FISCAL DECENTRALIZATION AND SUBNATIONAL EXPENDITURE POLICY

In most countries, the share of subnational outlays in total government spending has increased reflecting their more active role in service delivery, and, in some cases, greater policyma

I I NTRODUCTION

The need to improve the quality and efficiency of government spending has placed subnational governments at the forefront of service delivery, particularly in the social sectors In many countries, it has been recognized that subnational governments have an important role to play inthe implementation of public policies aimed at fostering social and human development The key argument for reliance on subnational governments for service delivery is that they are closer

to the intended beneficiaries of public programs For this reason, subnational governments are believed to be better equipped to extract information on local preferences and needs more effectively than the central government, and to be more accountable to local residents With greater voice in the design and implementation of public policies, local residents can also benefit from greater choice in the goods and services delivered to them by the government

Reliance on subnational governments, or the private sector, for the delivery of public goods and services is nevertheless not without pitfalls Fiscal decentralization may lead to the wasteful duplication of expenditure functions among the different levels of government and poor

governance, thereby reducing the quality of public services and the overall efficiency of the government An open policy question is whether the central government should rely more heavily on subnational governments for the design, rather than simply the implementation, of

1 The author is now at the Economics Department of the OECD Mailing address: Economics Department, OECD, 2, rue André Pascal, Paris 75775 Cedex 16, France Tel.: (33-1) 4524-8752,e-mail: luiz.demello@oecd.org

public programs in general, and in the social sectors in particular In this case, rather than acting

as agents of the center, subnational governments would enjoy greater policymaking autonomy

In most countries, the share of subnational outlays in total government spending has increased reflecting their more active role in service delivery, and, in some cases, greater policymaking autonomy in the design and implementation of public policies Nevertheless, the central

government still has an important role to play, even when subnational governments are key providers of public goods and services The central government is often called upon to finance provision, including in some cases the equalization of expenditure capacity among subnational jurisdictions; to avoid shortfalls in the financing of critical social services; and to provide technical assistance to subnational jurisdictions

In a nutshell, this paper is aimed at, first, reviewing the international experience with the

decentralization of government spending to subnational governments and to the private sector, with special emphasis on the social sectors, including education, health care, social assistance

and welfare, and housing A second objective of the paper is to discuss the main challenges

facing subnational governments in the design and implementation of public programs, includingthe earmarking of revenues, as well as economies of scale and spillover effects in the provision

of public services

The paper is organized as follows Section II reviews the conventional distribution of

expenditure functions among the different levels of government and discusses recent

developments in this area Section III presents some cross-country data on spending shares

among different levels of government Section IV focuses on education, health care, housing and infrastructure, and social assistance and welfare Section V discusses pending issues and policy implications.

II T HE F ISCAL R OLE OF S UBNATIONAL G OVERNMENTS : T HE “O LD ” AND T HE “N EW ”

Public finance theory has traditionally assigned a limited role to subnational governments in the design of social policies The key argument is that, based on the Musgravean functions of government, the fiscal role of subnational governments is limited to allocative efficiency, whereas the central government should focus on income redistribution and welfare, as well as

on macroeconomic stabilization Based on the principle of subsidiarity, the public goods and services that should be provided by subnational governments are those which generate benefits that can be internalized by those jurisdictions Accordingly, activities such as waste disposal, street maintenance, primary education, and urban transportation have been devolved or

decentralized to subnational governments in most countries Delegation/decentralization is considered the most effective way to deliver these services because their benefits are mostly local and there is little scope for economies of scale and interjurisdictional spillover effects in their provision

The role of subnational governments in the provision of public goods and services has been affected by a variety of factors These include demographics, technological changes in

communication and transportation, for instance, and the development of market institutions, which allows for more active private sector involvement in service delivery (Wildasin, 1998)

Moreover, it is often difficult to clearly define the benefits and beneficiaries of specific

programs For example, the benefits of public spending on primary education are known to be local but, from the point of view of poverty alleviation, they are national in scope Economies ofscale, particularly in administrative functions, may call for a more centralized management of service delivery even when services are provided locally

More recently, public finance theory has reassessed the assignment of expenditure functions among the different levels of government based on principal-agent considerations What matters

in this case is not allocative efficiency from the point of view of internalizing the benefits of provision within the jurisdictional boundaries of the provider, but the response of different layers of government to incentives for efficient, cost-effective service delivery The

consideration of the principal-agent nature of decentralized service delivery has highlighted a finer distinction between delegation and decentralization Delegation is preferred when

subnational governments are best suited as agents of the center and when clear incentives for efficiency gains can be put in place

