ABSTRACT

Studies on the relation between decentralization reforms and the performance of governments and their policies have presented a myriad of theoretical developments but inconclusive results on the empirically verifiable effects of these propositions. Many of the reviews carried out thus far in this field merely criticize these propositions without proposing alternatives for analysis. This article presents a review of this debate in a very specific way: it selects the propositions that points towards the causal factors explaining government's responsiveness to its citizens. Four types of interpretations are discussed: 1) allocative efficiency theories of fiscal decentralization, 2) the multiple dimensions of decentralization, 3) the institutional design of reforms, and 4) the qualitative aspects of governance. Despite the observed divergences in the use of concepts, analytical preoccupations and in relation to the empirical operationalization of variables, this article points to a series of hypotheses that could be extracted from this debate and that could be the object of an empirical test.

Keywords: Decentralization; Policy results; Theory review; Subnational governments; Explanatory and causal factors.

Introduction

The studies on decentralization reforms can be divided into two groups concerning their analytical objectives. The first group discusses the political-institutional context of the reforms and the relevant conditions which explain the adopted form of decentralization. The second group focuses on the expected effects of decentralization reforms on the performance of subnational governments and collective welfare.

This article deals with the second group of studies. That is, once decentralization reforms are adopted, what does the literature have to say about their effects and which factors (institutional, political, social, economic etc.) appear as relevant conditions to explain the observed results? After all, scholars and governments start from the assumption that decentralization can create positive effects in different national contexts: a political channel in ethnically divided societies; an instrument of democratization and participation; encouraging government responsiveness; improving the provision of public services, among other results discussed by a large group of authors.

On the other hand, there is a consensus among scholars that decentralization is not alone a necessary or even a sufficient condition to guarantee the expected results. There are other factors at play. What, then, would be the relevant conditions?

Studies on decentralization present a myriad of theoretical developments, but contradictory evidence in relation to the verifiable empirical effects of these propositions. Countless studies in the area cite the lack of this consensus (Shah, Thompson and Zou 2004; Treisman 1999; O'Dwyer and Ziblatt 2006; Schneider 2003; Oates 1999, 1128). Moreover, the very conception that subnational governments would be the appropriate locus to provide efficient policies - as opposed to centralized governments - is questioned by arguments that present the logical inconsistencies of such propositions. At an extreme, perhaps the debate on the favorable relationship between decentralization and the performance of governments (and their policies) may, without detriment to the theoretical or practical knowledge, be abandoned.

The same question raised by Martinez-Vazquez and McNab (2003, 1603) in relation to the assumed association between fiscal decentralization and economic growth - one of the topics associated with the debate - could be applied to the provision of public services in general:

Thus it is not clear, from a logical point of view, why local governments spending R$ 1 million on health would be more efficient and effective than, say, local agencies of the central government spending the same amount of resources.

The central issue of this criticism is that maybe the difference between more efficient and effective service provision in centralized or decentralized States actually does not exist, theoretically or empirically. As Treisman (2007, 279) rightly points out, the factors that are put forward as determinants in this debate are not exclusive to local administrations:

Despite these criticisms, the debate that associates decentralization with the increase in the responsiveness of local governments, on the one hand, and the performance of policies that determine collective welfare, on the other, has influenced and continues to influence both national and international government agenda. In academic terms, propositions that local provision of some public services can generate positive effects for communities, such as "bringing government and citizens closer together", "encouraging creative responses from local governments" or "capacity to attend to citizens' demands" remain.

This article has a very specific aim: to examine the theoretical propositions that form the basis of the principle of decentralization - understood as a form of organization of the State's functions -, in order to identify the mechanisms that are said to be relevant in explaining the performance of subnational governments responsible for policy-making. The organization of this debate is, in my opinion, essential in order to achieve an analytical advance in this field of knowledge. Identifying the propositions which suggest causal connections between decentralized policy-making and government performance can be understood as a first step towards a wide research agenda which establishes hypotheses for an empirical test.

The revision of the existent propositions shows that, in a first moment, very formal theoretical propositions emerge, far from the reality of countries and governments. In a subsequent moment, empirically oriented analyses appear, revealing multiple understandings of what decentralization is. Criticisms to the latter refer to the adopted forms of measuring the phenomenon in question. From the point of view of scientific knowledge production for the area, the result of this cacophony of interpretations is a permanent doubt on whether we are verifying the rejection of hypotheses derived from theoretical propositions or it is merely the use of different concepts that explains the discrepancy of results.

Such analytical pulverization is not an exclusive characteristic of the decentralization debate. Almeida (2001), for example, observes a similar scenario in the discussion of federalism and its effects: lack of minimum consensus on terms, concepts and theories being used as well as problems with the empirical foundations of some of these studies.

This article's assessment is that there are two main obstacles for reaching consensus about the relationship between decentralized policy-making and its effects. On one hand, this is explained by the very nature of the phenomenon in question. The factors that could explain the performance of subnational governments and their policies are not only numerous but also a result of complex interactions and combinations. This fact makes operationalizing empirical tests a challenge for studies in this area, especially for those that aim to make large comparisons between countries in which it is necessary to statistically control several dimensions that might vary across cases (Snyder 2001).

On the other hand, the lack of consensus concerning the definition of decentralization is also notable, which creates different classification criteria (of types or degrees of decentralization). What for some authors would be understood as highly decentralized, for others is exactly the opposite.

In Brazil's case, studies on the decentralization reforms point towards two important institutional features: the 1988 Constitution and the approval of a series of regulations by the federal government, which tried to induce subnational governments to provide new policies, especially from the mid-90s onwards (Arretche 2002; 2009; Souza 2006; Almeida 2005; Abrucio 2006; Gomes 2009). These studies explored in particular the political determinants of decentralization, on the one hand and, on the other, the effects of policy decentralization on access to public services. The analytical advance in the area is notable. The understanding of how the federal government introduces policies coordinated nationally introduced the role of institutions as an explanatory element of the limits and scope of action available for actors involved at the subnational level of government.

A relationship between decentralization reforms and an increase in the access to public services, especially universal ones, seems also to apply to other developing countries (for a summary of existing studies, see Shah, Thompson and Zou 2004). Faguet and Sánchez's (2008) study, for example, find considerable improvement concerning some social and urban infrastructure policies in Colombia and Bolivia after decentralization reforms. In Bolivia's case, political and fiscal decentralization would have allowed localities previously completely deprived of resources to use revenue to provide education services (Faguet and Sánchez 2008, 1300-1). In Brazil's case, Abrucio (2006, 107) sees the Fund for Primary Education and Teachers' Wages (FUNDEF, in Portuguese Fundo de Manutenção e Desenvolvimento do Ensino Fundamental e de Valorização do Magistério), - a policy with incentives to decentralize primary education - as one of the main reasons for the effective universalization of access to this level of schooling.

