POLICY RkESEARCH WORKING PAPER 2121 Fiscal Management Argentina and Brazil f~isaI ~ anag m entArgentina and Brazil -- two of the most decentralized reae ai 'ern Lra iespublic sectors in Latin Amenca and (along with Colombia and India) among the most decentralized democracies in the developing world - V1ilam D111inger faced similar problems in the Steven B. Webb 1 980s: excessive public deficits and high infiation exacerbated by subnational deficits. In the I 990s, Argentina was more successful at macroeconomic stabilization, partly because JmPosed harder budget constraints on the pubic sector nationally and partly because it had stronger party control of both national legislators and subrationai governments. The World Bank Latin America and the Caribbean Region Poverty Reduction and Economic Management MNlay 1999 P o itc i st ALH Won RM X P AP ER 2 121A Summary findings In shifting to decentralized public finances, a country's Dillinger and Webb investigate how, and how central government laces certain fiscal management successfully, Argentina and Brazil dealt with these problems. First, during and soon after the transition, problems in the 1990s. unless it reduces spending or increases its own tax In both countries, subnational governments account resources, the central government tends to have higher for about half of public spending and are vigorous deficits as it shifts fiscal resources to subnarional democracies in most (especially the largest) jurisdictions. governments through transfers, revenue sharing, or The return to democracy in the 1980s revived and delegation of tax bases. Reducing spending is hard not strengthened long-standing federal practices while only because cuts are always hard but because weakening macroeconomic performance, resulting In subnational governments might not take on expected unsustainable fiscal deficits, high inflation, sometimes tasks, leaving the central government with a legal or hyperinflation, and low or negative growth. Occasional political obligation to continue spending for certain tabilization plans failed within a few years. Then services. Argentina (in 1991) and Brazil (in 1994) Introduced Second, after decentralization, the local or state successful stabilization plans. government faces popular pressure to spend more and National issues were important in preventing and then tax less, creating the tendency to run deficits. This bringing about macroeconomic stabilization, but so were tendency can be a problem if subnational governments intergovernmental fiscal relations and the fiscal and their creditors expect or rely on bailouts by the management of subuational governments. State deficits central government. Econometric evidence from 32 large and federal transfers were often our of control in the industrial and developing countries indicates that higher I 980s, contributing to national macroeconomic subnational spending and deficits lead to greater national problems. Stabilization programs in the 1990s needed to deficits. establish control, and self-control, over subnational spending and borrowing. This paper - a product of Poverty Redu,,,ction and Economic Management, Latin America and the Caribbean Region - is part of the LCR regional studies program on fiscal decentralization in Latin America. Copies of the paper are available free from the World Banik, 18 18 H Street NW, Washington, DC 20433. Please contact Alex Lara, room 18-1 26, telephone 202-4.58-8148, fax 202-522-21119, Internet address alaraCc worldbank.org. Policy Research WorkingPapers are also posted on the Web at http :/'jwwvw.woridban-l. org/hzlt-,l/dec/Publications/Wn,'rkpapers/lhome.htinl. The authors may be contacted at wdillinger(mworldbank.org or sm(ebbolworldbank.tr. May 1999. (38 pages) I TI)e Polic' R2search IV " Paper ',eriesdsseminates the fiodings of work in progress to encourage the exrhaoge of ideas about ieteopineow issifes. An qbfectic ie seri-es is to get the findings outr quickly, enen ithe presentations are less than fwlly polished. The papers carry tenarnesofthe atithorsondshould he rited acrordingly. Thefindings, intec tretations, and conclusions expressed in this Paper are erloreN' of autho. They do not necessaril h represent the view nthe World Bank, its .xecugtie Directors, or the ctbiato tplans fiewtna Produced by the Policy Research Dissemination Center Fiscal Management in Federal Democracies: Argentina and Brazil William Dillinger and Steven B. Webb World Bank DRAFT comments welcome swebb@worldbank.org or wdillinger@worldbank.org We wish to thank Barry Ames, Suman Bery, Richard Bird, Mauricio Carrizosa Eduardo Gomez, Stephan Haggard, Paul Levy, Jenny Litvack, Joao Oliveira, David Rosenblatt, Barry Weingast, and Eliza Willis for their useful comments on earlier drafts; they bear no responsibility for any remaining errors. Cecilia Briceno provided assistance with the data. Elizabeth Forsythe edited the text. Fiscal Management in Federal Democracies: Argentina and Brazil William Dillinger and Steven B. Webb World Bank Decentralizing public finances has many advantages but also raises potential problems, especially in the area of fiscal management. One problem arises mainly during and soon after the transition to fiscal decentralization, when the central government, unless it reduces spending or increases its own tax resources, tends to have higher deficits as it shifts fiscal resources to subnational governments, via transfers, revenue sharing, or delegation of tax bases. Reducing spending is hard not only because cuts are always hard but also because subnational governments might not take on the tasks anticipated, leaving the central government with a legal or political obligation to continue spending for certain services. The second problem arises after decentralization, when the local or state government faces popular pressures to spend more and tax less, creating the tendency to run deficits. This universal tendency can be especially problematic if subnational governments and their creditors expect or rely on central government bailouts. Econometric evidence from 32 large industrial and developing countries indicates that higher subnational spending and deficits lead to greater deficits atthe national level (Fornasari, Webb, and Zou 1998; see also Treisman 1998). This report investigates how these problems arose in Argentina and Brazil as they sought to decentralize in the 1980s and how each country dealt with them. We seek to understand the institutional arrangements-fiscal and political-that contributed to success or failure. Argentina and Brazil have two of the most decentralized public sectors in Latin America and, along with Colombia and India, are the most decentralized democracies in the developing world. Subnational governments (SNGs) account for about half of public spending and are vigorous democracies in most jurisdictions, especially the largest. In both countries in the 1980s, the return to democracy revived and strengthened long-standing federal practices while weakening macroeconomic performance, resulting in unsustainable fiscal deficits, high inflation, sometimes hyperinflation, and low or negative growth. See table 1. The occasional stabilization plans each failed within a few years. Then in the 1990s-Argentina in 1991 and Brazil in 1994-both countries introduced stabilization plans that have had some success.' ,While national-level issues played the largest role first in preventing and then in bringing about macroeconomic stabilization, intergovernmental fiscal relations and the fiscal management of subnational governments were influential as well. State deficits and federal transfers often were out of control in the 1980s, contributing to the overall macroeconomic problems of both countries. Similarly, stabilization programs in the 1990s have sought to establish control and self-control over subnational spending and borrowing. 1 Table 1. Growth, Inflation, and Fiscal Balances in Argentina and Brazil, 1983-97 (percent) Country and indicator 1983-87 1988-90 1991-94 1995-97 Argentina Real growth (annual average) 1.77 -3.43 8.89 2.69 Inflation (annual average) 370 1,910 53 1.3 Overall balance to GDP Central -5.11 -0.87 0.27 -1.32 Subnational -4.90 -1.99 -1.33 -0.88 Primary balance to GDP Central -3.44 -0.07 1.40 0.28 Subnational -4.84 -1.76 -1.16 -0.56 Brazil Real growth (annual average) 4.27 1.61 2.73 3.52 Inflation (annual average) 186 1,670 1,350 30 Operational balance to GDP* Central 0.28 -1.69 Subnational -0.05 -2.16 Public enterprises -0.20 -0.46 Primary balance to GDP Central 1.68 0.24 Subnational 0.75 -0.51 Public enterprises 0.73 0.02 * Does not include the inflation component of nominal interest. Source: Argentine and Brazilian Ministries of Finance, International Monetary Fund GFS and staff estimates, and authors' estimates. In Argentina in the 1980s, provinces accounted for at least half of the public sector deficits that were fueling hyperinflation. By the end of the decade, hyperinflation had greatly eroded the real value of domestic debt, giving governments at all levels almost a clean slate on that side, although external debts remained substantial. Figure 1 shows the path of subnational debts following the stabilization in 1991. 2 Since then, the provincial debt has grown little in pesos (or dollars) and has declined as a share of gross domestic product (GDP) and of provincial revenues. A large share of the debt is owed to the private sector and is being serviced. The result of reforms made at the time of stabilization, this effort has helped to Figure 1. Provincial Debt Stocks C- i 15.00 "a 10.00 - ' - Argentina * 5.00 "- --- "- """'""Brazil I 0.00 motivate provinces to behave in a fiscally responsible manner. Source: World Bank and International Monetary Fund staff estimates. Argentina stock is calculated from 1996 base. Brazil stock is net domestic debt only. 2 In Brazil, in contrast, subnational debt as a share of GDP started the 1990s at a level similar to Argentina's but had more than doubled by 1997. Most of this debt was owed to the central government or to state banks, and up until the debt-rescheduling agreements in 1998, much of it was not being serviced by the states. Interest was being capitalized. Thus, it was a direct fiscal problem for the central government as well as part of the deficit of overall public sector. State debt accounted for one-third of the increase in domestic public debt from 1994 to 1997. The Brazilian stabilization program of 1994 left unchanged many rules and institutions that motivated the states to let their debt grow. A round of debt agreements completed in 1998 changed some of the rules, but it remains to be seen whether they will achieve a lasting resolution. I. Explanatory Variables The explanatory variables for the outcomes in Argentina and Brazil fall into two main groups: fiscal institutions and intergovernmental political relations. In each area we identify as hypotheses the conditions expected to lessen the likelihood that fiscal decentralization will lead to excessive deficits. A country will rarely meet all the conditions but will experience severe problems if too few of the conditions are met. The challenge of this research is to discover which combinations are adequate to prevent fiscal decentralization from leading to unsustainable fiscal deficits, and which are not. Fiscal institutions For fiscal institutions, key questions arise in three areas: the division of authority over revenues, spending, and borrowing. What authority do SNGs have to raise revenue and control their spending? What authority does the central government have to shift the responsibility for spending to subnational governments and to reduce the transfer of resources in order to balance the central government budget? What budget constraints do SNGs face, and how are they set? Under what conditions can SNGs issue debt? Revenue autonomy. With fiscal decentralization, SNGs usually obtain certain tax bases, but for reasons of politics, equity, and efficiency, these bases rarely cover all the expenses of SNGs, which also receive some federal transfers. One view is that subnational governments have smaller deficits if they rely more on their own tax bases (and have the power to change tax rates on the margin), because they have the ability to adjust to shocks by increasing revenue. Also, relying on one's own resources may strengthen the incentives to control spending, and of course it reduces the burden on central government. Thus unsustainable overall public sector deficits are less likely under condition 1, subnational governments raise much of their own revenue. Furthermore, if present, transfers from the center to the subnational governments are less likely to be problematic under condition 2, transfers are specified by legal formula, not ad hoc. Expenditure autonomy. Without SNGs' autonomy over their expenditure, there is no fiscal decentralization and no macro fiscal problem likely to come of it. Giving SNGs autonomy over spending is, of course, the way in which decentralization can improve efficiency in matching the needs and desires of a diverse population. Two aspects are usually important regarding the effects on macro fiscal management. One is whether the central government can dictate which functions the subnational governments must take on, at least in exchange for receiving transfers from the center. If the central government can do this, central spending and deficits are contained. If it cannot, as in Brazil, the central government can find itself with a constitutional obligation and political expectation to continue providing some service, even after revenues or tax bases have been turned over to subnational governments with the understanding that they will do the task. The other issue is whether the SNG has authority to spend less, particularly to cut personnel, salaries, and pension benefits, collectively the largest single item of subnational expenditure. If central rules constrain this authority, then it is more difficult to reduce deficits, and expectations of a central government bailout are higher. Thus unsustainable deficits should be less likely under condition 3, the central government can effectively delegate functions to subnational governments to go along with the delegation of revenue sources, and under condition 4, subnational governments have authority to cut their costs. Borrowing constraints. Although tax and spending policies create fiscal pressures, whether they cause problems for macro fiscal management depends on whether the SNGs face hard budget constraints, limiting their ability to borrow or to seek other ways to increase their resources by spending more. A hard budget constraint is essential for getting proper fiscal behavior (Tanzi 1995; Ter-Minassian 1997; IDB 1997; Weingast 1995; Wildasin 1997). Some borrowing may be sustainable and good for development, and many central governments try to control it (Ter-Minassian and Craig 1997). Unsustainable deficits are less likely under condition 5, the central government strictly controls subnational borrowing ex ante. But how to achieve this in practice is not always clear when the subnational governments have considerable political autonomy. Pseudo-strict controls could make matters worse if central government approval creates the impression, and perhaps the self-fulfilling expectation, that the central government also has made a guarantee. To run deficits, subnational government must find a source of financing, which potentially includes contractual borrowing from private foreign or domestic banks (especially banks owned by the subnational governments), issuance of domestic or foreign bonds, and the running up of arrears to suppliers and personnel. A creditor and the subnational government would only agree to finance unsustainable deficits if both sides expected to gain, most likely though some sort of federal bailout. The bailout can take many forms, including allowing the financial system (implicitly insured by the government) to count as an asset a debt that is not being serviced. So unsustainable deficits are also less likely under condition 6, the central government credibly commits not to have bailouts, prohibiting explicit bailouts and forcing subnational governments to service their debts, and under condition 7, regulators force creditors to accept the losses implied by any failure to service debt. It is still an open question whether ex ante regulation or ex post enforcement of debt service is more effective in preventing excessive SNG borrowing. Often financing from the central bank is what loosens the budget constraint for the subnational governments, either directly by discounting subnational debt or indirectly by easing the national government's budget constraint or allowing commercial banks to roll over bad subnational debts. Unsustainable deficits are less likely under condition 8, the central bank (and bank supervisors) is more autonomous and has a strong anti-inflation mandate. Intergovernmental political relations The rules for intergovernmental fiscal relations and the way in which they are implemented result from the political relations between levels of government (Riker 1964; Stepan 1997; Willis and others 1997). Why does the central government say yes or no to an SNG's request for more resources? Why do the SNGs take on additional tasks or accept reduced transfers in order to help the central government balance its budget? When the national government makes substantial transfers to politically autonomous subnational governments, its ability to take a firm stance toward subnational governments depends most on the power of the president vis i vis the governors and mayors themselves and their representation in congress.3 Systems with strong presidencies should be better able than systems with weak presidencies to fend off pressures to cover the states' deficits. The presidency is the only office with a national constituency and as such is more likely to take into account the interests of the overall economy, while the members of congress represent regional constituencies as well as their parties. We can analyze the power relationships between the national and subnational levels in four steps, corresponding to four conditions. Unsustainable deficits arising from fiscal decentralization are less likely 4 under condition 9, presidents are constitutionally strong at the national level, and under condition 10, governors have little constitutional autonomy. These conditions are not necessarily beneficial for all aspects of governance, and of course fiscal decentralization presumes that governors have enough political autonomy to be considered a politically separate level of government rather than just a field representative of the central government. Governors