In addition, when expenditure functions can be unbundled, rather than treated as a whole, each layer of government can perform the role that best reflects its preferences and needs or

comparative advantages For example, because primary education can be used as a tool of poverty alleviation, as mentioned above, national governments often provide financial support

to ensure a minimum level of provision in all jurisdictions, while day-to-day management responsibilities are assigned to local governments or school boards Transportation services may

be provided locally subject to regulations issued by the central government

In line with their changing role in service delivery and program design, subnational

governments have played a more active role in redistributive policies.2 Traditionally, it had beenargued that the effectiveness of subnational redistributive policies depended on taxpayer

mobility Factor mobility may prevent subnational governments from using the tax system for the purpose of redistribution, and locally-funded transfer payments may attract potential

claimants from neighboring jurisdictions, thereby discouraging local funding for these

programs More recently, the recognition that taxpayer mobility may be more limited than expected has provided some justification for subnational redistributive policies, particularly in terms of expenditure, rather than taxation, policies The central government should nevertheless remain at the forefront of redistributive policies In some cases, it may be better to use

intergovernmental grants to address differences in average income levels in various subnational jurisdictions than to use taxes and transfers to individuals (McLure and Martinez-Vazquez, 2000)

III S UBNATIONAL G OVERNMENT S PENDING : A N O VERVIEW

Subnational governments, including middle-tier jurisdictions, such as states and provinces, and local governments, are important providers of public goods and services Fiscal decentralizationindicators vary considerably among regions but, in more decentralized countries, such as large

2 The role of subnational governments in macroeconomic stabilization, another classical

function of government according to Musgrave, remains limited To avoid macroeconomic disarray due to decentralization, as highlighted in the literature (e.g., Ter-Minassian, 1997; de Mello, 2000), it has been recognized that the central government plays a crucial role in

imposing hard budget constraints at the subnational levels of government

federations, subnational governments account for over 36 percent of total governments spending

on average, or approximately 14 percent of GDP (Table 1) A sizeable share of subnational spending is financed through local revenue mobilization, but subnational governments also rely heavily on grants and transfers from higher levels of government, which account for nearly a third of their revenues on average Interestingly, subnational governments tend to rely on grants and transfers regardless of their size, measured in terms of ratios of expenditure to GDP and total government spending

An interesting question is how the changing role of subnational governments in service deliveryhas affected the composition of government spending among different tiers of government (Table 2) Internationally comparable information on public expenditure by function for the different levels of government is limited In general, education and health care account for a large share of subnational expenditures, particularly in the case of middle-tier jurisdictions Housing also accounts for a large share of local government spending, in addition to education and health care Over time, middle-tier governments have devoted a growing share of their total spending to health care programs Interestingly, although social security and welfare continues

to weigh more heavily on central government budgets, these programs have accounted for a growing share of local governments spending over time Local government spending on

housing, already high in relation to local government spending, has increased over time

Cross-country indicators of fiscal decentralization should be interpreted with caution

Subnational fiscal data are often less reliable than those of the central government and, more importantly, the quantitative indicators reported above do not allow for a more in-depth analysis

of the qualitative aspects of intergovernmental expenditure arrangements For instance, the rise

in local government spending on social security and welfare programs, as noted above, may be due to the delegation of service delivery functions to these jurisdictions, rather than greater autonomy in the design and implementation of their own welfare programs Moreover, based onthe data at hand, it is not possible to ascertain how much of subnational spending is financed through grants and transfers from higher levels of government, rather than local revenue

mobilization.3 Intergovernmental grants and transfers differ considerably in their nature, and hence the incentives they create at the local level For instance, local governments tend to have greater autonomy in service delivery and in program design when spending is financed through block, rather than matching or earmarked, grants

IV T HE I NTERNATIONAL E XPERIENCE

A Education

International experience varies on the role of subnational governments in the provision of education services In many countries, particularly in OECD member countries, the central government plays a key role in financing and standard setting, rather than service delivery, including personnel management, budget preparation, procurement activities, and

administration Centralized financing and standard setting is justified on the grounds that

education programs can be used as a poverty-alleviation instrument and as a means to equalize

3 The GFS does not provide data on earmarked grants and transfers to, and within, subnational governments

expenditure capacity among subnational units Outlays on education tend to account for a large share of subnational budgets, particularly at the local level Decentralization beyond the public sector is also common in many countries, where partnerships with the private sector have been encouraged and fully managed by subnational, particularly local, jurisdictions subject to the parameters set by the central government.4 In the developing world, unlike most industrial countries, the central government typically plays a more active role in service delivery due primarily to capacity constraints at the subnational level.5 Also, the decentralization of educationservices to subnational jurisdictions has been motivated by the need to take account of regional diversities, particularly of an ethnic and linguistic nature, in many industrial and developing countries.

