Report No. 22523-BR Brazil Issues in Fiscal Federalism June 4, 2002 Brazil Country Management Unit PREM Sector Management Unit Latin America and the Caribbean Region Document of the World Bank CURRENCY EQUIVALENTS Currency Unit - Real (R$) EXCHANGE RATE December 1997: R$1.12 US$ January 1999: R$1.21 US$ January 2000: R$1.80 US$ January 2001: R$1.96 US$ January 2002: R$ 2.37 US$ WEIGHTS AND MEASURES Metric System FISCAL YEAR January I - December 31 ABBREVIATIONS AND ACRONYMS BANESPA Sao Paulo State Bank (Banco do Estado de Sao Paulo) CEF Federal Saving Bank (Caixa Economica Federal) CLT Private Sector Labor Code (Consolidacao de Lei do Trdbalo) CMN National Monetary Council (Conselho Monetario Nacional) FGTS Unemployment Insurance Fund (Fundo de Guarantia do Tempo de Servico) FPE State Participation Fund (Fundo de Participacao dos Estados) FPM Municipal Participation Fund (Fundo de Participacao dos Municipios) FUNDEFE Education Fund (Fundo de Manutencao e Desenvolvimento do Ensino Fundamental e de Valorizacao do Magisterio) ICMS value added tax (Imposto Sobre Circulacao de Mercadorias e Servicos) INSS National Social Security Agency (Instituto Nacional de Seguridad Social) IPI Industrial Products Tax (Imposto Sobre Produtos Industrializados) ISS Municipal Services Tax (Imposto Sobre Servicos) LRF Law of Fiscal Responsibility (Lei de Responsibilidade Fiscal) PAB Basic Health Care Package (Piso Assistencial Basico) PSF Family Health Program (Programa de Saude da Familia) RCL Net current revenues (receita corrente liquida) RGPS Social Security System for Private Sector Employees (Regime Geral de Previdencia Social) RJU Public Sector Social Security System (Regime Juridico Unico) STN Federal Treasury Secretariat (Secretaria do Tesouro Nacional) SUS National Health Insurance System (Sistema Unico de Saude) TFG Health Care Ceiling (Teto Financeiro Globao Vice President LCR: David de Ferranti Director LCC5C: Vinod Thomas Director LCSPR: Ernesto May Lead Economist: Joachim von Amsberg Sector Manager: Ron Myers Task Manager: William Dillinger Foreword This report is based on the findings of a mission to Brazil in January, 2001. The report was prepared by Mr. William Dillinger. It was produced under the supervision of Gobind T. Nankani, Director, and Joachim von Amsberg, Lead Economist. The peer reviewers for this task were Jennie Litvack, Shahrokh Fardoust, and Robert Ebel. This report is based on extensive documentation produced by Brazilian scholars, including Fernando Luiz Abrucio, Jose Roberto Afonso, Rui de Britto Affonso, Ricardo Varsano, and Fernando Rezende. It also benefited from interviews with federal government officials (including Mssrs. Murilo Portugal (IMF), Fabio Barbosa (STN), Jorge Khalil (STN), and Carlos Eduardo de Freitas (Banco Central) as well as experts outside of government, including Mr. Mailson de Nobrega, and officials of state governments too numerous to mention. While this report has been discussed with institutions and individuals of the Brazilian Government, the views expressed in this report are exclusively those of the World Bank. Table of Contents EXECUTIVE SUMMARY ..................................... I 1. INTRODUCTION .....................................1 HISTORIC BACKGROUND .....................................2 OVERVIEW OF THE CURRENT STRUCTURE .....................................4 2. THE IMMEDIATE ISSUES ....................................6 SUBNATIONAL DEBT ....................................6 A Legacy of Debt Crises ....6...............................6 The New System of Fiscal Controls .....................................9 A Market-based Approach ................................... 19 PERSONNEL REGULATIONS AND PENSION REFORM ................................... 22 3. LONGER TERM REFORMS ................................... 25 EXPENDITURE ASSIGNMENT ................................... 25 Health care ................................... 32 Primary education .................................... 33 Water supply ................................... 34 REVENUE ASSIGNMENT .................................... 36 Tax Assignment .................................... 36 Transfer System ................................... 37 Macroeconomic impacts .................................... 46 PRIORITIES ................................... 49 ExECUTIVE SUMMARY i. With nearly half of public sector expenditure under their control and a dominant provider of education, -health care, infrastructure and public security, subnational governments are an important component of the Brazilian public sector. In the recent past, subnational deficits have threatened the stability of the macro economy. Subnational governments are now in the frontlines of 'second generation' reforms in management and public service delivery. ii. Brazilian federalism is not now in crisis. Brazil has a long experience with multi- party federalism, and the fundamental terms of the 'federal bargain' are well established. The principle of subnational political independence is undisputed. State threats to withdraw from the Union or withdraw from the national revenue sharing system are unknown. Nevertheless the federal system has persistent problems, with adverse implications for macroeconomic stability, poverty reduction, and efficiency in the delivery of public services. It is an opportune time to address them. Having achieved macro economic stability, Brazil is embarked on a wide range of reforms in public sector institutions. Reforms in the structure of intergovernmental relations should be a part of this agenda. Short Term Reforms iii. The immediate issue is fiscal. Over the last decade, state governments have shown a propensity to run deficits, and then seek-and obtain-federal bailouts. During the 1990's, Brazil experienced three major state debt crises; the last of which was large enough to threaten Brazil's macroeconomic stability. The federal government has responded by enacting a series of controls on subnational borrowing. The most recent, the Law of Fiscal Responsibility (LRF), not only imposes limits on borrowing, but attempts to change the budgetary practices that create the need to borrow in the first place. The net effect of the new system of controls to make the most overt forms of excessive borrowing extremely risky for the individuals responsible, and fiscally disadvantageous for the jurisdictions they serve. Together with federal controls on the banking sector, the new regime is likely to forestall subnational debt crisis in the near term. iv. But it faces longer term hurdles. One is the administrative burden that enforcement will impose, particularly on the courts and the legislatures. The imposition of fines and imprisonment require judicial hearings, which may strain an already overburdened court system. Impeachment requires a lengthy process of investigation and trial by state assemblies and municipal councils. The new system also places a heavy administrative burden on the Federal Treasury. v. There is also a risk that the federal government's political commitment to enforcing the restrictions will recede. Attempts to control subnational fiscal behavior have a mixed track record in Brazil. The explosion of state debt in the 1990's, for example, was facilitated by federal intermediaries. But there are some signs that this risk may be declining. The globalization of financial markets has increased international pressure on the federal government to maintain a hard budget constraint with respect to subnational governments. Because growth in subnational deficits undermines investor confidence, the federal government is under pressure to enforce the new debt control system, if only to keep the foreign investment flowing. Political support for enforcement of the fiscal rules may also have increased. There is evidence that Brazil has now developed a large, educated middle class, which derives its livelihood from the private economy (as opposed to the public sector, which was the mainstay of the middle class twenty years ago). Thus voters and opinion leaders may be less susceptible to populism and less supportive of the overextended state than they were twenty years ago. vi. Market-based .lending. Nevertheless, in the long run there is a case for shifting the system of subnational debt control from one that depends on central regulation to one that relies more on markets. There are several institutional models for doing so. They include bond markets (most common in the U.S.) and specialized banks (used in Europe). If the market model is to prevail in Brazil, several changes in the credit environment must occur. First, private long term funds must become available, at interest rates that are compatible with the returns to infrastructure investment. Brazil's long history of macro instability has made private lenders wary of long term commitments of any kind. In addition, the federal government's immense need for short term credit to finance its budget deficit has tended to crowd other potential borrowers--including subnational governments--out of the market. Continued macroeconomic stability and declining federal deficits will be required before the market model can be fully implemented. vii. Second, private lenders must have a level playing field. Historically, Government banks have provided the majority of financing for subnational capital works. Because they have access to forced savings schemes, they have been able to offer credit on below- market terms. This has discouraged potential private lenders from entering the market. Limitations on subsidized government lending may be necessary in order to attract private sector interest. viii. Third, the federal government will have to refrain from extending implicit guarantees on private loans to subnational governments. The federal government has a history of coming to the relief of over-indebted subnational governments. This has occasionally encouraged private lenders to over-extend credit, secure in the belief that the federal government will make good on any defaults. The federal government will have to disabuse them of this notion. "No bail-out" statements will not be sufficient. Instead, federal government will have to establish a pattern of non-interference in subnational defaults to private banks. ix. Finally, reforms will be required to remove the obstacles that prevent subnational governments from becoming--and remaining-creditworthy. Chief among these are various Constitutional restrictions that prevent subnational governments from controlling their wage bills. This topic is discussed below. x. Pensions. The second challenge to the present system of fiscal controls comes from the threat of growing state pension liabilities. Brazil's Constitution guarantees iii generous benefits to civil servants at all three levels of government, including a provision that retirement benefits will be set at 100% of exit salaries. Most pensions are paid out of general revenues. As the subnational governments' labor force ages and life expectancies increase, the burden of pension liabilities has increased. Rising pension obligations are on a collision course with the deficit controls imposed by the LRF. When the two collide, subnational governments will be forced to choose between violating the LRF or reneging on their Constitutional obligations to retirees. xi. To address this problem, recent changes in the Constitution have toughened the criteria for retirement Some states have also increased mandatory employee pension contributions and have used the proceeds from privatization to capitalize pension funds. States also now have the authority to hire new staff under private sector, rather than civil service, law. These measures, while helpful, are not likely to be sufficient. Ultimately, a means must be found to reduce the level of benefits. xii. Reducing the "acquired rights" of existing staff has proven extremely controversial in Brazil. Experience in other countries suggest a strategy. Government pension reforms normally distinguish between three populations: (1) existing retirees; (2) staff who are hired after the new rules go into effect, and (3) existing active staff. Existing retirees are normally protected in their existing benefits. Newly hired staff are subject to the new contribution and benefit parameters. In dealing with the existing active staff, governments typically adopt a transition rule, in which staff who are close to retirement are fully protected, but those who are more recently hired are not. In Brazil, the existing legislation guarantees full benefits to all existing staff, regardless of how long they have worked. A change in rules is needed, that would establish a cut off date for full benefits or sliding reductions in benefits depending on length of service. This would protect the benefits of staff who are close to retirement, while allowing the states to achieve critical savings on the pension benefits of staff who have joined more recently. Longer Term Reforms xiii. In the longer term, Brazil needs to confront structural problems in its intergovernmental relationship. What is most striking is the lack of accountability that arises from the absence of any clear definition of the responsibilities between states and municipal governments. Like Brazil's earlier democratic constitutions, the 1988 Constitution makes a general attempt to define the functional responsibilities of each tier of government. Among the responsibilities assigned to the federal government are national defense and social security, emission of currency, control of public debt, regulation of interstate and foreign trade, and the power to establish the "general norms of public employment." The Constitution also delineates certain concurrent responsibilities of the federal government and the states, including tax legislation, education, and social assistance, specifying that federal law will be limited to "general norms" but will prevail in case of conflict with state legislation. The federal Constitution grants the states "all powers not otherwise prohibited to them by this Constitution". This would appear to be a grant of plenipotentiary power, limited only by the powers specifically assigned to the federal government. But the Brazilian municipios also have a iv broad Constitutional mandate. They are assigned "the power to legislate over subjects of local interest" and to provide "services of local public interest". xiv. While fiscal constraints at the federal level now limit the reach of the federal government, the division of functions between states and municipal governments remains ambiguous. Either level is free to enter any field that it deems of public interest, while neglecting those it does not. With no clearly defined roles, neither states nor municipal governments can be held accountable for shortfalls in specific services. xv. This ambiguity is compounded a peculiarity of Brazilian federalism: the 'sovereignty' that was granted to municipios under Brazil 1988 Constitution. Unlike the constitutions of other federal countries, which typically define municipal governments as creatures of their respective states, the Brazilian Constitution establishes municipal governments as a separate, third tier of government. In principle, states therefore cannot compel or prohibit actions by the municipalities within their jurisdictions. Subnational government in Brazil thus consists of two equally "sovereign" levels of government operating in the same geographical space. xvi. Some critics argue that functional ambiguity is a necessary part of Brazilian federalism; that no rigid definition of municipal functions could cope with the diversity of Brazilian municipalities, and the small size of many of them. But these characteristics exist throughout the world. They have not prevented countries-particularly in Europe and North America-from assigning responsibilities to particular units of government. To overcome the constraints imposed by size, small municipalities can contract out- either with private firms (as is the case in France) or with higher levels of government (as is the case in Germany and some parts of the U.S.). Small municipalities can also form joint services districts. Yet another approach is to distinguish among classes of municipalities. Brazil has no tradition of using organizational distinctions to reflect the widely differencing circumstances of local governments. The country has only one form of local government: the municipio. The legislation that applies to an urban megalopolis such as Sao Paulo applies to a sparsely inhabited municipio in the Sertdo. Other countries-the U.K. and U.S. for example-permit legislative distinctions among classes of municipalities, tailoring their mandates to their circumstances. Brazil should consider doing the same. xvii. Efforts to define the functional responsibilities of municipal governments are in fact already underway on a sectoral basis. Health sector reforms in the 1990's established a distinction between municipios that are authorized to manage federal health funding directly, and those where federal health funds are allocated by the state (or in some cases by the federal government). More recent reforms aim at organizing groups of municipalities into referral hierarchies, focused on a single 'headquarters' municipality. Reforms in education-while less organizationally ambitious--are nevertheless forcing states and municipalities to pool their education spending and allocate it on a formula basis between state and municipal primary schools. But this may not be enough. In the long run, Brazil may need to consider changes in the structure of municipal government itself. v xviii. Reforms in Intergovernmental Transfers. Reforms in intergovernmental transfers and tax assignments also merit consideration. Brazil now employs an extensive system of revenue sharing to support state governments in poorer regions and municipal governments throughout the country. In administrative terms, the system works well. The transfers are formula based, and are made accurately and promptly. The scale and distribution of the transfer system is questionable, however. The principal federal transfers-the fundos de participacao-represent a considerable proportion federal expenditure. While the portion going to states is effective in targeting poor states, it is not immediately obvious that it is effective in reaching poor people. Funds are not earmarked and there is a ample scope for leakage to higher income groups. xix. The need for extensive transfer systems to support municipal governments is also subject to question. With earmarked transfers now available to finance primary education and health, the justification for large scale block grants to municipal governments is unclear. Their principal impact appears to be to suppress local tax effort. The Government might therefore consider scaling back revenue sharing in favor of a system that relies on more specific grants to finance functions with important distributional implications, while relying more on local taxes to finance services of more localized benefit. xx. Macro economic impacts offiscal federalism. The major increase in revenue sharing mandated by the 1988 Constitution, coupled with the Government's inability to offload a corresponding volume of expenditure responsibilities, caused some critics to fear for Brazil's macroeconomic stability. It appeared that this 'lop sided' decentralization would inevitably result in growing deficits at the federal level. This did not occur. The federal government managed to offset the growth in Constitutionally mandated transfers, largely by increasing taxes and social security contributions. The net macroeconomic effect of the 1988 reforms was therefore to increase the aggregate tax burden in Brazil. xxi. But Brazil's federal system may still constrain the federal government's ability to use fiscal policy as an instrument of macroeconomic management. Roughly one quarter of all taxes are collected by subnational governments and are therefore outside the direct control of the federal government. Half of the taxes collected by the federal government must be shared with subnational governments. Those taxes that are fully under the control of the federal government tend to be distortionary. As a result, the federal government faces a Hobbesian choice: if it increases its less distortionary taxes, half the proceeds must be shared with the states and municipios. If it increases its distortionary taxes, it aggravates their adverse economic impacts. Proposals to reform the tax system are currently under consideration. These are largely aimed at removing distortions in the tax system. In this respect, they merit serious attention. xxii. Priorities. Brazil's federal system does not require a drastic overhaul. The fundamnental terms of the 'federal bargain' are well established and work reasonably well. But it would benefit from repair. In the short term, there are two priorities. First, the Government needs to strictly enforce the new set of restrictions on subnational debt vi imposed in the wake of the recent debt crisis. This is needed not merely to control deficits themselves but to send an important signal to governors and mayors that the era of federal bailouts is at an end. xxiii. Second, the pension benefits of active subnational civil servants must be reduced. While the recent toughening of retirement criteria, combined with increased employee contributions and the use of privatization proceeds to capitalize retirement funds, will help reduce the scale of pension liabilities, no solution will be viable without a reduction in benefits. And without a solution to the pension problem, the prospects for enforcing the new system of debt controls are dim. xxiv. In the longer term, more fundamental change in the structure of subnational government should be considered. Despite substantial reforms over the last fifteen years, Brazil still retains some of the characteristics of an older, 'clientelistic' state at the subnational level. Governors and mayors can still dispense favors-or withhold them- without being held accountable for the performance of specific services. To clarify the assignment of functions, explicit distinctions between categories of municipalities may have to be made. To reduce the arbitrariness of intergovermnental transfers, better targeting and earmarking may be required. Over the long term, revenue sharing may be scaled back in favor of greater reliance on local benefit taxes. xxv. In pursuing such reforms, Brazil has a variety of international examples to draw on. It should also draw on its own recent innovations, particularly in the health and education sectors. Issues in Brazilian Fiscal Federalism 1. INTRODUCTION 1. With nearly half of public sector expenditure under their control and a dominant provider of education, health care, infrastructure, and public security, subnational governments are an important component of the Brazilian public sector. In the recent past, subnational deficits have threatened the stability of the macro economy. Subnational governments are now in the frontlines of 'second generation' reforms in management and public service delivery. But the position of subnational governments within Brazil's federal system has been the subject of ongoing controversy since the republic was founded. Like all federalisms, Brazil confronts tensions between national interests and* interests of individual states. The 'rules' of Brazilian federalism are therefore constantly subject to adjustment. Wns?s,t,~EUEL D 4-; l.4 2 2. Brazilian federalism is not now in crisis. Unlike Mexico or Russia, Brazil has a long experience with multi-party federalism, and the fundamental terms of the 'federal bargain' are well established. The principle of subnational political independence is well established--unlike in PRI-dominated Mexico or India, where federal supercession (i.e., the dismissal of local elected officials and their replacement by federal appointees) has been common. The revenue sharing relationship between tiers of government is transparent and predictable (unlike in Argentina). State threats to withhold taxes (as in Russia and China) or to withdraw from the national revenue sharing system (as in Mexico) are unknown. 