and the president indirectly contend for resources, especially via congress and the parties, whose effects on the intergovernmental balance of power we must also consider. The central government should be able to maintain a harder budget constraint under condition 11, electoral rules orient congress toward national, not local, issues, and under condition 12, party discipline is strong. The balance between these forces depends on the interaction of the constitution, electoral procedures, and party discipline. The constitution is a given at most times, as are electoral procedures. Therefore, party discipline is the dependent outcome. But constitutions and electoral procedures change with some frequency in Latin America, and they reflect the political balances and party characteristics at the time of each change. I. Empirical Framework This section describes Argentina and Brazil in terms of these variables or conditions. Section III then recounts the recent SNG fiscal crisis in each country and examines how these variables affected the central government's ability to force SNGs to adjust. A final section draws conclusions from the two experiences. Argentina Fiscal institutions. Although the 1853 constitution limited the federal government's taxing power to taxes on foreign trade, it granted the power to impose domestic direct taxes "for a determined time period" and subsequently allowed the federal government to impose domestic indirect taxes concurrently with the provinces (Murphy 1995). During the 1930s, in response to a drop in foreign trade taxes, the federal government abolished most existing provincial taxes and introduced national income and sales taxes. To compensate provinces for lost revenue, it established a system of revenue sharing (known as coparticipation). Revenue sharing has become more complex but remains the backbone of provincial revenues. In 1994, shared revenues accounted for 64 percent of total provincial revenues. General revenue sharing (coparticipation) is the largest single transfer, accounting for the bulk of transfers: 72 percent in 1991 and 57 percent in 1997. The pool of taxes subject to coparticipation consists of the federal income tax, value added tax (VAT), and excise and asset taxes; that is, of all the major federal domestic taxes other than social security and fuel. Before the transfer, several deductions are made from the provinces' share and are given to the national social security system and to several special funds for the provinces. Of those destined for special provincial funds, the largest deduction is to compensate for the costs of education and health services transferred to provinces (introduced in 1992) and for the Fondo Conurbano, which ostensibly assists Buenos Aires province in providing basic services in the suburbs of Buenos Aires city. Deductions from the pool before coparticipation are also made to finance transfers to provinces in financial difficulty, but such transfers are limited to a small percentage of the pool. In addition to coparticipation, the federal government also shares its fuel tax revenues, allocating fixed shares to finance housing (FONAVI) and provincial road and infrastructure projects. Oil-producing provinces also receive directly a share of the royalties from oil companies. Dependence on federal revenue sharing per se has not weakened provincial power. The provincial share of revenues is theirs by right, and there is little explicit federal discretion in determining the amount or distribution of transferred funds. In the vast majority of transfers, the volume of funds subject to sharing is determined either as a fixed share of specified taxes or as a fixed amount in pesos. The distribution of funds 5 among provinces is determined largely by formula, with coefficients fixed in the 1988 coparticipation law. Nevertheless, the dependence on intergovernmental transfers gives the federal government some leverage over the provinces. Because transfers can be created or altered through legislation, new transfers can be offered as a quid pro quo for provincial compliance with federal initiatives. This technique was employed extensively to achieve a series of fiscal agreements in 1991-94, discussed below. Transfers also provide private banks with instruments to ensure that debt service is paid, because debt service can be deducted from transfers at the source. The importance of revenue sharing varies among provinces, but it has at least some importance to all of them. In Buenos Aires city, it is only 9 percent of total revenue; in large provinces it accounts for about half: Buenos Aires, 46 percent; C6rdoba, 51 percent; Santa Fe, 57 percent; and Mendoza, 58 percent. In small provinces, the transfers typically account for more than 75 percent of revenues, reaching 95 percent in La Rioja and Tierra del Fuego. The provinces generate about 40 percent of revenue from self-administered taxes. Four taxes--on gross receipts, property, automobiles, and stamps-account for the majority of tax revenue. Of this total, the tax on gross receipts accounts for about 60 percent. Provincial taxes are largely the purview of large provinces. In 1996, the city and province of Buenos Aires accounted for two-thirds of total provincial tax collection, and the five large provinces (including Buenos Aires city) accounted for 87 percent of the total. So Argentina does not meet condition 1, revenue independence of provinces, but it does meet condition 2, the aggregate limit on transfers. The decentralization law and the fiscal pacts with the provinces in the early 1990s clarified which functions, such as primary and secondary education, the states and municipalities had to take on. The provinces also had authority to cut their costs, although the usual politics and rules governing public employment did not make it easy. Also, in 1998 teacher protests at the national level led the federal government to intervene, putting a special tax on autos to pay for teacher wage increases. So by the mid- 1990s, conditions 3 and 4 held, although there was some marginal slippage in the late 1990s. In Argentina prior to 1991 provinces borrowed a lot, much of it from their own provincial banks, which then discounted the loans to the central bank, effectively giving provinces a share in the seigniorage and inflation tax. There were over 20 provincial banks, including two each in Mendoza and C6rdoba. In 1990 they provided more than 60 percent of the credit needs of provincial governments, and the central bank lent massive amounts of new rediscounts to prevent the collapse of several provincial banks, due to poor loan recovery and massive overstaffing. At present there are no specific limitations on the ability of provinces to borrow from commercial banks, and a variety of measures have largely eliminated the provincial banks as sources of credit to the provinces. The 1991 convertibility law ended the ability of provincial banks to rely on the central government as a lender of last resort, forcing them to pursue delinquent loans vigorously (including loans to provinces), or to go out of business or to get capital infusions from the provincial government. Provincial governments that still own banks have to pay in additional capital if needed to meet the required ratios. The central bank can no longer discount any loans from provincial banks, and there is only limited deposit insurance, fully funded by the banks themselves. Since 1996, banks actually monitor each other through the requirement for each bank to issue subordinated debt that other banks will be willing to hold. Tight regulations limit the lending of a bank any one borrower, including a province, and require banks to provision against loans not being serviced. Provincial bonds and some provincial loans are subject to ex ante government controls. Such bonds have to be reviewed and registered by the Ministry of Economy. Bonds issued abroad or in foreign currency also need to be cleared with the central bank. Both reportedly exercise this role with a light touch. Although 6 this hands-off attitude seems at first like a serious gap in the hard budget constraint facing SNGs, it has the advantage of forestalling the expectation that the federal government will take responsibility for the quality of provincial debt and might provide an eventual bailout. So condition 5 is largely unmet. In addition, since 1993 a Ministry of Economy resolution explicitly prohibits any federal agency from using its resources to pay a creditor on behalf of a province. The federal government also, without hesitation, deducts debt service from the coparticipation transfers in cases where those transfers were used as collateral for provincial borrowing. This discourages provinces from borrowing in this way, although it makes creditors more willing to lend. So Argentina meets conditions 6 and 8, but not conditions 5 and 7. Intergovernmental relations. Argentina has a federal structure with 23 provinces and the city of Buenos Aires (which was accorded provincial status in 1994). Argentine federalism grew out of extended and bloody struggles in the nineteenth century between unitarists and federalists. In 1853, all provinces except Buenos Aires agreed to form a federation. The province of Buenos Aires remained outside the federation until 1859, when it joined and the city of Buenos Aires was designated as the national capital. The nineteenth century left a legacy of conflicts between center and periphery, imposed solutions, and special treatment for Buenos Aires. The original federal constitution of 1853 remains in force, although it was amended in 1860, 1866, 1898, 1957, and 1994 and was abrogated during periods of military rule. Thus Argentina's recent decentralization revived an old pattern in the country. The constitution establishes a presidential system with a bicameral legislature. The president is directly elected for a four-year term and since 1994 can succeed himself once. (Prior to 1994, the president was elected by an electoral college. This encouraged provincial identity in national politics and gave small provinces disproportionate representation relative to population.) The 1853 constitution provides for a strong central executive. The president has the power to submit legislation directly to congress and to approve or veto all legislative acts of congress. He can suspend civil liberties by declaring a state of siege (unilaterally, when congress is not in session, although members must confirm when they return). According to one observer, the "power of the president is so extensive that the presidency is the center of the political system. Patronage and participation in policymaking are solely a presidential prerogative. Once in office, the president is not legally required to seek wide backing for his policies. Historically, congress does not act as a check on the presidency, but spends most of its time debating the president's program, then passing a flurry of bills at end of session" (Rudolph 1985). Political parties are stable: the UCR (Radical party) was founded in 1890 and the Justicialista party (Peronist) was founded in 1946. As of 1995, the Justicialistas held the presidency and 15 of 24 governorships. Together, the Radicals and Justicialistas held more than 80 percent of seats in congress in the early 1990s. The two parties are more disciplined than parties in Brazil or the United States, but less so than in Mexico, United Kingdom, or Germany, Candidates are nominated on closed-party lists chosen by the provincial party organization, which helps to maintain party discipline. The party-line pattern of roll call votes masks considerable bargaining with provincial political interests prior to the vote (Eaton 1998). The upper chamber of congress (the senate) consists of three senators per province. Senators are directly elected, simultaneously, with two seats reserved for the party getting a plurality and one for the second vote getter. (Before the 1994 reforms, each province had two senators, elected by provincial legislatures, on staggered terms.) Members of the lower house (chamber of deputies) are directly elected, with the number of deputies determined by population, subject to a minimum per province. The minimum is five, which substantially increases the per capita representation of provinces with less population in the lower house as well as the senate (Jones 1998; Gibson, Calvo, and Falleti 1998). In elections for the lower 7 house, each.province serves as a single electoral district, with candidates running at large on party slates and with the number of seats assigned to each party based on its proportion of votes. The governor typically takes the lead in negotiating bargains between the state and federal governments. Because his power is greatly enhanced when his party is in power nationally, the governor has an incentive to be part of his president's team. The autonomy of the governors is also checked by the power of the federal government to intervene-that is, to dismiss a sitting governor and substitute a federal appointee. The 1853 constitution and all those since give the federal government broad powers of intervention. Federal government may intervene in a province in order to enforce federal laws or guarantee the republican form of government. Because the federal constitution defines the governor's duties to include upholding federal laws and the federal government has the power to define the meaning of republicanism, federal government has the power to assume control of a province at virtually any time. The province of C6rdoba, for example, has been under federal intervention for 24 of the past 50 years, including virtually the entire period from 1973 to 1983 (Province of C6rdoba 1997). Both military and civilian governments have used this power to ensure acquiescence with national policy. The 1994 constitution sought to increase the autonomy of the governors by requiring explicit congressional approval for federal interventions, and none has taken place since then. Still, condition 10--constitutional weakness of governors-is met in Argentina; indeed, Argentina seems to meet all of the political conditions: 9, 10, 11, and 12. Brazil Fiscal institutions. Brazil's 1988 constitution is explicit in dividing up revenues. It assigns specific tax bases to each level of government and creates a system of tax sharing that substantially redistributes revenue among levels of government and among regions. The state governments are assigned a value added tax, which is assessed and collected directly by each state government. Although the senate sets VAT rates, states have some flexibility to set the rates on intrastate sales, subject to a floor and ceiling set by the senate. The VAT is the highest-yielding source of revenue in Brazil, and it gives the states an independent base of power, particularly in the wealthy southeast, where it is the principal source of state revenues. The federal constitution requires states to transfer 25 percent of the proceeds of the VAT to the municipalities within their territory, partly on the basis of the "origin" of tax collections and partly according to formulas devised by each state legislature. States can also tax automobiles and property. Municipalities are assigned an urban property tax, a tax on services (the ISS), and a real estate transaction tax. The ISS is assessed and collected by the municipality at rates set by the municipality but subject to a maximum fixed by federal law. The property tax is also assessed and collected locally (Alfonso 1997; Bonfim and Shah 1994). Brazil's transfer system includes transfers from the federal government to states, federal transfers to the municipalities, and constitutionally mandated transfers from the states to the municipalities. The federal component of Brazil's intergovernmental tax sharing system consists of fixed shares of the federal government's two principal taxes: the income tax and the industrial products tax. The 1988 constitution requires the federal government to transfer 21.5 percent of the revenues from these two taxes to the states. Of this amount, 85 percent 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. Thus the states with the largest economies and tax bases of their own (south and southeast) rely very little on revenue sharing. Within each group of states, 95 percent of the funds is distributed among states on the basis of population and per capita income (with poorer states receiving proportionately more). The remaining 5 percent is distributed on the basis of geographic area. Another 22.5 percent of the income and industrial products taxes is distributed to municipalities. Of this, 10 percent is transferred to state capitals on the basis of population and per capita income of the surrounding state. The remaining 90 percent is distributed among all other municipalities on the basis of population and per capita state income. 