criticized for not being sufficient or equitable and are sel- of capacity development. dom related to road planning and management. This chap- * A system of monitoring, with physical and financial ter has identified several new instruments that are directed accountability, should be put in place, in order to have at making such transfers more transparent, and that incor- sufficiently accurate information about the progress porate the views of the users in a framework that holds road made and the difficulties encountered, so that correc- agencies at the national and local level accountable to the tions can be implemented promptly. public. The challenge will be to see how these various ini- tiatives perform under the strains of macroeconomic and The process of road decentralization should include for- political instability. mulating an implementation strategy to carry out the 98 References The word processed describes informally reproduced works Boeninger, Edgardo. 1991. "Governance and Develop- that may not be commonly available through libraries. ment: Issues and Constraints." Proceedings of the World Bank Annual Conference on Development Economics, Abrantes, Alexandre, and Alberto Diaz Legaspe. 1999. 267-287. "Contracting Public Health Care Services in Latin Bomfim, Antulio, and Anwar Shah. 1994. "Macroeconomic America." Paper Presented at the First Europe and Latin Management and the Division of Powers in Brazil: Per- America and the Caribbean Forum on Health Sector spectives for the 1990s." WorldDevelopment 22 (4):535-42. Reform. Paris, May 10-12. Draft. Bryk, Anthony, David Kerbow, and Sharon Rollow. 1997. Aedo, Cristian. 1998. "Difference in Schools and Student "Chicago School Reform." In D. Ravitch and J. Viteritti, Performance in Chile." In William D. Savedoff, ed., eds., New Schoolsfor a New Centuiry: The Redesign of Urban Organization Matters: Agency Problems in Health and Edu- Education. New Haven: Yale University Press, 164-200. cation. Washington, D.C.: Inter-American Development Bryk, A. S., Y M. Thum, J. Q. Easton, and S. Luppescu. Bank. 1998. "Examining Productivity: Ten-Year Trends in the Ahmad, Etisham, and Katherine Baer. 1997. "Colombia." Chicago Public Schools." Chicago: Consortium on In T. Ter-Minassian, ed., Fiscal Federalism: Theory and Chicago School Research, March. Practice. Washington, D.C.: International Monetary Fund. . 1997. "Assessing School Productivity Using Stu- Alesina, Alberto, and Roberto Perotti. 1998. "Economic dent Achievement: The Chicago Public Elementary Risk and Political Risk in Fiscal Unions." The Economic Schools". Chicago: Consortium on Chicago School Journal, (108):989-1008. Research. Processed. Arcias, Gustavo, and Humberto Belli. 1998. "Rebuilding Carey, John M. 1997. "Institutional Designs and Party Sys- the Social Contract: School Autonomy in Nicaragua." tems." In Larry Diamond, Marc E Plattner, Yun-han LCSHD Paper Series No. 40. Department of Human Chu, and Hung-Mao Tien, eds., Consolidating the Third Development, Latin America, and the Carribean Region. Wave Democracies. Baltimore: The Johns Hopkins Uni- Washington D.C.: The World Bank. versity Press. Barzelay, Michael. 1992. Breaking Through Bureaucracy: A Castefieda, Tarsicio. 1997a. "Health Sector Reforms in New Way for Managing in Government. Berkeley, CA: Uni- Columbia: Issues, Options and Lessons." Document pre- versity of California Press. pared for HDD of the World Bank, Washington, D.C. Bird, Richard M., and Ariel Fiszbein. 1996. "Fiscal Decen- .1997b. "Health Sector Reforms in Brazil: Decen- tralization in Colombia: The Central Role of the Central tralization to States and Municipalities." Document pre- Government." Washington, D.C.: World Bank. pared for HDD of the World Bank. Washington, D.C. Processed. - 1997c. "Health Sector Reforms in Chile: 99 BEYOND THE CENTER: DECE\TRALIZING THE STATE Deconcentration of Hospital Services and Decentraliza- Dillinger, William, and Steven B. Webb. 1999a. "Decen- tion of Primary Health Care." Document prepared for tralization and Fiscal Management in Colombia." Wash- HDD of the World Bank. Washington, D.C. ington, D.C.: World Bank. Processed. Catholic University (Pontifica Universidad Cat6lica de . 1999b. "Fiscal Management in Federal Democra- Chile). 1998. Informe Final: Evaluaci6n de la Lined de cies: Argentina and Brazil." Washington, D.C.: World Proyectos de Mejoramiento Educativo (PME). Convenio Bank. Processed. Ministerio de Educaci6n-Facultad de Educaci6n. Dominguez, Jorge. 1994. Authoritarian and Democratic August. Santiago, Chile. Regimes in Latin America. New York: Garland Pub. CEPAL (Comisi6n Econ6mica para America Latina y el Elster, Jon. 1995. "The Impact of Constitutions on Eco- Caribe). 