3. Nevertheless the federal system has persistent problems, with adverse implications for macroeconomic stability, poverty reduction, and efficiency in the delivery of public services. In the short term, the most striking is the propensity of subnational governments to run deficits, and then seek-and obtain-federal bailouts. While recent legislation attempts to restrain subnational borrowing, it does not sufficiently address its underlying cause: conflicts between unfunded mandates imposed by the federal constitution, and constraints on subnational revenues. 4. In the longer term, more fundamental problems in the federal relationship need to be addressed. The first is the lack of accountability that arises from the lack of a clear role for municipal governments. Second is the ineffectiveness of revenue sharing in addressing poverty. Third is the economic costs of an inefficient tax structure. Now appears to be an opportune time to address these issues. Having achieved macro economic stability, Brazil is embarked on a wide range of reforms in public sector institutions. Many of the problems with the existing federal structure are the subject of legislation currently under debate. 5. This report has two principal audiences. The first is the Bank, where it is intended to provide background for the Bank's subnational lending operations and national level policy dialogue. The second is the Brazilian government, where it is intended to synthesize the wide range of recent work on Brazilian federalism (by Brazilians and Brazilianists) and provide an external view of priority issues and directions for reform. HISTORIC BACKGROUND 6. Brazil is a territorially vast country which has experienced cycles of centralism and decentralism over its history. The roots of federalism can be traced back at least as far as the 19dh century monarchy, which maintained Brazil's territorial integrity by conceding considerable autonomy to regional elites. This strategy was institutionalized with the founding of Brazil as a federal republic in 1889.1 The new Constitution transformed the existing provinces into states, authorized the election of state governors and legislatures and called for the establishment of separate state constitutions. It severely 1 Dias, Jose Luciano de Mattos, 1991 A Federajdo Brasileira: Estructura e Desempenho, PhD dissertation, unpublished 3 limited federal taxing power, while granting major revenue sources to the states, and restricting federal government intervention in state affairs. 7. The first era of Brazilian democracy saw the capture of the central government by the political elites of the two most important states--Minas Gerais and Sao Paulo-which agreed to rotate the presidency between them. In 1930, a breakdown in the agreement provided an opening for a new political actor, Getulio Vargas, to seize power. His 1934 Constitution, vastly increased the power of the federal government. Vargas consolidated his power in an auto-coup in 1937, establishing an authoritarian state with a Constitution that explicitly broke the power of the states. Seeking greater political and administrative unity, Vargas appointed "intervenors" to replace popularly elected state governors. In 1939, state administrations were decreed to be administrative subdivisions of the federal government. In a symbolic act, Vargas publicly burned the state flags.2 8. Vargas was forced out of office in 1945. A democratic constitution was adopted in 1946, which resuscitated state governments, while leaving intact the federal powers established in the 1934 Constitution. This democratic period lasted nearly twenty years. It was brought down in 1964 by an economic and political crisis and an increasingly assertive, professionalized army. 9. Military rule in Brazil lasted from 1964 to 1985. Under the military, the forms of federalism were maintained intact, but emptied of their political content. Although state governors and the mayors of capital cities ceased to be directly elected, there were regular elections for Congress, state assemblies and city councils. Mayors of non-capital cities also continued to be directly elected. The military nevertheless maintained control over the political system through its ability to issue decree-laws (thus bypassing Congress) and its manipulation of the rules of political participation. This ensured military control in both Congress and most state assemblies throughout the military era.3 10. Ironically, the military era saw a rapid expansion in the resources and responsibilities of the states. In 1966, the military created Brazil's first system of revenue sharing4, thus ensuring the states--particularly in the poor Northeast--of a large and buoyant revenue source. The 1966 tax reform also converted the states' former sales tax into a modern value added tax, which went on to become the largest single source of tax revenue in Brazil. The state's administrative machinery also expanded under the military regime. As part of the military's modernization drive, states were encouraged to buy out private water and power utilities and consolidate them into state-level operating companies. These went on to become some of the largest public utilities in Latin America. 2 Souza, Celina; 1996; "Redemocratization and Decentralization in Brazil: the Strength of Member States" in Development and Change, Vol 27 (1996), Oxford, Blackwell Publishers 3 Lamounier, Bernard; 1990; '"Opening through Elections: Will the Brazilian Case Become a Paradigm?" in Politics in Developing Countries: Comparing Experiences with Democracy, edited by Larry Diamond, Juan J. Linz, Seymour Martin Lipset Boulder, Colo.; L. Rienner Publishers 4A earlier attempt had been made in 1961, but never functioned effectively. 4 11. In 1979, the military government announced its decision to surrender dictatorial powers. This was followed by a loosening of controls on the formation of political parties. In 1982, direct elections for governors were instituted. This was followed, in 1984, by indirect presidential elections. In 1985, direct elections were held for the mayors of capital cities. In 1986, a new Congress was elected with Constitution-making authority. In 1989, the first direct election for president was held. OVERVIEW OF THE CURRENT STRUCTURE 12. The 1988 Constitution establishes a federal structure of government, consisting of the federal government, 26 states (and a federal district with status of a state), and an undefined number of municipalities (now about 5,500). The federal executive branch is headed by a president, who is directly elected for a four year term. The legislative branch consists of two houses. The first is the Chamber of Deputies, consisting of 513 members, with the number of delegates per state determined on the basis of population (subject to a floor of 8 and a ceiling of 70). The second is the Senate, which is comprised of three senators from each state. 13.. State governments are headed by directly-elected governors. States have unicameral legislatures, whose members are elected at large by proportional representation. This structure is repeated at the municipal level, where mayors are directly elected and council members are elected at large by proportional representation. 14. Expenditure Assignment. Like Brazil's earlier democratic constitutions, the 1988 document explicitly reserves certain powers for the federal government, while providing a broad and general mandate to states and municipal governments. Among the responsibilities exclusively assigned to the federal government are national defense and social security, emission of currency, control of public debt, regulation of interstate and foreign trade, and the power to establish the "general norms of public employment." The Constitution also delineates certain concurrent responsibilities of the federal government and the states, including tax legislation, education, and social assistance, specifying that federal law will be limited to "general norms" but will prevail in case of conflict with state legislation. 15. States and municipalities have broad Constitutional mandates. States are granted "all powers not otherwise prohibited to them by this Constitution."5 Municipios are assigned "the power to legislate over subjects of local interest" and to provide "services of local public interest." Municipalities have a peculiar political status. Unlike other federal constitutions, which typically define municipal governments as creatures of their respective states, the 1988 Constitution establishes municipal government as a third tier of government with the same constitutional status as the states. States therefore cannot compel or prohibit actions by the municipalities within their jurisdictions. 16. Revenue assignment. In contrast to its vagueness in dividing expenditure responsibilities between levels of government, the Constitution is very explicit in 5Constituiqdo da Republica Federativa do Brasil. 5 dividing up revenues. It assigns specific tax bases to each level of government and creates a system of tax sharing which substantially redistributes revenue between each tier of government. The federal government derives virtually all its revenue from income, payroll, and turnover taxes (the latter two earmarked for social security). State governments are assigned a value added tax, which accounts for the majority of state revenue in the wealthier states of the southern part of the country. Poorer states, and most municipalities, derive the majority of their revenue from formula- based intergovernmental transfers. 17. Regulation. While the 1988 Constitution was largely decentralist in spirit, it does restrict subnational autonomy in some respects. The most important is personnel. The federal constitution defines the "rights" of public sector employees at all three levels of government, limiting subnational governments' ability to dismiss redundant staff or reduce salaries, and mandating expensive pension benefits. It also gives the federal senate the authority to control all subnational borrowing. 18. Electoral Rules. A key part of post-military Brazilian federalism-albeit an unwritten one--is embodied in the political relationship between state governors, members of congress, and the president. Brazil's political system is characterized by a high degree of party fragmentation and weak party discipline at the national level. Brazilian federal deputies are elected through a system of open list proportional representation.6 Candidates for the Chamber of Deputies run at large within each state (rather than facing off within individual districts). Except in the parties of the left, party discipline is weak. Brazilian politicians change their party affiliation frequently and bestow little power on national party organizations. To implement their programs, presidents must construct coalitions that satisfy the gamut of regional and local interests that are represented in Congress. But with comparatively weak party discipline and loyalty, these coalitions are loose and shifting rather than hard and fast.8 While individual candidates are free to assemble their own political bases, they often prefer to associate their campaigns with better-known political leaders. These tend to be govemors rather than presidents. Presidential candidates are often seen a virtual foreigners. Governors tend to be better known 'locals' who have developed a stronger and broader clientelistic network in their states. Moreover, governors control the perks--the public works and 6Under Brazil's system of proportional representation system, each state is a single electoral district, with the number of seats based on population, subject to the ceilings and floors noted in the text. Candidates or parties compete at large throughout the state. The number of seats won by each party in the state is based on its percentage of the total vote. In Brazil's open list proportional representation system, voters can choose among individual candidates within a party list, with the candidates winning the most votes occupying the seats won by the party. 7Ames, Barry, 1995; "Electoral Rules, Constituency Pressures, and the Pork Barrel: Bases of Voting in the Brazilian Congress" in The Journal of Politics, Vol. 57 No 2; University of Texas Press; Austin Texas Mainwaring, Scott; 1997, "Multipartism, Robust Federalism and Presidentialism in Brazil" in Presidentialism and Democracy edited by Scott Mainwaring, Matthew Soberg and Shugart Cambridge; New York: Cambridge University Press 6 jobs--that will enable the congressional candidate to get reelected.9 Sitting state governors thus command the loyalty of federal deputies, and can thwart or facilitate presidential designs. 19. The dispersed nature of political power in Brazil-and the 'politics of the governors' in particular--have been blamed for some of the recent problems in intergovernmental relations in Brazil. Critics of the system have argued that it grants disproportionate power to regional interests at the expense of national ones. State delegations can block legislation that would adversely affect them, in return for supporting other key items of the president's agenda.'0 But such simple causal relationships are hard to prove. As Argentina's recent difficulties attest, other countries with far more rigid political systems have also had their difficulties with subnational governments. Nevertheless, as Brazil moves toward a more tightly regulated system of intergovernmental relations, it is useful to keep in mind that it is the political will to enforce the rules as much as the political will to enact them that determines whether this new regime will be effective. 2. THE IMMEDIATE ISSUES SUBNATIONAL DEBT A Legacy of Debt Crises 20. Brazil is presently recovering from a severe subnational debt crisis that threatened to derail the national economic stabilization plan in the 1990's. While the crisis has abated, it is not clear that the measures taken to avoid a recurrence will stand the test of time. 21. Subnational governments have traditionally borrowed from a variety of sources. During the 1970's and 1980's, they borrowed from foreign multi-laterals, foreign commercial banks, and from specialized federal banking institutions (BNH, later CEF) to finance infrastructure investment. They borrowed by running up arrears to suppliers and contractors, and by running up arrears on salaries. These were often refinanced, through the issuance of bonds, which were underwritten by the states' commercial banks, and 9 Abrucio, Fernando and Claudio Couto; 1996 0 Impasse da Federag&o Brasileira; Cadernos CEDEC No 58 Centro de Estudos de Cultura Contemporanea: Sao Paulo 0 Burki, Shahid; Guillermo Perry and William Dillinger, 1999, Beyond the Center Decentralizing the State: Washington D.C.; World Bank 7 ultimately sold to private banks and investors. In addition, Brazil's largest state, Sao Paulo, borrowed directly from its own commercial bank, BANESPA." 22. Since the departure of the military, there have been three state debt crisis. The first was a legacy of the international debt crisis of the 1980's, when states--along with the federal government--ceased servicing their debt to foreign creditors. Once an agreement was reached between the federal government and the creditors at the national level, the Govermment attempted to induce the states to resume servicing their debt. In 1989, the federal government agreed to transform the accumulated state arrears and remaining principal into a single debt to the federal Treasury. US$ 19 billion was rescheduled under these terms. 23. The second crisis involved debt owed by the states to federal financial institutions. This was resolved in 1993 through a rescheduling of roughly US$ 28 billion of such debt. Again, the debt was transformed into debt to the federal Treasury. To close on this second agreement, the federal government conceded an escape clause. If the ratio of state debt service to revenue rose above a threshold fixed by the Senate, the excess could be deferred and capitalized into the outstanding stock of debt. The two agreements substantially reduced states' debt service obligations in cash terms. With principal rescheduled and debt service subject to a ceiling, the immediate burden of servicing debt was considerably reduced. The agreements, however, also established an unfortunate precedent. It created the perception that the federal government was prepared to provide debt relief to any state that required it. 24. With each debt workout, the federal government made an attempt to tighten the regulations on state borrowing. After the second debt crisis, the federal government prohibited itself from extending new loans to states currently in default. The Constitution was amended to allow the federal government to deduct debt service from intergovernmental transfers These federal regulations were not sufficient to forestall the most recent debt crisis. 25. The most recent-and largest-debt crisis was the result of the confluence of two factors: the personnel controls imposed by the 1988 Constitution and the Government's successful efforts to bring inflation under control. Prior to 1988, state staff could be employed under either of two legal regimes. The first--the statutory regime--conferred a wide range of civil service benefits and rights, including a protection against dismissal (except for cause) and pension benefits equal to 100% of exit salaries (termed the salario integral). The' second--the consolidated labor law (CLT) regime--allowed for dismissal without cause (although it established compensation requirements). The pensions of statutory staff were paid directly from state treasuries. State pension obligations to CLT staff were limited to a 21% payroll contribution to the national social security system (INSS). Until recently, twenty five of the 27 states--including the federal district-- owned commercial banks. While only Sao Paulo was a major direct borrower from its bank, many states, induced their banks to lend to favored clients, resulting in large volumes of non-perforning assets, and growing off-budget liabilities for the states that owned them. 8 26. The 1988 Constitution altered the picture in two important respects. First, it required states (along with all other government bodies) to adopt a single regime for their employees. In effect, states were required to absorb former CLTistas into the statutory regime, with all the benefits and rights pertaining thereto. Second, it expanded civil service benefits, in particular prohibiting states from lowering salaries. While these provisions have been subsequently modified, they continue to impose a burden-an unfunded mandate-on subnational governments. 27. During the initial post-Constitution years, the states managed to weather the personnel costs imposed by the 1988 Constitution through the use of inflation. Although the Constitution prohibited nominal cuts in salaries it provided no guarantee of real wages, which-in the absence of frequent nominal increases-fell rapidly during the high inflation of the period.'2 In 1994, this strategy became obsolete. A new stabilization plan, the Plano Real, was introduced in mid-1994 and had immediate and remarkable success. Annual inflation fell from 929 percent in 1994 to 22 percent in 1995 and nine percent in 1996. The plan, however, removed a past mechanism of state internal financial control: the ability to reduce real salaries and pensions via inflation. Without inflation, personnel costs soared. In the eight years prior to the Plano Real personnel costs had averaged 40% of net current revenues, with very little year-to-year variation. By 1998, that proportion had risen to 67%, with ratios as high as 85% in Rio Grande do Sul and 77% in Minas Gerais. 