8 Compared with previous arrangements, the 1988 constitution shared substantially more taxes with subnational levels, doubling the size of the participation funds (relative to their 1967 shares). The constitution also expanded the base of the state VAT at the expense of the federal government, by abolishing federal taxes on fuel, electric power, mining, and transport and incorporating their bases into the VAT. At the same time, the constitution also forced the states to sacrifice revenue in the interest of the municipalities, by increasing the municipios' share of the VAT from 20 to 25 percent. Between the last year before the constitution (1987) and the year in which the majority of tax sharing was fully implemented (1992), the federal share of tax revenues, net of transfers, dropped from 61 to 52 percent. The state share increased, due to the growth of federal transfers and the broadening of the VAT base. This increase was partly offset, however, by the states' obligation to increase their own transfers to the municipalities. The effect was most striking at the municipal level. The municipal share of net after-transfer revenues increased roughly 40 percent over the six-year period, from 12 percent in 1987 to 17 percent in 1992. See table 2. The participation funds Table 2. Federal and Subnational Shares of Tax Revenues in Brazil, substantially redistribute funds Before and After Transfers, 1987-92 among regions. On average, states (percent) of the north, northeast, and center- Share before transfers Share after transfers west receive twice the level of participation funds as states of the Year Federal State Municial Federal State Municial south and southeast. Because the 1987 70 27 3 61 28 12 majority of federal taxes are 1988 70 27 3 61 27 11 generated in the south and 1989 65 32 3 57 30 13 southeast, the extent of regional 1990 65 32 4 54 30 16 redistribution is even greater than 1991 62 32 6 51 30 18 this population-based calculation 1992 62 32 6 52 31 17 would suggest. Source: Afonso 1994. Taxes include federal social security contributions Transfers are mostly fixed by formula (not counting debt bailouts), so condition 2 is filled. Condition 1-revenue self-sufficiency-is met for the large states, but not the small ones. Following a pattern established in earlier constitutions, the 1988 constitution explicitly defines the powers of the federal government, while granting states "all powers not otherwise prohibited by this constitution." Among the responsibilities assigned to the federal government are national defense and social security. The federal government has exclusive control over the emission of currency, control of public debt regulation of interstate and foreign trade, and establishment of the "general norms of public employment." The constitution also delineates concurrent responsibilities of the federal government and the states, including tax legislation, education, and social assistance, and specifies that federal law will be limited to "9general norms" but will prevail in the event of conflict with state legislation. Although the 1988 constitution was largely decentralizing in spirit, it extended central control in areas such as personnel. The federal constitution defines the rights of public sector employees at all three levels of government. Underfthe constitution, governments can neither dismiss redundant civil servants nor reduce salaries in nominal terms. Public employees have the right to retire after only 35 years of employment (fewer for women and teachers), with a pension equal to their exit salary plus any subsequent increases granted to their previous position (some revisions were made in late 1998). These constitutional guarantees severely restrict the states' ability to adjust. State governors are directly elected for four-year terms (and as of 1998 can succeed themselves once). States have unicameral legislatures, whose members are elected at large by proportional representation. Members serve four-year terms with no limit on successive terms. This structure is repeated 9 at the municipal level, where mayors are directly elected and council members are elected at large by proportional representation. Unlike other federal constitutions, which typically leave the definition of municipal governments to the states, the 1988 federal constitution establishes both the state and municipal levels of government. The constitution provides no clear division of functional responsibilities between the states and municipal governments, merely reserving for the municipalities "the power to legislate over subjects [assuntos] of local interest, and to provide ... services of local public interest." Subnational governments thus consist of two equally "sovereign" levels of government with no clear functional boundary between them. Therefore, in the area of expenditures, Brazil meets neither condition 3 nor 4. Expenditure responsibilities are not clearly defined--often remaining at the federal level-and subnational government has little autonomy to cut costs. (Reforms in 1998-99 could change the latter.) In Brazil, the federal government has tried with increasing intensity to control state borrowing. The constitution gives the senate authority to regulate all state borrowing. It has traditionally set guidelines for state borrowing, based on the state's existing stock of debt, revenues, and debt service obligations, but the senate can grant exceptions to it and has done so. The executive branch imposes a variety of regulations and controls on borrowing according to the sources of credit. Borrowing from the domestic banking sector falls under the purview of the central bank. Since 1993, various regulations have restrained private banks from increasing their holdings of state debt, but the complexity of these regulations and their subsequent adjustments to accommodate deficits have weakened their credibility.4 Central bank regulations prohibit a state from borrowing from its own commercial bank, although this has not always been strictly observed and has been evaded blatantly in some cases. Contractors on state projects have been known to borrow from the state bank with a letter of comfort from the state government and then to default, by prior agreement, leaving the state bank with a "bad" loan that the state then assumes. The issuance of domestic bonds is controlled by the federal constitution, which since 1993 prohibits new bond issues until the year 2000 (other than to finance court judgments or to roll over the principal and capitalized interest on existing bonds). Apart from the Senate's overall supervision of states' debt, their external borrowing is largely exempt from federal regulation unless it requires a federal guarantee. In that case, the Ministry of Finance has the authority to grant or deny federal backing. An office in the federal Ministry of Finance monitors the finances of subnational entities and forwards its recommendations to the senate and the central bank, which have not always accepted the ministry's technical opinion, however. Another form of financing that Brazil has not yet been able to control is the capitalization of interest on existing debt. As discussed in the following section, this has resulted in a major increase in state indebtedness. Thus, in the borrowing area, Brazil meets condition 5, ex ante constraints, but not the others. Intergovernmental relations. Brazil, like Argentina, has 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 4,491). Like Argentina, Brazil has a long history of federalism going back to the nineteenth century, but that history has been less contentious. The federal republic was established through a peaceful overthrow of the emperor in 1889. The subsequent constitution transformed the existing provinces into states, authorized the election of state governors, and called for the establishment of separate state constitutions. The extent of state autonomy vis A vis the federal government has changed over time, with states eclipsed during the Estado Novo (1937-45) and a long period of military rule (1964- 85). Although state governors and mayors ceased to be elected directly, the military in the second period maintained the forms of democracy and federalism, using subnational politicians to legitimize military rule. Federalism revived with the return to democracy, which began in 1982 with the election of governors. 10 National democracy came more slowly, starting in 1984 with the first presidential election via an electoral college dominated by military appointees. In the next four years, the transition to democracy was completed: democratic elections for mayors of capital cities were held in 1985, a new congress with constitution- making authority was elected in 1986, and the first direct election of the president was held in 1989. Thus the democratic transition gave legitimacy to the state level of government before the other levels (Dias 1991; Hagopian 1996; Souza 1996). The constitutional commission completed its work in 1988. "The majority of the constitutional commission ... understood that democracy, in the area of public finance, demanded a weak federal government and strong subnational governments. In this way constitutional reform was occupied, on one hand, with the recovery of the powers of the legislative [branch] and financial autonomy of states and municipalities" (Afonso 1994). This view was reinforced by the political background of most of the delegates. Many of the senators were former governors. Most deputies were former mayors. Not surprisingly, the document was highly decentralist. Power at the federal level is constitutionally divided among the executive, legislative, and judicial branches. The executive branch is headed by the president, who is directly elected for a four-year term (and, as of 1998, can succeed himself once). In principle, the constitution gives the president considerable powers over the legislature. Under the 1988 constitution, the president has the exclusive right to initiate legislation in some policy areas, including those that create jobs or increase salaries in many parts of the public sector. The president prepares the annual budget, budget guidelines, and multi-year budget plan. Although the president must seek congressional approval of the budget, congress is restricted in the kinds of amendments it can propose. For example, it can only initiate programs or projects included in the president's budget. This gives the president power to block resources for new programs he opposes. The 1988 constitution also allows the president to adopt provisional measures (medidasprovisorias) that have the force of law for a 30- day period without congressional approval. The president can also can declare bills urgent (and push them to the top of the legislative agenda) and call special sessions of congress to consider issues determined by presidential initiative. In practice, the president's power is circumscribed by the difficulty of marshaling party support. Brazil's political system is characterized by a high degree of party fragmentation and weak party discipline at the national level. Even when the president seems to have strong support on a roll call vote, that support is the result of extensive prior negotiations and concessions to regional interests (Ames 1998). Fifteen political parties were represented in the national congress in 1998. The multiplicity of parties is due, in part, to Brazil's system of proportional representation. 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. Many of the actions of these parties are determined more by what goes on in their own states than by what goes on in national politics. State loyalties lead politicians to coalesce in support of projects that will benefit their own state, regardless of their party. Sitting state governors command the loyalty of federal deputies, because their support is more useful than the support of the president. But lists are open, unlike Argentina, so every candidate must compete against others from his party, depriving party leaders of a potential means for discipline. Few states have strong party machines to control this. Deputies also change parties frequently while in office, with 40 percent doing so in 1987-90 (Ames 1995a, 1995b). Because of their influence over deputies and senators of their state, regardless of party, governors can thwart or facilitate presidential designs:5 To implement their programs, presidents must construct coalitions that not only involve several parties, but also satisfy regional demands. They have several means of doing so. Once elected, presidents form de facto coalition governments by offering jobs and resources to different parties. This partitioning of government begins at the highest level 11 (cab inet positions and heads of major public enterprises and executive agencies) and continues on down to minor federal appointments in backward towns and remote regions. [But] with comparatively weak party discipline and loyalty, these coalitions are loose and shifting rather than hard and fast. A party's participation in the cabinet does not ensure the bloc support of legislators for the president. [Mainwaring 1997] Congress consists of two houses. The upper house (senate) is comprised of three senators from each state, elected for eight-year terms with no term limits. The lower house (chamber of deputies) consists of 513 members, with the number of delegates determined on the basis of state population, subject to a floor of 8 and a ceiling of 70 per state. Due to this structure, the largest state-Sao Paulo-with 21 percent of the national population, has less than 4 percent of seats in the senate and only 14 percent of seats in the chamber of deputies. The largest four states together account for nearly half of Brazil's total population but hold only 15 percent of seats in the senate and 39 percent of seats in the chamber of deputies. The floor of 8 delegates gives 11 of the 26 states higher representation per capita than the rest (Rosenblatt 1994). At all three levels of government, politics is a career profession. Individual politicians frequently rotate among levels of government and between the executive and legislative branches. Although politicians do not advance primarily through the party, individual positions are seen as stepping stones in a long-term career. City councilmen aspire to be state deputies, state deputies to be federal deputies, federal deputies to be state governors, governors to be senators, and ambitious senators to be president. Executive positions are preferred to legislative ones, because they provide the power to create jobs and spend money. Because of term limits in the executive branch, however, politicians make frequent if temporary moves to the legislative branches at all levels of government. Three-quarters of senators are former or future governors. For most deputies, the main route to career advancement is not by participating in any national or even statewide agenda, but rather by getting public works projects for selected municipalities within their state (Ames 1995a). They focus on one municipality if they plan to run for mayor and on several scattered ones if they look forward to statewide office (Samuels 1998). The constitution tightly restricts central government intervention in the affairs of the states, authorizing the use of federal troops only in case of foreign invasions, wars between states, threats to public order, and the need to guarantee the execution of federal laws and court judgments. This power has rarely been used, except during periods of dictatorship or military rule. States have traditionally used their militias to counterbalance the threat of federal intervention. States inherited the bulk of the former imperial police. By the early 1920s, the number of personnel in state militias was 70 percent of the number in the federal army (Dias 1991, p. 71). Sio Paulo's state militia is the fourth largest military force in South America. In Brazil, on the political side, condition 9 is met-a strong president-but conditions 10, 11, and 12 are not. Governors are constitutionally highly autonomous, electoral rules orient congress toward subnational interests, and party discipline is weak. Summary Table 3 contrasts the explanatory variables in the two countries in the 1990s. Argentina has stronger borrowing constraints in all categories except for ex ante rules, which are very sophisticated in Brazil. Argentina has a clearer delegation of spending responsibilities, but fewer federal controls on how subnational governments can spend in order to meet those obligations. Brazilian states have more taxing autonomy and generate more of their own resources. The large states generate almost all of their own resources in Brazil, but not in Argentina, where even the large states receive substantial transfers from the center. In both countries the sparsely populated and economically poorly developed states depend heavily on resources from the center. Although presidents in both countries have strong constitutional authority, Brazilian parties are much weaker internally and are not intrinsically loyal to the president. State 12 governments in Brazil have no fear of federal intervention, in contrast to those in Argentina, at least until the 1994 constitution. Macroeconomic conditions in Argentina since the stabilization have been more favorable for fiscal balance-growing income and tax revenue, lower initial debt stocks, and lower real interest rates. The following section shows the results of these differences. Table 3. Summary of Explanatory Variables for Argentina and Brazil in the 1990s Variable Argentina, 1991-98 Brazil, 1994-98 Fiscal authority 1. Subnational governments raise much of their No Yes for large states; own revenue No for small states 2. Transfers are specified by legal formula, not ad Yes; set by formula and limited in No; grants from the budget hoc aggregate, but special transfers and tax sharing set by were important for some small formula, but large debt provinces relief sometimes 3. Central government can effectively delegate Yes No functions to subnational governments to go along with the delegation of revenue sources 4. Subnational governments have authority to cut Yes; although central government No; legally difficult to cut their costs intervened recently on teacher labor costs wages Borrowing constraints 5. Central government strictly controls subnational No; federal government has some Yes, in principle; very borrowing ex ante authority but mainly requires sophisticated rules but also reporting significant loopholes 6. Central government credibly commits not to Yes No; cap on debt service have bailouts, prohibiting explicit bailouts and allows interest forcing subnational governments to service their capitalization debt 7. Regulators force creditors to accept the losses Yes, but rare because debt service No implied by any failure to service debt is deducted from transfers 8. The central bank (and bank regulators) is more Strong autonomy and commitment Limited autonomy; autonomous and has a strong anti-inflation to fixed exchange rate discretionary monetary mandate policy Intergovernmental political relations 9. Presidents are constitutionally strong at the Yes Yes national level 10. Governors have little constitutional autonomy Yes; federal intervention common No; no federal intervention 11. Electoral rules orient congress toward national, Yes No not local, interests 12. Party discipline is strong Yes; closed lists No; open lists II. Tests of the System and Outcomes in the 1980s and 1990s Federalism and other systems of intergovernmental relations unavoidably contain many tensions and contradictions, and the actual operation of the system becomes clearer when crises bring conflicts into the open, forcing politicians and other decision makers to reveal their priorities and to choose which compromises will set the precedents and establish reputations. 