1998. "La Descentralizaci6n de la Educaci6n y nomic Performance." Proceedings of the World Bank la Salud: Un Analisis Comparativo de la Experiencia Annual Conference on Development Economics 1994. Latinoamericana." Chile. Washington, D.C. Clavijo, Sergio. 1998. Termitas en la Estructura del Gasto Espinola, Viola. 1997. "Descentralizaci6n del sistema Publico y la Sostenibilidad Fiscal en Columbia: Educativo en Chile: Impacto en la gestion de las escue- 1990-2002. CEDE (Centro de Estudios sobre Desar- las." Informe No. 10. May. Department of Human rollo Econ6mico). Universidad de los Andes, Bogota, Development, Latin America, and the Caribbean Regi- Colombia. ion. Washington, D.C.: The World Bank. Creemers, Bert P M. 1994. The Effective Classroom. New Fiske, Edward B. 1996. "Decentralization of Education- York: Cassell, Politics and Consensus." World Bank, Directions in Creemers, Bert P M., Tony Peters, and Dave Reynolds, eds. Development Series. Washington, D.C. 1989. School Effectiveness and School Improvement: Proceed- Fornasari, Francesca, Steven B. Webb, and Heng-Fu Zou. ings of the Second International Congress. Rotterdam: Inter- 1999. "The Macroeconomic Impact of Decentralized national Congress for School Effectiveness. Spending Deficits: International Evidence." Latin- Dahl, Robert A. 1994. "A Democratic Dilemma: System American and Caribbean Sector, Poverty Reduction, Effectiveness versus Citizen Participation." Political Sci- Development Economic Research Group. Washington, ence Quarterly, 109 (1). D.C.: World Bank. Processed. Diamond, Larry. 1997. "Introduction: In Search of Consol- Fuller, Bruce, and Magdalena Rivarola. 1998. "Nicaragua's idation." In Larry Diamond, Marc F. Plattner, Yun-han Experiment to Decentralize Schools: Views of Parents, Chu, and Hung-Mao Tien, eds., Consolidating the Third Teachers, and Directors." World Bank Development Wave Democracies. Baltimore: The Johns Hopkins Uni- Economics Research Group, Working Paper Series on versity Press. Impact Evaluation of Education Reforms, No. 5, Febru- Dalin, Per, with Tekle Ayono, Anbesu Biazen, Birhanu ary. Washington, D.C. Dibaba, Mumtaz Jahan, Matthew B. Miles, and Carlos Galeotti, Gianluigi. 1991. "Decentralization and Political Rojas. 1994. How Schools Improve: An International Report. Rents." In David King, ed., Local Government Economics in New York: Cassell. Theory and Practice. London: Routledge. Darling-Hammond, Linda. 1997. The Right to Learn: A Garman, Christopher, and others. 1998. "The Politics of Blueprint for Creating Schools that Work. San Francisco: Decentralization in Latin America. Processed. Washing- Jossey-Bass Publishers. ton, D.C.: The World Bank. di Gropello, Emanuela. 1999. "Descentralizaci6n de la Gaynor, Cathy. 1998. "Decentralization of Education- Educaci6n en America latina: Un Analisis Comparativo Teacher Management." World Bank Directions in (Version preliminar)." Chile: Comisi6n Econ6mica para Development Series. Washington, D.C. America Latina y el Caribe (CEPAL). Germani, G., and Kalman Silvert. 1994. "Politics, Social Dillinger, William. 1995. Better Urban Services: Finding the Structure, and Military Intervention." In Jorge Right Incentives. Washington, D.C.: World Bank. Dominguez, ed., Authoritarian and Democratic Regimes in . 1997. "The Brazilian Debt Crisis." Washington, Latin America. New York: Garland. D.C.: World Bank. Processed. Gershberg, Alec Ian. 1999a. "Decentralization, Citizen Par- 100 REFER,N C,ES ticipation, and the Role of the State: The Autonomous Partnership for Educational Revitalization in the Americas. Schools Program in Nicaragua." Latin American Perspec- Hanushek, Eric A. 1994. Making Schools Work: Improving tives, forthcoming. Pesformance and Controlling Costs. Washington, D.C.: The . 1999b. "Fostering Effective Parental Participation Brookings Institution. in Education: Lessons from a Comparison of Reform Harding, April, and Alexander Preker. 1999. "Innovations Processes in Nicaragua and Mexico. World Development in Health Care Delivery: Organizational Reforms within 27 (4). the Public Sector. Chapter I: Conceptual Framework." . 1998a. "Decentralization and Re-centralization: Processed. Washington, D.C.: The World Bank. Lessons from the Social Sectors in Mexico and Harral, Clell, and Faiz, Asif. 1988. "Road Deterioration in Nicaragua." Inter American Development Bank, Office Developing Countries. Causes and Remedies." Policy of the Chief Economist, Working Paper No. 379. Wash- Studies, World Bank, June. Washington, D.C. ington, D.C Hartmann, Luis Felipe. 1999. "Decentralizing Health: The . 1998b. "Decentralization, Re-centralization, and Experiences and Policy Lessons in Bolivia." Washington, Performance Accountability: Building an Operationally D.C.: World Bank. Processed. Useful Framework for Analysis." Development Policy Haslam, M. Bruce. 1997. "How to Rebuild a Local Profes- Review 16 (4):405-31. sional Development Infrastructure." Getting Better by . 