However, more egalitarian access to state policies and public services is only a first step towards bringing about citizens' rights in the sense of promoting status equality among individuals or the right to equal opportunities (Marshall 1967). The current challenge in the Brazilian context (and possibly in Latin America) is to increase the efficiency and effectiveness of policies, or rather, the quality of public services. In this field, however, the academic consensus is not absolute. Which factors determine or influence a good educational performance? The teachers' training and salary? Schools' infrastructures? The available financial resources? The creative capacity of local governments? The monitoring and control of civil society? The institutional design of policies, containing incentives and constraints to behavior? Or, even, all of these elements together? There are many factors that could explain the result of policies and their complex and often little-known inter-relations (Torres et al. 2010).

Thus, in the context of the Brazilian debate, the question remains as to what extent decentralization would have effectively improved the quality of public services now being provided by subnational governments. If one departs from the central thesis of the studies on decentralization, it would be expected subnational administrations to have readapted or innovated their policies, producing positive results. Even though access to social policies has been considerably expanded in Brazil, for some authors the decentralization of these services is rather the result of a highly regulated policy led by the federal government (Arretche 2009). Take, for example, the case of education. Even though there are studies that confirm the increase in expenditure on education in Brazil in the last decade at all levels of government (Castro and Duarte 2008), it is still not clear to what extent this increase in resources has or has not impacted on the quality of this public service. This article proposes that a productive analytical alternative for this field is to start from the empirical test of existing propositions. There are, clearly, significant methodological decisions to be made in terms of research design. Even though I make some observations in relation to this issue, it will be treated only secondarily here. The central aim of the article, as pointed out, is to identify mid-range theoretical propositions that could be the object of systematic testing.

In the following sections, I present four types of interpretations that assert some kind of causal relationship between decentralization and the performance of subnational governments. These analyses are not free from criticism and, when relevant, I point out their impact on the debate. For the first type of interpretation, the mechanism that would guarantee more efficient and effective government refers to a large degree of freedom to define levels of taxation and spending by subnational governments. The second set of analyses presents the concept that decentralization comes in various degrees and types, as a criticism of the notion that fiscal decentralization would be the only dimension to be considered. In this case, despite the many conceptualizations of what decentralization is, administrative decentralization emerges as a relevant factor to explain the performance of governments in some studies. A third group of interpretations specifies that the success of a decentralization process is related to the institutional design of the reform. A fourth group of analyses identifies qualitative aspects associated with the performance of local governments. Finally, I revise the main explanatory factors of government performance derived from this review and propose a test of these propositions.

The Allocative Efficiency of Fiscal Decentralization

The first type of interpretation, economic in nature, presents theoretical models on the effects of decentralization inspired by the analytical arsenal of microeconomics and the notion of market equilibrium (Oates 1999; Tiebout 1956; Brennan and Buchanan 1980; Bretton 1996; Musgrave 1983; Huther and Shah 1998). It is the most criticized group of analysis and, therefore, has exerted much influence on this research agenda. The whole theoretical reflection of these authors centers on fiscal decentralization, understood as a sufficient and necessary dimension for producing government efficiency and effectiveness.

The central concept for these authors is allocative efficiency in the distribution of State functions. In other words, it is about finding the activities and services that each level of government, in taking on the responsibility of providing them, would produce maximum efficiency in terms of allocation of State resources. Similarly to the model of market equilibrium, which is reached in a decentralized way through the exchange of private goods between individuals, "free competition" among subnational governments would encourage efficient results in terms of the provision of public goods for the community. Allocative efficiency, therefore, would be the result of freedom to adapt local policies and services to citizens' preferences, including the decision on the rate of local taxation.

Take, for example, the words of one of the main authors, Oates (1999, 1135), on the "decentralization theorem". The analytical model is based on the premises of market efficiency:

In this model, the role of central governments should be restricted to the policies that need national coordination: macroeconomic equilibrium and compensation mechanisms for the disparities between jurisdictions such as the adoption of intergovernmental transfers.1 The remaining should be carried out locally. The reason that justifies this lies in the premise that only local governments can respond to the preferences of the citizens of their territories (or jurisdictions). The central government, in contrast, would be unable to define the "optimum" levels of taxation and allocation of resources in all local jurisdictions. The standard behavior of central governments is to apply uniform policies to the whole of the national territory.

The origin of this type of interpretation comes from Tiebout's proposition that it is only in the local arena that citizens can reveal their preferences (Tiebout 1956). In this abstract model, in which there is freedom to define taxation and public services, local governments supply their service (and tax) packages and wait for citizens to arrive. This mechanism would reveal (as well as attend to) citizens' preferences. It is assumed that the citizen can and wants to move to the locality which presents the "package" of benefits closest to their preferences. Thus, if a family moves to the American suburbs and has, for example, school-age children, it would consider the municipality that offers good local schools. Obviously, this family would not take into account only their children's education and Tiebout is aware of this. Non-economic factors, such as the status of living in a certain community, would also be part of this family's calculations. The important thing to retain here is that, for Tiebout, by choosing where to live, citizens automatically reveal their preferences for local policies and services and this is only possible at the local level since the central government would not be able to make local adjustments on such a detailed scale. Thus, follows the logic that taxation and the supply of public services for central and local governments are different.

Just as with other authors' propositions of this type of interpretation, citizens' mobility is central to Tiebout's argument - and also the object of most of the criticism. The whole model is based on the unrealistic assumption, as the author himself recognizes, that there are no restrictions on citizens' mobility. But, this does not stop him from suggesting the adoption of policies that lower the costs of mobility in the national territory, on the one hand, and increase access to information by the "consumer-voters", on the other hand (Tiebout 1956, 423).

So, what is the optimal equilibrium distribution of State functions? Decentralized policies would make sense in the case of public services "whose consumption is limited to their own jurisdictions" (Oates 1999, 1121), or rather, in those cases in which there are no spillover effects. For Oates (1999, 1120), the theory of "leaving local matters in local hands" would still hold up.

These propositions, in sum, sustain that freedom to tax and allocate resources by local governments is the key to meet citizens' demands. In other words, it produces allocative efficiency. Therefore, variation of service provision within the national territory would be nothing more than subnational governments' response to citizens' preferences.