13 Prior to economic stabilization-Argentina in 1991, Brazil in 1994-both countries were clearly caught in a vicious cycle of failed choices. Since then, Argentina has shifted to a virtuous cycle of mostly successes, especially after the convertibility plan and hard budget constraints on provinces passed some critical tests in the mid 1990s and increased their credibility. In Brazil, the stabilization has been more recent, and even up to 1998 it was still being tested, for instance by states testing the firmness of their budget constraints with the federal government. Argentina's story The past 10 years have seen dramatic macroeconomic events in Argentina. In the late 1980s macroeconomic instability reached the point of hyperinflation. This was followed by a highly successful stabilization plan (1991) and three years of rapid growth. In December 1994, events in Mexico (the Tequila event) caused a sharp recession. GDP fell 4.4 percent in 1995. The 1991 reforms followed a long series of failed economic stabilization programs. Their main failure was the inability to reduce the public sector's deficit. Since the 1940s, economic policy favored powerful interest groups. Unionized labor benefited from high wages, guaranteed employment, and rigid rules governing hiring and dismissals. Industry (capital) benefited from highly protected markets, tax exemptions, subsidized credit, subsidized inputs from public enterprises, and high prices on sales to public enterprises. Provincial governments could avail themselves of zero- or low-interest credit from provincial banks, which would rediscount to the central bank. Poor tax collection and the Tanzi-Oliviera effect reduced the real tax revenues of both provincial and federal governments.6 Indeed, inflation hurt provincial revenue more, because the shared taxes were domestic and collected with longer lags, while the exclusively federal trade taxes were denominated in foreign exchange. It did not seem strange that the provincial as well as federal governments came to depend on seigniorage and the inflation tax. By 1989 subsidies to subnational governments and others were estimated at about $8 billion-more than 5 percent of GDP and close to half of the overall public sector deficit: The growth of the state [deficit], along with capital flight provoked by an inconsistent exchange rate policy, had been financed during the late 1970s largely by external borrowing, though the expanding Eurodollar market. This permitted the government to run large deficits. However, the abrupt end to voluntary foreign commercial credit in the early 1980s, the decline in commodity prices, and the sudden rise in real international interest rate provoked a financial collapse. [World Bank 1996a, p. 1] In 1985 President Alfonsin introduced the Austral Plan. It consisted of a devaluation, tariff reductions, and a partial deindexation of wages. In the short term, this reduced the fiscal deficit and inflation. Political support for the plan weakened, however, as real wages dropped precipitously. In 1987 the Radicals lost control of the chamber of deputies and most governorships. The Justicialista party reemerged as the dominant political force but was split between reform and orthodox wings. The orthodox wing pinned its presidential hopes on Menem, governor of the poorest state, who had cofounded the reform wing but then fell out with its cofounder. With no support from the traditional party structure, Menem garnered popular support by advocating income distribution toward wage earners and a moratorium on foreign debt. The reform wing, in contrast, became increasingly indistinguishable from the Radicals who themselves were becoming increasingly discredited. In 1988 Alfonsin introduced a new plan, known as the Spring Plan, that failed almost immediately. Currency speculation increased. On the eve of elections, real wages dropped precipitously. Inflation in one week reached 17 percent, and a banking crisis was imminent. The government was virtually bankrupt due to its inability to collect taxes (Peralta-Ramos 1992). In 1989 Menem won overwhelmingly, and the Justicialista party took both houses of congress. Menem sought to reassure liberal sectors of the economy and the international community by, among other 14 measures, appointing a conservative businessman as economics minister. He raised gasoline taxes, cut tax concessions for exports, liberalized the foreign investment law, and sold the airline and telephone companies. But the stabilization plan failed, due to a fiscal deficit resulting from the large stock of debt that the government had to refinance continuously at extremely high interest rates. In 1990 Menem appointed a new economics minister, Domingo Cavallo. He took even more dramatic action-floating the austral, printing money to buy up foreign exchange, and unilaterally converting government domestic debt into 10- year dollar-denominated bonds. This succeeded in reducing government's immediate debt service obligations and the stock of debt in real terms. It also destroyed the government's credibility with financial markets, thus freezing access to new credit. Macroeconomic instability had become closely, and correctly, associated in people's minds with the overall decline of the economy since the early decades of the century. When hyperinflation in 1989-90 threatened to push the country further into underdevelopment, people grew desperate and were ready for strong policy medicine. Menem and Cavallo were able to muster political support for radical solutions. The keystone was the Convertibility Plan, introduced in April 1991. This fixed the exchange rate of the Argentine currency (renamed the peso) to the dollar and required that the monetary base not exceed the dollar value of international reserves. This in effect transformed the central bank into a currency board by mandating a 100 percent reserve requirement for the issue of high-power money (later, the law allowed up to one-third backing by federal government bonds). It also removed the power to devalue from the Ministry of Economy and placed it with congress, where the need to obtain majorities made changing the law relatively difficult, especially in a federal presidential system. A lot of difficult fiscal adjustment was necessary at the federal level (World Bank 1997). Reducing public employment and privatizing state-owned enterprises were the most important and politically difficult tasks. Extensive public relations explained how structural adjustment was necessary to prevent a revival of inflation, and in backroom political deals the executive took advantage of the federal system of representation. A winning set of provinces in both houses could consist of Buenos Aires province for its weight in the chamber of deputies and the low-population provinces for their weight in the senate. The Fondo Conurbano helped to obtain the support of Buenos Aires province, which in turn ran a surplus the first couple of years after the stabilization and then kept spending from outstripping the growth of revenues. For the small provinces, the flow of per capita coparticipation revenue, plus discretionary transfers in a few cases, was very high relative to their size, but not in the aggregate. Because of their low population and high representation per capita, small provinces were inexpensive places to buy support in the house and especially the senate (Gibson 1997). In this period, the federal government also renegotiated its outstanding debts with the provinces. Each side had claims against the other. Provincial governments were in default on long-standing loans from the federal government. The federal government, in turn, was in default on payments owed to the provinces. The exact value of these mutual obligations was subject to dispute, particularly since the hyperinflation of 1989 had distorted their value in real terms. The federal government emerged from the negotiations as the net debtor; the provinces emerged with virtually no debt to the federal govemment (interviews with government officials). This outcome proved advantageous not only for the states, but also for the federal government in the future. Because the states had no debt on which they might default to the federal government - the states were just getting transfers - the center had a stronger hand in pressuring states to adjust in response to subsequent fiscal shocks. To reinforce its autonomy, the central bank's charter was revised (in September 1992), establishing an independent board of directors (ratified by congress), with fixed terms of tenure. Two provisions in the charter had important implications for provinces. First, the charter dictated that the central bank could not take any new domestic assets. This meant that provinces could no longer count on the central bank to rediscount loans by provincial banks to provincial governments, ending their access to seigniorage and the 15 inflation tax. Second, the charter prohibited the central bank from guaranteeing bank deposits. Then provincial banks had to rely on depositor confidence to maintain liquidity. Both measures reduced the central bank's role as lender of last resort and hardened the financial budget constraint on provinces, limiting their ability to borrow (indirectly) from the central bank or from depositors. At first these constraints were not binding on provincial finances, because of the rapid growth of tax revenues. However, it proved to be important that the limits were entrenched in the midst of the post-convertibility boom, because they were firmly in place when the boom came to an end and pressure for deficit financing increased. With peso convertibility, inflation (wholesale price index) slowed to 3 percent in 1992, 0 percent in 1993, and 6 percent in 1994. Due to strong collection efforts, reverse-Tanzi-Oliveira effects, and economic growth, tax revenues increased dramatically, especially from shared taxes. This was reflected at the provincial level, where revenue jumped over 25 percent in real terms between 1991 and 1992, and over a percentage point in GDP, which was itself recovering strongly. See table 4. Table 4. Trends in Provincial Revenues and Expenditures in Argentina, 1991-97 (In percent of GDP) Type ofrevenue 1991 1992 1993 1994 1995 1996 1997 Current Revenues 8.19 9.62 9.59 9.45 9.34 9.39 9.52 Provincial taxes 2.59 3.25 3.41 3.42 3.27 3.27 3.21 Provincial nontax revenue 0.40 0.52 0.58 0.62 0.66 0.55 0.77 National transfers (aportes) 5.20 5.85 5.61 5.42 5.42 5.57 5.55 Current Expenditures 7.77 8.80 9.24 9.10 9.22 8.68 8.25 Capital Expenditures 1.21 1.07 1.28 1.40 1.52 1.58 1.55 Primary Surplus (deficit) (1.34) (0.65) (1.30) (1.34) (1.61) (0.20) 0.11 Overall Surplus (deficit) (1.48) (0.80) (1.48) (1.55) (1.86) (0.50) (0.29) Source: For 1991-94, World Bank 1996c; for 1996-97, Republic of Argentina, Inforie Econduico regional Taking advantage of the provinces' rapid and unexpected increase in revenues, the federal government negotiated a series of fiscal agreements in 1992-94. Pacto Fiscal I (August 1992) diverted coparticipation funds to the national social security system, which was poorly funded and required large subsidies. Under the agreement, provinces allowed 15 percent of the total taxes subject to coparticipation to be diverted to finance the transition costs of national social security reform. To close the pact, the federal government guaranteed a floor of $8.7 billion annually in coparticipation payments, expressed in terms of fixed amounts of pesos per month. In expressing the guarantee in terms of pesos, the federal government bet that revenues would grow (and that transfers would exceed the guaranteed amount). Over the next three years, the government won the bet, and the provinces did not collect on their insurance. Although they did collect in 1995 and early 1996, the national executive paid a modest price in exchange for important structural reforms. As additional quid pro quo, the government established a Fondo de Desequilibrio Fiscal for small poor provinces. Because political power of the small provinces is disproportionate to their population, due to their overrepresentation in both houses of congress, the fund bought political support at relatively low cost. Government also called in political chits from the delegation of Buenos Aires province, demanding support for the fiscal pact as reward for the recently enacted Fondo Conurbano.7 Pacto Fiscal I was followed by the transfer of federal health and secondary school programs, including 284,000 federal employees, to the provinces. In retum, the federal government guaranteed payment equal to its expenditures in each province in 1992 (totaling $1.318 billion). The payment was 16 financed not from central government revenues but from the provinces' share of coparticipation taxes. This allowed the federal government to exit from a major area of expenditure without any increase in transfers to the provinces. In August 1993, the federal government negotiated Pacto Fiscal II, which reformed provincial pension funds. Prior to the reforms, each province maintained a separate pension fund for its employees. Benefit payments were funded out of current revenues. These included, in addition to earmarked salary deductions, subsidies from general treasury revenues. The scale of treasury subsidies was significant, totaling $773 million in the 16 provinces expected to transfer their pensions to the federal government. (Data for other provinces, including Buenos Aires, are not available.) Actuarial projections suggest that most provincial pensions were financially unsustainable. Although the pensions were legally a concern of individual provinces, the Ministry of Economy feared that a major collapse would ultimately redound on the federal government itself. Under the second fiscal pact, the federal government agreed to take over the pension system of any province that passed a law authorizing it to do so. Once transferred, the federal pension system would eventually confer on federal and provincial civil servants the same retirement eligibility conditions, contributions, and benefits that were available to the private sector. Provinces were slow to implement the reforms, however. Because the transfer would reduce the benefits of active employees, it was opposed by public employee unions. Not all provinces would benefit from the transfer in any case. Although federalizing a provincial pension plan would eliminate the need to subsidize pensions from the general treasury, it would also increase employer contributions in some cases. The small provinces and C6rdoba would be the primary beneficiaries (World Bank 1996b and 1996c). The provinces of Buenos Aires and Santa Fe would not benefit, because savings from the elimination of subsidies would be more than offset by increases in direct payments (from employers) into the national system.9 In return for signing the second fiscal pact, the federal government offered to raise the guaranteed minimum of coparticipation funds. During the same period, government also offered to finance the privatization of provincial banks, but prior to 1995 only four small provinces (Corrientes, La Rioja, Chaco, and Entre Rios) had done so. While provincial revenues increased rapidly in the 1991-94 period, expenditures increased at the same rate or faster in some provinces. Between 1991 and 1994, revenues increased by 1.4 percentage points of GDP and expenditures by 1.5 point s. Personnel costs accounted for about half the increase in expenditure. Growth in transfers to social security, private schools, and municipalities accounted for another 16 percent, and growth in capital spending, for 13 percent. Interest costs remained small, about 2 percent of total outlays. Provinces with precarious finances at the start of the post-convertibility boom continued to have precarious finances despite dramatic increases in revenue. Smaller provinces, along with C6rdoba and Buenos Aires city, were in greatest jeopardy. In December 1994, events in Mexico (the Tequila crisis) ended the post-convertibility revenue boom. Argentina was particularly vulnerable to the Tequila crisis, due to its heavy reliance on foreign capital inflows, the inflexibility of its exchange rate regime, and the need for a strong financial system under the Convertibility Plan. GDP fell 4 percent in real terms in 1995, and provincial revenues dropped 8 percent. The Tequila crisis also prompted a run on most provincial banks, except the Bank of the Province of Buenos Aires. With the prospect of central bank assistance closed off by the convertibility law and the central bank's new charter, provinces were forced to use their own resources to prevent provincial banks from failing and to recapitalize them. Through a project financed by the World Bank and the Inter-American Development Bank, the federal government refinanced the liabilities of the provincial banks taken over by provincial governments, but it made the assistance conditional on privatization of the banks. The Tequila crisis tested the hard budget constraint put in place by the convertibility law and the new central bank charter. Initially the provinces responded to the fall in revenues by borrowing -either 17 directly from the treasury or from their provincial banks--or by obtaining emergency grants from the federal government. As a result, provincial debt expanded dramatically, totaling $17.3 billion by mid-1996, more than two-thirds of which had been run up since 1991. They used mainly three sources. First, they ran up arrears to suppliers and personnel. Second, after exhausting the tolerance for delayed payment, they resorted to forced lending, paying staff and suppliers with bonds. These bonds could be converted to cash at a discount at provincial banks and then used to pay taxes. These two kinds of forced lending accounted for more than half of the debt incurred in the first year and a half of the crisis. Third, despite their financial difficulties, the provinces were able to continue borrowing from private banks, pledging their coparticipation transfers as guarantee. The federal government was a party to this arrangement in that Banco de la Naci6n would deduct the debt service from federal tax receipts and transfer only the remainder to the provinces. Even after the Tequila crisis, the government adhered to this practice. This sharply reduced net transfers to provinces with high levels of bank debt, taking as much as one-third of coparticipation to pay creditors. The extent of adjustment differed among provinces. Buenos Aires province adjusted the least of the major provinces. In part, this is because it had less need to adjust. While Buenos Aires province began the post-convertibility period with an overall deficit of 10 percent of revenues, subsequent increases in expenditures, although large, were less than the 83 percent real increase in revenues between 1991 and 1994. The province increased non-interest recurrent expenditures by "only" 50 percent in real terms. (Capital spending quadrupled, however, and interest costs increased more than fivefold, raising total spending 71 percent.) The province also Table 5. Provincial Debt in Argentina, 1996--97 bnftdfo h raino h od benefited from the creation of the Fondo (as of November 1998) Conurbano in 1992, and without it 1995 1996 1997 revenues would have increased more Total Provincial Debt 15,393 17,292 16,249 slowly than expenditures. The