28. As their finances became increasingly precarious, states resorted to deferring debt service, by rolling over debts at maturity and capitalizing interest. Eventually, the private market declined to hold state debt even on the overnight market. At this point, the states sought relief from the federal government. In this, they had a sympathetic partner in the form of the federal Congress, which authorized the Central Bank to make a market in state bonds, exchanging them for federal bonds. Under these domestic bond exchange agreements, the Senate had the authority to determine the proportion of the bonds that would have to be liquidated at maturity. The Senate used this authority liberally, authorizing 100% rollovers not only of principal but of accumulated interest. Sao Paulo pursued a similar strategy with respect to its debt to BANESPA. It ceased servicing its debt to the bank, rolling over principal at maturity and allowing interest to capitalize to the extent that it soon became the bank's principal asset. 29. With interest capitalizing at real rates of over 20%, the stock of state debt grew explosively. The stock of bonds grew by Rs$ 12 billion between 1994 and 1995, and another Rs$ 8.5 billion in the following year. At the end of 1996, the total stock of state (and municipal) bond debt stood at Rs$ 52 billion. The heavy interest obligations on this growing stock of debt, combined with the states' inability to reduce personnel costs or raise revenues, resulted in deficits that were large enough to have macroeconomic consequences. 2 The shift of CLTistas to statutory status also generated cost-savings in the short run, as states no longer had to contribute to the INSS. They did, however, acquire the obligation to pay pensions out of revenue once the forner CLTistas retired. 9 30. Federal negotiations with individual states began in mid-1995. It was not until December 1997 that the first major debtor state--Sao Paulo--signed a binding agreement with the federal government.13 The other debtor states followed over the following nine months. While the terms of the agreements vary among states, they generally followed the pattern of the two previous debt agreements. Debt was rescheduled (rather than written off) and a debt service ceiling was imposed, allowing debt service above the threshold to be capitalized into the stock of debt. The principal innovation of the new debt agreements was a large interest rate subsidy. Rather than bearing the existing rate on federal bonds, the federal government agreed to impose a fixed real interest rate of six percent. The macroeconomic impact of these agreements was rather limited, however. Although the agreements lowered the interest rates paid by the states, the federal government continued to be the states' principal creditor and continued to pay the overnight rate at the marginal cost of borrowing funds. As a result, the agreements did not reduce the aggregate interest cost paid by the public sector. They merely shifted more of it explicitly onto the federal treasury. The terms of the agreements, moreover, were not sufficient to forestall the capitalization of interest on debt owed to the federal government. As shown in figure 1, state debt continued to grow. Figure 1 Trends In State Debt The New System of 250.0 * extemal debt Fiscal Controls 200.0 El .[ other domestic 200.0 - ~~ other domestIc ~31. Lei 9496 Targets eJ 150.0 _* bank debt To forestall such a 100.0- Lei 7976/Aviso 3C 5. Lei8727 crises in the future, the 50o. *Le 82 federal government 0.0 -_3a Lei 9696/PROES enacted a battery of 1997 1998 1999 2000 0 bond (net) controls on state fiscal ___________________________ behavior. The first was a direct product of recent state debt negotiations. As a condition of debt relief, each state was required to agree to a package of adjustment targets. As specified in Law 9496/1997, these include (1) scheduled declines in the debt:revenue ratio, (2) increases in the primary balance, (3) limits on personnel spending, (4) growth in own-source revenues, (5) ceilings on investments and (6) a list of state enterprises to be privatized or concessioned. Targets are specified as a percent of revenues and are adjusted annually. 14.15 The targets vary among states, according (in part) to initial conditions. For example, as shown in Table 1, Sao Paulo's current agreement calls for it to achieve a primary surplus of 8% of revenues 3 Sao Paulo's debt to BANESPA was also included in its refinancing package. 14 Revenues (receita corrente liquida) are defined, for purposes of Lei 9496, as receipts in the previous twelve months, excluding those arising from borrowing operations, sales of assets, discretionary transfers, and in the case of states, those received for legal or constitutional transfer to municipios. 15 180 municipios also benefited from the recent debt rescheduling, of which the municipios of Sao Paulo and Rio de Janeiro were the largest. The municipal agreements do not include fiscal targets. 10 in 2001. Goais must achieve a surplus of 15%. At the same time, Sao Paulo must achieve a debt: revenue ratio of 2.16:1 by 2010. Goias is committed to a more lenient figure: 3.44:1. Table 1 Fiscal adjustment targets selected states As % of RLR Sao Paulo Goias debt: in 2010 as % RLR 2.16 3.44 Primary balance, 2001 (Rs bn) 8% 15% maximum personnel spending, 2001 as % 61.7% 60% RLR growth in own source revenue (increase 3.7% 2.1 2001/2000) 1 1 ceiling on investment (% RLR) 10.5% X 12% Sources: Govemo do Estado de Sao Paulo, Programa de Reestructuracao e de Ajuste Fiscal do Estado de Sao Paulo 2000-2002; Estado de Goias, Programa Apoio a Reestructuracao e de Ajuste Fiscal do Estado de Goais 2000-2002 32. In principle, the Government has two instruments to enforce compliance. First it can deny federal guarantees on new state borrowing. This would largely restrict state borrowing from multilateral banks, as domestic borrowing does not normally require a federal guarantee. Second, it can impose an interest penalty on the rescheduled debt. Rather than the six percent real interest rate imposed under the Lei 9496 agreement, a state in violation of its targets would be charged a rate based on the federal government's current cost of funds (currently about nine percent in real terms) plus one percent. In addition, the proportion of revenues to be paid as debt service would be raised br four percentage points. These remedies would be enforced through deduction at source. 6 33. Senate Resolution. The Senate, under the Brazilian Constitution, has the authority to approve or reject any subnational borrowing proposal, and has traditionally maintained a (frequently-revised) resolution that establish criteria for making this decision. The latest resolution is considerably stricter than its predecessors. The current version, Senate Resolutions 40 and 43, consist of a broad set of controls on subnational demand for credit, designed to control all forms of subnational borrowing-any form of contract that implies payment at a future date. They do so through several different stipulations. First, they impose a ceiling on new borrowing, debt service, and debt stock. New borrowing is prohibited if: (1) the total volume of borrowing in the budget year exceeds 16% of net current revenue (RLR); (2) debt service exceeds 11.5% of RLR; or (3) the total stock exceeds two times RLR in the case of states, and 1.2 times RLR in the case of municipalities. (Each level of government has 15 years to achieve the latter targets.) Note 16 Because the states were required to pledge their VAT and federal revenue sharing as collateral as a condition of debt refinancing, the federal government can deduct penalty interest from intergovernmental transfers or garnish the principal source of state tax revenues. 11 that unlike past Senate debt restrictions, these calculations are made on the basis of projected revenues and expenditures, rather than performance over the previous twelve months. While in principle this can be a more accurate means of assessing creditworthiness, it also increases the amount of discretion involved in making the calculation. 34. The resolutions also prohibit specific kinds of borrowing. They forbids the emission of bonds by states or municipalities through 2010, except to finance rollovers of principal on existing bonds and to finance court judgments (precatorios). They explicitly prohibit borrowing to refinance arrears on existing debt. Resolutions 40 and 43 also single out and prohibit certain practices that have been used to evade federal regulations in the past: borrowing from decentralized entities and assuming obligations to suppliers in the form of promissory notes, credit card debt, etc. Finally, they forbid borrowing in violation of debt reduction schedules imposed under Lei 9496. Table 2 Senate Debt Regulation Resolution 4043 35. CMN Resolutions. In addition States municip io to these controls on the demand for new borrowing/RLR <16% <16% credit, the federal government has debt service/RLR <11.5% <11.5% also acted to restrict its supply. The stock of debt/RLR <2* 1.2* passage of new borrowing * target to be achieved in increments of 1/15 restrictions was accompanied by a resolution of the National Monetary Council imposing a complementary set of restrictions to be imposed by the Central Bank (Res. 2653). CMN resolution 2653 authorizes the Central Bank, in its capacity as supervisor of the domestic banking system, to control the supply of credit to subnational governments by domestic banks. Under the resolution, outstanding loans to the public sector may not exceed 45% of any bank's equity. "Public sector' in this case includes the federal government, states, and municipios (as well as the public enterprises controlled by them). "Bank" means both private and public banks. 36. When it was first promulgated, this resolution was particularly binding on the principal source of long term credit to subnational government: the national savings bank, Caixa Economica. Because the Caixa Economica was already close its maximum level, the 45% limit on bank exposure virtually froze its net lending at its existing level. The recent recapitalization of CEF has given the bank additional room to lend. In addition, Caixa does much of its lending to subnational government as an agent of the national unemployment insurance fund, FGTS. As such its lending is not subject to the CMN 2653 ceiling. BNDES, similarly, has additional room to lend to subnational governments. Because BNDES lends through private banks, its principal constraint, at present, is a lack of private banks willing to accept the risk of doing so. 37. In addition, the CMN resolution specifically prohibits banks from lending to any state or municipio in violation of the Senate debt ceilings. It also prohibits banks from lending to any public sector entity that (1) is in default to any other bank or (2) has disputes with the public sector debt registry (CADIP). Finally, it prohibits banks from refinancing any form of arrears to suppliers or contractors. Any bank that extends credit 12 in violation of these rules is required to deposit (within 15 days of the violation's discovery) an amount equivalent to the loan in a non-interest bearing account at the Central Bank, where it is to remain until the violation is corrected. 38. Privatization of state banks: Another measure to limit the supply of credit was the closure of state commercial banks. As noted earlier, Sao Paulo was the only state that borrowed directly from its bank on a significant scale. Although few of the other state banks were direct sources of state credit, they nevertheless facilitated state borrowing, by underwriting state bond issues. They were also a major source of off-budget liabilities. Due to a history of politically motivated lending and lax management, most were insolvent. (The state bank of Rio de Janeiro, for example, was found to have a negative net worth in excess of US$ 8 billion.) At the start of the crisis, there were 33 deposit- taking state banks in 25 states (including the federal district). Under the federal government PROES program, all but five of the banks have been privatized, put under federal control pending privatization, or transformed into development agencies (which are not permitted to take deposits.) The remaining five consist of the state banks of Espirito Santo, Para, Rio Grande do Sul, Sergipe, and the smaller of Sao Paulo's commercial banks, Nossa Caixa/Nosso Banco.) 39. LRF. This series of controls and enforcement measures was capped, in May 2000, by the Law of Fiscal Responsibility (Lei de Responsibilidade Fiscal-LRF) and its companion legislation (approved in October, 2000) amending the penal code and laws governing impeachable offenses. The LRF addresses three major areas of finance: budgeting, personnel management and debt. 40. With respect to budgeting, the LRF requires subnational governments to establish, at the outset of each budget exercise, annual targets for revenues, expenditures, the primary balance and changes in the stock of debt17 This is to be accompanied by a report on compliance with the previous year's targets, an evaluation of the state's current financial and actuarial position (including the position of any pension funds) and an evaluation of fiscal risks, including contingent liabilities and the measures that will be taken to address them should they materialize. The annual budget law itself (Lei Orcamentaria Annual) is to be compatible with the state's multi-year budget (required under existing legislation) and must contain an annex demonstrating: (1) its compatibility with federal monetary, credit, and foreign exchange policies, (2) the principal assumptions used in projecting the major budgetary aggregates,18 (3) the fiscal impact of tax exemptions and subsidies, and (4) a contingency fund for debt service. 41. With respect the budget execution, the law requires the executive (governor or mayor) to prepare a monthly cash flow program within 30 days of the publication of the budget. If, within any 2-month period, revenues are incompatible with the annual fiscal 17 Municipios with populations under 50,000 have five years before they are required to comply with this provision. 18 The law includes detailed instructions on revenue estimation (Arts. 11-13) and expenditure justification (Arts. 15-17). 13 targets, each branch of government (i.e., the executive, legislative and judicial branches) must limit new appropriations and disbursements (except for Constitutional obligations and debt service) over the following thirty days so as to bring spending into compliance with fiscal targets. Compliance with the fiscal targets is to be evaluated and made public every four months. 42. With respect to personnel, the law elaborates existing ceilings on personnel costs. Existing legislation sets a global limit on personnel spending for each level of government. The LRF sets separate ceilings for each branch of government at each level. If personnel costs exceed 95% of the limit for any branch, that branch is prohibited from (1) conceding salary increases, (2) creating new positions, or (3) hiring new staff (other than to replace retirees) , and must reduce the excess within eight months. In the interim (and as long as the target is not achieved) the jurisdiction is ineligible for discretionary federal transfers or federal guarantees and is prohibited from contracting new debt. The law furthermore declares that any increase in personnel spending that has not been justified or will take effect within six months of the end of current term is null and void. 43. With respect to debt, the LRF imposes controls on both the existing stock and new borrowing operations. States and municipalities are required to maintain debt stocks below ceilings established by the Senate. If, at the end of any four month period, a state or municipio exceeds its ceiling, it must reduce the excess within one year. In the interim, the jurisdiction is prohibited from any new borrowing (other than bond rollovers), must maintain a primary surplus sufficient to achieve the target within the required time frame, and is ineligible for discretionary transfers. The law declares any borrowing in excess of the Senate ceiling to be null and void and requires the immediate repayment of principal without interest. 44. The LRF prohibits one level of government from lending to another (except through federal banks) and subjects debt rescheduling to the same criteria that are imposed on new borrowing. In effect, this prohibits the federal treasury making loans directly to subnational governments, and sharply limits the federal government's ability to reschedule subnational debt. The law also prohibits governors and mayors from contracting obligations to pay within the last six months of their administrations, unless these can be retired during the remainder of their term. Finally, it authorizes the federal government to garnish intergovernmental transfers or subnational tax receipts as collateral for loans. 45. The LRF also make provisions for transparency. It reiterates the existing Constitutional provision requiring states and municipalities to issue bimonthly cash flow statements and end of year final accounts, and mandates an interim report (relat6rio de gestaofisca) every four months19, demonstrating compliance with ceilings on personnel and debt (including liquidation of revenue anticipation debt and additions to accounts payable). 19 Municipios with populations below 50,000 are required to produce the report every six months. 14 46. The LRF and its companion law (Lei 10.028/2000) make extensive provisions for enforcement. In total, there are five types of enforcement mechanisms. The first consists offiscal sanctions, including ineligibility for discretionary transfers, federal guarantees, or new loan approvals. As shown in Table 3, these apply to jurisdictions that fail to comply with personnel ceilings, ceilings on debt, or transparency requirements. Nullification applies to contracts or administrative decisions that violate the terms of the LRF. Salary increases, for example, are null and void if they would result in a violation of ceilings on personnel spending. New debts are null and void (and must be immediately repaid) if they would result in a violation of the ceilings on debt. Individuals responsible for violations of the LRF are subject to fines. Personnel spending or borrowing in excess of LRF ceilings can result in a fine equal to 30% of the annual salary of person responsible. Several violations also render a governor or mayor subject to impeachment. Lei 10.028 adds violations of the debt ceilings to the list of grounds for the impeachment of governors (Lei 1.079/1950) and mayors (Decreto Lei 201/1967).20 Violations of prohibitions on election year hiring and borrowing can result in prison terms for periods ranging from three months to four years. 21 20 Impeachment does not, of course, necessarily result in dismissal. Under the applicable federal law (Lei 1079/1950) governors are to be tried by a tribunal composed of five members of the legislature and five high court judges (desembargadores), under the leadership of the president of the local Tribunal de Justica. Their decision is final. At the municipal level, impeachment charges against a mayor are first investigated by a commission consisting of three city councilmembers. Their findings are then reported to the council as a whole, which can dismiss a mayor by a two thirds majority vote. 21 Under the penal code (decree law 2848/1940 as amended by law 7209/84) imprisonment may take place in a medium- or maximum security prison, a prison farm, or in regime aberto, an arrangement that leaves convicts at liberty during the day but requires nights and weekends to be spent in prison. 15 Table 3 Selected LRF Infractions and Their Consequences Offense Consequence fails to establish fiscal targets in LDO x Exceeds ceiling on personnel spending x x x authorizes increases in personnel spending within last six months of x x term __n__ _ Exceeds senate ceiling on debt x x contracts new debt in violation of ceiling x x x contracts debt during last 6 month of term x issues bond not authorized by law and registered with federal govt. x fails to liquidate revenue anticipation loan by end of year = x fails to publish accounts x fails to send annual accounts to the federal govermnent by deadline x x - source: Lei 10.028/00 * suspension of voluntary transfers, prohibition on new borrowing **e.g. cancellation new positions, new hiring, new debt 47. In the short term, the LRF, together its enforcement legislation, the Lei 9496 debt agreements, and CMN resolution 2653, represent an important step in the control of subnational deficits in Brazil. This is as much due to its symbolic significance as to its detailed provisions. In effect, the LRF sends a signal to future governors and mayors: the era of easy federal bailouts has ended. 48. It is too early to determine whether these measures have had an impact on fiscal performance. The first of the 9497 debt rescheduling agreements--Sao Paulo's--went in effect in mid-1997. The majority were not signed until 1998. Resolutions 40 and 43 date from 200122; CMN resolution 2653, from 1999. The LRF went into effect in mid-2000. Most of the impact of this battery of controls will not be observable for several years. At present, aggregate data on state fiscal performance is available only through 2000. This data nevertheless suggests a substantial improvement in fiscal performance between the pre-Lei 9496 period and 1999. The aggregate state primary deficit, which averaged -8% of net revenues in 1995-97, fell to -23% in 1998, largely due to election year capital spending, financed by the sale of state enterprises. But it rebounded in the following two years, reaching a surplus equal to four percent of net current revenues in 1999 and remaining in positive territory in 2000. 49. Changes in the overall deficit are more difficult to determine. Brazilian fiscal statistics do not report deferred (capitalized) interest as a current expense. The growth of interest obligations on state bonds and BANESPA debt during the mid-1990s is therefore not reflected in the figures, nor is interest capitalized under the 9496 debt refinancing agreement reflected in the data for the post-Lei 9496 period. In principal, it is possible to estimate the size of the overall deficit by applying prevailing interest rates and inflation 22 An earlier version of the resolutions, Res. 78, went into effect in 1998. 