Many criticisms to these propositions have emerged. First, there is an underlying assumption that any political decision is far from an optimal equilibrium or, put more directly, is far from achieving allocative efficiency.2 In this type of interpretation, political elements are absent, such as the motivation of politicians, the role of local bureaucracies, local elite's hold over the decision-making process and, above all, the political consensus that is possible to be achieved given the circumstances. Even though this article does not present a review of this type of criticism, it is worth mentioning the existence of a group of authors that have developed analytical models which incorporate the political motivations (or rationalities) that would influence decentralization reforms (Treisman 1999; Panizza 1999) as well as the influence of institutional designs on these choices, especially the electoral system and federalism (Garman, Haggard and Willis 2001; O'Neill 2003).3

The second criticism refers to the fact that, despite the far-reaching theoretical intentions, such propositions are seen as specific to the American case. As Treisman (2007, 284-85) points out, a historical rhetoric of distrust of central authority and, therefore, a strong anti-centralization ideology, is characteristic of the school of thought of these authors, which does not occur so intensely in other countries. As we will see further on, the central government is invariably a key actor both in relation to the degree of supervision or regulation of the subnational governments - even in the north-American case (Sellers and Lindström 2007).

Furthermore, one of the most important criticisms refers to the consequences of this model: a variation on service provision becomes a mechanism for inequalities concerning the collective welfare among citizens of the same national territory. As well as reasons of substantive social justice, this State model would inevitably create inefficient and ineffective policies, given that localities would have different capacities (fiscal, administrative etc) to respond to the "preferences" of their citizens. As we will see, the institutional design of a reform is understood by some authors as the corrective mechanism for these negative effects.

The last criticism refers to the notion that fiscal decentralization is a sufficient (as well as necessary) condition for local governments to efficiently allocate resources. However, fiscal decentralization, as we shall soon see, would only be one of the existing dimensions.

Despite such criticisms, the arguments originally presented by these authors continue with remarkable force, especially when it comes to the positive effects of bringing governments and citizens together. As I highlighted in the Introduction, there are studies which suggest that fiscal decentralization reforms would have produced a more egalitarian access to public services. However, part of the criticisms of this type of interpretation is that fairer access to fiscal resources would actually be a consequence of redistributive policies guided by central government and not isolated initiatives from local governments themselves.

The Multiple Dimensions of Decentralization

The common characteristic of this type of interpretation is a rupture with the notion that fiscal decentralization is a sufficient condition for subnational governments' autonomy (Schneider 2003; O'Dwyer and Ziblatt 2006; Falleti 2006; Treisman 2007; Martinez-Vazquez and McNab 2003). The main concepts, thus, are fiscal, political and administrative decentralization.4 For some authors, moreover, decentralization cannot be seen as a dual concept ("centralized" or "decentralized") since these forms would not exist in a pure state but, on the contrary, we would observe varying degrees of decentralization/centralization (Sellers and Lindström 2007; Stegarescu 2005). Following this line of thought, subnational governments come to be understood not as autonomous entities, but inserted into a context in which central government regulates the scope and form of decentralization (Arretche 2010). As a consequence, the nature of processes of decentralization makes identifying the mechanisms that would produce responsiveness, efficiency and effectiveness more complex. Maybe for this reason it is not surprising the low level of consensus within this group with regard to possible cause-effect relations between the dimensions of decentralization and policy outcome.

These analyses also debate on the difficulties in adopting indicators that reflect the level of decentralization of a country. It is in this field, therefore, that one finds plenty of methodologies of classification. The central issue refers to the definition of measures that can identify types or degrees of decentralization in a way that reflects the concepts derived from the analytical propositions. Therefore, the answer to the question "which of these two countries is more decentralized?" becomes more complex since there is no consensus on the actual definition of the concept of decentralization.

Different concepts of decentralization

Despite the large number of studies that uses the different dimensions of decentralization for empirical tests, such concepts are not used in the same way.

For some authors (Schneider 2003; Falleti 2006; O'Dwyer and Ziblatt 2006) the dimension of administrative decentralization is what defines the autonomous capacity of local governments and, thus, is the most relevant factor in explaining government performance. For Schneider (2003), administrative autonomy refers to the government's capacity to decide which policy package to offer as well as autonomy to decide how their goals will be achieved. What is specific about this conceptualization is that spending capacity is not understood as part of the fiscal dimension, as is normally the case in the studies in this area, but instead as part of the administrative one. For this author, the fiscal dimension refers only to the allocation of "(...) a greater proportion of fiscal resources at a level other than the centre" (Schneider 2003, 10) and are not, therefore, conceptually associated with decision-making autonomy. Thus: "(...) a system can be fiscally decentralized without administrative decentralization if the use of local funds is set at the central level" (Schneider 2003, 10). Therefore, the capacity to decide where and in what ways public revenues will be allocated is the essence of decentralization in administrative terms.

The political dimension of decentralization for Schneider refers to the different forms of representation of local interests in local instances, not being exclusively about the existence of elected bodies. Political functions can be understood as decentralized when there are channels of local participation and contestation that are translated into public finance decisions (Schneider 2003, 18).

For Treisman (2007), in contrast, administrative decentralization has no relation to decision-making autonomy. The administrative dimension, for this author, refers to actions carried out by local officials that are politically subordinate to the central government:

Decision-making autonomy for subnational governments would be included in the political dimension of decentralization in this author's analysis. Thus, Treisman identifies political decentralization as the dimension that involves subnational governments' autonomy to decide how to allocate public resources. For this author, political decentralization:

And this decentralized decision-making authority must exist if "(...) at least one subnational tier of government has exclusive authority to make decisions on at least one policy issue" (Treisman 2007, 24). This criterion can present some difficulties relating to empirical data. On one hand, it seems that few policies are attributed exclusively to subnational levels of government (Rodden 2005; Almeida 2001). On the other hand, if it is true that in most cases (or countries) responsibilities for some policies are shared, so it would be possible, at least in theory, for a subnational government to retain some decision-making autonomy in relation to a policy that is shared with other levels of government even if not in an exclusive way. In relation to fiscal decentralization, Treisman does not give any definition in particular but highlights the lack of consensus with regard to this dimension.

For other authors, the appropriate concept to define decentralization regarding decision-making autonomy would be devolution. O'Neill (2003, 1070) defines decentralization as "devolution of autonomous political and fiscal power to subnational officials", but nothing is mentioned regarding the administrative dimension, so central and important in Schneider's analysis, for example. Could it be inferred that these two dimensions account for administrative autonomy, as Treisman puts it? For Litvack, Ahmad and Bird (1998), devolution occurs when central government transfers decision-making authority, funding and administration to subnational governments and these then "(...) raise their own revenues, and have independent authority to make investment decisions" (Litvack, Ahmad and Bird 1998, 6). This notion of autonomy points towards a model in which subnational governments would be practically free of any restriction from the central government. For Faguet and Sánchez (2008, 1296), decentralization is the process of devolution by central government of specific functions to democratically elected governments. This last definition is contrary to that presented by Schneider, since for this author, as we have seen, decision-making autonomy (administrative) can coexist with no political autonomy.