1995. "Fiscal Decentralization and Intergovern- Design Series No. 4. Arlington, VA: New American mental Relations: An Analysis of Federal Versus State Schools. Education Finance in Mexico." Review of Urban and Heggie, Ian, and Piers Vickers. 1998. Commnercial Manage- Regional Development Studies 7 (2): 199-142. ment and Financing of Roads. World Bank Technical Paper, GFS (Government Finance Statistics Yearbook). 1991. Washing- No. 409, May. Washington, D.C. ton, D.C.: International Monetary Fund. Henevald, Ward, and Helen Craig. 1996. Schools Count. Gonzalez-Prieto, Guillermo, and Manuel Alvarez. 1999. World Bank Project Designs and the Quality of Primary Edu- "Descentralizaci6n en Salud en la Republica Argentina. cation in Sub-Saharan Africa. Washington, DC: World Washington, D.C.: World Bank. Processed. Bank. Guedes, A., A. L. Amaral, R. Walker, and T. Lobo. 1995. Hirschman, Alberto. 1970. Exit, Voice and Loyalty. Cam- "Brazil-Decentralized Management of Education in bridge, MA: Harvard University Press. Minas Gerais." World Bank, Education Sector, July. Humplick, Frannie, and Azadeh Moini-Araghi. 1996a. Washington, D.C. "Decentralized Structures for Providing Roads. A Cross- Gyamfi, Peter, and Ruan, Guillermo. 1996. "Road Mainte- Country Comparison." Policy Research Working Paper nance by Contract: Dissemination of Good Practice in No. 1658, World Bank, September. Washington, D.C. Latin America and the Caribbean Region." Regional . 1996b. "Is There an Optimal Structure for Decen- Studies Program, Report No. 44, World Bank, Decem- tralized Provision of Roads?" World Bank Policy ber. Washington, D.C. Research Working Paper No. 1657. Washington, D.C. Haggard, Stephen, and others. 1995. The Political Economy IDB (Inter-American Development Bank). 1994. "Special of Democratic Transitions. Princeton: Princeton University Report: Fiscal Decentralization." In Economzic and Social Press. Progress in Latin America: 1994 Report. Washington, D.C. Hagopian, Frances, and Scott Mainwaring. 1987. "Democ- . 1997. Latin America after a Decadle of Reforms: Eco- racy in Brazil: Problems and Prospects." World Polity nonzic and Social Progress in Latin America 1997 Report. Journal. 4 (3): 485-514. Washington, D.C. Ham, Chris, and Loraine Hawkins. 1999. "Innovations in Inman, Robert P, and Daniel L. Rubenfeld. 1997. Rethink- Health Care Delivery: Reform within the Public Sector. ing Federalism. "Journal of Economic Perspectives 11(4): 43- A Framework for Implementation. Processed. Washing- 64. ton, D.C.: The World Bank. Javed Burki, Shahid, and Guillermo E. Perry. 1998. Beyond Hanson, E. Mark. 1997. "Educational Decentralization: the Washington Consensus-Institiltions Mattes: Washing- Issues and Challenges." Occasional Paper No. 9. Chile: ton, D.C.: World Bank. 101 BEYOND THE CENTER: DECENTRALIZING THE STATE Jimenez, Emmanuel, and Yasuyuki Sawada. 1998. "Do 95." In Potter, D., and others, Democratization. New Community-Managed Schools Work? An Evaluation of York: Polity Press. El Salvador's EDUCO Program." World Bank, Develop- Litvack, Jennie, Junaid Ahmad, and Richard Bird. 1998. ment Research Group, Working Paper Series on Impact Rethinking Decentralization in Developing Countries. Evaluation of Education Reforms, No. 8, February. Washington. D.C.: World Bank. Washington, D.C. Macias, Claudia, and Marisa Valdes. 1999. "Decentraliza- Jones, Mark P, Pablo Sanguinetti, and Mariano Tommasi. tion of the Health Sector in Mexico." Washington, D.C.: "Politics, Institutions, and Public Sector Spending in the World Bank. Processed. Argentine Provinces." Forthcoming in James Poterba McEwan, Patrick, and Martin Carnoy. 1998. "Public and Juergen von Hagen, eds., Buidget Institutions and Fis- Investment or Private Schools? A Reconstruction of cal Outcomes. Chicago: National Bureau of Economic Educational Improvements in Child." Processed. Wash- Research and the University of Chicago Press. ington, D.C.: The World Bank. Karl, Terry Lynn. 1996. "Dilemmas of Democratization in McKinnon, Ronald. 1997. "Market-preserving Federalism Latin America." In Roderic Ai Camp, ed., Democracy in in the American Monetary Union." In M. Blejer and T. Latin America. Wilmington: SR Books. Ter-Minassian, eds., Macroeconomic Dimensions of Puiblic Kellor, Eileen M., and Allan Odden. 1999. "Cincinnati: A Finance: Essays in Honor of Vito Tanzi. London: Routledge. Case Study Design of a School-Based Performance Mohrman, Susan Anlbers, and Priscilla Wohlstetter, eds. Award Program." Paper presented to the American Edu- 1994. School-Based Management: Organizing for High Per- cation Finance Association, Seattle, WA, March 18-2 1. formance. San Francisco: Jossey-Bass. King, David. 1984. Fiscal Tiers: The Economics of Multi-Level Musgrove, Philip. 1999. "Public Spending in Health Care: Government. London: George Allen and Unwin. How are Different Criteria Related?" Washington, D.C.: King, Elizabeth, and Berk Ozler. 