16 factors to the stock of debt. The debt service ceiling fixed under Lei 9497 changes the nature of the state's debt service obligations, however. Rather than being obligated to pay interest on the entire stock of debt--with the presumed option of refinancing the principal- the states are required to pay a fixed percentage of their revenues to service their rescheduled debt. Any debt service obligation in excess of the ceiling is automatically refinanced by the federal government. Any remaining deficit, however, must be financed from new borrowing. Under these circumstances, the deficit, net of the debt service ceiling, is a more accurate measure of a state's ability to meet its financial obligations. On the basis of this calculation, the states' overall deficit-i.e., their deficit net of automatic federal refinancing--equaled about six percent of net revenues in 2000. (SeeTable 4) 50. The stock of debt continued to grow in the post 9496 period. From December 1997 (when the 9496 rescheduling began) to December 2000, the stock grew by nearly 40% in real terms. This does not represent a violation of the 9496 agreements and the Resolution 40/43 ceilings, however. Table 4 Trends in Major Components of State As shown in Table 5, virtually all the Primary and Overall Balances growth was due to new borrowing as percent RCL 1995 1996 1997 1998 1999 2000 required to recapitalize state banks Personnel 62% 65% 60% 62% 56% 58% prior to their privatization (as IAnctive 19% permitted by Lei 9496 and not subject capital investment 15% 14% 32% 3 17% 14% to Senate Resolutions) and the Primary balance -7%1 -8% -9%° -23 +4% +1% capitalization of interest on Overallbalance -12% -13% -21% -30%° -1% -6% rescheduled debt (as permitted under source: STN website: execucao orcamentaro dos Estados 1995- the 9496 agreements). Such data as exists suggests that the flow of new contractual debt, other than borrowing to finance privatization, has declined. Table 5 Trends in Stock of State Debt 51. Will the new set of federal (B197 1998 1999 2000 agreements, laws, and regulations state governments 117.6 130.6 156.7 161.2 forestall subnational debt crises in domestic debt 113.2 124.2 147.7 151.5 the long run? The net effect of the bond (net) 35.8 13.6 1.9 1.7 new regime is to make the most Lei 9696/PROES 58.1 100.0 135.3 136.7 overt forms of excessive borrowing Lei 8727* 0.01 10.5 12.3 25.7 . 05 1.3 25. extremely risky for the individuals Lei 7976/Aviso 30 2.8 2.8 3.4 3.2 bank debt 18.5 27.1 22.2 4.0 respons other domestic 0.0 0.0 6.5 12.6 disadvantageous for the jurisdictions less deposits, conta grafica -2.0 -29.8 -33.9 -32.4 they serve. A state attempting to extemal debt 4.4 6.4 9.1 9.6 borrow in excess of its Lei 9496 source: Banco Central Boletim limit incurs penalty interest on its rescheduled debt, which is automatically recoverable through deductions from intergovernmental transfers. If the borrowing exceeds the Senate ceilings, the state faces the loss of discretionary transfers. Its governor is subject to a fine and the possibility of impeachment. Most sources of credit supply are also restrained. Lending from the Caixa Economica is severely restricted. Bonds may not be issued until 2010. The Central Bank 17 is prohibited from refinancing state debt. Lending by multilateral organizations is restricted by the provisions governing federal guarantees. The system also brings some of the more subtle forms of borrowing under control. In placing limits on personnel spending and election year commitments, the 9496 agreements and the LRF restrain some of the traditional sources of state deficits. Overall, it would appear that while existing debt can continue to grow through the capitalization of deferred debt service, the current battery of laws and resolutions eliminate the possibility of excessive new borrowing by subnational governments. 52. Will the new system of debt controls be enforced? There are two reasons for concern. The first is the administrative burden that enforcement will impose. Resolutions 40/43 require that all subnational borrowing operations-including short term revenue anticipation loans-obtain the approval of the Treasury The Treasury, in turn, is required to confirm that the borrower (1) satisfies the three fiscal ratios described earlier; (2) has not engaged in prohibited forms of borrowing (including borrowing from its own enterprises or from suppliers); and (3) is in compliance with the terms of all federal debt rescheduling agreements.23 To date, the workload has been of manageable size. In 2000, there were only 385 requests for borrowing authorizations. In the first eight months of 2001, that number shrank to ten. 53. The small level of requests is reportedly due to two factors. First, 2000 was a municipal election year. Mayors were therefore forbidden to borrow during the last six months of the year. Second, there is a general perception that requests would be fruitless, due to the federal government's decision to clamp down on lending from the Caixa Economica and federally guaranteed external loans. But in a universe of 27 states and over 5,500 potential municipal borrowers--and a scope of work that includes not only long term project lending but also short term cash management debt and suppliers' credits--this trickle could turn into a flood. If these requests cannot be processed quickly, pressures may arise to either perform the work superficially or to abandon the procedure entirely. 54. The penalty aspects of the LRF and its accompanying enforcement legislation also place a heavy burden on the courts and the legislative branch. The imposition of fines and imprisonment require judicial hearings, which may strain an already overburdened court system. Impeachment requires a lengthy process of investigation and trial by state assemblies and municipal councils. The prospect of long delays or eventual dismissals may reduce the risks perceived by governors or mayors who are tempted to violate the law. 23 This responsibility has since been transferred to the Treasury. 18 55. The second risk is that the federal government's political commitment to enforce the system will recede. Attempts to control subnational fiscal behavior through federal administrative and legal controls have a mixed record in Brazil. Efforts to control borrowing date back at least as far as the military era, and were not successful in forestalling the three state debt crises that followed Brazil's return to democracy. In fact, it is the federal government, in its capacity as lender, that has historically contributed to subnational debt crises. As noted earlier, the 1993 crisis was largely the product of excessive lending by the Caixa Is the LRF self-enforcing? Economica and other federal banks. And while the more It could be argued that the LRF, by placing part recent crisis originated in of the risk of illegal borrowing on lenders, will enforce private bond placements and itself., obviating the need for a federal credit analysis. brrong fromestate Bank loans in violation of the debt ceilings, for example, corrowmg it sthe are null and void and are to be repaid without interest. In commercial banks, it was the principle, this should discourage banks from making federal government's them. Contractors and suppliers, by the same token, willingness to make a market should be reluctant to provide unauthorized credit to in state bonds and its delay in subnational government, as they run the risk of having addressing BANESPA's the credit nullified and losing the value of whatever portfolio problems that allowed goods or services they have supplied. These mechanisms the debt to grow to its ultimate will only be effective if lenders believe they will be proportions. caught, however. 56. The risk of federal In the case of bank loans, the odds are high. in backtracking may have been order to be legally enforceable, all contracts for bank reduced by changes in the loans must be registered on the national credit monitoring reducednby ch a l and system, CADIP. Suppliers credits, however, may be more institutional, political and difficult to catch. And if the past is any guide, arrears to economic environment. The personnel and to the federal social security system may recent privatization of most escape detection entirely. state banks, for example, will I prevent states from borrowing from their banks or accumulating off budget contingent liabilities to bank depositors. Brazil has adopted the Basel accords, which require the Central Bank to impose minimum provisioning requirements on bank loans to subnational governments. 57. Globalization is also increasing pressure on the federal government to maintain control on subnational deficits. Because growth in subnational deficits undermines investor confidence, the federal government is under pressure to enforce the new debt control system, if only to keep the foreign investment flowing into the country. 58. The electorate has also changed. There is evidence that Brazil has now developed a large, educated middle class, which derives its livelihood from the private economy (as opposed to the public sector, which was the mainstay of the middle class twenty years ago). Voters may hold their political leaders to a higher standard of fiscal probity than they did in the past. This is said to be reflected in the last gubematorial elections, where fiscally conservative governors were largely re-elected and populists generally lost. 19 59. But a system that is dependent upon the political climate is vulnerable, particularly when it is as administratively cumbersome as the one currently in force. In the long run, there is an argument for reducing the role of the federal government in controlling subnational fiscal behavior and access to credit, and increasing the use of market mechanisms to perform this role. Such a system would rely on private lenders, rather than the government, to determine which subnational governments are acceptable credit risks. A Market-based Approach 60. Market based mechanisms could take a variety of forms. In the U.S., municipal bond markets have been the primary vehicle for subnational borrowing. While the U.S., is perceived as having highly sophisticated issuers, in fact many of the 40,000 individual jurisdictions that issue bonds are quite small. They rely on advisors, whose function is to provide technical competence and financial expertise. In the U.S. subnational governments typically do not sell directly to investors. Instead, they select an underwriter who will purchase the bonds and then reoffer the issue to the public. The underwriter is a risk-assuming middleman or broker who guarantees the issuer the face- or contracted amount of the bonds. Underwriters, in turn, offer the bonds to potential investors. Underwriters make their profit from the spread, which is the difference between the price they pay the issuer and price at which they sell to investors. The largest owners of municipal bonds in the U.S. are individual investors, mutual and money market funds, property and casualty insurance companies, and commercial banks. In recent years, individual participation in municipal bonds has expanded through investments in mutual and money market funds. Municipal bonds are considered relatively safe from default, despite some adverse results in recent years. After they have been issued, they can be sold to other investors on the secondary market. 61. The 'European' model, in contrast, relies on specialized banks to finance subnational debt. Some of these are collectively owned by the municipalities themselves (as in Sweden and Finland). Others were founded by national governments and have since been privatized. Credit Local de France, for example, began life as a lending window of the Government-owned Caisse de Depots et Consignations, a national savings bank. It provided inexpensive loans to local governments, financed from low interest- paying savings deposits. Financial deregulation in 1987 reduced the volume of funds available from the Caisse de Depot, requiring the Credit Local to rely increasingly on private capital markets to finance its operations. The fund was spun off as a separate government-owned enterprise in 1987. It was privatized in 1993. Under its current charter, Credit Local (now known as Dexia) cannot accept deposits, and instead funds its operations by borrowing in the international and domestic capital markets. It operates on a commercial basis, with portfolio risks borne by the company's shareholders. 62. Both models bear serious consideration in Brazil. Five sets of reforms would be required, however, for either system to work. First, the federal government would have to remove the obstacles that prevent subnational governments from being good credit risks. Chief among these is a Constitutional mandate requiring subnational governments to offer unaffordable levels of pension benefits. (This is discussed below.) In addition, 20 legislative changes may be required to improve the quality of subnational collateral. National credit legislation now makes enforcement of security in private sector loans nearly impossible. Judicial discretion invariably favors continuation even in the event of defaults. Bankruptcy rules put secured credits after workers' payments, taxes, social security and expenses. Private banks may find that subnational governments are as invulnerable as private firms. 63. Second, the federal government would have to limit its role as a direct lender to subnational governments. The supply of finance-particularly project finance-to subnational governments has historically been dominated by federal banks. Because the federal banks are heavily subsidized, they have tended to crowd potential private lenders out of the market. Both the Caixa Economica and BNDES have access to low cost federal funds collected from social welfare contributions. These low-cost funds are passed on to subnational- and other authorized borrowers with a margin to cover the institutions' operating expenses. In both cases, the resulting subnational loans are made at less than one-half the interest rate of the most nearly comparable market-rate loans to private borrowers. The loans are for longer periods than are available on the commercial market and contain other favorable provisions, such grace periods on amortization payments, that are generally unavailable commercially. As a result, private lenders are unable to compete in this market. 64. Third, the federal government would have to limit its role as a borrower. Historically, large federal deficits have created a voracious demand for credit, along with the high interest rates required to satisfy it. As a result, a large share of private savings is invested in federal debt. Government securities now constitute about 30 percent of bank assets. This is also the fate of what would otherwise be a logical source of long term project finance: pension funds, mutual funds, and insurance companies. Pension fund and mutual fund assets in Brazil are very large, not only in absolute terms but also relative to bank assets and relative to GDP. Total investment fund assets total US$120 billion, compared to banking system assets of US$494 billion. The size of institutional investors is significantly larger in Brazil than in all other Latin American countries except for Chile. But, like banks, Brazilian institutional investors prefer to invest in short term federal securities, rather than long term subnational debt. To reduce the price of capital to a level compatible with subnational project financing, the federal government will have to reduce its own demand for credit. 65. A fourth required reform would be a reduction in the transaction costs of private lending to subnational governments. Private banks presently have little experience in municipal credit analysis. While Brazil has a thriving credit rating industry directed at private sector borrowers, credit rating for subnational government remains undeveloped. 66. Finally, the Government will have to remove the implicit federal guarantee that lies behind private sector lending. It is tempting to argue that the federal government could withdraw from the subnational credit market by simply refusing to lend to subnational governments. The facts are not so simple. As described earlier, private banks were intimately involved in the most recent debt crisis.. State bonds were initially sold to private banks. AROs, too, were extended by private banks. Private depositors and 21 interbank lenders were the initial sources of savings that financed BANESPA's lending to Sao Paulo. What was implicit in private loans to the states, however, was a federal guarantee. While some lenders may have believed their borrowers were creditworthy, it is. likely that they also assumed that the federal government would make good on state obligations, once the obligations were large enough to threaten the stability of the financial system or provoke a breakdown of services in a major state. In this, they were not disappointed. The Central Bank made a market in state bonds once private banks were unwilling to hold them. The federal Treasury refinanced the states' ARO debt. Central Bank backing kept BANESPA in operation, despite its non-performing debt portfolio. 67. Experience suggests that legal prohibitions on federal debt relief--as contained in the LRF--are insufficient to remove the perception of an implicit federal guarantee. Instead, the federal government will have to demonstrate through its actions that it is no longer willing to come to the aid of states that have overborrowed or the lenders that have extended them credit. Can Brazil's federal government do so? In the case of a moderately sized municipio with a moderately sized domestic loan, it may be possible. But a large jurisdiction with a large loan -or any form of external borrowing-raises a more complicated set of problems. In the past, the federal government has proven vulnerable to three forms of subnational pressure. 68. The first is the threat that widespread state defaults could cause the collapse of the financial system. This was clearly a factor in the federal government's decision to keep BANESPA operating and to ultimately finance the recapitalization of the largest state banks. The second is the threat that a subnational default on an external loan could blacken the federal government's reputation in international capital markets. Foreign capital markets are not adept at distinguishing between the default of a subnational borrower and default of its national government (or even to distinguish between a default by one country and another in the same region.) Default by any state on a eurobond, for example, could substantially raise the risk premium on federal external borrowing. The third is the threat of a collapse in key public services. A breakdown in public order in any Brazilian state would generate immense pressure for federal relief. 69. The federal government cannot entirely ignore these threats. But there are ways to address them short of the administrative micro-management that characterizes the present system. Threats to the banking system can be forestalled through tougher prudential regulation (which can prohibit individual bank from overexposure to any single borrower, and mandate special provisioning requirements for particularly risky classes of borrowers.) Threats to Brazil's credit reputation in international markets can be addressed by prohibiting subnational governments from borrowing abroad. The threat of a breakdown in public services is more difficult to address. In this respect, Brazil might benefit from the example of the New York City financial crisis. In 1975, after years of mounting debt, New York's creditors declined to roll over the city's loans. The city sought relief from the federal government and was initially turned down. Ultimately, some federal (and state) financial relief was provided (in concert with concessions by the city's creditors and public sector unions). But the city's political leadership was forced to cede all control over fiscal matters-include tax rates, contracts with suppliers, wage 22 levels, and the allocation of the budget-to a state-run financial control board. This not only gave the state the power to force an adjustmnent package on the city. It also served as an object lessons to future city administrations-in New York and throughout the country-of the consequences of fiscal recklessness. PERSONNEL REGULATIONS AND PENSION REFORM 70. Even if the current system of debt controls is rigidly enforced, fiscal pressures may arise from another quarter: personnel costs. At present, personnel benefits guaranteed in the federal Constitution are on a collision course with the personnel and deficit ceilings imposed by the Lei de Responsibilidade Fiscal. When they collide, states will be forced to choose between violating the LRF or the federal Constitution. 71. The source of the problem is a series of unfunded mandates imposed by the federal Constitution. Prior to the 1988 Constitution, subnational staff could be employed under either of two legal regimes. The first--the statutory regime--conferred a wide range of civil service benefits and rights, including protection against dismissal (except for cause) and pension benefits equal to 100% of exit salaries (termed the salario integral). Pensions of statutory staff were paid directly from state treasuries and were unfunded. The second regime-the consolidated labor law (CLT)--allowed for dismissal without cause (although it established compensation requirements). State pension obligations to CLT staff were limited to a 21% payroll contribution to the national social security system (RGPS). 72. The 1988 Constitution altered the picture in two important respects. First, it required states (along with all other government bodies) to adopt a single regime for their employees. In effect, states were required to absorb former CLTistas into the statutory regime, with all the benefits and rights pertaining thereto. Second, it expanded civil service benefits, in particular prohibiting states from lowering salaries. While these provisions have been subsequently modified, they continue to impose a burden-an unfunded mandate-on subnational governments. 