As we have observed, there is a lack of consensus in defining the dimensions of decentralization and which of them would be associated with decision-making autonomy, despite the common understanding that such distinctions are necessary. Let us now see the relationship that has been found in some studies between these dimensions of decentralization and the result of policies carried out by subnational governments.

Effects of the different dimensions of decentralization on policy results

Based on the premise that there are different dimensions or degrees of decentralization, some studies turn to test empirically their relation to policy results.

The strongest case is presented by Schneider (2003), who suggests that the administrative dimension is more significant in explaining social spending. Nevertheless, in another empirical study based on Schneider's premises (O'Dwyer and Ziblatt 2006), such results are refuted.

Analyzing a group of 68 countries in relation to their degree of fiscal, administrative and political decentralization, Schneider (2003) comes to the conclusion that administrative autonomy, more than the other dimensions, produces positive effects in relation to the adoption of policies that corresponds to citizens' demands. For Schneider, public spending on "pro-poor" policies - i.e. expenditure on health, education, welfare etc -, is a legitimate measure of these demands (Schneider 2003, 25). As control measures, variables of the economic level of a country Gross Domestic Product (GDP) and the proportion of the elderly population (assuming that these countries could have higher social spending due to their age structure) are added. In order to measure fiscal decentralization, Schneider adopts the proportion of subnational governments' expenditure on total expenditures and for the administrative decentralization, the proportion of subnational own-tax revenue as an indirect measure of autonomous decision-making capacity.

In his results, the following associations are found through regressions: the more politically decentralized, the lower the fiscal capacity and the lower the pro-poor expenditure, on one hand. On the other, the higher the administrative decentralization, the higher the pro-poor expenditure. In his analysis, the degree of fiscal decentralization does not appear to be associated to any pattern of expenditure (Schneider 2003, 27).

The author's final argument is that political decentralization - as in federal States - disperses forces interested in a pro-poor expenditure agenda. In systems with dispersed political power, constructing alliances that would support social spending would be more difficult to carry out as it would have to be done in all local jurisdictions in a coordinated way. In the end, he suggests that the relevant factors that explain higher social expenditure would be the combination between political (national) centralization and administrative decentralization. In other words, units politically subordinated to the central government.

Similar results are presented by Sellers and Lindström (2007), one of the comparative studies between countries that also attempts to capture the different degrees and types of decentralization, but with more emphasis on the degree of supervision (or regulation) of central government. The study is limited to developed countries. By adopting a different form of classifying degrees of decentralization, the authors come to conclusions that refute long-standing theses of studies analyzing the relationship between degrees of centralization/decentralization of the State and the scope of their social protection. For example, despite the fact that the literature analyzing the emergence of the Welfare State in Nordic countries traditionally features the high degree of centralization of these States, the analysis of the qualitative indicators organized by the authors shows that local governments in these countries have a high administrative capacity, that is, own-tax revenue capacity and decision-making autonomy concerning local spending. This result essentially refutes the widely accepted proposition that subnational governments in "social-democratic" Welfare States have low administrative autonomy, or rather, that they would only act as units subordinated hierarchically to the central government. Moreover, for these authors, the existence of a local capacity previous to the implementation of the Welfare State in these countries, as well as a relatively homogenous national party system, would have produced not only support for the national agenda of expanding social services but also a greater commitment to local implementation of this agenda.

On the other hand, O'Dwyer and Ziblatt (2006) come to different conclusions when analyzing the effects of decentralization in 69 selected countries. The authors use the same concept distinctions proposed by Schneider (2003) and described above. Interestingly, the study does not find any significant statistical association among the three dimensions of decentralization (fiscal, political and administrative). In other words, the degree of decentralization in one of the dimensions does not explain the position of a country in the other two, reiterating the theory of multiple forms of decentralization.

For their empirical test, the authors create two indicators which serve as dependent variables: an indicator of governmental efficiency, which is the ratio between Social Expenditure and Number of Public Employees, and Government Effectiveness, a compound indicator created by the World Bank, which assesses the population's perceptions/evaluations of the services (O'Dwyer and Ziblatt 2006, 3). These two indicators together would gauge the quality of a country's governance. Indicators of the degree of fiscal, administrative, political decentralization, economic development, type of political regime and the openness of the economy (indirect measure of globalization) are introduced as independent variables.

Using bivariate analysis techniques first, the authors conclude that fiscal decentralization is always associated with higher quality of governance; in the political dimension case, they find an association with effectiveness (but not with efficiency) and no significant association in the case of the administrative dimension. As observed, the authors find different results from those presented by Schneider (2003). However, when the authors then analyze all the variables together, using multivariate analysis techniques, an unexpected result emerges: the level of fiscal, political or administrative decentralization of a country would explain neither governmental efficiency nor effectiveness. Instead, they would be explained simply by the socioeconomic development level of a country. Based on these results, the core notion of the positive effects that could be generated through decentralization reforms is challenged. The positive effects that would normally be attributed to decentralization can simply be a product of the level of a country's development (O'Dwyer and Ziblatt 2006, 14). If this applies, the authors continue, contrary to what is normally claimed, maybe centralization produces better governance in "late-developers" countries (O'Dwyer and Ziblatt 2006, 14). Similar arguments are presented by the group discussing the role of institutional design on decentralization: developing countries might be a case apart, although they do not share the notion that centralization is an alternative.

The main criticism of this group of analyses is in relation to the operationalization of the measures that should reflect analytical concepts on the one hand, and the difficulties inherent in comparative studies on the other, especially when the distinction between degrees and forms of decentralization renders the analyses more complexity. The variation in the ways of classifying degrees and types of decentralization also ends up producing discrepant results (Sellers and Lindström 2007; Stegarescu 2005).

In relation specifically to Schneider's analysis, despite the relevant analytical distinction between the dimensions of decentralization, the operationalization of the measures used is questionable. To take "the proportion of own-tax revenue" as a measure of the administrative dimension - a variable traditionally used to measure the degree of fiscal decentralization - can be misleading. In other words, the indicator that should represent the administrative dimension - identified by the author as more important to explain the results - does not effectively reflect such phenomenon. If this applies, then, the positive results found for the administrative dimension would in fact be effects of fiscal decentralization. Moreover, Schneider in the end adopts the same assumptions as theorists of allocative efficiency, the object of a large number of criticisms: he assumes that own-tax revenue is an indirect measure of autonomous decision-making capacity to allocate resources.