1998. "What's Decen- World Bank. Processed. tralization Got to do with Learning? The Case of Musgrave, R., and P Musgrave. 1984. Public Finance in The- Nicaragua's School Autonomy Reform." World Bank, ory and Practice. New York: McGraw Hill. Development Economics Research Group, Working Navarro, Juan Carlos, and Rafael de la Cruz. 1998. "Fed- Paper Series on Impact Evaluation of Education eral, State and Nonprofit Schools in Venezuela." In Reforms, No. 9, June. Washington, D.C. William D. Savedoff, ed., Organization Matters. Agency King, Elizabeth, Laura Rawlings, Berk Ozler, Patricia Problems in Health and Education. Washington, D.C.: Callejas, Nora Gordon, and Nora Mayorga. 1996. Inter-American Development Bank. "Nicaragua's School Autonomy Reform: A First Look." Nellis, John. 1991. "The Reform of Public Sector Manage- World Bank, Working Paper Series on Impact Evalua- ment: Lessons from Experience." World Bank Policy and rion of Education Reforms, No. 1, October. Washing- Research Series, No. 18. Washington, D.C. ton, D.C. Nickson, R. Andrew. 1995. Local Government in Latin Amer- Ladd, Helen, ed. 1996. Holding Schools Accountable.' Perfor- ica. Boulder: Lynne Rienner Publishers. mance-based Reform in Education. Washington, D.C.: North, Douglass. 1990. Institutions, Institutional Change and Brookings Institution. Economic Performance. Cambridge: Cambridge University Lawler, Edward E. 1992. The Ultirmate Advantage. Creating Press. the High Involvement Organization. San Francisco: Jossey- Oates, Wallace E. 1972. Federal Fiscalism. New York: Har- Bass. court, Brace, and Javonovich. Linz, Juan J., and Alfred Stepan. 1997. "Toward Consoli- O'Donnell, Guillermo, Philippe C. Schmitter, and Laurence dated Democracies." In Larry Diamond, Marc F. Plat- Whitehead. 1986. Transitions from Authoritarian Rule: tner, Yun-han Chu, and Hung-Mao Tien, eds., Consoli- Prospects for Democracy. Baltimore: The Johns Hopkins clating the Third Wave Democracies. Baltimore: The Johns University Press. Hopkins University Press. OECD (Organization for Economic Cooperation and Lipset, S.M. 1983. Political Man. London: Heinemann. Development). 1997. Managing Across Levelf of Govern- Little, W 1997. "Democratization in Latin America, 1980- ment. Paris. 102 RF F ERh-\N,ES 1998. Education at a Glance-OECD Indicators. "Selection and Evaluation of Locally-Developed Versus Paris. Externally-Developed School-wide Projects." The Uni- Ostrom, Elinor, L. Schroeder, and S. Wynne. 1993. Institu- versity of Memphis, September. Memphis, TN: Center tional Incentives and Suistainable Development. Boulder: for Research in Educational Policy. Processed. Westview Press. Ross, Steven M., William L. Sanders, S. Paul Wright, and Over, Mead, and Naoko Watanabe. 1999. "Evaluating the Samuel Stringfield. 1998. "The Memphis Restructuring Impact of Organizational Reforms in Hospitals." Initiative: Achievement Results for Years 1 and 2 on the Processed. Washington, D.C.: The World Bank. Tennessee Value-Added Assessment System (TVAAS)." Paes de Barros, Ricardo, and Rosane Silva Pinto de Men- The University of Memphis, June. Memphis, TN: Cen- don,a. 1998. "The Impact of Three Institutional Inno- ter for Research in Educational Policy, vations in Brazilian Education." In William D. Savedoff, Roubini, Nouriel, and Jeffrey D. Sachs. 1989. "Political and ed., Organization Matters: Agency Problems in Health and Economic Determinants of Budget Deficits in the Indus- Edocation. Washington, D.C.: Inter-American Develop- trial Democracies." Ezuropean Economic Revieuw 33: ment Bank. 903-938. Paul, Samuel. 1991. "Strengthening Public Service Ruiz, Meir, and Guissani. 1997. La Descentralizacion y Accountability: A Conceptual Framework." Washing- Financiamientrio de los Servicios cde Salztd y Eduicacion en ton, D.C.: World Bank. Bolivia. Serie Reformas de Politica Publica No. 48 Perry, Guillermo, and Jorge A. Rodriguez. 1991. "Las CEPAL, Santiago No. 1997. Finanzas Intergubermentales en la Constitucion de Sanguinetti, Pablo, and Mariano Tommasi. 1997. "Los 1991." Coyuntura Economica 21 (4):65-83. Determinantes Econ6micos, Institucionales y Politicos Potter, D. 1993. "Democratization in Asia. In D. Hold, ed., del Comportamiento fiscal de las Provincias Argentinas: Prospects for Democracy. Palo Alto: Stanford University 1983-1996." Washington, D.C.: Inter-American Devel- Press. opment Bank. PREAL (Partnership for Educational Revitalization in the Savedoff, William, ed. 1998. Organization Matters: Agency Americas). 1998. The Futznre at Stake: Report of the Task Problems in Health and Education. Washington, DC: Inter- Force on Education, Equity and Economic Competitiveness in American Development Bank. Latin America and the Caribbean." Chile. Schmelkes, Sylvia. 