Subnational governments may play an important role in personnel management in the educationsector The decentralization of education programs typically begins with the delegation of administrative functions, particularly personnel management, to subnational governments In the United Kingdom, the United States, and Canada, among other countries, emphasis has been

4 Among OECD countries, the main education decentralization programs are

Grant-Maintained-Schools (United Kingdom); Charter Schools (United States, also in New Zealand and Canada), and Future Schools (Australia) According to the General Accounting Office of the United States, based on the study of Dade County (Florida), Prince William County (Virginia),

and the Edmonton school district (Alberta, Canada), these decentralization programs have a positive impact on the adoption of innovative managerial and teaching practices, as well as the responsiveness to local preferences and needs (Rondeau, 1997)

5 For instance, in Mexico, 65 percent of all schools and 75 percent of those in basic education (preschool and primary education) were federally funded and controlled until 1992 (Juarez, 1999) Federal involvement in education was unequally distributed across the country, but the federal government designed education policies, set the academic calendar, and assigned

textbooks and other activities After 1992, when a comprehensive decentralization program was implemented, the states were granted autonomy to control and operate all schools in basic education, subject to federal funding and regulation

placed on school districts, rather than local governments for the decentralization of personnel management functions When greater personnel management autonomy is granted to

subnational governments, it is important to avoid the risk of overstaffing and excessive

discretion over pay in the course of decentralization, particularly if governance is poor at the subnational level In some countries, unionization and issues related to civil servants’ benefits and entitlements have created resistance to decentralization efforts (i.e., Venezuela).6 Moreover,

in a labor-intensive sector such as education, subnational budgets become excessively

dependent on the central government’s personnel policy when teacher’s salaries are set by the center In many countries in Latin America, this situation created unfunded mandates at the locallevel and led to the implementation of special transfer programs to finance subnational payroll outlays (i.e., Colombia) In Brazil, income differentials among subnational units resulted in sizeable pay inequality between state and municipal schools and sometimes within the same state This problem has nevertheless been mitigated, at least in part, due to the earmarking, since

1998, of 60 percent of primary education spending to teacher’s salaries.7

Financing arrangements vary across countries In general, education programs are financed by the center, even when implemented by subnational governments When subnational co-

financing is required, some countries earmark subnational revenues to avoid financing

shortfalls, in addition to the grants and transfers received from higher levels of government In

6 In 1994, Venezuela implemented a decentralization program in the education sector by

unifying the federal, state, and municipal school networks (pre-school, primary and secondary education) and transferring the respective assets and personnel to the states, which would be responsible for service delivery See Tracy (1996), for more information

7 For more information on the decentralized service delivery in Latin America, see ECLAC (1998), and Burki, Perry and Dillinger (1999)

Brazil, for example, 15 percent of state and municipal revenues are earmarked to finance

primary education programs In other countries, local education budgets may be financed entirely through grants and transfers from higher levels of government In many cases, transfers are formula-based and take expenditure needs into account, particularly when the equalization

of subnational expenditure capacity is pursued Functions are costed subject to the minimum standards set by the center.8 Other countries also rely on the decentralization of service delivery

to strengthen cost recovery through the collection of user charges and fees for the provision of some services (i.e., contingency fees, and (subsidized) school fees, among others) (Litvack and Seddon, 1999) Willingness to pay for service delivery is believed to increase when the services are provided by local governments Other cost recovery schemes such as community financing have also emerged as means for the central government to reduce the fiscal burden of education service provision.9

A critical policy question is whether subnational governments are more efficient than the center

in the provision of education services Efficiency gains can be assessed when education

attainment indicators (i.e., literacy rates and school enrollment rates, among others) are

improved for the same consolidated spending level as a result of decentralization, or,

8 In Brazil, for example, a national minimum spending level per primary school student is set annually with a federal top-up grant to subnational governments that cannot afford the minimumspending level set by the center See de Mello (2001), for more information