A last point in relation to Schneider's argument is worth mentioning. The idea that national decision-making spaces produce an environment that is more favorable to the organization and union of political forces interested in a social expenditure agenda could well be applied to the organization of pro-market interests, for example. The point that I would like to highlight is the direction of such causal effects: the proposition that "the nationalization of political forces would produce higher social expenditure" could well be read as "in countries where forces interested in increasing social expenditure won national representation, the conditions for implementing a pro-poor agenda are more favorable". Thus, perhaps the "path dependence" theory could be more useful analytically: if organized interests achieved national representation and, after this, the approval of pro-poor policies, maybe it is more difficult to return to the status quo ante, or rather, to dismantle a network of social protection. Despite the statistical associations and correlations, there is no causal relationship between the manner of distribution of State functions per se and the adoption of a pro-poor political agenda. The central element could well be simply which groups manage to hold power.

In short, there is no consensus as to the effects that different types and degrees of decentralization produce on the performance of subnational governments. Thus, if the studies in the area consider it necessary to distinguish between the dimensions of decentralization as an explanatory factor, then a first step in this direction would be to create a common conceptual reference. For Arretche (2010), for example, the relevant distinction is between decision-making autonomy and decentralized execution, defined according to the degree of central government's regulation of decentralized policies. For the author, policies regulated by central government limit the subnational government's choice of the policy "package" and, in contrast, those of low regulation allow subnational governments a greater deal of decision-making autonomy.

Borrowing Arretche's (2010) conceptual reference, we can say that maybe the main interest of the studies on decentralization is not the decentralized delivery of services but rather the degree of decision-making autonomy on local services. In this sense, the degree of autonomy that subnational governments retain in each dimension of decentralization is what would become analytically relevant to understand the impacts of decentralization on local governments' performance. In this way, we would have: the fiscal dimension (degree of autonomy that subnational governments have to define local taxation), the administrative (degree of autonomy to choose their public services "package") and the political dimension (degree of autonomy in the local political representation). An empirical test among countries in these molds could explore how subnational governments administrate their own policies and whether or not greater autonomy effectively generates the effects expected by the existing propositions.

The Institutional Design of Decentralization Reforms

A third type of interpretation asserts that the positive effects of a process of decentralization are related to the institutional design of the reform. This form of interpretation has its main advocates in the World Bank's studies. It must be emphasized that the role of "institutions" as an explanatory factor of social and political phenomena is not exclusive to this group of authors, being observed in different studies of the so-called neo-institutionalism approach (see, for example, Peters (2005) and Immergut (1998)). However, my review here is restricted to the studies that sustain - in a normative way - that adopting specific institutional designs can produce the desired results.

For this third type of interpretation, the essential mechanisms that guarantee governance are: existence of participation channels and accountability. In order to achieve this, the institutional design of the reform should contain positive incentives for government responsiveness as well as constraints for irresponsible behavior. This would be especially true in the case of developing countries, characterized by weak institutions and little capacity for mobilization and monitoring by the local population.

Such diagnoses identify various problems which compromise the efficiency and effectiveness of policies and that are somehow related to an inappropriate institutional design, but in this review I will discuss only three of them: regional disparities, low government responsiveness and the incentive for irresponsible fiscal behavior.

It is interesting to notice that the importance attributed to the institutional design demonstrates both a change in the diagnosis of the effects of decentralization reforms and also concerning policy recommendations of the World Bank. If at first decentralization was heralded as a solution for greater governing efficiency, decreasing inequalities of access and improving quality of public services - by bringing closer together the decision-making process to the needs and demands of the users of the services (The World Bank, 1994) -, gradually, new interpretations of these processes emerge. With inadequate institutional designs, decentralization reforms can produce negative effects for governance:

Another change of interpretation that can be noted is the recognition that central governments have a strategic role in the formulation of reforms. All the recommendations of this group of studies involve creating control and supervision mechanisms by the central government.

The concept that local governments would have more appropriate conditions to respond to local demands, on the other hand, remains the institution's trademark. These studies are based on the assumption that, unless there are contrary reasons (such as economies of scale and spillover effects), the decision on the allocation of resources could, in theory, be more efficient when carried out by local governments, since they would be in a more privileged position to adapt public expenditure to their citizens' preferences. As it can be seen, the assumptions for local efficiency are the same as those presented by the theorists of allocative efficiency, discussed above. The difference in this case is the diagnosis that the institutional conditions verified in developed countries would not exist in the developing world (Shah, Thompson and Zou 2004; Prud'Homme 1995; Litvack, Ahmad and Bird 1998; Rondinelli, Nellis and Cheema 1983). Essentially, the community's "voice" would not be effective in local governmental decisions in developing countries, hindering the institutionalization of an essential mechanism for governance: accountability. As Litvack, Ahmad and Bird (1998, 2) summarize:

Thus, the reason that would explain the varying results observed in relation to government performance after decentralization reforms - or the discrepancy between the theoretical concept and the observed results - is attributed to inappropriate institutional design in the case of developing countries (Shah, Thompson and Zou 2004). And what would be an inadequate institutional design? Litvack, Ahmad and Bird (1998) list a number of considerations. Among them, the importance of clear definitions on what level of government will be responsible for financing and providing the service and regulations that limits subnational endebtedness and creates incentives to improve their own-tax revenue collection in order to curb dependency on intergovernmental transfers.

The authors call the attention to one of the negative effects that decentralization reforms can generate: increasing internal disparities. In this respect, the role of intergovernmental transfers becomes part of this debate: some defend a design that prevents dependency from this type of revenue and others see them as an important mechanism to compensate fiscal disparities. For the latter, a fiscal decentralization reform in which local governments depend almost exclusively on their own-tax revenue to form their budgets will punish localities with low economic dynamics.

Prud'Homme (1995, 7-10) shares the notion that the problems associated with decentralization are related to the institutional design and that institutions tend to be weak in developing countries. The author criticizes the theories of fiscal federalism and allocative efficiency as they describe only the positive effects of decentralization reforms. For him, there are no theoretical reasons to sustain the idea that all functions are carried out in a more efficient way by local governments. There are "dangers of decentralization", as the title of his article indicates, such as horizontal and vertical disparities, imbalances from the macroeconomic point of view (fiscal stabilization), inefficiency and ineffectiveness.

The assumption that local governments will pursue the interests of the local community would not apply in the case of developing countries. In these countries, argues the author, there are scarcely enough resources to respond to demands, the vote is personal, charismatic and would not work as a mechanism where voters reveal their preferences; the electoral platform are fragile and unrealistic, the mayor's career does not necessarily depend on observing local preferences and bureaucratic implementation of the mayor's programs is not automatic - "principal-agent" problem (Prud'Homme 1995, 7-10).