province ofachieved an overall surplus in 1992. The Banks hi balance began to decline in subsequent Naian l (1) men 5,)771 43 years, reaching a negative 4 percent of National government (2) 768 4731994. International organizations 1,525 1,997 Provincial Bonds 1,124 2,248 3,640 To finance this deficit, the "Consolidated debt" (3) 5,142 5,582 province relied on two forms of (1) Including private and provincial banks. Bank debt and thus borrowing. The first consisted of so- total declined from 1996 to 1997 largely because BA Province called debt consolidation bonds-forced renegotiated some of its debt with its provincial bank. loans from suppliers and staff. Although (2) Loans to provinces under BOTE 10 and BOCEP. (3) Debt contracted with federal fiduciary fund to finance thes ha d noia ari i ar privatization of commercial banks and bonds issued to suppliers and the cud beueo ay onial other debtors in lieu of payment and bonds issued to refinance such debt. shorter maturities in practice. The second Source: For 1996, Ministerio de Economia IEFE; For 1997, consisted of medium-term (three-year) Republic of Argentina, Informe econ6mico regional, December Eurobonds. The first Eurobond was 1997. November 1998 issued (for $100 million) in July 1994. The Tequila crisis caused a 3 percent real drop in the province's revenues. The province responded by reducing capital expenditures. This adjustment was sufficient to maintain the overall deficit at its 1994 level. Further adjustment was not forthcoming, however. Although revenues rebounded 23 percent in real terms between 1995 and 1997, expenditures increased 36 percent (largely because of higher subsidies and more capital works). As a result, the overall deficit increased and gross new borrowing averaged $400 million annually over te 1995-97 period. As of June 1997, the stock of debt consolidation bonds had 18 reached $150 million; the stock of Eurobonds, $520 million. Further growth of debt in 1997 was only forestalled by selling the province's power company. The total stock of debt remained relatively small, however. Total debt equaled 32 percent of revenues in 1997, and more than half of this was a long-term zero-coupon bond. The city of Buenos Aires offers a different story. When the convertibility law went into effect, the city was an administrative arm of the national government. Its overall deficit stood at 12 percent of revenue. Between 1991 and 1994, revenues increased 50 percent in real terms. Expenditures increased 43 percent. This permitted some shrinkage in the deficit-to 8 percent of revenue in 1994. The deficit was financed by running up arrears to suppliers and contractors and later paying them with bonds that could be redeemed at a discount at the municipal bank. The Tequila effect caused revenues to fall 7 percent in real terms. The city adjusted partially, reducing expenditures 6 percent. The deficit nevertheless remained at 8 percent of revenues, still financed largely through arrears. Real adjustment did not occur until two years later. In 1996, a new administration took office, including the first elected mayor and council in the city's history. On taking office, the city council imposed a hiring freeze and an attrition policy, freezing personnel spending at the 1995 level in real terms. Capital spending was cut in half. Revenues rose 14 percent in 1997. The city used the increase to improve its overall balance, which attained a surplus in 1997. The city also took steps to confront its large stock of short-term debt. As of mid-1996, the value of floating debt (accounts payable) totaled $538 million. To address these debts, the city established a procedure to evaluate and reschedule claims, offering three alternatives for debts contracted before the administration took office: (a) 35 percent reduction in stock with 18 months to pay beginning in 1998, (b) a 25 percent reduction with 30 months to pay, and (c) a 0 percent reduction with 48 months to pay. As of September 30, 1997, these arrangements had reduced accounts payable to $363 million. The city also faced $457 million in outstanding bonds (bocons) and contractual debt to suppliers and staff. To address these, the city began to liquidate bocons as they came due and to replace them with longer-term lower-interest-rate bonds (Tango bonds). In 1997 four sets of Tango bonds were issued totaling $500 million (of which $150 million was issued in pesos and the rest in U.S. dollars and Italian lira). Of the $500 million, roughly $400 million was used to retire bocons and floating debt; the remainder was used to recapitalize the city bank so as to give it a positive net worth as a precursor to privatization. At the end of 1997, the stock of bocons and contractual debt to suppliers had shrunk to $191 million. It had been replaced, however, by a growing stock of Eurobonds. C6rdoba is another case of post-Tequila adjustment. C6rdoba began the first year of the Convertibility Plan with a significant overall deficit (16 percent of revenue). Although revenues increased 75 percent in real terms between 1991 and 1994, expenditures increased even faster (83 percent), largely due to higher spending on personnel. To finance its deficit, C6rdoba borrowed from the Banco de la Provincia de C6rdoba. By the end of 1993, borrowing from the Banco de la Provincia de C6rdoba (gross of province of C6rdoba deposits at the bank) was $436 million. A year later, it had jumped to $725 million (World Bank 1996b, p. 146). When the crisis struck, revenues fell 5 percent in real terms. More important, the provincial bank experienced a run on deposits. This ended the province's ability to borrow from its bank, which became instead a financial drain. The province was forced to pay off some of its own debt to the bank and to borrow externally in order to provide liquidity. The provincial power company also deposited funds in the bank, with interest paid in accrual. A new administration took office in 1995 (five months early, due to Cordoba's fiscal crisis) ending three consecutive mandates by the outgoing governor. It implemented drastic measures. To reduce the wage bill, it reduced salaries 30 percent (in return for a 40 percent cut in hours of work), dismissed more than 6,000 nonstatutory staff, and transferred 1,500 health workers to the municipalities without compensation.0 It cut capital spending 40 percent. To regularize arrears to personnel, the new administration issued $800 million in certificates of cancellation of debt (CECORs). CECORs could be used to pay taxes at par or 19 discounted on the secondary market at 80 percent of par. The certificates had a 12 percent yield, with a one- to two-year grace period on interest and principal and a two-year maturity. In the two subsequent years, the province's overall balance continued to improve. The overall deficit was 3 percent of revenues in 1996, which became a surplus (2 percent of revenues) in 1997. This performance was aided by the province's signature of the second Pacto Fiscal (in 1996), which made available $116 million in retroactive coparticipation guarantees. The province also maintained its 30 percent cut in wages throughout 1996 and the first five months of 1997. Two-thirds of the cut was restored in May 1997, and the full amount was restored in 1998. Nevertheless, the experience left a distinct lesson that excess spending and deficits could be politically painful. The province also converted much of its short-term CECOR debt into longer-term obligations: $400 million in CECORs matured in July 1997 and another $200 million in January 1998. As the CECORs matured, the province retired them, partly with cash and partly with longer-term loans from private commercial banks. In 1997 the province borrowed $130 million from Banco de Rio de la Plata and Banco de Galicia to pay off the first group of CECORs (liquidating the rest from cash). It borrowed another $120 million from the Banco de Galicia to pay off the second group. The outstanding stock of CECORs was down to $367 million by the end of 1997 (and was less than $250 million once the second group of CECORs was liquidated in January 1998). CECOR debt was largely replaced by debt to private banks. Debt to the Banco de la Provincia also declined. In April 1996, the province signed an agreement with the bank, fixing the amount owed by the province and establishing a four-year amortization schedule for liquidating it. By the end of 1996, the debt was down to $181 million. A year later, it was down to $100 million. The province's debt to its bank recently increased, however, in connection with the province's efforts to improve the bank's financial condition. Under a plan sanctioned by the central bank," the provincial bank's nonperforming debt (along with the debt of the Banco Social de C6rdoba) was transferred to a fiduciary fund under the control of the province. The province floated a bond equal to the expected value of the nonperforming portfolio, transferring the proceeds to the bank as payment for nonperforming assets. The province intended to sell the bond to private investors. When no interested investors emerged, it was sold to the provincial bank. The bank thus traded its nonperforming portfolio for a provincial bond. Other provinces followed a pattern of fiscal restraint where the provincial government was of the same party as the national president. Indeed this pattern seems to go back to the 1980s as well (Jones, Sanguinetti, and Tommasi 1997). Next to the province of Buenos Aires, Santa Fe was the most important case of a Justicialista province adjusting promptly after the start of stabilization. Getting those two to adjust before the crisis provided a critical mass of fiscally sound provinces and allowed the central government to take a hard line in forcing the other provinces to adjust. The success of the Menem-Cavallo team in forcing the provinces to go along with the national adjustment program resulted from an unusual conjuncture of historical opportunities, and the capital markets and political circumstances may again test the firmness of the anti-inflation, anti-deficit position. Nevertheless, the establishment of sound rules and institutions and the precedent of upholding them in the face of political pressure have improved the likelihood that good policies will withstand future tests. Brazil's story In Brazil, two stories concerning fiscal decentralization and macroeconomic management occurred simultaneously, each with its own plot but with occasional interaction between them. The first involved the major increase in revenue sharing mandated by the 1988 constitution. The second involved the federal government's response to the growing stock of state debt. 20 The large increase in tax sharing mandated by the 1988 constitution prompted some observers to fear for Brazil's macroeconomic stability. There was concern that the increase in tax sharing would provoke federal deficits, unless the federal government could cut its ordinary (nontransfer) expenditures by an equivalent amount or raise federal taxes proportionately. At the outset, this situation seemed likely to happen. President Collor's initial attempts to forestall this scenario failed. Under the rubric of Operagao Desmonte, Collor proposed a program of expenditure decentralization intended to match the new division of revenues. This was rejected by congress. In fact, it had very little support anywhere in the public sector. "Reducing the size of the federal government ran aground on the corporative resistance of the federal employees, the sectoral agencies, the executive branch, and principally the federal congress. All, in the end, had no intention of giving up the power to fill positions, contract staff, manipulate budget allocations, distribute subsidies, build works, and to lure the support of government, mayors, congressmen, and local leaders" (Afonso 1994). Nevertheless, fiscal adjustment ultimately occurred at the federal level. Part of the adjustment took place on the expenditure side in the early 1990s. Having failed to decentralize federal functions formally, the federal government succeeded in off-loading the responsibility for certain expenditures 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. In Rio de Janeiro, federal hospitals were transferred to the state and municipality. The federal government also managed to off-load some health care costs onto subnational governments by reducing the amount of federal compensation payments. Federal discretionary transfers to subnational governments also fell. Table 6. Trends in Subnational Revenues and Expenditures in Brazil, 1991-97 (In percent of GDP) Type ofrevenue 1991 1992 1993 1994 1995 1996 1997 Federal Government Current Revenues 15.9 14.8 16.5 18.6 19.2 17.8 18.6 Current Expenditures 15.5 14.1 14.3 15.9 18.6 17.6 18.0 Capital Expenditures 2.7 1.0 1.4 1.0 0.8 0.8 0.9 Primary Surplus (deficit) 0.9 1.3 1.4 3.1 0.6 0.4 (0.3) Operational Surplus (deficit) 0.3 (0.8) 0.0 1.6 (1.6) (1.7) (1.8) State and Municipal Governments Current Revenues 14.2 12.9 12.8 13.6 13.6 13.0 14.3 Taxes 8.5 7.9 7.6 8.5 7.2 7.3 7.0 Nontax revenue 2.4 1.6 1.7 1.5 2.8 2.0 3.3 National transfers 3.3 3.4 3.5 3.6 3.6 3.7 4.0 Current Expenditures 12.7 12.2 12.4 11.5 11.4 12.1 Capital Expenditures 2.3 2.1 2.1 1.9 1.9 1.6 Primary Surplus (deficit) 1.5 0.4 0.6 0.5 (0.2) (0.6) (0.7) Operational Surplus (deficit) 1.3 (0.8) 0.3 (1.0) (2.3) (1.9) (2.3) Source: IMF, Brazil Recent Economic Developments Feb 1997 and April 1998; 1996-97 Bank staff estimates. The politics of this process were complex and relied heavily on federal inducements. To persuade SAo Paulo and Rio to accept the suburban railway systems, for example, the federal government had to 21 undertake a massive upgrading of the assets, financing these expenditures through loans that remained obligations of the federal government even after states took ownership. In the same way, the city of Rio was induced to take over federal hospitals only after the federal government agreed to continue paying the salaries of existing staff until they retired. Local political opportunism also played a major role. As federal spending on operations and maintenance declined, the quality of federal highways, hospitals, and other facilities deteriorated conspicuously. Some governors and mayors found it in their interest to take over these assets, in order to demonstrate their superior management ability to voters. Taking credit for improving public services was a politically attractive strategy for governments with the capacity to make the improvements. For places lacking capacity or an electorate appreciative of efficiency, in contrast, it was still more attractive to use the money to put political supporters on the payroll or to award them contracts. The largest part of fiscal adjustment at the center occurred not through cuts in expenditure but rather through adjustments in revenue. From the late 1980s to the early 1990s, federal spending as a share of GDP declined less than one percent of GDP, and after that it rose again. On the revenue side, at first the federal government used the guise of economic stabilization measures to cut back some of the increase in constitutionally mandated transfers. Constitutional Amendment 4 authorized the federal government to reduce by 20 percent the portion of taxes (on income and industrial products) subject to sharing. Although intended to be temporary (January 1994 to June 1996), the provision was subsequently extended. The federal government also increased the rates of taxes that it was not required to share with subnational governments. For example, it dramatically raised social security taxes, from 45 percent of federal tax revenue in 1988 to 60 percent in 1991 (Afonso 1993). The federal government also increased its tax on financial operations and introduced a new tax on checks, initially temporary. When agreements required federal and state governments to share the most efficient taxes, they created a systemic incentive to increase inefficient taxes. As a result, the federal deficit remained roughly constant over this period, contrary to initial expectations. At the same time, state and local governments expanded strongly, resulting in a larger public sector overall. At the state level, expenditures increased 33 percent in real terms between 1986 and 1995, while GDP grew only 14 percent. A subnational fiscal crisis emerged from the indebtedness of the states. States and municipalities in Brazil traditionally borrowed from a wide variety of sources. They issued bonds on the domestic market and borrowed from domestic private commercial banks. They borrowed from various federal intermediaries, including the federal housing and savings bank (CEF) and the federal development bank (BNDES). A majority (24) of the 26 states owned commercial banks, from which they occasionally borrowed (or, more frequently, which they instructed to lend to favored clients). States also borrowed abroad, both from the World Bank and the Inter-American Development Bank (which demanded a federal guarantee) and from private lenders (which did not). Finally, states borrowed through informal mechanisms, such as arrears on payments to suppliers and on salaries to state employees. From the beginning of Brazil's political opening through 1998, there were three major subnational debt crises (Dillinger 1997). The first had its roots in the international debt crisis of the 1980s. Faced with foreign exchange constraints (even when they had the wherewithal to pay in domestic currency), states ceased servicing their foreign debt. The federal government had guaranteed much of the debt, which thus became a federal liability once the states defaulted. In 1989 the federal government agreed with the states to transform the outstanding stock of federally guaranteed external debt into a long-term debt to the federal treasury. The second crisis involved debt owed by the states to federal financial institutions, principally CEF. This was resolved in 1993 by rescheduling roughly $28 billion of such debt. Again, the debt was transformed into debt to the federal treasury, with 20 years to maturity and interest rates based on the original contracts. To close on this second agreement, the federal government conceded an escape clause, which applied retroactively to the rescheduled foreign debt and certain other debts to the federal 22 government. If the ratio of state debt service to revenue rose above a threshold fixed by the senate, the excess could be deferred. Deferred debt service would be capitalized into the stock of debt, to be repaid when debt service fell below the threshold amount or in the 10 years after the original 20-year repayment period had expired. With principal rescheduled and debt service subject