9 Community financing has been an important aspect of decentralization in countries where governments have been unable to meet the demand for education services In this case,

community-funded schools have coexisted with the public school network to offer schooling with particular religious, ethnic, political, or other orientations With the exception of Bolivia and Nicaragua, community financing has been less widespread in Latin America than in Africa and Asia (Bray, 1996; ECLAC, 1998) See also de Mello (2004), for cross-country evidence on the association between fiscal decentralization and social capital

alternatively, when the same indicators can be achieved through lower spending Simple

correlations between education attainment and decentralization indicators are typically weak and should be interpreted with caution because they do not control for other determinants of education attainment.10 Neither do these indicators capture the more qualitative nature of

decentralization, which could explain differences in attainment for countries with similar

subnational spending shares For efficiency to be boosted, it is important to know whether service quality improves through the decentralization of service delivery, access to education services is facilitated, and spending reaches the target population.11

Anecdotal evidence suggests that the quality of services provided by subnational governments tends to be lower than those provided by higher-level jurisdictions, particularly in the

developing world, but it is hard to control for all determinants of quality, in addition to

decentralization In industrial countries, where empirical evidence is more readily available, pupil performance, measured in standardized test scores, tends to improve through greater local government autonomy in education policymaking Unfortunately, data inadequacies often prevent the use of this more sophisticated methodology in the developing world, where more ambitious decentralization programs are being, or have been, implemented

10 See Baldacci, Guin-Siu, and de Mello (2003), for cross-country evidence on the efficiency of government spending on education and health care

11 More sophisticated methods of measuring efficiency gains in the education sector consist of comparing test scores in a given school district relative to a control group in another district thathas not been affected by decentralization, after controlling for other observable determinants of school performance, such as class size, and socio-economic indicators

In Brazil, recent research has shown that the states in which public schools are free to manage their budgets and appoint the headmaster, as well as to implement participatory management schemes, tend to have lower repetition and dropout rates, and their students tend to have better test scores.12 There is some empirical evidence that quality can be improved when subnational governments, particularly at the local level, enjoy greater autonomy in program design and implementation with increased use of demand-driven, result-oriented, participatory

administration in public schools.13 In Latin America, this trend has been referred to as the municipalization of the education sector, as has been particularly noteworthy in Brazil and Chile(de Mello, 2001).14 Overall, it has been suggested that improvements in school systems depend heavily on empowering local authorities to adapt programs to local conditions (Dalin, 1994)

Information on access and the incidence of public spending is scarce, particularly at the

subnational level A key determinant of efficiency in the provision of education services is targeting, or whether public outlays on a given program actually reach the intended

12 A study of the Curumim Program (Paes de Barros, Mendonça and Soares, 1998), implemented

in the state of Minas Gerais, Brazil, suggests that test scores, enrollment and repetition rates, as well as age-grade gap indicators have improved in a sample of pupils in primary schools that participated in the program, after controlling for other determinants of school performance, such

as parents’ educational background and occupation, and number of books at home, among others

13 The most common participatory administration schemes in Brazil are the School Council, available in 1997 in over 37 percent of the nearly 42,000 schools funded by state governments, and the Parents and Teachers Association, available in nearly one-third of states schools See Parente and Lück (1999), for more information

14 Chile is an example of increased private-sector involvement in the education sector in the developing world, where subsidized private schools now account for over one-third of total school enrollment Competition among schools to attract students has increased because funding

is based on grade-adjusted school attendance However, there is evidence that decentralization has not by itself removed inequalities between jurisdictions of different incomes (Parry, 1997)

beneficiaries Benefit incidence studies indicate that that the poor reap a disproportionately small share of the benefits of public spending on education, further underscoring the need to target these outlays more accurately Information on incidence rates by government level is evenscarcer, and the correlation between decentralization parameters and incidence rates tends to be weak The information reported in Table 3 shows that, for instance, in the case of primary education, for the same level of expenditure decentralization—measured as the ratio of

subnational spending to central government spending—the share of public spending accruing to the lowest income quintile is almost twice as high in Argentina as it is in Pakistan In the case ofpublic spending on secondary education, which tends to be less pro-poor than spending on primary education, both countries have similar incidence rates This suggests that the design of education programs is more important than intergovernmental spending levels per se, and that other factors, such as the share of private spending on education and school improvement programs, also play an important role in determining incidence rates across countries

B Health Care

As in the case of education, decentralization in the health care sector may enhance the

efficiency of public spending through improved service quality and coverage Decentralization

of health care tends to be more complex than in other sectors because diseconomies of scale tend to discourage subnational provision of costly curative health care services and spillover