From this point of view, Prud'Homme presents the argument that intergovernmental transfers are essential mechanisms for minimizing one of the "dangers" of decentralization: increasing disparities. Intergovernmental transfers tend to be criticized for encouraging irresponsible fiscal behavior. That is, subnational governments would not have to engage in unpopular tax-rise policies or more efficient revenue collection to finance their actions. According to this logic, the dominant strategy becomes to knock on the central government's door to ask for more transfers. Prud'Homme (1995, 11) criticizes this negative view:

Another institutional factor that would interfere in the efficiency, effectiveness and responsiveness of governments is connected to administrative capacities, including the disparities of capacity among governments. This issue is also dealt with by authors from different analytical perspectives (Sellers and Lindström 2007; Falleti 2006; Faguet and Sánchez 2008; Litvack, Ahmad and Bird 1998; Fizbein 1997). Prud'Homme (1995) also sees local administrative capacity as a potential impediment to efficient and responsive policy-making. For this author, these problems may be related, for example, to the difficulties in attracting qualified personnel to local administrations, since central governments tend to offer best wages as well as best career development prospects, mobility etc.

For Litvack, Ahmad and Bird (1998), one of the considerations that central governments should take into account, before designing decentralization reforms, is if local governments have the fiscal and administrative capacity to provide services. Again, this problem would be more acute in the case of developing countries. For the authors: "(...) [fiscal and administrative] capacity building should precede decentralization" (Litvack, Ahmad and Bird 1998, 28). Otherwise, the chances of services being inefficient and ineffective increase considerably.

The main criticism of this type of interpretation relates to the "power" of institutional design. Even though it is generally accepted that institutions matter, it is not clear which set of institutions will guarantee positive results. More specifically, there is no guarantee that institutional rules will produce similar effects in different contexts. Moreover, even though regulations that punish a high level of indebtedness by subnational governments or that define maximum levels of expenditure can be done, this type of rules does not guarantee alone an improvement in the quality of public services.

Finally, the suggestion that political institutions are "weak" in developing countries is controversial. What electoral system could guarantee an increase in government responsiveness? There is no consensus, in the academic debate, on which electoral systems are "weak" and which are "strong". This is still an open debate. Why the "personal and charismatic vote" - seen as problematic in the above arguments - could not produce a leader who responds to local interests? Obviously, the assumption that political institutions in developing countries are so weak that impede governments to be closer to their citizens is a hypothesis yet to be empirically tested.

In summary, for this group of interpretations, decentralization reforms should take into account an institutional design that contains appropriate incentives and constraints. Among the mechanisms that should be established are: the creation of institutional sets that promote accountability, responsible fiscal behavior and a guarantee to a minimum level of fiscal and administrative capacities. Intergovernmental transfers would be one of the institutional ways to compensate internal disparities.

Qualitative Aspects of Governance

A fourth type of analysis identifies aspects that could be understood as qualitative and are associated with the performance of local governments. In this case, factors such as political actors' capacity for leadership as well as the role of bureaucracy in policy implementation would be crucial to explain the performance of subnational administrations. The results of policies adopted after decentralization reforms, these studies claim, would be also explained by the behavior of key actors and their capacity to institutionalize changes in the administrative machine, including the creation of effective channels of community participation. These studies do not ignore the effects of other structural elements, such as the role of central government, the type and scope of decentralization, fiscal conditions and others previously discussed. For this type of study, though, if decentralization reforms aim to transfer some degree of decision-making autonomy to local governments, then it becomes central to observe which strategies were adopted and their effects.

Grindle (2007) analyzes whether the "promise of good governance" was realized after the implementation of decentralization reforms. Public sector entrepreneurship is identified as an important factor to explain governments' results in the case of Mexican municipalities. Incentives and constraints put in place by the institutional design of the central government's reform have also played a significant role according to her analyses. The question of how local governments reacted to the opportunities given by a decentralization reform is a central concept in this interpretation. The more specific question refers to which factors could explain the observed variation in the quality of governance among some municipalities selected for a case study. As the author analyzes the municipalities within one country, there is a minimum control of exogenous variables (degree of economic development, insertion in the international economy) and endogenous variables (institutional design, relevant political forces etc.) which could directly or indirectly influence the observed results.

The conclusions point to a set of important factors that explains why some Mexican localities were successful in their local reforms and others were not. A combination of the following factors was identified: an increase in local elections competition - with the end of the predominance of a single party Partido Revolucionário Institucional (Institutional Revolutionary Party (PRI)) -, the pressure from central government so local administrations would implement the expected local administrative reforms, the participation of civil society and the existence of a local (elected) political leader committed to the reform. More specifically, this leader showed a capacity to create consensus and support for the administrative reform from the local bureaucracy, to guarantee financial support from state and federal bodies and a commitment to the creation of effective communication channels with the local community. For Grindle (2007, 22), public sector entrepreneurship:

Fizbein (1997) finds similar results concerning the role of local leaders. In a case study that analyses the performance of Colombian local governments, Fizbein highlights the importance of administrative capacity in the production of allocative efficiency and effectiveness. Such capacity, however, does not emerge as a direct result of the institutional design but instead from the existence of a local leader who is committed to build channels of communication with the local population. According to the author, a "customer-oriented" government is able to create institutional channels to listen to the demands of communities and to translate them into government actions (Fizbein 1997, 1031).

But how are these capacities created? For Fizbein, it is an endogenous construction, a local administration initiative. A central factor for this construction, based on the analysis of Colombian municipalities, is the existence of leadership committed to the participation of local population:

Assuming the same logic that leads to market equilibrium - where by attending consumer preferences companies produce more efficient and effective products -, the participation of the population in local administration would also produce a similar result in local administrations. The local leader, as an entrepreneur of the local administrative reform, exerts significant influence over the way the demands of the local population will be addressed.

The role of bureaucracy in the implementation of public policy has already been dealt with extensively in the social sciences and, therefore, is not exclusive to the debate on decentralization reforms (see, for example, the review presented by Hill 2009). However, Grindle and Fizbein's position is that local leaders who have the ability to create consensus and are committed to a participative administration play a key role in the success of local service provision in the aftermath of decentralization reforms. Thus, such interpretation differs from the so-called "top-down" propositions, in which the challenge of public officials is to guarantee the control of the bureaucracies responsible for the implementation of policies (Barret 2004). This is the case, for example, of the analyses based on the principal-agent theory, in which the central problem is guaranteeing that bureaucracies carry out in an efficient manner what was approved by politicians (Shepsle and Boncheck 1997, 345-80). But maybe endorsement, and not only providing institutional mechanisms of control, is also an important element to guarantee the implementation of reforms that involve the interests of bureaucracies and changes in the status quo.

The lack of professional community support in the process of decentralization of some health services in Zimbabwe was identified as the main problem explaining the failure of this policy, according to Wang et al. (2002). The authors attribute this result to the lack of effective communication between the policy formulators and the bureaucrats responsible for implementing the services. A series of top-down rules controlling the way the service had to be delivered was approved by the formulators without presenting or justifying the reasons or even the objectives of such measures. The result was a lack of support from the bureaucracy responsible for the implementation, since nurses and senior and middle-ranking managers did not agree with the goals of the reform (Wang et al. 2002, 449).