to a ceiling, the immediate burden of debt servicing was considerably reduced. With debt service subject to a ceiling, however, excess debt service was capitalized into the stock of debt, expanding it at a rate that would accelerate whenever real interest rates increased. The agreements established three precedents that influenced subsequent debt agreements. First, they reinforced the perception that the federal government was prepared to provide debt relief to any state requiring it. Second, they set a precedent for providing such relief in the form of rescheduling, rather than forgiveness. Through the combination of grace periods, rescheduling, and debt service caps, the agreements reduced the debt service burden of sitting administrations, leaving the fiscal consequences to their successors. Third, for heavily indebted states, the limits on debt service largely eliminated the expected future cost of current borrowing and interest capitalization. The third state debt crisis was precipitated by the success of the government's stabilization plan. The Plano Real, introduced in mid-1994, consisted of an initial fiscal and wage adjustment, an exchange rate anchor (modified later in 1995 and made less rigid), and a tight monetary policy. The plan had remarkable success. Annual inflation fell from 929 percent in 1994 to 22 percent in 1995 and to 9 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. As a result, state governments found themselves with payrolls equal to 80 or 90 percent of revenues. The impact of the Plano Real on interest rates also adversely affected state finances. Because the state debt was indexed, it was still large in real terms when triple-digit inflation ended. Moreover, much of the existing stock of state debt was vulnerable to fluctuations in short-term interest rates. The principal on rescheduled domestic debt was indexed to the average rate paid on deposits. The rate on state bonds was linked to the overnight rate on federal bonds. Domestic real interest rates had been high for several years preceding the Plano Real, due to the risks associated with inflation. While reducing those risks, the plan's tight monetary policy perpetuated high real interest rates. Faced with inflexible personnel costs and persistently high interest costs, the states chose to default. Three main categories of obligations went into default and were picked up by the central government: medium-term state bonds, short-term advances and arrears, and losses of state banks. The largest single objects of default were state bonds. Although 15 states and two municipalities had issued bonds, bond financing was dominated by four states-Slo Paulo, Rio de Janeiro, Minas Gerais, and Rio Grande do Sul-and the municipalities of Sao Paulo and Rio. With interest at floating rates due at maturity, bonds were traditionally underwritten by the states' commercial banks and ultimately sold to private banks and investors. The states had had difficulty marketing the bonds since the late 1980s, and when the Plano Real tightened credit markets, private banks declined to hold state debt, at any price, even on the overnight market. Unable to liquidate the bond debt (which had reached Rs$3 1 billion by the end of 1994), the states sought relief from the federal government. Under the so-called troca (exchange) arrangement, the federal government authorized states to exchange their bonds for more readily marketable federal or central bank bonds. The central bank then held the state bonds in its portfolio.2 23 Table 7. Growth in the Stock of Under the exchange agreement, the senate had the Subnational Bonds in Brazil, 1990-97 authority to determine the proportion of bonds that would have ___________________________ to be liquidated at maturity. The senate, presumably reflecting Stock ofbonds various state interests rather than party discipline, used this (billions ofreales as Percent authority liberally, authorizing 100 percent rollovers in the Year ofDecember 1996 increase initial years (although this was later reduced to 98 percent). 1990 17.3 More important, the senate allowed states to roll over not only 1991 17.2 - the original principal but also accumulated interest. Interest 1992 23.2 35 charges on the exchanged bonds were based on the yield 1993 25.9 12 obtained by the federal government at periodic bond auctions. 1994 313 21 Due to the prevailing tight monetary policy, the real rate on 1995 43.2 38 federal bonds was high: 22 percent in 1994 and 25 percent in 1996 51.7 20 1995, With interest capitalizing at the rate on federal bonds, the stock of bond debt grew explosively. As shown in table 7, the stock of bonds grew by Rsa12 billion between 1994 and 1995 and another Rs$8.5 billion the following year. At the end of 1996, the total stock of state (and municipal) bond debt stood at Rs$52 billion. During the high-inflation years, states regularly used revenue anticipation loans (AROs) and arrears to personnel and suppliers as tactics to cut the real value of outlays, but once inflation declined permanently and real interest rates rose, states were unable to pay these short-term obligations. (Revenue anticipation loans were due within a month of the end of the fiscal year.) As state ARO debt and arrears to personnel and suppliers mounted, pressure for federal relief increased. In November 1995, the federal government responded. Under Resolution 162 of the National Monetary Council, the government established the Program for State Restructuring and Fiscal Adjustment. This provided two lines of credit to states: the first to pay off arrears to employees and contractors; the second to refinance AROs. The loans offered only brief respite. Each had to be folly amortized within the term of the current state administration. Interest rates were based on the CEF's cost of funds plus a spread. For a loan approved in early 1996, this implied a real interest rate of 27 percent a year and an amortization period of slightly less than 24 months. Under the terms of the loans, the states also had to agree to 35-45 specific reform measures, covering personnel management, state enterprises, tax administration, debt reduction, and aggregate expenditure control. In practice, the government had little power to enforce these conditions because funds were disbursed before any policy conditionality was imposed. The only ex post penalty for failing to comply with these conditions was a three-month reduction in the maturity of the loan or mandatory prepayment of the entire amount. The third component of state debt originated from the collapse of state banks, which had made bad loans to the private sector (sometimes as an indirect channel for credits to their states) and directly to the government, in the case of io Paulo, which had begun to default on its debt service in the early 1990s. The tight credit and high real interest rates after Pano Real forced the problems into the open. The state's debt to the State Bank of S o Paulo (BANESPA) grew from Rs$5 billion at the end of 1993 to Rs$21 billion at the end of 1996 Poor portfolio performance and high administrative costs prompted heavy operating losses at other state banks and resulted in a steady decline in their net worth. The banks stayed in business by borrowing foro the central ank. The central bank took over BANESA, as well as the state bank of Rio de Janeiro, but returned them a year later with their underlying problems still unsolved. By assuming control and then relinquishing it back to the state government, the central bank strengthened the perception that the banks' liabilities carried an implicit federal guarantee. This guarantee, together with direct liquidity support from the centra bank, permitted BANESPA, Bank of Rio de Janeiro, and others to remain in operation and to continue capitalizing the unpaid interest owed by borrowers. 24 Table 8. Stock of State Debt in Brazil, Taken together, these temporary solutions in effect December 31, 1996 federalized the state debt. The central bank brought into its (millions of reales) portfolio the bonds formerly held by state and private banks. Where Typethe debt to ANESPA had been the concern of the bank's Typeof dbt Aount shareholders and depositors, it 'was now implicitly assumed by the Bonds 44,953 central bank. Where the AROs and arrears had been owed to Contractual debt private banks and individuals, the restructured debt was now owed Rescheduled external 15,090 to the federal treasury. As of December 1996, the debt totaled Rescheduled domestic 32,584 Othedea dmsi32,18 Rs$12 8 billion, of which all but Rs$5 billion had been assumed by Other extemal 3,235 the federal government (see table 8). Other domestic 3,235 BANESPA 27,699 Resolution 162 2,772 Negotiations with individual states began in mid-1995. Total 128,440 There were three parties to the negotiations. The first was the federal congress, because any debt refinancing required an act of legislation. Although the large debtor states accounted for nearly half the population of Brazil (47 percent), their representation in congress was much less. With three senators per state, the big debtor states held only 15 percent of seats in the senate. In the chamber of deputies, due to the constitutional minimum and maximum of seats per state, the debtor states held only 39 percent of seats. As a group, the debtor states therefore had to accommodate the interests of the legislative majority. This did not produce a tough line against the major debtors, however. It merely meant that any relief given to the major debtors was shared with the others. The second party to the negotiations was the president and his economic team. In principle, they were the one constituency with a clear interest in bringing state debt under control. Even though the debt was a legal obligation of the states, the federal government had to float the ever-increasing volume of bonds required to fnance it. But the president himself was constrained. He was constrained, first, by the weakness of party discipline in Brazil, which meant that he could not compel support from his coalition partners in congress. He was constrained further by the number of competing items on his legislative agenda. In addition to debt workouts, President Cardoso sought constitutional reforms that were critical to sustaining his victory over inflation. These included an administrative reform (to allow dismissal of public servants) and a pension reform (to tighten retirement eligibility criteria). Each required him to negotiate votes in congress, and he would not exhaust his political capital on the state debt issue alone. The third party consisted of the individual states themselves, as represented by their governors. The debtor states were comfortable with the status quo and resolutely opposed to any agreement that would require them to resume servicing their debt, at least during the current administration. With interest on bonds being capitalized and state bank liabilities accumulating off budget, the major debtors were servicing very little of their debt. They preferred to continue doing so. The largest debtor states had a strong bargaining position for two reasons. By and large, they were fiscally self-sufficient, owing to their VAT revenues. Federal threats to withhold debt service from intergovernmental transfers therefore were ignored. Second, the sheer size of the state debt, particularly in the case of So Paulo, discouraged the federal government from forcing states into default. In principle, the central bank could have brought the capitalization of bond interest to a halt by refusing to exchange federal bonds for state bonds. Had it done so, the states would have been forced to default. The government feared that such a default would threaten the stability of the domestic capital market. Cutting off central bank support to BANESPA or the state banks of Rio de Janeiro and Fo Grande do Sul seemed to pose a similar threat. Such action, the Ministry of Finance feared, could have prompted a liquidity crisis at the banks and a run on deposits, undermining confidence in the banking system as a whole. The major debtor states, in the eyes of both the president's team and the states, were too big to fail. 25 At the outset, there was some expectation (particularly at the World Bank) that a Brady-style agreement might be reached. Under such an approach, each state would commit itself to internal fiscal reforms: cuts in personnel, subsidies, and capital spending and increases in taxation. In retum, the federal government would provide debt relief in the form of rescheduling and interest rate subsidies. The agreed combination of internal fiscal adjustments and federal relief would suffice to permit the state to service its remaining debt. The requirement of state fiscal adjustment as a quid pro quo for debt relief would discourage lax fiscal behavior in the future. The Ministry of Finance, opposed an explicit reduction in the stock of debt, arguing that it would merely clear the decks for a new round of state borrowing in the future. The debtor states, for their part, were disinclined to undertake any internal fiscal adjustment and saw no reason to raise taxes or cut expenditures simply to begin servicing debt on which they were currently paying nothing. Not surprisingly, negotiations proceeded intermittently, and the stock of debt continued to grow. A set of agreements-in-principle signed between September 1996 and January 1997 did not yield binding contracts. In December 1997 the first major debtor state-SAo Paulo-signed a binding agreement with the federal government. Under the agreement, the federal government agreed to assume all of Sao Paulo's bond debt and debt to BANESPA. This totaled $50 billion. Of this, $40 billion was refinanced as a loan to the state government, with 30 years to maturity and a real rate of interest of 6 percent, well below prevailing domestic rates. Another $6.2 billion (12.5 percent) was to be amortized immediately through the transfer of stock in state enterprises, including BANESPA, two large power companies, and the state railroad. The remaining $3.8 billion (7.5 percent) was to be forgiven by the federal government. The most important element of the agreement was a ceiling limiting debt service to 13 percent of net current revenues. Details in the agreement reduced this to 8.86 percent in 1997 and provided that the full 13 percent would not become effective until 2000. This ceiling covered not only the newly refinanced debt but also the debt refinanced under the two previous reschedulings. For Sgo Paulo, it implied virtually no increase in actual cash debt service. Because interest and principal on the debt subject to the ceiling were already close to 8.86 percent of current revenues, the majority of debt service on the newly refinanced debt was deferred. The Sgo Paulo agreement did not constitute a sustainable Brady-style solution to the state's debt, although it contained some Brady elements. Some of the debt stock was forgiven by the federal government (Rs$3.8 billion), and some was to be liquidated immediately through the transfer of assets. The 6 percent interest rate subsidy was consistent with the Brady approach, in permanently reducing the carrying costs of the debt. The agreement differed from a Brady approach in two important respects, however. First, it required no immediate adjustment on the part of state governments. No expenditure cuts or revenue increases were required as prior conditions of debt relief. Second, it did not provide enough debt relief to permit the state fully to service the remaining amount. According to World Bank projections, debt service was likely to exceed the debt service ceiling by a substantial margin. Thus the agreement could set in motion a further accumulation of deferred debt service and another debt crisis in the future. After Sgo Paulo, Minas Gerais signed a similar agreement in February 1998. The state transferred Rs$11.8 billion in debt to the federal government, of which Rs$9.2 billion was refinanced for 30 years; Rs$920 million was to be amortized immediately through the transfer of assets (chiefly proceeds from the sale of the larger of the two state banks), and the rest was to be forgiven. Like SAo Paulo, Minas Gerais was granted a debt service ceiling, set at 6.7 percent of current revenues in 1998, to be phased up to 13 percent in 2000. The Minas Gerais agreement did include an innovation, intended to address the state's off-budget liabilities as the owner of two banks with negative net worths. As part of the debt agreement, the state was required to recognize the negative net worth of its two state banks and to borrow Rs$4.1 billion from the federal government to pay off their net liabilities. This was included in the total package of debt to be refinanced, raising the total to Rs$15.9 billion. Because the debt was subject to the debt service ceiling, most of the debt service was likely to be deferred. Rio Grande do Sul signed a debt rescheduling contract on similar terms in April of 1998, and Rio de Janeiro signed one later in the year. In the course of legislative 26 debate, congress offered the generous refinancing terms to all the remaining states in Brazil, even those that had been servicing their debt. After Sao Paulo, Minas Gerais signed a similar agreement in February 1998. The state transferred $11.8 billion in debt to the federal government, of which $9.2 billion was refinanced for 30 years; $920 million was to be amortized immediately through the transfer of assets (chiefly proceeds from the sale of the larger of the two state banks), and the rest was to be forgiven. Like Slo Paulo, Minas Gerais was granted a debt service ceiling, set at 6.7 percent of current revenues in 1998, to be phased up to 13 percent in 2000. The Minas Gerais agreement did include an innovation, intended to address the state's off-budget liabilities as the owner of two banks with negative net worth. As part of the debt agreement, the state was required to recognize the negative net worth of its two state banks and to borrow Rs$4.1 billion from the federal government to pay off their net liabilities. This was included in the total package of debt to be refinanced, raising the total to Rs$15.9 billion. Because the debt was subject to the debt service ceiling, most of the debt service was likely to be deferred. Rio Grande do Sul signed a debt rescheduling contract on similar terms in April of 1998, and Rio de Janeiro signed one later in the year. In the course of legislative debate, congress offered the generous refinancing terms to all the remaining states in Brazil, even those that had been servicing their debt. These agreements had limited macroeconomic impact in the near term. Although they lowered the interest rates paid