1996. The Quality of Primacy Eduication: A Preker, Alexander, April Harding, and Navin Girishankar. Case Studdy of Puebla. Mexico. Paris: UNESCO, Interna- 1999. "The Economics of Public and Private Parricipa- tional Institute for Educational Planning. tion in Health Care: New Insights from Institutional Shipps, Dorothy, Joe Kahne, and Mark Smylie. 1997. Economics." Washington, D.C.: World Bank. Processed. "Been Down So Long, It Looks Like Up to Me: Decen- Prud'homme, Remy. 1995. "The Dangers of Decentraliza- tralization in the Chicago Public Schools, 1988-1997." tion." The World Bank Research Observer, 10 (2):201-220. Paper presented to the annual meeting of The University Przeworski, Adam. 1995. "Comment on 'The Impact of Council for Educational Administration, Orlando, FL, Constitutions on Economic Performance,' by Elster." October 31-November 2. Proceedings of the Annual World Bank Conference on Skidmore, Thomas E., and Peter H. Smith. 1997. Modern Development Economics 1994. Washington, D.C. Latin Amzerica. New York: Oxford University Press. Qian, Yingyi, and Barry R. Weingast 1997. "Federalism as Slavin, Robert E. 1996. "Salas de Clase Efectivas, Escuelas a Commitment to Preserving Market Incentives."Jour- Efectivas." Occasional Paper No. 3. Chile: Partnership nal of Economic Perspectives. 11 (4):83-92. for Educational Revitalization in the Americas. Riker, William H. 1964. Federalism. Origin, Operation, Sig- Smith, B. C. 1996. "Sustainable Local Democracy" Public nificance. Boston: Little, Brown and Company. Administration and Development, 16 (1) :163-178. Rojas, Fernando. 1998. "LAC Regional Study on Decen- Stein, Ernesto. 1997. "Fiscal Decentralization and Govern- tralization." Background Paper for World Development ment Size in Latin America." In K. Fugasaku and R. Report 1999/2000, September. Washington, D.C. Hausman, eds., Democracy, Decentralization and Defi cits in Ross, Steven M., Marty Alberg, and John Nunnery. 1998. Latin America. Paris: Organization for Economic Cooper- 103 BEYOND THE CENTER: DECENTRALIZING THE STATE ation and Development. erations: Issues for Policy and Research." Policy Research Stepan, Alfred. 1997. Toward a Comparative Theory of Feder- Working Paper No. 1884. Public Economics, Develop- alism. All Soul's College, Oxford University Press. ment Research Group. Washington, D.C.: World Bank. Sterman, Luiza, and others. 1999. "A Descentraliza,cao do Processed. Sistema de Saucde no Brasil. Nuicleo de Investigacao em Willis, Eliza, Christopher Garman, and Stephan Haggard. Servi,os e Sistemas de Saude do Instituto de Saucde." Sec- 1997. "The Politics of Decentralization in Latin Amer- retaria de Estado da Sauide de Sao Paulo. Sao Paulo, ica." Department of Political Science, University of Cal- Brazil. ifornia, San Diego, La Jolla, CA Summers, Anita A. and Amy W Johnson. 1996. "The Winkler, Donald R. 1993. "The Design and Administra- Effects of School-Based Management Plans." In Eric A. tion of Intergovernmental Transfers." Regional Studies Hanushek and Dale W Jorgenson, eds., Improving Amer- Program, Report No. 29, World Bank, July. Washing- ica's Schools. The Role of Incentives. Washington, D.C.: ton, D.C. Board on Science, Technology, and Economic Policy, . 1994. "The Design and Administration of Inter- National Research Council. governmental Transfers: Fiscal Decentralization in Latin Summit of the Americas II. 1998. Plan of Action. Santiago, America." World Bank, Discussion Paper, No. 235, July. Chile, April 1-3. Washington, D.C. Talvitie, Antti P 1996. "International Experiences in Wong, Kenneth K., Robert Dreeben, and Laurence E. Restructuring the Road Sector." World Bank, January. Lynn, Jr. 1997. "Integrated Governance as a Reform Washington, D.C. Strategy in the Chicago Public Schools." Office of Edu- Tanzi, Vito. 1996. "Fiscal Federalism and Decentralization: cational Research and Improvement, Department of A Review of Some Efficiency and Macroeconomic Education, Laboratory for Student Success Publication Aspects." Annual World Bank Conference on Develop- Series, Report No. L97-10. ment Economics 1995. Washington, D.C. World Bank. 1994. "El Salvador-Community Education Ter-Minassian, Teresa. 1997. "Brazil." In Teresa Ter-Minass- Strategy: Decentralized School Management." World ian, ed., Fiscal Federalism in Theory and Practice: A Collec- Bank Education Sector, Final Sector Report, No. 13502, tion of Essays. Washington, D.C.: International Monetary December. Washington, D.C. Fund. . 1994. "Venezuela. Health Services Reform Pro- Ter-Minassian, Teresa, and Jon Craig. 1997. "Control of ject." Staff Appraisal Report, No. 13474-VE. Washing- Subnational Government Borrowing." In Teresa Ter- ton, D.C. Minassian, ed., Fiscal Federalism in Theory and Practice: A . 1994. World Development Report 1994. New York: Collection of Essays. Washington, D.C.: International Oxford University Press. Monetary Fund. . 