73. As noted earlier, rising personnel costs were an important contributor to the most recent debt crisis. Since then, states have generally managed to reduce the costs of active (i.e., non-retired staff) by freezing nominal salaries, limiting hiring, and dismissing staff who were employed on short term contracts or had been hired irregularly. In 1988, Congress passed the 19t Amendment, which will increase states' control over the wage bill. The 19th amendment grants subnational governments temporary authority to dismiss existing civil servants, provided certain conditions are met: (1) personnel costs must exceed a threshold established in complementary legislation; (2) at least 20% of positions filled by political appointment (cargos de confianza) have been eliminated; and (3) non-confirmed civil servants have been dismissed. The Amendment also abolishes the requirement of a single employment regime, opening the door for an eventual return to a mix of private sector and public sector regimes. In principal, this should give subnational governments more flexibility in hiring and dismissing existing staff. As it only applies to new staff, it would have little immediate impact, however. 23 74. The main threat comes now comes from retirees. Retirement benefits for statutory employees are quite generous in Brazil. As noted earlier, pension benefits are Constitutionally fixed at 100% of exit salaries (the so-called salario integral) and are indexed to increases in the position vacated by the retiree. Eligibility criteria for retirement are also quite liberal. Until recent amendments, male non-teaching staff were allowed to retire after 35 years of service, regardless of age. Female non teaching staff could retire after 30 years. Teachers of either sex were given an additional five years' dispensation. Thus, not atypically, a female teacher could retire after 25 years of service at age 45 and expect to collect pension benefits for another thirty years. "Service", in this case, meant employment in any formal sector position. Thus a staff member could join the state civil service after spending most of her career elsewhere, and still expect to have her pension fully paid by the state. 75. In 1998, Congress toughened some of the conditions for retirement. Under the 20th amendment, a staff member must (1) have ten years of service in the public sector, (2) have five years in the position from which he or she is retiring, (3) meet minimum age criteria (60 years of age for men, 55 for women) and (4) meet a years-of-contribution criterion (35 for men, 30 for women). The new criteria apply to both new and existing staff (with transitional rules for staff who are close to retirement age). Congress has also enacted legislation requiring the RGPS to compensate states and municipal governments for pension benefits accrued while a staff member was employed under the CLT. Under Law 9796/99, and Decretos 3112 and 3217, the RGPS is required to pay a proportional share of former CLTistas' retirement costs, based on the RGPS schedule of benefits. In the long run, the 19th amendment will also reduce pension costs, as staff hired as CLTistas would be subject to the lower benefit regime of the national RGPS system.24 76. But the impact of these measures will be not be sufficient. They do not address the fundamental source of the pension problem, the salario integral. In guaranteeing the salario integral to even the most highly paid staff, the states have taken on an impossible financial burden. As of this writing, Congress was considering legislation that would cap pension benefits for higher-wage staff. Under the current version of the legislation (PL09), any unit of government would be allowed to establish a fund to supplement the benefits provided by the RJIU. Benefits would be based on defined contributions rather than defined benefits. Most importantly, staff participating in the complementary fund would have their RJU benefits capped at the RGPS ceiling. As a result, governments would no longer have an open ended obligation to pay pension benefits above the RGPS ceiling. They would merely be required to operate a complementary system that would provide additional pension payments to staff earning over the RGPS ceiling. The Constitutionality of such a reform has yet to be tested. Advocates of the legislation argue that it would pass Constitutional muster as long as the combination of the RJU benefit and the likely yield of the complementary fund would reasonably approximate the beneficiary's exit salary. 24 Despite this, governments still have a strong incentive to hire staff as estatutarios. Because CLT contributions must be made while the staff is working, they fall on the administration currently in office. Pension costs of estatutarios, in contrast, are only paid after retirement, and therefore fall on subsequent regimes. 24 77. In its current form, the impact of the legislation would be limited. It applies only to staff whose salaries are above the maximum RGPS benefit, and only to that part of the their salaries that is above the limit. Only 15% of state workers and 12% of municipal employees25 would see any reduction in benefits under the current version of the proposed law. In addition, participation in a complementary fund would be voluntary for existing staff. The cap can, however, be imposed on new staff. In the long run, this device could have a significant impact on the public sector's pension obligations. 78. For the present, pension costs are therefore high and rising. In 1999, aggregate state payments to retirees averaged 20% of net current revenues (and nearly 30% of the total wage bill). In the largest seven states, it averaged 23% of net revenues and one third of the wage bill. The figures are expected to increase. A recent World Bank Study26 estimates that state pension payments will triple between 1998 and 2010. This is in part due to the aging of the labor force. Staff who began employment during the boom years of the Brazilian miracle are now reaching retirement age. It also reflects the impact of the Constitutionally-mandated unification of employment regimes. Because the 1988 Constitution required states to combine statutory and CLT regimes, benefits formerly confined to statutory employees now extend to all state personnel. The states therefore face the prospect not only of increased pension obligations to a growing number of long- term statutory employees , but also a rapid increase in obligations to former CLTistas. A recent, study by the Fundacao Getulio Vargas estimates the net present value of state pension liabilities to be R$ 288 billion, or about three times total net current revenues in 1999. (This includes liabilities to all existing staff-active and retired-but excludes liabilities to future staff who would replace existing staff as they retire.27) As shown in table 6, the burden is particularly high in the South, where it equals 4.37 times net current revenues. (While the pension burden in the Northeast is a high percentage of GDP, it is a relatively small proportion of net revenues, as net revenues are derived more from intergovernmental transfers than from the local economy). Table 6 Estimates of State Pension Liabilities 79. Some states have attempted to address the looming pension crisis by as % GDP as ratio to net increasing employee contributions and/or ratio to net establishing pension funds capitalized current revenues through privatization proceeds. Bahia, for Norte 36.1% 2.39 example, recently increased the Northeast 45.1% 2.46 contribution level for active staff from Southeast 3.4 3.22 8.7% to 12% (to take full effect in 2004) South 32.2 4.37 and allocated over R$ 400 million from Center west 37.8 3.80 Source: FGV (pension and GDP); STN (revenues) the sale of its power company to 25 The 12% estimate applies only to municipalities that are state capitals. PL09's impact on other municipalities would be even more limited. 26 World Bank; 2000 Brazil-Critical Issues in Social Security; Report No. 1964 1-BR; Washington DC:World Bank 27 Of the total, R$ 169 billion are liabilities to existing retirees, R$86 billion to existing active staff and R$ 32 billion to dependent survivors in case of death while in service. 25 capitalize a pension fund. (It also attempted to increase the contribution level of retired staff, in effect reducing net benefits. This was declared unconstitutional by the Supreme Court.) These are steps in the right direction. But any solution that focuses solely on increasing resources is likely to be untenable. What is required is a reduction in the pension benefits of currently active staff.28 80. Reducing the "acquired rights" of existing staff has proven extremely controversial in Brazil. Experience in other countries suggests a strategy. Government pension reforms normally distinguish between three populations: (1) existing retirees, (2) staff who are hired after the new rules go into effect, and (3) existing staff. Existing retirees are normally protected in their existing benefits. Newly hired staff are subject to the new contribution and benefit parameters. In dealing with the existing active' staff, governments typically adopt a transition rule, in which staff who are close to retirement are fully protected, but those who are more recently hired are not. In Brazil, the existing legislation guarantees full benefits to all existing staff, regardless of how long they have worked. While it permits states to hire new staff as CLTistas, it requires states to continue to guarantee the salario integral to newly hired estatutarios. A change in rules is needed, that would establish a cut off date for full benefits or sliding reduction in benefits depending on length of service. This would protect the benefits of staff who are close to retirement, while allowing the states to achieve critical savings on the pension benefits of staff who have joined more recently. Over the long run, it would produce a pension system that is actuarially sound, not only for CLTistas, but for statutory staff. 3. LONGER TERM REFORMS EXPENDITURE ASSIGNMENT 81. In the longer term, Brazil needs to confront more fundamental problems in its federal structure. What is most striking is the lack of accountability that arises from the absence of any clear definition of the responsibilities between states and municipal governments. The debate over federalism in Brazil has traditionally focused on the division of resources among the three levels of government. In a sense, it has been a debate about the division of 'spoils'-federal and state taxes-among the three tiers of government. This reflects a long tradition of paternalism and clientelism in Brazilian politics-indeed in the politics of much of Latin America. Politicians retain considerable discretion to grant or withhold favors to selected groups in return for their political support. In this respect, the debate over federalism has been a debate over the division of 28 World Bank; 2000 Brazil-Critical Issues in Social Security; Report No. 1964 1-BR; Washington DC:World Bank 26 discretionary power among politicians at each tier of government. But as Brazil completes its transformation to a modem state, this model of government appears increasingly anachronistic. Government is no longer a granter of discretionary favors. It is an entity obligated to provide certain non-market functions and services to its constituents. This requires a change in the terms of the federal debate. The issue is no longer how to divide up revenues, but rather how to define the responsibilities of each tier of government and the working relationships between them in a way that is efficient and transparent to their clientele-the electorate. 82. The Brazilian Constititution does make some attempt at defining the responsibilities of each tier of government. Among the responsibilities assigned to the federal government are national defense and social security, emission of currency, control of public debt, regulation of interstate and foreign trade, and the power to establish the "general norms of public employment." The Constitution also delineates certain concurrent responsibilities of the federal government and the states, including tax legislation, education, and social assistance, specifying that federal law will be limited to "general norms" but will prevail in case of conflict with state legislation. 83. The limits on the functional responsibilities of the federal government are nevertheless relatively clear. In matters of security, it is responsible for national defense (i.e., the armed forces) but not domestic security (the police). In education, it is responsible for most public universities.29 Although it plays a role in financing primary education, it does not own or manage public schools. Although it is responsible for financing a large proportion of health care (through the national insurance system, SUS) it relies on state, municipal and private facilities to actually deliver health care services. The national highway system has been apportioned between the federal and state governments. Metropolitan rail transport-once the purview of the federal government-has now been decentralized to the states. 84. But the division of responsibilities between states and municipalities remains ambiguous. The federal Constitution grants the states "all powers not otherwise prohibited to them by this Constitution." 30 This would appear to be a grant of plenipotentiary power, limited only by the powers specifically assigned to the federal government. But the Brazilian municipios also have a broad Constitutional mandate. They are assigned "the power to legislate over subjects of local interest", and to provide "services of local public interest." While the Constitution does assign municipalities responsibility for primary education and health, it leaves open the possibility of technical and financial aid by higher levels of government. In practice, state governments continue to play a predominant role in both services. 85. The ambiguity in the division of state and municipal functions is compounded by peculiar Constitutional status of Brazilian municipalities. Unlike other federal constitutions, which typically define municipal governments as creatures of their 29 With the important exception of the University of Sao Paulo, which is state-funded. 30 Constituicao da Republica Federativa do Brasil, Article 25 section I 27 respective states, the 1988 Constitution establishes municipal government as a third tier of government with a Constitutional status equal to the states. States therefore cannot compel or prohibit actions by the municipalities within their jurisdictions. Subnational government in Brazil thus consists of two equally "sovereign" levels of government with no clear functional boundary between them. 86. Functional amnbiguity clearly undermines the accountability of subnational government to its citizens. With no clearly defined functions, the performance of each level of subnational government cannot be assessed. In effect, each level of government is free to enter any field that it deems of public interest, while neglecting those it does not. As one Brazilian observer has noted (in the case of education) "we have a tower of Babel, protected under the politically convenient concept of "collaboration". Under this concept, all three levels of government can operate schools systems (or not); finance education (or not); choose where they want to operate (or not). The result: no level of government is responsible (or accountable) for the provision (or not) of primary education. Each level does what it can or wants to, under the guise of collaboration."3' The same could be said about the other services that subnational governments provide. The absence of a clear division of functions between levels of government often leads to unproductive competition between governors and big city mayors and clientelistic relationships between governors and the political leadership in smaller towns. In much of small-town Brazil, it is said, mayors are not so much chief executives responsible for the performance of specific services, but rather intermediaries between citizens and higher levels of government, responsible for transmitting demands upwards, and bringing such largesse as they are able to obtain back to their constituents. 87. Which functions should be assigned to which level of government? The literature on fiscal federalism provides a useful framework for thinking about this question. It begins by assigning three roles to the public sector as a whole: macroeconomic stabilization, income redistribution, and resource allocation (in the case of market failure).32 The model assigns the first two roles to the central government. The central government is assigned responsibility for stabilization on the grounds that local economies have no access to an independent monetary policy and are too open for effective countercyclical measures to be effective. The income redistribution function is also assigned to the central government, on the grounds that local attempts to address income disparities are likely to provoke inefficient migration: higher-income groups will move to low-tax areas and low-income groups to move to high-benefit areas. (Even with their populations in situ, local governments in poor regions would have little income to redistribute anyway.) 88. Subnational governments enter the picture only with respect to the third function: resource allocation. Theory argues that if the benefits of particular services are largely confined to local jurisdictions, welfare gains can be achieved by permitting the level and 31 Abrucio, Fernando and Valeriano Costa; 1998; Reforma do Estado e o Contexto Federativo Brasileiro; Sao Paulo: Fundacao Konrad-Adenauer-Stiftung e.V. 32 Oates 1972; Imman and Rubenfeld 1997 28 mix of such services to vary according to local preferences. Local consumers, confronted with the costs of alternative levels of service (according to this viewpoint) will reveal their preferences by voting for rival political candidates or moving to other jurisdictions. In this respect, local politics can approximate the efficiencies of a market in the allocation of local public services by "pricing" municipal services and relying on the local political process and household mobility to clear the market. 89. There are a number of limitations to this approach, however. As a practical matter, it is difficult to define the scope of benefits of a specific service and match it to a specific jurisdiction. Although the benefits of defense may be unambiguously national and the benefits of public lighting unambiguously local, the largest sectoral expenditures of government-education, health, and transportation-fall somewhere in between. Education, for example, can be considered to have extremely localized benefits, reflecting the impact of education on the future income of pupils. But it also has national benefits as a vehicle for poverty alleviation and political acculturation. 90. There are also administrative constraints to the model. Public services are subject to economies of scale, particularly in the use of technical staff. Not every village can use a full-time homicide detective, even if the benefits of his services would be entirely local. Nor can a town with 200 students afford to offer 12 years of grade- differentiated education. The problem of scale economies can be particularly acute where municipalities are small in size. (See box, below.) 91. By the same token, a central government may find it cumbersome and expensive to operate a local office in every municipality in order to carry out its functions. Even programs that are unambiguously central have to be administered on the ground. A program of direct income transfers to the poor, for example, requires field offices to determine who is eligible. But creating a separate unit of field administration in each municipality would result in unnecessarily high administrative costs. 29 Size Counts-But Is Not an Absolute Constraint It has been argued that Brazil cannot devise a clear functional assignment for local governments because too many of them are too small. (See Table 7.) Brazilian legislation does not distinguish among municipalities on the basis of size. As a result, the only roles that can be specifically assigned to the biggest unit are those that can be performed by the smallest. In a situation where the smallest municipio may have a population of only a few hundred, this could be extremely limiting. In fact, Brazil's municipios are not particularly small by intemational standards. In Brazil, the average population of a municipio is about 33,000. In continental Europe, average populations are an order of magnitude smaller, ranging from 1,600 to 7,000. The U.S. has 39,000 general purpose local govemments (excluding special purposes districts) with an average population of 7,000. Among the OECD countries, large average sizes are found only in Japan and the United Kingdom. Where OECD governments have attempted reform, the direction of reform has nevertheless been toward reduction in the number of local governments. The number of school districts in the U.S. declined dramatically in the 1950s, as jurisdictions sought to combine enough students in a single jurisdiction to run grade-differentiated primary schools. Germany has reduced the number of gemeinden by half. (Under the consolidation reforms of 1965-1977, small gemeinden were either merged into larger units (in some Lander) or grouped into associations of municipalities with under joint administration.) The United Kingdom has eliminated a tier of subnational government in Scotland, Wales, and metropolitan areas of England. In Brazil, the number of small municipios is proliferating. In 1950, Brazil had only 1,889 municipios. By 1980 this-had increased to nearly 4,000. As of 1998, the total had reached 5,500. Seventy five percent have populations under 20,000. Proliferation continues to be encouraged by the formula used to distribute revenue sharing. The criteria used to allocate the principal federal transfer to municipal governments--the FPM-- sets a minimum transfer amount for small municipios, regardless of population. As a result, a small jurisdiction can double its revenues (in per capita terms) by dividing in two. Two recent changes in federal law will impose some restraints. First, the total amount of FPM transferred to the municipios of any given state is now fixed. Thus proliferation does not permit a state to increase its share of national FPM resources; it merely spreads the existing state total among a larger number of municipalities. Second, the process for approving municipal proliferation has been made more difficult. Formerly, voters of a given district within a municipality could unilaterally vote to form a new jurisdiction. Now secession must be approved by both the voters of the proposed new municipio, and those residing in the jurisdiction that would remain. These measures have yet to halt proliferation. European and U.S. experience suggests that small size is not an absolute constraint on efficient, accountable local government. Nevertheless, Brazil would be advised to toughen the criteria for creating new municipios. 