by the states, state debt continued to grow through capitalization of interest. The federal government continued to be the states' creditor and continued to pay the overnight rate as the marginal cost of borrowing funds, so the interest paid by the consolidated public sector did not decline but instead shifted more of the cost explicitly onto the federal treasury. Although the large Brazilian states all ran excessive deficits, a few smaller states had good fiscal performance, such as Ceari and Parana. Places with strong fiscal reform could benefit from attracting more private investment so that a few chose to abstain from the low road of excess deficits and federal bailouts. Thus the individual states chose between extremes of good or bad fiscal behavior. The intermediate course of struggling through without substantial bailouts but also without strong enough performance to attract the attention of investors was not as attractive as the two extremes. The lack of a critical mass of states with good behavior, which might have blocked bailouts in the early 1990s, changed the incentive structure and caused some states to switch from good to bad fiscal behavior. IV. Conclusions Although political and fiscal decentralization are popular innovations in developing countries in the 1990s, both have very long standing in Argentina and Brazil. Indeed, it is difficult to imagine what the restoration of democracy would have been like without substantial political decentralization-federalism- and some concomitant form of fiscal decentralization. The details of decentralization were not foreordained, however. The two countries chose paths different from each other and from their pasts, and those differences in detail mattered, in particular for macroeconomic management. Even though macroeconomic instability dates back to the previous undemocratic and more centralized periods, dealing with a decentralized fiscal system has been one of the problems with which the architects of stabilization in the 1990s have had to deal. Decentralization has made solutions more complicated in general but not impossible. In Argentina after 1991, the provinces (not always enthusiastically) helped the central government to reduce its own deficit, agreeing to give back a portion of their coparticipation funds and taking over secondary education and health with no increase in central government transfers. Since 1995 they have kept their deficits under control. In Brazil, the stabilization started later and, as in Argentina in the early 1990s, it 27 was unclear even at the beginning of 1998 how Brazil would bring the state fiscal deficits under control. Although the central government managed to recover from the fiscal impacts of revenue-led decentralization (by reducing the increase in transfers and introducing new, unshared taxes), it was unable to contain the growth of subnational deficits, especially after the end of high inflation in 1994. SNG deficits--due largely to unpaid interest obligations-grew to equal more than half the overall public sector deficit. The factors discussed at the outset of this report influenced the outcomes of the two cases to different degrees, although on the basis of only 10 years of experience in two countries, any generalizations are preliminary. Putting in place as the status quo the rules and institutions for a hard budget constraint (intertemporally) seems to be the most important (conditions 2 and 6). The political system and the economic situation assisted in making this happen in Argentina and in making it persist, and Argentina needed every possible advantage to overturn a decades-long tradition of bad macroeconomic management. The assignment of expenditure functions and tax bases and the design of intergovernmental transfers and tax sharing (conditions 1, 2, and 3) played a passive role, adapting to the overall macro policy or amplifying its effects, but not being an independent factor. Hard budget constraint The absence of a hard budget constraint at the federal level clearly played a role in the deficit problems of both countries before their stabilizations and allowed Brazil's state debt problem to continue growing under the Plano Real. The turning point for intergovernmental fiscal relations in Argentina was when Menem and Cavallo used the post-hyperinflation fiscal boom and mandate for reform to establish a set of rules and institutions for a hard budget constraint (conditions 5 through 8, especially the last three). The strong anti-inflation commitment and tight limits on central bank credit to the public sector in Argentina after 1991 limited subnational spending and deficits in two ways. First, it allowed the federal government to reject provincial pleas for more resources after the Tequila shock with the rationale that it could not increase transfers without endangering the stabilization. Second, it constrained the provinces' ability to borrow from their own banks by eliminating their access to the central bank rediscount facility and tightening bank regulation. After the Tequila shock, most provinces had to help recapitalize their banks, rather than being able to borrow from them. For the institutionalization of the hard budget constraint in Argentina, it seems to have been important to erect the wall for states before the crisis made them hit it and forced their adjustment, because by then the deficit hawks in the central government needed to have just enough political leverage to veto a change in the status quo, not impose a new one. In fiscal affairs, the timing of institution building is critical. The test of the institutions and the establishment of their reputations as either strong or weak typically come when a shock reduces revenues, and something has to adjust. If the rules and institutions for a hard budget constraint are in place and enforced before the crisis, then enforcement through the crisis requires only enough political strength to block a change in the rule. But if the rules are not in place ahead of time, then getting the right rules requires enough political strength to overcome all potential vetoes. Federalism makes a veto easier, which is good in the first case but bad in the second. Furthermore, with rules in place ahead of time, it is more likely that the crisis will come to a head sooner, when there is still a politically and economically significant group of non- bailout states that would veto a bailout to others. Access to grants and credit Brazil and Argentina have different approaches to controlling SNG borrowing (conditions 5 to 8, especially the first two). During the 1980 and 1990s, there was much more federal lending to SNGs in Brazil than in Argentina. This included direct loans by federal financial intermediaries, federal guarantees on external loans, and the assumption and refinancing of bonds and private contractual debt. The Brazilian 28 states were still borrowing from federally owned BNDES in the late 1980s, whereas Argentina had closed the equivalent institution, BANADE, in the mid-1980s. The first Brazilian state debt crisis derived from federal guarantees on SNG external borrowing; the second derived from loans by federal financial intermediaries to SNGs. The third crisis, although it originated in bonds sold to private financial institutions, only reached dangerous magnitudes after it was federalized through the troca arrangement with the central bank. With the federal government as creditor, as in Brazil, problems with debt servicing were more politicized. Also, they came immediately at the expense of the federal government, putting it in the position of trying to recover losses. With most provincial debt held by the financial sector, as in Argentina, even if it was state banks, the federal governor could play an ostensibly more neutral role politically, enforcing contracts and regulations to allocate any losses between the states and their creditors. The repeated cycle of federal government debt refinancing coupled with caps on debt service had the perverse incentive effects that one would expect. For states above the debt service ceiling, it put a 13 percent federal tax on their own tax effort at the margin, since any extra revenue to a state raised its debt- service ceiling. Furthermore, for the most indebted states with no reasonable expectation of ever paying down the debt to the point where service would be below 13 percent of revenue, the debt service ceiling put a 100 percent tax on efforts to reduce the stock of debt, e.g. through amortization with privatization proceeds. Thus, for a governor who was not risk averse, the history of federal refinancing created an incentive to find ways to borrow more and an incentive for creditors to lend, taking a gamble that the federal government would eventually add that debt on to the conta of the state. Therefore, although the federal government was the states' principal creditor, it was unable to take advantage of this position. By the time some consensus for action had been reached, the group of bankrupt states appeared too big to fail. This strengthened the hand of the debtor states and perhaps explains the disadvantageous terms on which the federal government finally agreed to refinance the debts. Argentina has been more sparing in the use of federal lending and more inclined to liquidate nonperforming provincial debts than to refinance them. Much of the provincial debt that existed before the convertibility law was liquidated through cross-debt negotiations after 1991. Although Argentina maintains a system of emergency grants for provinces with deficits, there is a low legal ceiling on the total amount of such relief. The two countries also differed in their approaches to the regulation of new borrowing (condition 5). Brazil used ex ante controls, requiring senate approval for all SNG borrowing operations, prohibiting the issuance of bonds and borrowing from state-owned banks, and fixing a limit on total domestic banking credit to SNGs. This proved ineffective in the aggregate. Senate ceilings tended to be honored in the breach, as were prohibitions on borrowing from state banks. The states observed the prohibitions on new bond issues only in a narrow legal sense: they did not halt the capitalization of interest on existing bond debt. Nor did these regulations stop new "emergency" lending by the federal government, as shown by the Resolution 162 emergency relief program and by continued new lending from federal financial intermediaries. All the federal involvement seems to have enhanced the expectations of state borrowers and their creditors that the federal government guaranteed the debt and would take it over if necessary. Argentina, in contrast, relied on more indirect means to control SNG debt (conditions 5 through 8). Borrowing from provincial banks, for example, was not prohibited. The government exercised pro forma control over the issuance of domestic or external bonds, which formed the most vulnerable part of the otherwise hard budget constraint that the federal government imposes on the provinces. The most effective part since 1991 has been forcing the provinces to pay their debts without inflation or other forms of bailout. The convertibility law and the new charter of the central bank constrain the provinces' ability to borrow from their own banks by eliminating the central bank rediscount facility, tightening bank regulation, and eliminating most deposit insurance. Borrowing from domestic private banks is not controlled; instead, the federal government collaborates in the enforcement of debt service, deducting debt service from 29 coparticipation." This assures political support from the financial sector and gives provincial governments an incentive not to borrow excessively. It does, however, make private creditors less sensitive to risk and inclined to lend excessively. Revenue-led decentralization Increasing intergovernmental transfers without a corresponding mandate to provide services (violation of condition 3) seems to lead to larger and less efficient government sectors, as in Brazil, but not necessarily to higher deficits. A deficit problem could have been expected at the federal level, but it was averted by higher taxes-in Brazil by increasing the tax rates and in Argentina by rapidly increasing the real collection rates after 1991. In Argentina the federal government and at least the larger states disagree about whether the increased transfers were more or less than enough to cover the increased costs. In any case, the hard budget constraint from the central bank forced both levels of government to adjust their expenditures to fit within their revenues, and this was possible without blatant declines in the level of service. Revenue autonomy Contrary to conventional wisdom, large formula-driven intergovernmental transfers of revenue were not a cause of unsustainable deficits (condition 1). With the provinces of Argentina and the smaller states of Brazil, being dependent on the federal government for resources gave the federal government the leverage to force states to service their debt. In Argentina creditors could obtain priority access to the transfers, which helped to keep the primary deficits down and to motivate SNGs to avoid further buildup of debt. In contrast, the largest Brazilian debtor states received very few recurrent federal transfers, relying almost entirely on local taxes. This independence seemed to fortify their bargaining position, enabling them to refuse paying debt service to the federal government. Expenditure autonomy The more important difference seems to have been in the power to reduce costs (condition 4). Constitutionally imposed salary and benefit guarantees limited the states' ability to adjust in Brazil. Personnel was by far the largest item of state expenditure, and states' inability to cut these costs made default on debt hard to avoid.14 Being able to blame uncontrollable wage bills on the federal constitution shifted some of the political onus for deficits onto the federal government, weakening its case for state adjustment. In Argentina, in contrast, the freedom to reduce salaries and dismiss staff (often increasing efficiency) allowed provinces to adjust to the crisis by reducing the wage bill, as exemplified by the C6rdoba case. Power of the presidency Although the constitutional powers of the presidency are strong in the two countries, the Argentine president also derived important benefits from his role as leader of a relatively unified, disciplined party system (conditions 9 and 12). Party cohesion in Argentina, although less pronounced than in many parliamentary systems, sufficed to get important provincial governors to support the program of fiscal adjustment required by the Convertibility Plan, starting with the two largest Justicialista provinces, Buenos Aires and Santa Fe. While the Argentine governors acted as players on their national party teams, the Brazilian governors acted as leaders of their provincial teams, for whom party affiliation was only a flag of convenience. And national legislators acted mainly as members of their provincial team, with loyalty to the national party depending on the current stance of their governor. Thus Brazilian president, despite being the 30 nominal leader, could not depend with as much certainty on the governing coalition for support in putting through a program. The Argentine president's authority to adjust was enhanced by the trauma of hyperinflation. The economic collapse and civil unrest that accompanied the hyperinflation of the late 1980s prepared the political ground for radical measures. By invoking Justicialista party discipline and raising the specter of a return to hyperinflation, Menem was able to marshal broad support for his program. The two fit well together, for it was the Justicialistas' first return to power since the military coup ousted and then persecuted them. Getting reelected and avoiding another takeover from the right required the Justicialistas to demonstrate that they could end hyperinflation. To reduce the deficit in Brazil, the president required a means to force states to adjust, meaning to cut non-interest spending and raise taxes in order to reduce their contributions to the aggregate deficit. To do this, he needed leverage over the state governors (conditions 9 and 10). But Brazilian federalism made this difficult. The "politics of governors" in the congress made it difficult to assemble a consensus behind concerted action against any major debtor. Also, although he took office on the heels of quadruple-digit inflation-indeed as a hero for having stopped it- the popular and institutional support for anti-inflation policies in Brazil was not as strong as in Argentina. Due to widespread indexation, inflation had not left as traumatic memories in Brazil as it had in Argentina. Critical mass of reformers The contrasting fiscal performance of the subnational government sector in the two countries, at least up through mid-1998, illustrates the importance of getting a critical mass of fiscally prudent states to help the national executive enforce a hard budget constraint. Slo Paulo and the other large Brazilian states ran big deficits, threatened to wreck the banking system by defaulting, and then got the federal government to bail them out. With Sao Paulo setting the standard for leniency, all the other states demanded similar treatment, and the federal executive lacked a critical mass of states to support resistance. The situation was different in Argentina after 1991. Although Buenos Aires province used its weight to maximize its share of federal transfers, Justicialista party discipline and a sense of responsibility got the province to stay within its budget constraint, starting at the time of stabilization, setting a standard toward which the mid-size and small provinces could be pushed.'5 In the early 1990s, generous subsidies bought the support of the low- population provinces for the reform program, putting the rules of the hard budget constraint in place. Thus Menem combined party discipline with a strategy taking advantage of the size disparities to build a coalition for fiscal control. Federal intervention In Argentina the primacy of national over provincial laws is spelled out in the constitution and has been demonstrated in repeated interventions by the federal government throughout this century and the last, by civilian as well as military governments (condition 10). Although interventions have been rare since 1983 and absent under the 1994 constitution, there was no doubt of the federal government's legal and military capacity to intervene, and this power seems to have been a latent factor in the ability of the national government to enforce the states' debt obligations. In Brazil, by contrast, the constitutional limits on federal intervention were taken seriously, and strong state militias assured governors that the federal government could not intervene easily. Therefore, they acted like sovereign debtors. Economic conditions and timing 31 Economic conditions and timing were also instrumental to success of fiscal reform in Argentina's provinces, which benefited from rapid growth of the economy and of provincial revenues during the initial years of the stabilization. This made provinces amenable to Menem and Cavallo's efforts to shift part of the costs of the government's own adjustment onto them and to limit the opportunity to finance deficits through borrowing from provincial banks. It seems to have been important to enact these measures while SNG revenues were still rising and well before the crisis hit. Then at the time of crisis, the deficit hawks in the central government merely needed enough political leverage to veto a change in the status quo, not impose a new one. Menem also benefited by enacting his reform when the aggregate stock of SNG debt was relatively low. This reduced pressure for federal bailouts and created wide options for the provinces. Brazil, in contrast, had slower economic growth in the period leading up to the state debt crisis of the late 1990s (slower than Argentina's growth in the early 1990s and only slightly faster than Brazil's before 1994), and the stocks of debt were already fairly large by the time the government turned its attention from achieving price stability under the Plano Real to confronting the problems of state debt. Still learning The success and failure thus far in Argentina and Brazil must be viewed in a long-run perspective. Policymaking in a democracy, especially a new one, is a continuous learning process. In addition to the changes of rules and players, the reputations of institutions as well as individuals are being established. Those with mainly regional interests are learning to get more out of the center, changing and pushing the rules in their favor, while those with nationwide interests are learning to defend them. The cases of Argentina and Brazil illustrate the problems and also some of the solutions in establishing fiscal restraint in federal systems, especially when the restraint has been broken down for a long time. They show that the quality of macro fiscal management by subnational governments depends not only on the contract dividing revenue and responsibility between the national and subnational governments but also on the nature of the subnational entity's representation in national politics, particularly whether it occurs through a strong party system. 