From the review of this group of interpretations, an additional factor emerges as important to explain the performance of policies carried out by local governments: the role of local political actors leading the necessary administrative reforms. This is related, on the one hand, to the capacity of producing consensus and support for the administrative reform and, on the other hand, to the commitment of institutionalizing effective channels of communication with the local community.

Final Considerations

The review I presented here reveals the coexistence of different explanatory propositions on the relationship between decentralization reforms and government performance. The debate begins from a rather simple assumption, which is still present in the current context: local governments would have better conditions to adapt the policies under their responsibility to the preferences of their citizens in comparison to centralized States. From this reasoning, it follows that the mechanism guaranteeing efficient results is to award local governments sufficient decision-making autonomy in relation to local taxation and freedom to decide on the "package" of services to be offered. The development of this discussion, however, produced many criticisms to this type of proposition, some of them discussed in detail in this article.

Despite the divergences observed in concepts, forms of analysis and conclusions, from the point of view of a research agenda it is possible to list explanatory factors that could be the subject of an integrated empirical analysis. These can be summarized in the following items and could be read as concepts' input to raise hypotheses:

Fiscal capacity

A factor considered elementary for the operationalization of policies by local governments. If there are no financial resources, governments will not be able to act. The fiscal capacity of subnational governments is determined by their own tax revenue and resources for intergovernmental transfers. Central considerations in this discussion are: to what extent is the local government autonomous to tax and define its own tax rates and the design of intergovernmental transfers (whether transfers are regular or not, earmarked or not, tax-shared or redistributive). From the point of view of the theoretical proposition of fiscal federalism, fiscal autonomy would be a central mechanism to guarantee the allocative efficiency of resources which responds to citizens' demands. On the other hand, for the studies that understand that the central government plays an important role concerning the redistribution of resources, the institutional design of the reform must incorporate equalizing measures, such as intergovernmental transfers. Even so, such design would have to contain a limit or necessary balance between creating positive incentives for "fiscal effort" and discouraging total dependence on the transfer of resources.

Administrative capacity

This factor refers to minimum administrative conditions for effective provision of public services by local governments. It includes the existence of an organizational structure, with qualified personnel and efficient administrative routines, which can respond quickly and appropriately in the process of implementing policies. Many of the studies discussed here highlight this factor as a determinant for government performance.

The institutional design of policies

Including the incentives and constraints of national legislation or regulations. The issue, in this case, relates to the lack of consensus on which institutions allow the creation of a positive structure of incentives and, in contrast, those which widely and universally produce negative incentives. In other words, the interaction with other institutional factors specific to each country could overturn the positive effects expected in theory. The theories presented on this issue are fairly wide-ranging, from the proposition that it is up to the central government to approve a regulatory boundary that prevents the irresponsible fiscal behavior of governments - for example, indebtedness - to the institutionalization of channels of participation, communication and accountability for the local population. For part of these studies, this would be especially valid in the case of developing countries, characterized by weak institutions, low budgets and unresponsive politicians. Obviously, this latter assumption about developing countries' institutions would also be a hypothesis for empirical test.

The role of local political leadership

In some studies, local leaders played an important role by convincing others on the need for policy and administrative reforms as well as guaranteeing the support of key actors in the implementation process, such as the local bureaucrats, guaranteeing legitimacy and support to the actions.

Based on the propositions listed above, it seems clear to me that a research design that is based exclusively on quantitative analysis (large N) is not sufficient for a comprehensive empirical test. Many these propositions' assumptions demand a qualitative inquiry, or rather, studies that empirically observe how local governments manage policies under their responsibility. Verifying how local governments interact with the local community or how (and if) they respond to the local demands at the stages of agenda setting and formulation of policies, to check the administrative capacity and if there was innovative policies introduced locally are all questions that would benefit from qualitative studies, maybe adopting a comparative strategy or analysis that combine quantitative and qualitative aspects. The interaction between quantitative and qualitative analyses, in a complementary way, would allow several hypotheses derived from the debate reviewed here to be tested, for example, by better informing the construction of variables for statistical tests based on qualitative information. In the same way, regularities found in the quantitative analyses could complement and help to in-depth case studies.

The assumption that local governments are in a better position to adopt an agenda of policies that respond to collective welfare - in comparison to the more centralized forms of organization - remains a hypothesis to be tested. There are no conclusive results. But the fact is that decentralization processes, even though in varying forms and degrees - have been and continue to be adopted in several countries, so any research that advances the understanding of how local governments operate can help to identify the mechanisms that explain the increase in collective welfare.

Notes

1 Oates recognizes in his 1999 article that intergovernmental transfers should contain criteria related to the "social values" of a society, not only for reasons of allocative efficiency but also political ones. For the author, however, it remains a fact that an increase in the amount of transfers does not automatically guarantee increase in collective welfare.

2 This perception is not exclusive to Oates and is present in many other studies, all of them of an economic nature. Political decisions are always suboptimal (for example, the famous and influential Arrow's Theorem of 1951 on the unpredictability of political decisions, or McKelvey on the Chaos Theorem of political decisions).

3 Note that even in relation to this debate, the consensus is not absolute. For analysts of the Brazilian case, for example, voting systems can, in theory, contain a series of incentives for the behavior of political actors, but the rules of the decision-making process end up cancelling out these effects (Figueiredo and Limongi 1999). The same apply to federalism. The proposition that federalism, because of its institutional characteristics, make the approval of reforms more difficult is also contested in various studies (Obinger, Leibfried and Castles 2005). See also, for the Brazilian case, the studies of Arretche (2009).

4 Another distinction used in some studies is between three forms of decentralization: deconcentration, delegation and devolution. It was originally proposed by Rondinelli apud Litvack, Ahmad, and Bird 1998). It is a similar understanding to the degrees of decentralization/centralization, with different meanings depending on the author in question. For some authors (Schneider 2003), these would be typical dimensions of administrative decentralization while for others (Martinez-Vazquez 2003) they can also be applied to the degrees of fiscal decentralization.

Bibliographical References

Abrucio, F. L. 2006. "Para além da descentralização: os desafios da coordenação federativa no Brasil". In Democracia, descentralização e desenvolvimento: Brasil & Espanha, org. S. Fleury. Rio de Janeiro: Editora FGV.

Almeida, M. H. T. 2005. Recentralizando a federação? Revista de Sociologia e Política 24: 29-40.

___. 2001. Federalismo, democracia e governo no Brasil: ideias, hipóteses e evidências. BIB (Revista Brasileira de Informação Bibliográfica em Ciências Sociais), no. 51:13-34.

Arretche, M. 2010. Federalismo e igualdade territorial: uma contradição em termos? Dados 53 (3): 587-620.