33 IBGE, Anudrio Estaistico do Brasil, 1998, 1999, Rio de Janeiro. 30 92. Contracting Out. Brazil is hardly alone in facing this problem. Experience elsewhere-particularly in Europe and North America-provides some useful examples of how it can be solved. To achieve scale economies without losing local control over decision making, small municipalities can contract out. Such contracts can be with private firms. Small communes in France, for example, rely on large private firms to provide water services. Small municipalities can also contract with larger neighboring jurisdictions or with higher levels of government. In Los Angeles County, U.S., small municipalities contract with the county government for police, fire, and other services according to a fixed schedule of service levels and fees. In Germany, small gemeinden (municipalities) contract with county governments to provide primary and secondary schools. 93. Service districts. Special districts can also be formed to provide specific services. One option is for municipalities to form joint service companies. Brazil already has some experience with such companies in the Table 7 Size Composition of Municipios health care sector. (See box.) Special Population size category % of % oi districts can also be formed as independent municipios population units of local government. This approach Less than 20,000. 75% 19.6% is particularly popular in the U.S., where 20.000 a 100,.000 inhab. 21.5% 30.6% there are 48,400 special districts, 100.000 a 500.000 inhab. 3.2% 23.4% providing such services as primary and 500.000 a 1.000.000 0.3% 6.8% secondary education, fire protection, and More than 1.000.000 0.2% 19.7% pest control in areas that lie outside the inhab. boundaries of general purpose municipal Total 100% 100% governments. This model is not particularly attractive for Brazil. With a separate unit of government responsible for each function, coordination across functions can be difficult. Moreover, small specialized units of government have proven difficult to hold accountable. While their functions are clearly defined, their profile is too low to capture the attention of voters. The Politics of Municipal Health Consortia The Ministry of Health is currently encouraging the formation of intermunicipal health care consortia to operate joint facilities. As of 1997, there were 109 health consortia, involving 1,386 municipios, covering 22.66 million people. Municipal health consortia are organized under private law as voluntary, not-for-profit organizations and have no autonomy from the prefeituras and municipal secretariats of health that comprise them. According to a recent study (Abrucio and Costa, Reforma do Estado e o Contexto Federativo Brasileiro, 1998) this legal fragility is an impediment to the consortias' stability. It is never certain how long the union will last or whether a change in policy by one member will render the union unviable. As the report notes, mayors are accustomed to acting as political enemies, competing for seats in the state assembly or federal congress as a means of advancing their political careers. Long term cooperation is therefore often difficult. 31 94. Distinguish among types of local governments. Yet another approach is distinguish among classes of municipalities. Brazil-like other Latin American countries-has no tradition of using organizational distinctions to reflect the widely differing circumstances of local governments. In rural parts of Brazil, a municipio may cover a vast expanse of territory with only a few villages within it. In urban areas, a municipio may have a population of millions. No legal distinction is made between rural or urban municipios or among municipios of different size. The legislation for municipios in Brazil applies equally to the iMunicipio of Sao Paulo (with a population over 10 million) and the municipio of Pirapora de Bom Jesus, with only 4,585. This degree of uniformity can make it difficult to define the roles of subnational governments with any degree of specificity. An alternative is to adopt different legislation for different classes of municipalities. England34, for example, has one form of government for services that must be provided in both urban and rural areas (such as education and health) and another for services that are distinctly urban (such as solid waste management and urban transport.) The U.S., with its proliferation of special purpose districts, is an extreme example of this approach. 95. Central governments, by the same token, can reduce the costs of administering programs of national interest by relying on local governments to act as field agents on their behalf. This can take the form of explicit contractual relationships. In Germany, for example, Lander (states) act explicitly as agents of the federal government in the construction and maintenance of federal highways. In France, the departments are paymasters in the administration of national social assistance. U.S. counties perform the same role on behalf of the federal government. 96. Financing. Central governments can also induce local governments to perform functions on their behalf by providing financing. A national government, for example, may have an interest in ensuring a minimum level of primary education throughout the country, in furtherance of its role in income distribution. But the central government can achieve this objective through financial support to local authorities, while leaving local government responsible for day to day management; e.g., personnel management (decisions on the hiring and dismissal of staff, salary schedules, conditions of work) and detailed budgetary allocations (between teaching- and administrative personnel, teaching materials and supplies, maintenance of school buildings, and other programs). 97. Regulation. Central governments can achieve the same end through regulation- requiring local governments to take national interests into account, rather than paying them to do so. National interests in water quality for example, can be pursued by imposing outfall standards on individual municipalities. 98. These contractual, financial, and regulatory relationships can, of course, go badly wrong. Private firmns can renege on their contracts. Joint service agreements can founder on the political rivalries of individual mayors (See box above.) One of the most common breakdowns occurs when a central government assigns a function to local governments but then fails to relinquish enough management control for the recipient government to 34 Different legislation applies to each of the component parts of the United Kingdom. 32 carry it out. In Colombia, the central government--while nominally decentralizing education to the provinces-maintained its authority to determine teacher salaries. It then granted a major salary increase to the teachers. As the provinces lacked the resources to pay the increase, the central government was ultimately required to create a separate transfer program to enable them to meet their payrolls. But these are issues of design. They are not grounds to dismiss all efforts to define the functional role of local governments in Brazil. Health care 99. While such reforms have long been resisted in principle, they are in fact already proceeding on a sector by sector basis. The Brazilian Constitution of 1988 established a universal right to free health services to be provided by the public sector. To operationalize this mandate, Law 8080 of 1990 created the Unified Health System (Sistema Onico de Saude), or SUS. SUS was originally entirely a prospective payment system that reimbursed private and public providers for specific treatments based on a fee schedule set by the Ministry of Health. Since then, the system of open-ended direct reimbursements to providers has been replaced by one based on annual budget ceilings (tetosfinanceiros globais-TFGs) assigned to each municipality (or to state governments acting on their behalf). While a large proportion of hospital and outpatient care continues to be provided by private firms, SUS payments must now be authorized by subnational governments. 100. The TFG has two principal components. The first consists of per capita grants for basic health care. This includes the basic health benefit (Piso Assistencial Bdsico-PAB), which finances a list of cost-effective basic interventions, and funding for a family health program (Programa de Satude da Familia-PSF) and community health agents (Programa de Agentes Comunitdrios de Sauide). Funding for the three programs is provided directly to municipios on a per capita basis (with additional incentives for expanding coverage of the PSF and PAC). Where municipalities have not yet been declared competent to manage health care, funds are provisionally transferred to states to be used on their behalf. 101. The second component of the TFG consists of funding for hospital admissions (AIH) and high cost ambulatory,care (APAC). Budget ceilings for each type of treatment are fixed for each municipio, based in part on spending in the previous year, population, and installed capacity. Payments under the AIH and APAC components may be made to any registered health care provider-state or municipal, public or private. Each payment must be authorized by. the municipal secretariat of health, however, or by the state secretariat of health on a municipality's behalf, if the municipality has not yet been declared competent. 102. In imposing a competence test on municipalities, the SUS system establishes a hierarchy among municipalities in the provision of health care. Municipios are classified in three groups. The highest, Gestao Plena do Sistema Municipal, allows a municipality to manage the entire range of health care-basic, outpatient, and hospitalization-within its jurisdiction, set norms for PAB services in subordinate municipios, and administer 33 SUS payments for ambulatory and hospital services to public and private providers. The second level (Gestdo Plena de Atencdo Basica) permits municipios to provide basic health care, including administration of the PAB subject to norms fixed by the state or by a municipio approved for gestao plena, and to administer SUS payments to their own hospital and outpatient services.35 At the lowest level are the municipalities which have not been declared competent for either status. In these cases, management of all aspects of health care is assigned to the state or to a municipio authorized for gestao plena. 103. Recent revisions to the SUS system (NOAS/SUS 01/2001) will take these hierarchical distinctions one step further. Under the proposed Portaria 95, each state health secretariat is to establish a system of health micro-regions-- geographical areas where all health providers (primary, secondary, tertiary, public and private) will be organized into a coordinated network with clear referral systems. Each region is to have a designated regional capital (municipio-sede), which will have the authority to manage the referral process in its regions. Primary education 104. Efforts to define the respective responsibilities of state and municipal government in education date as far back as 1971. Law 5692/71 required each state to "establish the (respective) responsibilities of the state and its municipalities with respect to different grades of education" and called for a progressive decentralization to the municipal level, particularly first five years (primeiro grau). This had little impact. Municipal schools, which accounted for 26% of enrollment in grades 1-5 in 1970, saw their share increase to only 30% in 1986. The 1988 Constitution gave further support to municipalization, explicitly assigning municipios responsibility for primary education. It left the door open for continuing state involvement, however, by including the proviso that the service would be provided with the "financial and technical assistance of the federal and state governments". In freeing municipalities from state control, it also rendered municipalization subject to the voluntary compliance of mayors. Not surprisingly, the municipal share of enrollment in primary school as a whole ensino fundamental-grades 1- 8) remained the same in 1996 (33%) as it was in 1980. 105. Individual states attempted to force municipalization during the 1990s. In 1990, the state of Ceara began a donor-supported program of municipalization directed at rural areas. It was based on the presumption that once physical facilities were repaired and training and technical assistance were provided by the state, municipios would voluntarily assume responsibility for school operations. This assumption proved ill- founded. In 1992, the state of Parana began an more aggressive program, freezing enrollment in state primary schools and setting up a system of per-pupil grants for municipalities willing to take over primary schools. This also had limited success.36 Sao 3 Municipios at this level can also administer SUS payment to state and federal facilities, subject to the latter's approval. 36 Arretche, Marta and Vicente Rodriguez; 1999; Decentralizacao das Politicas Sociais no Brasil, Sao Paulo: FUNDAP. 34 Paulo's efforts to municipalize education were thwarted by a coalition consisting of the teachers' union, the bureaucracy of the state govermment, and the mayors. 106. Recent reforms in the financing of primary education have provoked a wave municipalization. The 14th amendment to the federal Constitution made fundamental changes in the national system of primary education financing. Under the Amendment, each state government must assign 15% of its two principal revenues sources-the ICMS and the FPE--to a special fund for primary education, termed FUNDEF. The municipalities within the state must contribute 15% of their principal transfer revenues (a share of the state ICMS and the FPM). If the sum of the state and municipal contributions, divided by the number of primary school students, is less than a standard figure (R$315 in 1998), the federal government is required to make up the difference. The total amount of Fundef funds is then distributed among the state and its municipalities on the basis of enrollment. The effect of this reform is to create a strong financial incentive for municipios to take over the state schools (or risk losing FPM and ICMS revenues.) This is apparent in the shift in the distribution of enrollment. As shown in the table below, the municipal share of total public enrollment increased from 33% in 1996 to 44% in 1998. Table 8 Trends in Primary School Enrollment 1980-1998 107. The 14th amendment total ede state muf. satatdoes not, of course, permit 1980 22598 .7% 52.8% 33.6% 12.8% states to abandon primary 1985 24769 .5 57.2 30.2 12.1 education nor does it force 1991 29204 .3 57.2 30.0 12.4 municipios to assume it. Due to 1996 3131 .01 55.7 33.0 11.2 the high degree of autonomy 1999 36060 0.1 46.0 44.8 91 granted to municipios by the Source: MEC, Censo Escolar 1988 Constitution, states cannot force any organizational solution on subnational government; they can only encourage it. Water supply 108. Water supply and sanitation is a third area where efforts to define the respective responsibilities of states and municipal governments have shown some success. The water supply sector has undergone several organizational changes over the last forty years. The first arose in connection with PLANASA, a centralized program established under the military. Under PLANASA, each state was required to establish a single state water utility with a mandate to absorb the existing municipal and private water companies in its territory. Investment financing for the new state companies was provided through the national housing bank (BNH) drawing on the savings generated by a mandatory unemployment insurance scheme (FGTS). In principal, municipalities had the option of remaining outside the system, and there were some important holdouts. (Sao Paulo's state water utility serves only about half the municipios in the state, for example.) Most chose to join, signing 30 year concession contracts with the state companies. 35 109. PLANASA began to go into decline in the mid 1980s. BNH was abolished in 1986, with its banling functions transferred to Caixa Economica. Responsibility for technical management of PLANASA was transferred to a series of ministries, which were subject to frequent reorganizations and changes in leadership. The recession of the 1980s sharply reduced FGTS resources. Investment fell accordingly. A revolt of the municipios began in 1989, when municipio of Novo Hamburgo created its own municipal water company, and attempted to break its concession contract with the state water company. It was later joined by other municipalities. 110. Government adopted a new policy in 1995, opening the door for private participation, external fmancing, and cofinancing with state and local government. This did not resolve the controversy over municipalities' ability to opt out of the PLANASA system, however. New disputes arose over the amount of compensation municipalities would owe to state water companies if they chose to opt out. In addition, jurisdictional issues that had been formerly been addressed centrally emerged: Which level of government would set water quality and pollution treatment standards? Who would allocate water resources? Who would coordinate water supply and sewerage in contiguous urban areas comprised of several municipal jurisdictions? 111. The federal govermment proposes to address these questions under a proposed new law on the water sector (Diretrizes Nacionais para o Saneamento Basica). The law authorizes the federal government to set general parameters for water quality and sewage treatment standards, and coverage targets for water, sewerage and sewage treatment. It declares that water and sanitation will be provided by the level of government that is able to do so 'de forma completa', thus preserving the authority of isolated municipios to provide their own services while permitting states (and in principle the federal government) to assume responsibility where water resources must be apportioned among competing jurisdictions. It explicitly gives states the power to regulate and monitor (but not necessarily provide) water and sewerage in all metropolitan areas, urban agglomerations, or any other group of municipios that share common facilities. In addition, states are given the power to take over municipal systems when there is an immediate risk to public health, grave danger to environment, or when the absence of such services would put at risk the use of natural resources under the control of the state. 112. As a whole, these efforts represent an important move toward a system that more precisely defines the functional role of municipios within the federal system. While Fundef does not force a definition of the municipal role in the provision of primary education, it does tie financing to enrollment. SUS reforms recognize the distinctions among municipios with different characteristics, and establishes the principle of municipal hierarchy. The new water law would explicitly define the respective roles of federal, state and municipal governments in the provision of water supply and sewerage. It is not clear that this will be enough. In the long run, Brazil might consider changes in the structure of municipal government itself, along the lines suggested earlier. 36 REVENUE ASSIGNMENT 113. In contrast to its vagueness in dividing expenditure responsibilities between levels of government, the Constitution is very explicit in dividing up revenues. It assigns specific tax bases to each level of government and creates a system of tax sharing which substantially redistributes revenue from higher tiers to lower ones, and among the jurisdictions at each tier. Tax Assignment 114. The system of tax assignment is spelled out in the 1988 Constitution. It gives the federal government exclusive power to tax personal and corporate income, foreign trade, and financial transactions, as well as tax certain manufactured goods (IPI),and establishes a system of social security taxes based on payrolls or gross receipts. As shown in the table below, federal taxes (including social security contributions) account for about 70% of the total tax receipts of all three levels of government. Roughly half of federal tax receipts are explicitly earmarked for social security. 115. The state governments are assigned three taxes: an estate tax, a vehicle ownership tax, and tax on value added (ICMS). The ICMS is by far the most important of these Table 9 Sources of Tax Revenue three and is, in fact, the highest yielding tax in Brazil. It is imposed Acronym base J %rtotal on the sale or distribution of receipts, merchandise and on selected services Ordinary Federal Taxes (principally electricity, IR income, profit 18.0 telecommunications, and inter city WPI gross value of sales of 5.3 bus transport). It is subject to two manufactured goods, imports basic rate regimes. The first one, set I. value of imports 2.6 by the federal government, governs CPMF financial transactions 2.6 interstate transactions. The second CSLL corporate profits 2.2 IOF financial transactions 1.6 governs intrastate sales. Intrastate ITR value of rural property 0.1 rates are determined by the state Federal social security taxes . legislature, subject to parameters set Cofins tunover 10.1 by the federal government and PIS/PASEP turnover 3.1 agreements among the finance INSS payroll 15.5 secretariats of the 26 states and the FGTS payroll 5.7 federal district. ICMS value added production and 22.2 distribution of merchandise, 116. Despite its high yield, the selected services ICMS is a controversial tax. As IPVA value of vehicles 1.5 discussed below, it is subject to a Municipal Taxes high degree of tax exporting. It has ISS gross value of services .1.6 _55 gross value of services 1.6 also been used by states (other than IPTU value of urban property 1.2 Other . . 