32 Epilogue - Brazil In Brazil, the situation has evolved in several ways since the onset of the state debt crises, many of them positive. First, the central bank has tightened limits on the domestic financial system's aggregate exposure to subnational debt. It has also taken steps to limit lending by federal financial intermediaries to subnational governments. Although these controls are not air tight, they tend to limit state borrowing to unpaid service on outstanding debt. Second, the agreements of 1998 are written in a way that allows and requires the federal government to withhold debt service from transfers, as in Argentina. Thus, even when the much-publicized refusal of Minas Gerais in January 1999 to pay debt service resulted in the federal government covering the state's debt service in order to protect the national credit rating, the state was effectively forced to pay, because the Federal government deducted their payments from their transfers. Third, Sao Paulo has become one of the reform states, making it more technically feasible to impose budget constraint on the states as a group. If this strategy works, the recent crisis may end up being of the same importance for Brazil as the 1994-95 crisis was for Argentina - testing the system and setting precedents. Some underlying causes of the states' fiscal problems have also been addressed. In 1998, the Brazilian Congress approved a Constitutional amendment that will allow states to dismiss staff, provided their personnel spending exceeds a threshold percentage of state revenues. Congress also approved some amendments to pension legislation, which help but are not yet sufficient to provide states with the means to respond to fiscal pressures without resorting to eventual default. Congress is currently (April 1999) debating a comprehensive 'law of fiscal responsibility' that would, following the New Zealand model, introduce stronger requirements for fiscal responsibility and restrict borrowing at all levels of government. 33 References The word "processed" describes informally reproduced works that may not be commonly available through library systems. Afonso, Jose Roberto. 1993. "Decentralizagao fiscal e financiamento da saude: Algumas ideias ou provocagoes." CEPP Discussion Paper 16. CEPP. October. Processed. 1994. "Decentralizaglo: Um estudo de caso sobre o Brasil." Banco Nacional de Desenvolvimento Economico e Social, Rio de Janeiro. Processed. Alfonso, Rui de Britto Alvares. 1997. Os estados e a descentralizaqdo no Brasil. Serie Politica Fiscal 93. Santiago de Chile: CEPAL/GTZ. Ames, Barry. 1995a. "Electoral Rules, Constituency Pressures, and Pork Barrel: Bases of Voting in the Brazilian Congress." Journal of Politics 57 (May): 324-43. 1995b. "Electoral Strategy under Open-List Proportional Representation." American Journal of Political Science 39 (May): 406-33. 1998. "Toward a Theory of Legislative Parties in Brazil." Department of Political Science, University of Pittsburgh. Processed. Argentina, Republic of. Various years. Informe econdmnico regional. Buenos Aires. Bomfim, Antillio, and Anwar Shah. 1994. "Macroeconomic Management and the Division of Powers in Brazil: Perspectives for the 1990s." World Development 22 (4):535-42. C6rdoba, Province of. 1997. Gobernadoresy legisladores de C6rdoba, 1820-1996. C6rdoba. Dias, Jose Luciano de Mattos. 1991. "A federagao brasileira: Estructura e desempenho." Ph.D. diss., Rio de Janiero. Dillinger, William. 1997. "The Brazilian Debt Crisis." World Bank, Washington, D.C. Processed. Eaton, Kent H. 1998. "Fiscal Policy Making in the Argentine Legislature." Department of Political Science, Yale University, New Haven, Conn. Processed. Fornasari, Francesca, Steven B. Webb, and Heng-Fu Zou. 1998. "Decentralized Spending and Central Government Deficits: International Evidence." LCSPR/DECRG World Bank, Washington, D.C. Processed. Gibson, Edward L. 1997. "Federalism and Electoral Coalitions: Making Market Reform Politically Viable in Argentina" Department of Political Science, Northwestern University, Evanston, IL. Processed. Gibson, Edward L., Ernesto F. Calvo, and Tulia G. Falleti. 1998. "Reallocative Federalism: Overrepresentation and Public Spending in the Western Hemisphere." Department of Political Science, Northwestern University, Evanston, [L. Processed. 34 Hagopian, Francis. 1996. Traditional Politics and Regime Change in Brazil, Cambridge, U.K.: Cambridge University Press. IDB (Inter-American Development Bank). 1997. Latin America after a Decade of Reforms: Economic and Social Progress in Latin America 1997 Report. Washington, D.C. Jones, Mark P. 1998. "Federalism and Party Discipline in the Argentine Congress." Department of Political Science, Michigan State University, East Lansing, MI. Processed. Jones, Mark P., Pablo Sanguinetti, and Mariano Tommasi. 1997. "Politics, Institutions, and Public Sector Spending in the Argentine Provinces." Forthcoming in James Poterba and Juergen von Hagen, ed., Budget Institutions and Fiscal Outcomes. Chicago: National Bureau of Economic Research and the University of Chicago Press. Mainwaring, Scott. 1997. "Multipartism, Robust Federalism, and Presidentialism in Brazil." In Scott Mainwaring and Matthew Shugart, ed., Presidentialism and Democracy. New York: Cambridge University Press. Murphy, Ricardo. 1995. Fiscal Decentralization in Latin America. Baltimore, Md.: Johns Hopkins University Press for the Inter-American Development Bank. Peralta-Ramos, Monica. 1992. The Political Economy ofArgentina. Boulder, Colo.: Westview Press. Riker, William H. 1964. Federalism: Origin, Operation, and Significance. Boston: Little, Brown. Rosenblatt, David. 1994. "States' Voting Power and the Federal Budget: Preliminary Results from Brazil." LCSPR, World Bank, Washington, D.C. Processed. Rudolph, James. 1985. Argentina: A Country Study. Washington, D.C.: U.S. Government Printing Office. Samuels, David J. 1998. "Progressive Ambitions, Federalism, and Pork-barreling in Brazil." Department of Political Science, University of California, San Diego, La Jolla. Processed. Souza, Celina. 1996. "Redemocratization and Decentralization in Brazil: The Strength of the Member States." Development and Change 27: 529-5 5. Stepan, Alfred. 1997. "Toward a Comparative Theory of Federalism." All Souls' College, Oxford University. Processed. Tanzi, Vito. 1995. "Fiscal Decentralization." Annual Bank Conference on Development Economics 1994. Washington DC. Ter-Minassian, Teresa. 1997. "Brazil." In Teresa Ter-Minassian, ed., Fiscal Federalism in Theory and Practice: A Collection ofEssays. Washington, D.C.: International Monetary Fund. Ter-Minassian, Teresa, and Jon Craig. 1997. "Control of Subnational Government Borrowing." In Teresa Ter-Minassian, ed., Fiscal Federalism in Theory and Practice: A Collection of Essays. Washington, D.C.: International Monetary Fund. Tiebout, Charles. 1956. "A Pure Theory of Local Expenditures." Journal ofPolitical Economy 64: 416-24. 35 Treisman, Daniel. 1998. "Decentralization and Inflation in Developed and Developing Countries." Department of Economics, University of California, Los Angeles. Processed. Weingast, Barry A. 1995. "The Economic Role of Political Institutions: Market-Preserving Federalism and Economic Development." Journal of Law, Economics, and Organization 1 l(1): 1-3 1. Wildasin, David E. 1997. "Fiscal Aspects of Evolving Federations: Issues for Policy and Research." Policy Research Working Paper 1884. Policy Research Department, World Bank, Washington, D.C. Processed. Willis, Eliza, Christopher Garman, and Stephan Haggard. 1997. "The Politics of Decentralization in Latin America." Department of Political Science, University of California - San Diego, LaJolla, CA. Processed. World Bank. 1996a. Argentina: The Convertibility Plan: Assessment and Potential Prospects. Washington, D.C. 1996b. Argentina: C6rdoba Public Sector Assessment: Proposals for Reform. Washington, D.C. 1996c. Argentina: Provincial Finances Study. Washington, D.C. 1998. Argentina: The Fiscal Dimension of the Convertibility Plan. Washington, D.C. 36 Endnotes 'The debt moratorium by Minas Gerais in January 1999 and the ensuing currency crisis occurred after this paper was completed. An epilogue will be added after the episode becomes clearer. The Argentine data are too confused before 1991 to obtain estimates of provincial debt stock. Actual debt stocks are publicly available only starting in 1996, but the early 1990s can be extrapolated backward using data on deficits, because the currency of account has a stable value. 3 In classic decentralized fiscal systems, local voters are taxpayers who pay the bill and therefore exercise restraint on spending (Tiebout 1956; Weingast 1995). 4 Resolution 2008 of the National Monetary Council prohibited private banks from increasing their holdings of state debt, excluding bonds. This was revised in 1998 by Resolution 2461, limiting the maximum exposure of the entire banking system to the level of state debt in September 1997, adjusted for inflation at 80 percent of a certain price index. This ceiling is binding now, but the resolution does not say how credit opportunities should be allocated to states and banks when the debt declines. 5 For presidents, limited party discipline and loyalty have an equivocal impact, however. Presidents cannot fully count on the support of their party or coalition, as they might with highly disciplined parties. If the president's party had a majority in congress, disciplined parties would be a clear asset. In a system of fragmented parties, weaker discipline also has advantages for presidents, however, enabling them to entice other parties to support them. It means that presidents rarely face disciplined majorities determined to block them. 6 The Tanzi-Oliveira effect is effect of inflation reducing the real value of tax revenues because of lags in collection. 7Based on an interview with Luis Antonio Zapata. Provinces later claimed to have been blindsided by this agreement. In the 1994 revisions to the constitution, they insisted on added a clause prohibiting the transfer of federal programs without adequate compensation. 9 Absorbing the net losers would cost the federal government, however. The federal government would have to not only pick up the provincial subsidy, but also pay more. This is because the contribution rate to the federal pension system is lower than in provincial plans. Thus the amount of contributions coming in will fall. Because benefits paid to existing retirees will remain unchanged, the gap between contributions and pension payments will widen. Over time, the net cost to the federal government is likely to fall, however. This is partly because pensioners retiring after the transfer will receive the national benefit regime, which is less generous than the provincial plans. In addition, the number of retirees will fall, because the minimum retirement age is higher in the national system than in the provinces. Finally, once transferred to the federal scheme, pensions will be subject to the Solidarity Law of 1995, which eliminated automatic wage indexation for pensioners and imposed ceilings on the size of the monthly pension. '0 The new administration discovered large numbers of statutory employees on fraudulent long-term sick leave. It forced these employees back to work, dismissing 2,256 substitutes and 4,264 contract workers. n1 Central Bank Resolution 747,23 December 1997. 12 The central bank, on orders from the federal treasury, conducted such swaps a few times on a temporary basis in the 1980s, but the exchange in 1994 was more permanent. Initially, the swaps were for treasury debt, but when there were no longer enough treasury bonds, the central bank issued its own notes. 13 These deductions were made to exercise collateral guarantees demanded by creditors in the early years after hyperinflation. Recently, some provinces have been able to borrow in the market without such guarantees. 37 " In 1998, the Brazilian Congress did, in fact, approve a Constitutional amendment that would allow states to dismiss staff, provided their personnel spending exceeded a threshold percentage of state revenues. The congress was also considering amendments to pension legislation. If approved, these amendments would go a long way to providing states with the means to respond to fiscal pressures without resorting to default. 1s While Sao Paulo was too big to fail, Buenos Aires would have been too big to rescue, because it generates half of the nation's GDP and even more of its tax revenue. If a federal bailout were necessary, the province would bear most of the cost of economic crisis and higher federal taxation. 38 Policy Research Working Paper Series Contact te Author Date for paper WPS2097 The Theory of Access Pricing: Tommaso M. Valletti April 1999 G. Chenet-Smith An Overview for Infrastructure Antonio Estache 36370 Regulators WPS2098 Reconsidering the Evidence Madhur Gautam April 1999 M. Gautam on Reiurns to T&V Extension Jock R. Anderson 30442 in Kenya WPS2099 Mutuai Funds and Institutional Estelle James April 1999 M. Leenaerts Investments: What is the Most Gary Ferrier 84264 Efficient Way to Set Up Individual Jamnes Smalhout Accounts in a Social Security Diniti Vittas System? WPS2100 Decentralization in Regional Fiscal Lev Freinkman April 1999 Z. Nekaien-Nowrouz Systems in Russia: Trends and Plamen Yossifov 39057 Unks to Economic Performance WPS2101 Are Wages and Productivity in Dorte Verner April 1999 H. Vargas Zmbabwe Affected by Human Capital 37871 Investment and internLtional Trde? WPS2102 Self-Employment and Labor William F. Maloney April 1999 T. Gomez Turnover: Cross-Country Evidence 32127 WPS2103 Was Credit Channel a Key Monetary Hyun E. Kim April 1999 D. Peterson Transmission Mechanism following 32692 the Recent Financial Crisis in the Republic of Korea? WPS2104 The Relative Effects of Skill Taye Mengistae April 1999 A. Bonfield Formation and Job Matching on 31248 Wage Growth i. Ethiopia WPS2105 Wage Rates and Job Queues: Tave Mengistae April 1999 A. Bonfield Does the Public Sector Overpay 31248 n Ethiopia? WPS2106 Subjective Economic Welfare Martin Ravailion April 1999 P. Sader Michael Lokshin 33902 WPS21 07 Approaches to Uberalizing Services Sherry M. Stephenson May 1999 L. Tabada 36896 WPS2108 Public Goods and Ethnic Divisions Alberto Alesina May 1999 K. Labrie Reza Baquir 31001 Wiliam Easterly WPS2109 When Is Fiscal Adjustment an William Easterly May 1999 K. Labrie illusion? 31001 Policy Research Working Paper Series Contact Title Author Date for paper WPS2110 Life during Growth: International William Easterly May 1999 K. Labrie Evidence on Quality of Life and 31001 Per Capita Income WPS2111 Agricultural Land Reform in Postwar Toshihiko Kawagoe May 1999 P. Kokila Japan: Experiences and Issues 33716 WPS2112 industrial Policy after the East Asian Ashoka Mody May 1999 S. Kpundeh Crisis: From "Outward Orientation' 39591 To New Internal Ca,abilities? WPS2113 Wage Determination and Gender Stefano Paternostro May 1999 N. Nouviale Discrimination in a Transition David E. Sahn 34514 Economy The Case of Romania WPS2114 Economic Reforms and Total Factor Pablo Fajnzylber May 1999 S. Holt Productivity Growth in Latin America Daniel Lederman 37859 And the Caribbean (1950-95): An Empirical Note WPS2115 Foreign Investment and Productivity Simeon Djankov May 1999 R. Vo Growth in Czech Enterprises Bernard Hoekman 33722 WPS2116 Does Child Labor Displace Martin Ravallion May 1999 P. Sader Schooling? Evidence on Behavioral Quentin Wodon 33901 Responses to an Enrollment Subsidy WPS2117 On the Size and Number of Regional Soamiely Andriamananjara May 1999 L. Tabada Integration Arrangements: A Political 36896 Economy Model WPS2118 Developing Countries and the Next Anne 0. Krueger May 1999 L. Tabada Round of Multilateral Trade 36896 Negotiations WPS2119 Tax Deductions, Environmental Ian Parry May 1999 R. Yazigi Policy, and the "Double Dividend' Antonio Bento 37176 Hypothesis WPS2120 Can Africa Export Manufactures? Ibrahim A. Elbadawi May 1999 C. Bernardo The Role of Endowment, Exchange 31148 Rates, and Transactions Costs