1995a. "Analyzing Capacity of Local Govern- Tiebout, Charles. 1956. "A Pure Theory of Local Expendi- ments: A Case Study of Municipalities in North East tures."Jozernal of Political Economy 64(5): 416-424. Brazil." Draft Report. Washington, D.C. Treisman, Daniel. 1998. "Decentralization and Inflation in . 1995b. "Argentina. Provincial Health Sector Developed and Developing Countries." Department of Development Project." Staff Appraisal Report, No. Economics, University of California, Los Angeles. 14399-AR. Washington, D.C. Processed. . 1996. "Sustainable Transport. Priorities for Policy United States Central Intelligence Agency. U.S. CIA World Reform." Development in Practice Series, May. Wash- Fact Book. 1997. www.odci.gov/cia/publications/fact- ington, D.C. book/country-frame.html. -. 1997. "El Salvador's Community-Managed Weingast, Barry R. 1995. "The Economic Role of Political Schools (EDUCO) Program: A First Report." Develop- Institutions: Market-Preserving Federalism and Eco- ment Research Group, Working Paper Series on nomic Development." ThejJournal of Law, Economics, and Impact Evaluation of Education Reforms, No. 4, July. Organization, 7 (1). Oxford University Press. Washington, D.C. Wildasin, David E. 1997. "Fiscal Aspects of Evolving Fed- . 1998. Argentina: The Fiscal Dimension of the Con- 104 REFERENCES vertibility Plan, A Background Report. Washington, D.C. . 1998. "Colombia: Economic and Social Develop- ment Issues for the Short and Medium Term." Report No. 18394-CO, November. Washington, D.C. . 1999. "Peru Improving Health Care for the Poor." Report No. 18549-PE. Draft. Washington, D.C. . 1999. World Development Report 1999/2000; Enter- ing the 21'X Century: The Changing Development Landscape. New York: Oxford University Press. 105 DISTRIBUTORS CHINA The Middle East Observer INDONESIA Chinese Corporation for Promotion and 41, Sherif Street Pt. Indira Limited OF WrVnORLD BANK Humanities Cairo Jalan Borobudur 20 15, Ding Hui Dong Li, Kun Lan Hotal Tel: (20 2) 392 6919 PO Box 181 GROUP Haidian District 100036 Fax: (20 2) 393 9732 Jakarta 10320 Beijing E-mail: mafouda@meobserver.com.eg Tel: (62 21) 390 4290 PUBLICATIONS Tel: (86 10) 88117711 URL: www.meobserver.com.eg Fax: (62 21) 390 4289 Fax: (86 10) 88129871 Prices and credit terms vary from E-mail: wangjiang99@yahoo.com FINLAND IRAN country to country. Please con- Akateeminen Kirjakauppa Kowkab Publishers sult your local distributor before Ms. Zhou Rui FIN-00101 Helsinki Tehran placing an order. P.O. Box 2825 Tel: (358 9)121 4418 Tel: (9821)2583723 Beijing Fax: (358 9)1214435 Fax: (98 21) 258 3723 E-mail: akatilaus@akateeminen.com E-mail: kowkabpub@tavana.net ARGENTINA China Financial & Economic Publishing URL: www.akateeminen.com World Publications SA House Ketab Sara Co. Publishers Av. Cordoba 1877 8, Da Fo Si Dong Jie FRANCE P.O. Box 15745-733 1120 Buenos Aires Beijing Editions Eska; DJB/Offlib Tehran 15117 Tel: (5411) 815 8156 Tel: (86 10) 6401 7365 12, rue du Quatre-Septembre Tel: (98 21) 8716104 Fax: (5411) 815 8156 Fax: (86 10) 6401 7365 75002 Paris Fax: (98 21) 8712479 E-mail: wpbooks@intovia.com.ar Tel: (331) 42 86 58 88 E-mail: ketab-sara@neda.net.ir COLOMBIA Fax: (331) 42 60 45 35 AUSTRALIA, PAPUA NEW GUINEA, Infoenlace Ltda./An IHS Group Company E-mail: offilib@offilib.fr IRELAND FIJI, SOLOMON ISLANDS, VANUATU, Calle 72 No. 13-23 - Piso 3 URL: www.offilib.fr Government Supplies Agency AND SAMOA Edificlo Nueva Granada 4-5 Harcourt Road DA. Information Services P.O. Box (A.A) 34270 GERMANY Dublin 2 648 Whitehorse Road Santalk de BogotA, D.C. UNO-VERLAG Tel: (353 1) 661 3111 Mitcham 3132 Tel: (571)2558783 Poppelsdorfer Allee 55 Fax: (3531)4752670 Victoria, Australia Fax: (571)2480808 D-53115 Bonn URL: www.opw.ie/govt.htm Tel: (61 3) 9210 7777 E-mail: infoenlace@gaitana.interred.net.co Tel: (49 228) 949020 Fax: (61 3) 9210 7788 Fax: (49 228) 217492 ISRAEL E-mail: service@dadirect.com.au CtTE D'IVOIRE E-mail: unoverlag@aol.com Yozmot Literature Ltd. URL: www.dadirect.com.au Centre d'Edition et de Diffusion Africaines URL: www.uno-verlag.de P.O. Box 56055 (CEDA) 3 Yohanan Hasandlar St. AUSTRIA 04 B.P. 541 GHANA Tel Aviv 61560 Gerold and Co. Abidjan 04 Epp Books Services Tel: (972 3) 5285 397 Weihburggasse 26 Tel: (225) 24 6510 Post Office Box 44 Fax: (972 3) 5285 397 A-i 01 0 Wieno Fax: (225) 25 0567 TUC Tel: (431)51247310 E-mail: info@ceda-ci.com Accra R.O.Y. International Fax: (431)512473129 URL: www.ceda-ci.com Tel: (233 21) 778843 P.O. Box 13056 E-mail: buch@gerold.telecom.at Fax: (233 21) 779099 Tel Aviv 61130 CYPRUS E-mail: epp@africaonline.com.gh Tel: (972 3) 649 9469 BANGLADESH Center for Applied Research Fax: (972 3) 648 6039 Micro Industries Development Assistance 6, Diogenes Street, Engomi GREECE E-mail: royil@netvision.net.il Society (MIDAS) P O. Box 2006 Papasotiriou