___. 2009. Continuidades e descontinuidades da federação brasileira: de como 1988 facilitou 1995. Dados 52: 377-423.

___. 2002. Federalismo e relações intergovernamentais no Brasil: a reforma de programas sociais. Dados 45(3): 431-58.

Barret, S. 2004. Implementation studies: Time for a revival? Personal reflection on 20 years of implementation studies. Public Administration 82 (2): 249-62.

Brennan, G., and J. M. Buchanan. 1980. The power of tax. Cambridge: Cambridge University Press.

Bretton, A. 1996. Competitive governments. Cambridge: Cambridge University Press.

Castro, J. A., and Duarte, B. C. 2008. Descentralização da educação Ppública no Brasil: trajetória dos gastos e das matrículas. Texto para Discussão IPEA, n. 1352. Brasília.

Faguet, J.-P., and Sánchez, F. 2008. Decentralization's effect on educational outcomes in Bolivia and Colombia. World Development 36(7): 1294-316.

Falleti, T. 2006. Efeitos da descentralização nas relações intergovernamentais: o Brasil em perspectiva comparada. Sociologias no. 16:46-85.

Figueiredo, A., and F. Limongi. 1999. Executivo e Legislativo na nova ordem constitucional. Rio de Janeiro: Editora FGV.

Fizbein, A. 1997. The emergence of local capacity: Lessons from Colombia. World Development 25 (7): 1029-1043.

Garman, C., Haggard, S., and Willis, E. 2001. Fiscal decentralization: A political theory with Latin American Cases. World Politics 53: 205-36.

Gomes, S. 2009. Políticas nacionais e implementação subnacional: uma revisão da Descentralização Pós-Fundef. Dados 52 (3): 659-90.

Grindle, M. S. 2007. Going local: Decentralization, democratization, and the promise of good governance. Princenton: Princenton University Press.

Hill, M. 2009. The public policy process. Harlow, UK: Pearson Longman.

Huther, J., and Shah, A. 1998. Applying a simple measure of good governance to the debate on fiscal decentralization. Policy Research Working Paper 1894, The World Bank.

Immergut, E. 1998. The theoretical core of the new institutionalism. Politics & Society 26 (1): 5-34.

Litvack, J., J. Ahmad, and R. Bird. 1998. Rethinking decentralization in developing countries. The World Bank Sector Studies Series. Washington, DC.

Marshall, T. H. 1967. Cidadania, classe social e status. Rio de Janeiro: Zahar Editores.

Martinez-Vazquez, J. 2003. Fiscal decentralization and economic growth. World Development 31 (9).

Martinez-Vazquez, J., and R. M. McNab. 2003. Fiscal decentralization and economic growth. World Development 31 (9): 1597-616.

Musgrave, R. A. 1983. Who should tax, where, and what? In Tax assignment in federal countries. C. McLure. Canberra: Centre for Research on Federal Financial Relations, Australia National University.

Oates, W. E. 1999. An essay on fiscal federalism. Jounal of Economic Literature 37 (3): 1120-149.

Obinger, H., S. Leibfried, and F. Castles, eds. 2005. Federalism and the Welfare State. New world and european experiences. Cambridge: Cambridge University Press.

O'Dwyer, C., and D. Ziblatt. 2006. Does decentralization make government more efficient and effective? Commonwealth & Comparative Politics 44 (3): 1-18.

O'Neill, K. 2003. Decentralization as an electoral strategy. Comparative Political Studies 36 (9): 1068-91.

Panizza, U. 1999. On the determinants of fiscal decentralization: Theory and evidence. Journal of Public Economics 74: 97-139.

Peters, G. B. 2005. Institutional theory in political science: The 'New Institutionalism'. London: Continuum.

Prud'Homme, R. 1995. The dangers of decentralization. World Bank Observer 10.

Rodden, J. 2005. Federalismo e descentralização em perspectiva comparada: sobre significados e medidas. Revista de Sociologia e Política, no. 24:9-27.

Rondinelli, D. A., Nellis, J. R., and Cheema, G. S. 1983. Decentralization in developing countries: A review of recent experiences. World Bank Working Paper 581.

Schneider, A. 2003. Who gets what from whom? The impact of decentralisation on tax capacity and pro-poor policy. IDS Working Paper 179. Brighton: Institute of Development Studies.

Sellers, J. M., and Lindström. 2007. Decentralization, local government, and the Welfare State. Governance: An International Journal of Policy, Administration, and Institutions 20 (4): 609-32.

Shah, A., T. Thompson, and H.-F. Zou. 2004. The impact of decentralization on service delivery, corruption, fiscal management and growth in developing and emerging market economies: A synthesis of empirical evidence. CESifo DICE Report,World Bank.

Shepsle, K. A., and M. S. Boncheck. 1997. Analysing politics: Rationality, behavior, and institutions. New York, USA: W. W. Norton and Company.

Snyder, R. 2001. Scaling down: The subnational comparative method. Studies in Comparative International Development 36 (1): 93-110.

Souza, C. 2006. "Desenho constitucional, instituições federativas e relações intergovernamentais no Brasil pós-1988". In Democracia, descentralização e desenvolvimento: Brasil e Espanha, org. S. Fleury. Rio de Janeiro: Editora FGV.

Stegarescu, D. 2005. Public sector decentralisation: measurement concepts and recent international trends. Fiscal Studies 26 (3): 301-33.

The World Bank. 1994. World development report: Infrastructure or development. Oxford University Press.

Tiebout, C. M. 1956. A pure theory of local expenditures. The Journal of Political Economy 64 (5): 416-24.

Torres, H., S. Gomes, T. Pavez, and E. Fusaro. 2010. Perfis do professorado da rede pública de São Paulo: a interação entre espaço, regras institucionais e escolhas individuais no resultado de uma política pública. Dados 53 (1): 125-58.

Treisman, D. 2007. The archictecture of government: Rethinking political decentralization. New York: Cambridge University Press.

___. 1999. Political decentralization and economic reform: A game-theoretic analysis. American Journal of Political Science 43 (2): 488-517.

Wang, Y., C. Collins, S. Tang, and T. Martineau. 2002. Health systems decentralization and human resources management in low and middle income countries. Public Administration and Development 22: 439-53.

* This paper is an extract from the post-doctoral research "Decentralization and Autonomy of Subnational Governments in Brazil" funded by the São Paulo Research Foundation (FAPESP) and the Centre for Metropolitan Studies (CEM/CEBRAP). A preliminary version was presented at the 7th Meeting of the Brazilian Political Science Association (ABCP) in Recife. I would like to thank comments from Vera Schattan, Marta Arretche, the two anonymous referees of the BPSR and Eduardo Marques, coordinator of ABCP Public Policy Session. Written originally in Portuguese.