The objectives of decentralization in the health care sector also vary significantly between developing and developed countries Few developing countries have long-term experience with health sector decentralization and reform packages have seldom been evaluated The focus of decentralization in the developing world tends to be on increasing the coverage of the public health care system, particularly in rural areas and in less developed regions, and on securing additional sources of finance from subnational jurisdictions This has been the main elements of decentralization in Latin America, for instance, since the early 1990s (i.e., Brazil, Mexico, Colombia, Bolivia) Private-sector providers coexist with the public health care system, private-public service quality differentials are sizeable, and the incidence of public spending tends to bebetter than that of education because the better-off tend to rely on private health care providers Developed countries, on the other hand, typically focus on efficiency gains through

decentralization outside the public sector This includes the increased use of co-payment

schemes, delisting of entitlements, shifts toward the public contracting model, separation of provision and management from purchasing functions, contracting out of hotel functions and other supporting services, and private capital investments

15 In Brazil, for instance, municipal joint ventures have been developed in recent years to generate economies of scale in administrative activities, including hiring new staff, licensing private health care providers, and procurement More recently, efforts have been made to strengthen the institutional framework within which these ventures are created (Ribeiro and Costa, 1999) According to the Brazilian Ministry of Health, as of July 1999, there are 143

health care reforms in many OECD countries have focused on increasing private-sector

involvement in service delivery, as well as institutional and managerial restructuring,

consolidation and regionalization of purchasing functions, and consolidation of health care institutions through mergers and vertical integration of services (Ranade, 1998).17 In many countries, there is widespread public opinion support for national health insurance schemes (i.e.,Canada, Germany, the Netherlands, New Zealand, Sweden, the United Kingdom) As a result, proposals to privatize the funding of health care have faced opposition and attention has focused

on improving the allocation of funds to health care providers within the framework of public funding Efficiency gains have also been sought by improving consumer choice as an incentive toward greater competition and cost-effectiveness among service providers.18

Financing plays a crucial role in the decentralization of health care Most health care services provided publicly are financed by the central government through general taxation, often in conjunction with the social security system As in the case of education, many countries resort intermunicipal ventures in Brazil, covering a total of 1,740 municipalities

16 See OECD (2003), for further discussion

17 The New Zealand government’s proposal to institute “managed competition” in the

purchasing and provision of care—undertaken in 1991 at the same time as reforms in the UnitedKingdom—was motivated by a relatively fragmented funding system, rising levels of user charges and ensuing barriers to access, long waiting lists, and overcapacity of rural hospitals The reforms were also motivated by public opinion support for greater private sector

involvement in service delivery, and by the overall pro-market reforms that were implemented

at the time in New Zealand

18 See Hurst and Jee-Hughes (2000), for a discussion of performance indicators in health care systems in OECD countries

often based on historical input costs, rather than on equalizing grants based on per capita

expenditure needs and standardized costs Input-oriented budgeting tends to bias public

spending towards curative, rather than preventive, care In the developing world, financing shortfalls are common, and intergovernmental grants are often aimed at financing capital

outlays and fail to take into account the need for recurrent spending associated with expensive curative care programs (i.e., operations and maintenance) Moreover, despite the effort to increase spending on health care in the course of decentralization, financing formulas have seldom included equalization components in many countries.20 Because of the need for

alternative sources of finance for the provision of health care, many countries have introduced fees and user charges, which are more widespread in the case of health care than education In many countries, these revenues are collected in separate funds, not often consolidated in the government accounts (i.e., transition economies).21 Increased cost recovery through user

19 In 2000, legislation was passed in Brazil requiring the states and municipalities to earmark

12 percent and 15 percent, respectively, of their revenues (net of intergovernmental transfers) to finance outlays on health care (Afonso and de Mello, 2002) Per capita transfers to subnational governments were also introduced to finance the provision of preventive health care In

Colombia, a share of central government transfers to the municipalities is earmarked to finance health care spending In Bolivia, 3.2 percent of central government transfers to municipalities are earmarked to finance the provision of a priority benefits package for mothers and children (Bossert, 2000)

20 In an attempt to equalize spending capacity in the course of decentralization, a horizontal equity fund (Municipal Common Fund) was set up in Chile reassigning up to 60 percent of the revenues of wealthier municipalities to their poorer counterparts using a formula based on population and municipal income (Bossert, 2000)

21 Bossert (2000) shows that per capita spending on health care increased during

decentralization in Chile and Colombia Although wealthier municipalities spent more on a per capita basis than poorer municipalities, the gap between them narrowed over time, resulting in