6.8 Sao Paulo) to lure industrial Total ____________________ 100.0 investment. The state of Bahia, for example, offers a 6-year partial tax 37 deferral on the ICMS to firms starting new operations (or expanding existing operations by at least 35% of installed capacity) in manufacturing, agriculture, mining, tourism or electricity generation. The proportion of tax subject to deferral starts at 75% and declines in two-year increments. The deferral is tantamount to a grant: the deferred amount is subject to a 3% nominal interest rate with no monetary correction. This terms are matched by other states of the Northeast. 117. To municipal govermments, the 1988 Constitution assigns exclusive power to impose taxes on personal and professional services (ISS), urban property (IPTU), and real estate transfers (excluding estate taxes.) The ISS is consistently the highest yielding of these three taxes. As defined by federal law, it is imposed on all services except communications and interstate/intermunicipal public transportation. The tax is generally imposed as a fixed percentage of the retail price of the service provided, although altemative, simpler methods of assessment are used for self employed individuals. 37 Tax rates are set by individual municipalities, subject to ceilings set by the federal government. Rates can vary considerably across sectors. In the municipio of Sao Paulo, for example, private schools are subject to a tax rate of 2%. Hotels and restaurants, banking services, and security services are taxed at 5%. Nightclubs pay 10%. 118. The urban property tax is imposed on the capital value of all land and buildings with the legally-designated urban areas of each municipio. Like ISS, the IPTU is administered by municipal governments. Property valuations are based on the physical characteristics of each property, converted to an estimate of market value using construction cost data and surveys of neighborhood land values. These valuations are adjusted each year on the basis of an inflation index. Bills are typically payable at the beginning of each year and are thus relatively resilient in the face of inflation during the billing year. (Where taxpayers have the option of paying in installments, each installment is also §ubject to indexation.) Yields of the property tax are nevertheless extremely low. As shown in Table 9, they account for only 1.2% of total national tax collections. Transfer System 119. The Constitution also mandates an extensive system of intergovernmental transfers. The principal federal to state transfer-termed the Fundo de Participacao dos Estados (FPE) is derived from fixed shares (currently 21.5%) of the federal govermment's two principal non- social security taxes: the income tax and industrial products tax. Of this amount, 85% is distributed to the states of the North, Northeast and Center West regions, with the remainder going to the states of the South and Southeast. Within each group of states, 95% of the funds are distributed among states on the basis of population and per capita income (with poorer states receiving proportionately more). The remaining five percent is distributed on the basis of geographic area. The FPE is purely 37 Tax legislation is designed to avoid overlap with federal corporate income and state value added taxes. In the banking sector, for example, the ISS is imposed on non-lending services such annual credit card fees, ATM fees, and check issuing, but not on the interest charged on loans. In the case of bus transportation, the tax is imposed on intra-municipal buses, but not on intercity or interstate bus transport, which are subject to the state value added tax.) 38 redistributive, in the sense that the richest state-Sao Paulo-receives virtually nothing from it. FPE revenues, on the other hand, constitute well over half of the current revenues of the poorest states of the Northeast 120. The principal federal-to-municipal transfer is the Fundo de Participacao dos Municipios (FPM). Like the FPE, it is funded from fixed shares (currently 22.5%) of the federal income and industrial products taxes. Of this, ten percent is transferred to state capitals on the basis of population (as a share of the combined population of all state capitals) and the per capita income of each state. The other 90% is transferred to the municipios of the interior (i.e. non-capitals). Of this, 96% is distributed the basis of population. The distribution occurs in two steps. First an allocation is made to each state, based on its share of the national population. Each state's total is then distributed among its municipios on the basis of population, using a formula that favors smaller jurisdictions. (A minimum share is provided for municipios with populations under 10,188, and a maximum for municipios with populations over 156,216. As a result, a municipio with a population of 156,000 receives only 6.6 times as much with one with only 10,000, even though its population is 15 times greater.) The remaining four percent of the 90% (i.e. 3.6% of the total) is distributed to municipios with populations over 156,216, on the basis of population and the income per capita of their respective states. 121. The states are required to share part of their own tax revenues with their respective municipios. The principal state to municipal transfer is derived from a 25% share of state ICMS revenues. Under the federal Constitution, seventy five percent of this must be distributed among municipios on the basis of the origin of tax receipts. The remainder is distributed on the basis of formulas devised by each state legislature. Due to the high yields of the ICMS, this transfer is an important source of municipal revenues, particularly in the larger, more industrialized jurisdictions. 122. There are other, smaller transfers from the federal government to the states and municipios. Some are earmarked for specific services. These include federal contributions to Fundef and SUS which (as discussed earlier) are earmarked for primary education and health care respectively. It also includes the salario educacao, a 2.5% tax on the payroll of enterprises, collected by federal government, and earmarked for primary education. Discretionary transfers-at least those recognized as such-are a relatively small part of the intergovernmental fiscal relationship. In 1999, discretionary transfers totaled about two percent of the net disposable revenue of states and municipios. Finally, there are certain federal compensation transfers, which figure primarily in the revenues of the wealthier states. These include compensation payments for federally imposed ICMS exemptions on exports and capital equipment (under the Lei Kandir) and the compensacao IPI dos estados exportadoras. States are also permitted to retain the federal income tax obligations of their employees, an important factor in the higher income states in the southeast. 123. In evaluating the present system of transfers and tax assignment, it is useful to return to the basic principals of federal finance. The principles of revenue assignment 39 follow from the principles of expenditure assignrment. Finance should follow flmction.38 The structure of subnational finance-the mix of user charges, taxes, transfers-depends on the functions that have been assigned to each tier of government. This is because different sources of revenue have different effects on behavior and different patterns of incidence. User charges, for example, impose costs directly on individual consumers and can ration consumption by price. Therefore, there is an argument for assigning user charges to finance services whose benefits are confined largely to individual consumers, such as water supply and bus transportation. 124. By the same token, certain forms of direct taxation are appropriate for financing services whose benefits cannot be confined to individual consumers but do not extend beyond the boundaries of individual subnational jurisdictions. In the same way that user charges allow individuals to express their demand for services whose benefits are largely private, such taxes enable taxpayers to express their demand for local services that are consumed collectively. 125. In the traditional fiscal federalism formulation, intergovernmental transfers have a very limited role to play. They exist only to ensure that local spending decisions take wider benefits and costs into account. Left to their own devices, local governments would be expected to base their budget decisions only on the benefits captured by their constituents. This would cause them to "under provide" services that confer benefits on the nation as a whole. Transfers, in this view of the world, can be used to improve allocative efficiency by inducing local governments to take these wider benefits into account. 126. Equity impacts. In practice, transfers play two other important roles. The first is as a vehicle for national income distribution policies. As noted earlier, theory assigns the income redistribution function to central government, on the grounds that local attempts to address income disparities are likely to provoke inefficient migration. In principle, the national government would reduce income inequalities to a socially acceptable level through a combination of progressive taxes and direct transfers to the poor, leaving municipalities to focus strictly on resource allocation. Many municipal services have important distributional implications, however. Primary education, for example, is often the largest redistributive expenditure that the public sector undertakes. Primary health care, similarly, tends to benefit the poor disproportionately. One way to ensure that people in poor areas receive these services is to have the central government provide them directly. In France, for example, the salaries of public school teachers are paid directly by the central government. But another approach is to assign the education function to local governments, and provide central government transfers to poorer jurisdictions to ensure that all schools are able to achieve a minimum standard. (This is the practice, within individual states, in the U.S.) 38 Bird, Richard; 1993; "Threading the Fiscal Labyrinth: Some Issues in Fiscal Decentralization" in National Tax Journal Vol 46 40 127. The Brazilian system does achieve a substantial degree of equalization among states. Figure 2, below, compares variations in state GDP-an indicator of tax base- with state revenues per capita. Without transfers, variations in revenues would parallel variations in state GDP. Per capita revenues in Sao Paulo would be 6.6 times those of Maranhao. The result of the existing system of transfers, is a remarkable flattening of regional variations. As shown in the chart below, Sao Paulo's per capita revenues, after Stats GDP and Revenues 2.50- 2.00 m rcts/cap x 1.00 gdfa Ilw - 0.50 - 0.00 transfers, are only 2.3 times those of Maranhao. While the standard deviation in per capita income among all states is $2108 (50% of the mean), the standard deviation in per capita revenues is only Rs 210 (36% of the mean). The correlation between state GDP and state per capita income is only .44. 3 Thus, while there is some correlation between per 'capita revenue and per capita GDP, it is not the case that poor states have poor governments. 128. In principle, this degree of equalization would appear justifiable on equity grounds. Regional income disparities are wide in Brazil. The Northeast, with less than 30% of Brazil's population, has nearly two third's of its poor. State governments, moreover, are heavily involved in services with important distributional impacts, including primary education and health care. Left to their own devices, states in the Northeast would be hard pressed to provide these services from their own tax bases. It is not clear, however, that these funds find their way to the poor. If the poor lack influence over the allocation of state spending-a particular issue in the more traditional states of the Northeast-- governments may devote a disproportionate share of their funds to programs that benefit their more comfortable citizens. Thus the FPE may be an effective means of targeting poor states but an ineffective means of targeting poor people. 129. Federal earmarking may increase' the pro-poor tilt of state and municipai expenditures. While the fundos de participacao are not explicitly earmarked, aggregate subnational spending is subject to federally-mandated sectoral requirements. Under the 39 These statistics are strongly affected by the presence of three outliers: two sparsely populated former territories (Amapa and Roraima) and the Federal District, which is both a state and a municipio. The correlation between state per capita GDP and per capita revenues is .13 if only the DF is excluded. It is .44 if Amapa and Roraima are also excluded. 41 1988 Constitution, state and municipios are required to spend 25% of total recurrent revenues on education. (The 14th amendment of 1997 requires them to spend at least 60% of that on primary education during the period 1997-2006. ) The 29th amendment requires states to spend 12% of recurrent revenues on health. If spending in these sectors disproportionately benefits the poor, then transfers to the poorer states of the Northeast may have some distributional impact. Even so, they are subject to considerable leakage. The remaining 63 percent of state expenditures are not subject to earmarking. And not all spending on education and health benefits the poor. Subsidies to the tuition-free state universities-an expense that largely benefits the middle class-fall under this category, for example. 130. The FUNDEF and SUS transfers would appear to be more effective in targeting the poor. As described earlier, FUNDEF funding is explicitly earmarked for primary school operating costs. SUS transfers are explicitly earmarked for health care, with significant proportions set aside for programs (such as PAB and agents communitarios) that benefit the poor. In the long run, Brazil might do better at achieving its income distribution objectives through such programs as FUNDEF and SUS than through open- ended revenue sharing. Does the Northeast Subsidize SAo Paulo? The redistributive impact of the federal transfer system is offset, to a degree, by at flow of resources from poor states to rich ones that are implicit in the state tax system. When the ICMS is imposed on a manufacturer, the proceeds accrue to the state where the manufacturer is located. But the burden of the tax is incorporated in the price of the good and passed along through the subsequent stages of production and distribution until falls upon the final consumer.40 If the consumer is located in anther state, the tax can be exported. In Brazil, this works to the advantage of states in the South and Southeast, which have a positive trade balance (in terms of ICMS- taxable goods) with the poorer states of the North and Northeast. In effect, Sao Paulo can collect taxes from the residents of Piaui. To offset this tax exporting, the Constitution provides for a complex system of federally mandated differentiated rates. Rates on exports from wealthier states to poorer ones are taxed at a lower rate than goods moving in the opposite direction. This reduces the extent of inter regional incidence shifting but does not eliminate it. 131. The existing system of transfers to municipios is much more difficult to justify on distributional grounds. Only a small part (13.6%) of the federal transfer---the FPM--is based on income. The majority is instead based on population, with a strong bias in favor of very small municipalities. The state-to-municipal transfers system is largely based on the origin of ICMS tax collections. The overall effect of the transfers is to favor municipios with small populations, regardless of the level of poverty. As illustrated in Figure 3, the correlation between per capita municipal income and GDP is strongly positive.41 40 Other patterns of incidence are of course possible, depending upon the supply elasticities of capital and labor and the demand elasticities of consumers. 41 Excluding the federal District, Roraima and Amapa, the correlation coefficient is 0.8 42 Regional Equalization in other Federal Countries. Countries vary in their degree of concern with regional fiscal equalization. This is as much a matter of history and culture as it is of public finance principles. Among industrialized federal countries, Germany and the US represent the two extremes. The German fiscal system is explicitly aimed at reducing disparities in per capita state (Lander) spending. 42 It is based on shared taxes and a system of interstate transfers. All major taxes-including personal incorme, corporate income, and the VAT on domestic transactions-are administered by the states, at fixed, nationally-uniform rates. (Only the customs duties and the VAT on imports are directly administered by the federal government.) In performing this role, the states act as agents of the federal government, remitting revenues to the federal government before redistribution. (The states are permitted to impose only a few minor taxes on their own account.) Each tax is subject to a specific sharing formula. The division of the income tax between the federal and state level (i.e., the primary distribution) is fixed, with the states' share distributed among individual states (the secondary distribution) made on the basis of origin (with special rules for the distribution of the corporate income tax). The primary division of the VAT between the federal and state levels is adjusted annually. Of the states' share of the VAT, 75% is distributed among individual States on a per capita basis, with the remainder allocated to poorer states, to enable them to reach 92% of the average revenue per capita of all states. The German system also includes a direct transfer of revenue from rich states to poor ones, the so-called finanzausgleich. States with above-average per capita revenues are required to transfer funds to states with below- average per capita fiscal capacity, such that no state is more than five percent below the national average. The U.S. system, in contrast, makes no attempt to equalize state revenues. There is no explicit system of federal revenue sharing. Nor is there any explicit division of tax bases among levels of government (other than a Constitutional prohibition on federal export taxes). As a result, both levels, of government are free to tax their own bases using the instruments and rates they prefer. At present, personal income and payroll taxes are the primary sources of federal revenue. At the state level, retail sales taxes and personal income taxes (separately administered by the states) are the principal taxes.43 Variation in state per capita revenues are nevertheless smaller than might be expected. Per capita state revenues in the poorest state are only 20% below the national average. Per capita revenues in the richest state are 37% above the norm. This is in part because regional variations in GDP are relatively small. Per capita income in the richest state (Connecticut) is only 60% higher than in the poorest (New Mexico). To an extent, it also reflects the net effect of a vast array of program-specific grants that the federal government makes to states. There are approximately 600 of these, accounting for about 20% of average state resources. To a small degree, these are geographically redistributional. On average, poor states receive a slightly higher proportion than richer ones. But the correlation is weak. The overall effect of federal transfers is to reallocate variations in state per capita income, rather than reduce them. 44 42 Spahn, Paul Bernd; 1998; 'Policy Coordination and Control with Decentralized Government' Frankfurter Volkswirtschaftliche Diskussionsbeitrage (Germany); No. 88:1-67. 43 Individual state government make large scale transfers to local government, however, particularly to equalize school spending. 44 Data excludes two outliers (the federal district and the state of Alaska). Income data is from Harvard University/Office of Senator Moynihan, The Federal Budget and the States, (December, 1999); fiscal data for 1997 from US government, 1997 Census of Governments 43 132. Efficiency Impacts: Transfers can also be justified on the grounds that they reduce the economic distortions or high administrative costs that would arise if subnational governments relied only on their own tax bases. The characteristics that make for a good local tax also make for high costs of administration.45 To function as a local tax, the incidence of a tax must be borne locally. To be cheaply administered, a tax must be imposed on a large volume of taxable activity flowing through a small number of collection points. Few taxes can meet both criteria simultaneously. Taxes that meet the test of localized incidence (such as property or retail sales taxes) tend to involve large numbers of small-scale taxpayers. Taxes that meet the ease of administration test (such as taxes on manufacturing and imports and origin-based value-added taxes tend to involve large-scale inter-jurisdictional incidence shifting (as noted earlier). For these reasons, transfers may be justified as a substitute for local taxes. 