S.A., URL: www.royint.co.il HouseS5, Road 16 Nicosia International Technical Bookstore Dhanmondi R/Area Tel: (357 2) 59 0730 35, Stournara Str. Palestinian Authority/Middle East Dhaka 1209 Fax: (357 2) 66 2051 106 82 Athens Index Information Services Tel: (880 2) 326427 E-mail: ttzitzim@sting.cycollege.ac.cy Tel: (301() 364 1826 P.O.B. 19502 Jerusalem Fax: (880 2) 811188 Fax: (301) 364 8254 Tel: (972 2) 6271219 CZECH REPUBLIC E-ma,l: gprekas9papasotirIou.gr Fax: (972 2) 6271634 BELGIUM USIS, NIS Prodejna Jean de Lannoy Havelkova 22 HAITI ITALY Av. du Roi 202 130 00 Prague 3 Culture Diffusion Licosa Libreria Commissionaria Sansoni 1060 Brussels Tel: (42 2) 2423 1486 Mr. Yves Cldment Jumelle S.PA. Tel: (32 2) 538 5169 Fax: (42 2) 24231114 5, Rue Capois Via Duca di Calabria 1/1 Fax: (32 2) 538 0841 E-mail: pospisilovaj@usiscr.cz C.P. 257 50125 Firenze E-mail: jean.de.lannoyWinfoboard.be URL: : www.usiscr.cz Port-au-Prince Tel: (39 55) 645 415 Tel: 1509) 23 9260 Fax: (39 55) 641 257 BOSNIA AND HERZEGOVINA DENMARK Fax: (509) 23 4858 E-mail: licosa@ftbcc.it Book Trading Company "Sahinpasic" Samfundslitteratur URL: www.ftbcc.it/licosa Marsala Tita 29/Il Rosenoerns AIle 11 HONG KONG, CHINA; MACAU 71000 Sarajevo DK-1970 Frederiksberg C Asia 2000 Ltd. JAMAICA Tel: (387 71) 64 48 56 Tel: (45 35) 351942 Sales & Circulation Department Ian Randle Publishers Ltd Fax: (387 71) 64 48 56 Fax: (45 35) 35 7822 302 Seabird House 206 Old Hope Road E-mail: tajib@btcsahinpasic.com E-mail: ck@sl.cbs.dk 22-28 Wyndham Street, Central Kingston 6 URL: www.btcsahinpasic.com URL: www.sl.cbs.dk Hong Kong, China Tel: (876) 927 2085 Tel: (852) 2530 1409 Fax: (876) 977 0243 BRAZIL ECUADOR Fax: (852) 2526 1107 E-mail: irpl@colis.com Publicacoes Tecnicas Internacionais Ltda. Libri Mundi-Libreria Internacional E-mail: saleosasia2000.com.hk Rua Peixoto Gomide, 209 Juan Leon Mera 851 URL: wwm.asia2000.com.hk JAPAN 01409 Sao Paulo, SP P.O. Box 17-01-3029 Eastern Book Service (EBS) Tel: (5511) 259 6644 Quito HUNGARY 3-13 Hongo 3-chome, Bunkyo-ku Fax: (55 11) 258 6990 Tel: (5932)521606 Euro Info Service Tokyo 113 E-mail: postmaster@pti.uol.br Fax: (593 2) 504209 Margitszgeti Europa Haz Tel: (81 3) 3818 0861 URL: www.uol.br E-mail: librimul@librimundi.com.ec H-1138 Budapest Fax: (81 3) 3818 0864 Tel: (36 1) 350 80 24; 350 80 25 E-mail: orders@svt-ebs.co.jp CANADA CODEU Fax: (36 1)3509032 URL: www.svt-ebs.co.jp Renouf Publishing Co. Ltd. Ruiz de Castilla 763, Edif. Expocolor E-mail: euroinfo@mail.matav.hu 5369 Canotek Road Primer piso, Of. #2 KENYA Ottawa, Ontario K1J 9J3 Oulto INDIA Legacy Books Tel: (6131 745-2665 Tel: (593 2) 507-383 Allied Puhlishers Ltd. Loita House Fax: (613) 745-7660 Fax: (5932)507-383 751 Mount Road P.O. Box 68077 E-mail: order.dept@renoufbooks.com E-mail: codeu@impsat.net.ec Madras 600 002 Nairobi URL: www.renoufbooks.com Tel: (91 44) 852 3938 Tel: (254 2) 330853 EGYPT, ARAB REPUBLIC OF Fax: (91 44) 852 0649 Fax: (254 2) 330854 Al Ahram Distribution Agency E-mail: allied.mds@smb.sprintrpg.ems. E-mail: legacy@form-net.com Al Galaa Street vsnl.net.in Cairo Tel: (20 2) 578 60 83 Fax: (20 2) 578 68 33 Africa Book Service (E.A.) Ltd. Oasis Official SINGAPORE; TAIWAN, CHINA, THAILAND Mr. Talat Lone P.O. Box 3627 MYANMAR; BRUNEI Central Books Distribution Co. Ltd. Quaran House, Mfangano Street Wellington Hemisphere Publishing Services 306 Silom Road P.O. Box 45245 Tel: (644) 4991551 Golden Wheel Building Bangkok 16500 Nairobi Fax: (64 4) 4991972 41 Kallang Pudding Road, #04-03 Tel: (66 2) 233-6930-9 Tel: (254 2) 223 641 E-mail: oasisWactix gen.nz Singapore 349316 Fax: (66 2) 237-8321 Fax: (254 2) 330 272 URL: www.oasisbooks.co.nz Tel: (65) 741 5166 Fax: (65) 742 9356 TRINIDAD & TOBAGO AND KOREA, REPUBLIC OF NIGERIA E-mail: info@hemisphere.com.sg THE CARIBBEAN Dayang Books Trading Co. University Press Plc Systematics Studies Ltd. International Division Three Crowns Build ng Jericho SLOVENIA St. Augustine Shopping Center 954-22, Bangbae-Dong. Socho-ku Private Mail Bag 5095 Gospodarski vestnik Publishing Group Eastern Main Road Seoul Ibadan Dunajska cesta 5 St. Augustine Tel: (82 2) 582 3588 Tel: (234 22) 411356 1000 Ljubljana Tel: (868) 645 8466 Fax: (82 2) 521 8827 Fax: (234 22) 412056 Tel: (386 61) 133 83 47 Fax: (868) 645 8467 E-mail: dico3@chollian.net E-mail: suike@hotrrail.com Fax: (386 61)133 80 30 E-mail: tobe@trinidad.net E-mal: repansekl9gvestnik.si Eulyoo Publishing Co., Ltd. PAKISTAN URL: www.gvestnik.si/EUROPA/index.htm UGANDA 46-1, Susong-Dong Oxford University Press Gustro Limited Jongro-Gu 5 Bangalore Town, Sharae Faisal SOUTH AFRICA, BOTSWANA P.O. Box 9997 Seoul P.O. Box 13033 For single titles: Madhvani Building Tel: (82 2) 734 3515 Karachi 75350 Oxford University Press Southern Africa Plot 1 6/4. Jinja Road Fax: '82 2) 732 9154 Tel: (92 21) 446307: 449032; 440532 P.O. Box 12119 