133. At the state level, this argument might apply to the ICMS. It could be argued that the interstate tax exporting arising from the ICMS is so egregious that the tax should be abolished and replaced by an increase in federal transfers. In this respect, Brazil would be following the example of other federal countries. The other Latin American federations- Argentina and Mexico-impose a VAT at the federal level, sharing the proceeds with state governments on the basis of population and other factors. The major European federations-Germany and Russia--similarly impose the VAT at the federal level. Germany redistributes part of the VAT among the Lander on a non-origin basis. 46 Among large federal countries, only Canada permits subnational governments to impose a value added tax (in conjunction with the federal VAT). 134. Proposals to federalize the ICMS have been under consideration in Brazil since the mid-1990's. The most recent proposal of the Ministry of Finance would abolish all existing taxes on sales (including, at the federal level, the IPI, COFINS, and PIS/PASEP, as well as the state ICMS and the municipal ISS) and create a new federally-administered value added tax. Roughly half (49.4%) of the new VAT would be retained by the federal government. A total of 48.8% would be transferred to the state governments-35% on the basis of origin, 1.8% on the basis of the FPE criteria, and 12% through a fiscal equalization fund. The remaining 1.8% would be transferred directly to municipios under FPM criteria. (Municipios would also receive 25% of the states' share of the federal VAT.) To maintain some degree of fiscal autonomy at the subnational level, states would be assigned a new tax on selected goods and services, including fuel, electric energy, tobacco, beverages, cars, and telecommunications. It would be imposed only on sales to final demand, and would incorporate the federal VAT in its basis for calculation. Municipios would be assigned a broad based sales tax, which would exclude items subject to the new state tax. 47 The Ministry of Finance proposal would not eliminate tax 45 Bird, R.; "Rethinking Tax Assigmnent, The Need for Better Subnational Taxes" mimeo, April 1999. 46 Although the Russian VAT is legally a federal tax, in practice provinces (oblasts and republics) retain part of what is collected in their territories. Note that the U.S. has no VAT. 47 Projeto de Emenda Constitucional Art 155 inciso I. 44 exporting entirely. Because 35% of the federal VAT would be distributed to states on the basis of origin, states that are net exporters would continue to be able to export part of their tax burden onto other jurisdictions. It would, however, reduce the scale of tax exporting, while substantially increasing the scale of intergovernmental transfers. 135. But there are other, less dramatic solutions to the problem of tax exporting. One interesting solution has been put forth by Brazilian economist Ricardo Varsano. Rather than eliminate the state VAT and substitute a selective sales tax, it would retain the state VAT, but eliminate interstate tax exporting by imposing a zero rate on interstate sales. Manufactured goods consumed within the state of origin would continue to be taxed in the state of origin. Goods destined for other states would be zero rated. Because registered traders or final consumers in the state of destination would have no prior VAT to credit against their tax liabilities, all prior stages of manufacturing and distribution would be subject to the VAT in the state of destination. 136. One drawback to this solution is the strong incentive for fraud that it creates. Because interstate sales are zero-rated (while intra-state sales face rates as high as 25%), firms have a strong incentive to declare intrastate sales as interstate sales. A solution to this problem has been proposed Varsano. Under the Varsano proposal, the federal government would impose a an additional federal VAT on interstate sales-a so-called compensating VAT (CVAT)-at a rate equal to the average state rate on intrastate sales. State GDP and Nkincipal Fvenues 2.5 2.0 ~~~~~ 1.5 *~~~~~~~~~~~~~~~~~~~~~ revlicap t: IDD E a~~~~~~~~~~~~~~~~~~~~~~~ gdpfcap 01.0 Thus firms would face roughly the same rate of VAT on all sales, whether inter-or interstate. The CVAT would be (indirectly) transferred to the destination state once the good was shipped across state lines. 48 48 To simplify administration, the federal government would not directly transfer the CVAT. Instead, an importing firm would be allowed to credit the CVAT against its (newly created) federal VAT. The importing firm would therefore pay a state VAT based on the gross value of its sales, with no credit for the VAT paid on interstate purchases. Because it would deduct the CVAT from its federal VAT liability, its combined tax burden would be the roughly same as if it were still able to deduct a VAT on interstate purchases, as would the net revenues accruing to the state. 45 137. Over the long term, the ICMS could be phased out and replaced by more direct form of taxation. U.S. state governments, for example, rely heavily on retail sales taxes. Retail sales taxes have generally not performed well in developing countries, due to the difficulty of enforcing compliance by the multitudes of small scale, often unregistered, retailers. 9 Some parts of the base are easier to tax than others, however. Fuels, electric energy and telecommunications have proven to be the most productive sectors of the ICMS, for example, largely because they are relatively easy to administer. (Fuel taxes are imposed at the refinery level and then passed onto consumers in the form of higher prices. The ICMS on power and telecommunications, similarly, is collected from large scale utility companies. Beverages and tobacco taxes are collected at the manufacturing level, on the basis of imputed retail prices. Assuming that states would be allowed to continue administering these taxes through so-called tax substitution, the prospects for a selective sales tax are fairly good. Over the longer term, the basis for a state retail sales tax could be expanded to include other forms of retail sales, as retailers become increasingly formalized, and tax administration improves. 138. Over the long term, a state income tax also bears consideration. At present, the coverage of the federal personal income tax is too narrow to serve as a principal revenue source for both the federal and state governments. Its burden tends to fall disproportionately on employees in the formal sector-including the government-where it can be imposed through withholding. It falls lightly on self employed professionals and those employed in the informal sector. As the coverage of the personal income tax improves, however, it could serve as a replacement for the state VAT. 139. At the municipal level, the case for transfers on efficiency grounds is even more questionable. The tax bases already assigned to municipal governments are under- exploited. As shown in Figure 4, tax collections in all but the largest Brazilian municipalities amount to a derisory sum. Total property tax collection account for only 1.2% of national tax collections. Excluding Sao Paulo and Rio de Janeiro, yields averaged just R$24 per capita in 1997. Ostensibly, this is due to problems in tax administration. Years-and sometimes decades-can pass before new construction or changes in property use are reflected on property tax rolls. The system used to index property valuations fail to capture increases in local real estate prices. Collection inefficiency is also poor. In principle, delinquent accounts are subject to fines and interest penalties. Municipalities can impose a variety of administrative tools to enforce payment, including liens on title. If these measures are unsuccessful, delinquent accounts can be referred to the courts, which have the authority to order the sale of delinquent properties for back taxes. This ultimately remedy is not used widely enough or consistently enough to have an impact on taxpayer behavior, however. The risk of losing a property is not sufficiently credible to induce taxpayers to pay taxes voluntarily. 140. The technical solutions to these problems are not far to seek. Municipios can survey properties more frequently to capture new construction and changes in use. They 49 Bird, R. Rethinking Tax Assignment, The Need for Better Subnational Taxes (mimeo, April 1999; McClure Charles, "The Brazilian Tax Assigmnent Problem: Ends, Means and Constraints" in Proceedings of International Monetary Fund Seminar on Decentralization, Washington DC, 2000. 46 can regularly revise the tables used to convert physical characteristics into market values, in order to capture relative changes in land prices. And they can improve collection, by aggressively pursuing major delinquents in the courts. These remedies are widely known in Brazil and there has been long standing federal program of technical assistance to local government to Figure 4 Composition of Municipal Revenues by Population Size implement them. But Class the obstacle to 500 increasing property tax 450 revenues is ultimately 300 political. As in many 8350 ImE 300 M 1other current countries, the property rX 250 E transfers tax raises political _ 12500 ii m 0 taxes opposition far out of 100 _ J= f proportion to its 50 _ _ - -f- _ _ _ revenue yield. As a Less 20.000 - 50.000 -Over Rio de Sao result, property tax than 49.999 199.000 200,000 Janeiro Paulo administration tends to 20,000 Ske Clags be neglected except in times of fiscal crisis or when all other sources of revenue have been exhausted. 141. Under these conditions, it is not clear that the financing needs of municipios are so onerous as to justify the large scale of transfers that the FPM and the ICMS sharing provide. Where municipios provide primary education, FUNDEF is available. Where they are authorized to manage basic health care, SUS funding exists. State governments, not municipios, are responsible for public security. Under these conditions, the scale of transfers to municipalities appears to be excessive. 142. Itwill be difficult to improve the targeting of the transfer system until the broader issue of expenditure responsibility is resolved, however. As noted earlier, finance should follow function. With the division of responsibilities between states and municipios undefined, it is difficult to determine the appropriate objectives of the municipal transfer system. In this respect, FUNDEF and SUS show the way forward. By tying revenues to functions rather than jurisdictions, they help ensure adequate funding for these two sectors, regardless of which level of government is responsible. Macroeconomic impacts 143. One of the hallmarks of the 1988 Constitution was an increase the amount of revenue sharing. The 1988 Constitution increased the size of the participation funds (from 31% of federal income and industrial products taxes in 1985-88 to 44% in 1993). It also eliminated certain federal excise taxes--on fuels, electric energy, and telecommunications--and incorporated them in the base of the state ICMS, indirectly shifting revenues from the federal to the state level. Between the last pre-Constitution year (1988) and the year in which the majority of tax sharing was fully implemented 47 (1992) the federal share of tax revenues, net of transfers, dropped from 67% to 60%.5° The result, at the outset, was a significant shift in the distribution of net-after-transfer revenues among the three levels of government. Between 1988 and 1992, the federal share fell from 67% to 60%. 144. As the new Constitution made no commensurate devolution of expenditure responsibilities, this caused some observers to fear for Brazil's macro stability. It appeared that the growth in mandatory tax sharing would lead to deficits at the federal level. To forestall this, then-president Collor proposed a program of expenditure decentralizations intended to match the new division of revenues. This was rejected by Congress. 145. Nevertheless, a fiscal adjustment-albeit an ad hoc one--ultimately did occur at the federal level. First, the federal government increased the rates of taxes that it was not required to share with subnational governments. Social security taxes, for example, were increased dramatically. The federal government also increased its tax on financial operations and introduced a new temporary tax on checks. Second, the government clawed back some of the increases in revenue sharing mandated by the Constitution. Under the rubric of an emergency fund, it froze participation funds at their 1993 level through mid 1997. Federal discretionary transfers to subnational governments also fell. Between 1988 and 1990, the level of discretionary transfers, as a percent of GDP, dropped from 0.69% of GDP to 0.28%.? There were also some adjustments on the expenditure side. Although the government failed to gain legislative approval for a systematic transfer of functions, it was able to offload certain expenditure responsibilities onto subnational governments on an ad hoc basis The federal suburban railways in Sao Paulo and Rio de Janeiro were transferred to their respective state governments, for example. Certain federal highways were also transferred to the states. 146. The net effect of these measures was to largely neutralize the budgetary impact of the increase in revenue sharing. Contrary to initial expectations, the federal deficit remained roughly constant over the initial adjustment period. Although the federal share of net available resources declined over the period (from 67% of net resources in 1988 to 62% in 1993) the fall was relative, not absolute. Total federal resources in 1993 were roughly equal to their level in 1988, in real terms. But the result was a growing public sector. Because much of the adjustment was made through tax increases, the imbalance between revenue and expenditure decentralization was resolved not by redistributing expenditure but increasing the overall size of government. 147. The structure of Brazilian fiscal federalism has nevertheless raised doubts about the federal government's ability to conduct fiscal policy. Three characteristics of the system constrain the federal government's fiscal powers. First, a significant proportion of taxes are administered by subnational governments. As shown in the Table 9, 50 Figure excludes revenues from municipal taxes. 51 Serra, Jose and Jose Roberto Afonso, 1991, "Financas Publicas Municipais: Trajectoria e Mitos" in Conjuntura Economica, Rio de Janeiro: Fundac,o Getfflio Vargas. 48 subnational taxes as a group account for about one quarter of total taxes. State ICMS revenues alone account for about 22% of the total. While the federal government has some control over ICMS revenues-through its control on interstate tax rates-rates on intrastate sales, as well as the quality of state tax administration, remain outside its authority. Thus in principle state governnents could run expansionary fiscal policies while the federal government was trying to contract, or vice versa. 148. Second, roughly half of Fig. 6 Characteristics of National Tax Structure federal tax collections are subject to (% of total taxes) formula-driven revenue sharing. As noted above, 45% of the federal income and industrial products taxes, as well as 100% of the income tax on i subnational subnational employees, must be _ shared transferred automatically to state and o rdistortonary municipios. This again dilutes the \ remainder federal government's control over fiscal policy: half the proceeds of any increase in the income or industrial products must be transferred to other levels of government. 149. Finally, those taxes that are fully under the control of the federal government are undesirable on economic grounds. The taxes in this group largely consist of social security contributions, which are imposed in the form of turnover- or payroll taxes. Turnover taxes discriminate in favor of vertically integrated firms.52 Payroll taxes place the burden of taxation on workers in the formal sector-thus encouraging informality and/or pricing Brazilian labor out of international markets. 150. As a result, the federal government faces a Hobbesian choice: if it increases the personal or corporate income tax, half the proceeds must be shared with the states and municipios. If it increases payroll or turnover taxes, it increases the economic costs of those two taxes. 151. The Ministry of Finance's tax reform proposal would address this problem By assigning the federal government a VAT to replace the existing turnover and industrial products taxes, it puts a more a Figure 6 Who Collects the Taxes? efficient tax instrument in the hands Intemnational Comparisons of the federal government. By 120% substituting a selected sales tax for *2 80%- _ = i Dlocalthe state value added tax, it increases e 40%- _ lWt t iit | t Osate federal control over tax policy. But ie 20% 'A 0% _ Cltt the need to increase the federal sermany us brazil canada government's control does not appear urgent. Although a large proportion 52 Because the federal IPI can be deducted from a firm's ICMS obligation, it is not a tumover tax. 49 of taxes are under subnational administration, it is not clear that this cripples the federal government's ability to make fiscal policy-certainly no more so than in other large federal countries. Figure 6 compares the share of total taxes administered by each level of government in selected federal countries. Subject to the caveats in the footnote below, it suggests that the federal government in Brazil has about as much control over aggregate fiscal policy as Germany, the U.S. or Canada. 53 PRIORITIES 152. Brazil's federal system does not require a drastic overhaul. The fundamental terms of the 'federal bargain' are well established and work reasonably well. But it would benefit from repair. In the short term, there are two priorities. First, the Government should strictly enforce the terms of the LRF (and related controls on the banking sector). This is needed not merely to control deficits themselves but to send an important signal to governors and mayors that the era of federal bailouts is at an end. 153. Second, the pension benefits of active subnational civil servants must be reduced. While the recent toughening of retirement criteria, combined with increased employee contributions and the use of privatization proceeds to capitalize retirement funds will help reduce the scale of pension liabilities, no solution will be viable without a reduction in benefits. And without a solution to the pension problem, the prospects for enforcing the LRF are dim. 154. In the longer term, more fundamental change in the structure of subnational government should be considered. Despite substantial reforms over the last fifteen years, Brazil still retains some of the characteristics of an older, 'clientelistic' state at the subnational level. Governors and mayors can still dispense favors-or withhold them- without being held accountable for the performance of specific services. Reform will require action on many fronts. To clarify the assignment of functions, new approaches to organization, including explicit distinctions between categories of municipalities, may have to be made. To reduce the arbitrariness of intergovernmental transfers, better targeting and earmarking may be required (along the lines already imposed by FUNDEF.) Over the long term, revenue sharing may be scaled back in favor of greater reliance on local benefit taxes. In pursuing such reforms, Brazil has a variety of 53 Note: The federal share of national tax collections does not necessarily reflect the extent of federal control over national tax policy. Federal governments can control subnationally-administered taxes through controls over their rates. By the same token, federal governments can be stymied in their efforts to change federal tax rates if a large proportion of the receipts are shared with subnational government. Notes to data: Revenues include non-tax revenues (operating surpluses of enterprises, administrative fees and charges, employee contributions to the affected governments pension fluds) which are particularly important at the state level in the US (accounting for 40% of own source revenues) and at the local level in all four countries, accounting for 32% of local own source revenue in Germany, 38% in the U.S., 42% in Brazil, and 28% in Canada. Note that for Germany, state own source revenues include Lander shares of the value added tax, which is administered by the states under uniform federal legislation, and redistributed among states according to federally-determined formulas. It also includes the Lander's share of the income tax, which-though retained on the basis of origin-is subject to redistribution under the finanzausgleich. The proportion of own source revenues not subject to distribution is insignificant. 50 international examples to draw on. It should also draw on its own recent innovations, particularly in the health and education sectors. 7 VENEZUELA G _ F40 COLOMBIA \ L ) ( SURINAE1GUIAN BRAZIL -W A A ~ ~~~~~A TL AN T IC OC EA NBRAZIL 0' A\L > \' RIVERS Manou 0% L ~~~~~~~~~~~~~~~~~~~~~~~@ STATE CAPITALS * NATIONAL CAPITAL AMAZONAS ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~STATE BOUNDARIES X J~~~~~~~~~~~~~~~MARANHAO>I AINLvlA CEARA tal INTERNATIONAL BOUNDARIES rt ~~~~~~~~~~~~~~~~~~~~~EN4$U Recife -10° 2ot S A This map was produced by tse P E R U g N RONDONIAX 8 MA70 fTOC NTINS Reservoir Map Design Unit of The World Bank. ROND IA 10, ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~The boundaries, color, denominations 5A I A ~~~~~~~~~~~~~and anyother infomnatian shawn an P E R U ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~this map da nat imply, an the port of \ ! / 0 | GROSSO / | > ef r3/oIvDdor The World Bank Gmup, any judgment Slaor an the legal ntotus af any territory, or any endonsement ar acceptance of such boundaries. P \l \8 \; ? W SA~~~~~~~~~~~~~~~~GOnd HAUO %vrizonte EPE2 )1BRAZI LTO Ct/iab6E) BRASiLl ~ ~ ~ ~ ~ PeL't,i BRZI )CHIm A s ooneirOIA?UVIAS M N S~ ~ ~ ~~~~~OEN " ' P A C ~~~~~~~~O IV R I C (FS R)OCEAN OCAN H X DO SUL ErAiNzon5 9 d y jRorism6pol(AUGENINA ATLANTIC PARAG~~~~~~~A~~~Ylj T L A N T~~~~~~ A OEA CHIE! A 250 500 750 1000UKLOMETERSI j ) ! > RIO R5A~~~ULDt aAlegre 0 200 400 OMILE_ )(' t ( If" E' w1 OCEAN \.4' IMAGING Report No.: 22523 BR Type: SR