Kampala E-mail: eulyoo@chollian.net Fax: (92 21) 4547640;449032 Ni City 7463 Tel: (256 41) 251467 E-mail: usmanm@oup.net.pk Cape Towvn Fax: (256 41) 251468 LEBANON Tel: (27 21) 595 4400 E-mail: gus@swiftuganda.com Librairie du Liban Pak Book Corporation Fax: (27 21) 595 4430 P,O. Box 11-9232 Aziz Chambers 21 E-mail: oxford@oup.co.za UKRAINE Beirut Queen's Road LIBRA Publishing House Tel: (961 9) 217 944 Lahore For subscription orders: Ms. Sophia Ghemborovskaya Fax: (961 9) 217 434 Tel: (92 42) 636 3222; 636 0885 International Subscription Service 53/80 Saksahanskoho Str. E-mail: hsageh@cyberia.net.lb Fax: (92 42) 636 2328 P.O. Box 41095 252033. Kiev 33 URL: www.librairie-du-liban.com.lb E-mail: pbc@brain.net.pk Craighall Tel: (7 44) 227 62 77 Johannesburg 2024 Fax: (7 44) 227 62 77 MALAYSIA Mirza Book Agency Tel: (27 11) 880 1448 University of Malaya Cooperative 65, Shahrah-e-Quaid-e-Azam Fax: (2711) 8806248 UNITED KINGDOM Bookshop, Limited Lahore 54000 E-mail: iss@is.co.za Microinfo Ltd. P.O. Box 1127, Jalan Pantai Bars Tel: (92 42) 7353601 P.O. Box 3, Omega Park 59700 Kuala Lumpur Fax: (92 42) 576 3714 SPAIN Alton Tel: (60 3) 7565 000 E-mail: merchant@brain.net.pk Mundi-Prensa Libros, s.a. Hampshire GU34 2 PG Fax: (60 3) 755 4424 Castello 37 Tel: (44 1420) 86 848 E-mal: umkoop@tm,net.my PERU 28001 Madrid Fax: 441420) 89889 Editorial Desarrollo SA Tel: (34 91) 436 37 00 E-mail: wbankrIukminfo.demon.co.uk MEXICO Apartado 3824 Fax: (34 91) 575 39 98 URL: www.microinfo.co.uk INFOTEC Ica 242, OF. 106 E-mail: libreria@mundiprensa.es Av. San Fernando No. 37 Lima 1 URL: www.mundiprensa.es The Stationery Office Col. Toriello Guerra Tel: (51 14) 285 380 51 Nine Elms Lane 14050 Mexico D.F. Fax: (51 14) 286 628 Mundi-Prensa Barcelona London SW8 SDR Tel: 1525) 6242800 Consell de Cent No. 391 Tel: (44 171) 873-8372 Fax: (525)6242822 PHILIPPINES 08009 Barcelona Fax: (44 171) 873-8242 E-mail: infotec@rtn.net.mx International Booksource Center. Inc. Tel: (34 3) 488 3492 E-mail: chris.allen@theso.co.uk URL: www.rtn.net.mx 11 27-A Antipolo St. Fax: (34 3) 487 7659 URL: www.tsonline,co.uk Barangay, Venezuela E-mail: barcelona@mundiprensa.es Mundi-Prensa Mexico, S.A. de C.V. Makati City VENEZUELA c/Rio Panuco, 141 - Colonia Cuauhtemoc Tel: (63 2) 896 6501 SRI LANKA, THE MALDIVES Tecni-Ciencia Libros, S.A. 06503 Mexico DF Fax: (63 2) 896 6497 Lake House Bookshop Sr. Luis Fernando Ramirez, Director Tel: :52 5) 533 56 58 P.O. Box 244 Centro Cuidad Comerc al Tamanaco Fax: (52 5) 514 67 99 POLAND 100, Sir Chittampalam Gardiner Mawatha Nivel C-2 E-mail: 1015245.2361@compuserve.com International Publishing Service Colombo 2 Caracas Ul. Piekna 31/37 Tel: (941)32104 Tel: (58 28 959 5547 NEPAL 00 677 Warsaw Fax: (94 1) 432 104 Fax: (58 2) 959 5636 Everest Media International Services (P.) Tel: (48 2) 628 6089 E-mail: LHL@sri.lanka.net E-mail: IfrgO01@ibm.net Ltd. Fax: (482) 621 7255 GPO Box 5443 E-mail: books%ips@ikp.atm.com.pl SWEDEN ZAMBIA Kathmandu URL: Aww.ipscg.waw.pl/ips/export Wennergren-Williams Informations Service University Bookshop, University of Zambia Tel: (977 1)416026 AB Great East Road Campus Fax: (977 1) 224 431 PORTUGAL P.O. Box 1305 P.O. Box 32379 Livraria Portugal S-171 25 Solna Lusaka NETHERLANDS Apartado 2681 Tel: (46 8) 705 9750 Tel: (260 1) 252576 De Lindeboom/lntemationale Rua Do Carmo 70-74 Fax: (46 8) 27 0071 Fax: (2601) 253952 Publikaties b.v 1200 Lisbon E-mail: mail@wwi.se E-mail: hunene@admir.unza.zm M.A. de Ruyterstraat 20A Tel: (351 t) 347 4982 7482 BZ Haaksbergen Fax: (351 1) 347 0264 SWITZERLAND ZIMBABWE Tel: (31 53) 574 0004 Librarie Payot S.A. Academic and Baobab Books (Pvt.) Ltd. Fax: (31 53) 572 9296 ROMANIA Service Institutionnel 4 Conald Road E-mail: lindeboo@worldonline.nl Compani De Librarii Bucuresti s.a. Cbtes-de-Montbenon 30 Graniteside URL: home.worldonline.nl/-lindeboo Str. Lipscani nr. 26, sector 3 1002 Lausanne P.O. Box 567 Bucharest Tel: (41 21) 341 3229 Harare NEW ZEALAND Tel: (40 1)313 9645 Fax: (41 21)3413235 Tel: (2634) 755 035 EBSCO NZ Ltd. Fax: (401() 312 4000 E-mail: institutionnel@payot-libraire.ch Fax: (263 4 ) 759 052 Private Mail Bag 99914 E-mail: Academic@Africaonline.Co.Zw New Market RUSSIAN FEDERATION ADECO Van Diermen Editions Techniques Auckland lzdatelstvo << Ves Mir >> Ch, de Lacuez 41 Tel: (64 9) 524 8119 Moscow 101831 CH-1807 Blonay Fax: (64 9) 524 8067 Tel: (7 95)917 8749 Tel: (41 21) 943 2673 E-mail: WGent%ess-nz.ebsco@iss. Fax: (7 95)917 9259 Fax: (41 21) 943 3605 ebsco.com E-mail: ozimarin@glasnet.ru URL: www.vesmir.tsx.org THE WORLD BANK 1818 H Street, N.W. Washington, D.C. 20433 USA Telephone: 202-477-1234 Facsimile: 202-477-6391 Telex: MCI 64145 WORLDBANK MCI 248423 WORLDBANK World Wide WVeb: http:IIwww., worldbank.org/ E-mail: books@Ccworldbank.org ISBN 0-8213-4521-4 DESIGN: JEFFREY KIBLER/THE MAGAZINE GROUP/ILLUSTRATION: FERRUCCIO