ElEctions and local govErnmEnt pErformancE in Brazil

The 1988 federal Constitution introduced a complex and innovative institutional arrangement that not only reestablished political rights and democratic procedures, but also reinforced decentralization as a fundamental guideline for policy implementation in Brazil. As a result, municipalities have become pivotal actors in the policymaking process. Scholars of Latin American politics have given much emphasis to the causes and determinants of decentralization, but not much has been done toward a more general understanding of how this increased decentralization has affected policymaker behavior and policy outcomes. This paper aims to do exactly that. Specifically, it investigates how institutional arrangements and electoral competition affect local government performance. The theoretical basis is the electoral democratic theory that broadly highlights elections as instruments of citizen control in retrospective and prospective voting approaches. The research employs a large-N cross sub-national analysis based on a dataset of electoral, partisan, socioeconomic and public financial information collected from over 5500 municipalities. Local governments’ performance, our dependent variables, are synthetic indicators formulated from 2009 nationwide surveys on public education, health, housing and welfare services. The OLS regression results confirm the hypothesis that politics variables do matter in how politicians make decisions and implement policy under the new Brazilian democratic Era. The empirical evidences suggest that electoral competition does not present a direct effect on government performance, however, ideology and citizen participation do. Therefore, this paper helps to expand our understanding of a political system’s impact on public policy outputs, which is extremely important not only for academic purpose but also to support policymakers’ decisions.


ElEiçõEs E dEsEmpEnho do govErno local no Brasil
Resumo A Constituição Federal de 1988 introduziu um arranjo institucional complexo e inovador que não só restabeleceu direitos políticos e procedimentos democráticos, mas também reforçou a descentralização como uma diretriz fundamental às políticas públicas no Brasil.Como resultado, os municípios se tornaram atores fundamentais no processo de formulação e implementação de políticas públicas.Os estudiosos da política latino-americana deram muita ênfase às causas e determinantes da descentralização, mas pouco avançaram em entender como o aumento da descentralização tem afetado o comportamento dos formuladores e os resultados das políticas, em termos mais gerais.Este artigo pretende fazer exatamente isso.Mais especificamente, investiga-se como arranjos institucionais e competição eleitoral afetam o desempenho do governo local.A base teórica é a teoria democrática eleitoral que destaca os instrumentos de controle do cidadão em abordagens de votação retrospectiva e prospectiva.A pesquisa emprega a análise transversal Introduction T he 1988 federal Constitution introduced a complex and innovative institutional arrangement that not only reestablished political rights and democratic procedures, but also reinforced decentralization and citizen participation as fundamental guidelines for policy implementation in Brazil.As a result, municipalities have become pivotal actors in the policymaking process.
Scholars of Latin American politics have given much emphasis to the causes and determinants of decentralization, but not much has been done toward understanding how politics affects policymaker behavior and policy outputs under this new democratic setting.This paper aims to do exactly that.Specifically, it investigates how a political system's aspects, such as electoral competition and citizen participation, among others, impact local government performance.This issue became even more relevant in comparative policy analysis due to the fact that policy results in Brazil are extremely diverse.Put differently, despite the advances provided by the new legal framework and the governance improvements after two decades, inequality is the word that best defines policy implementation in this country.
It is noteworthy that the Brazilian cities are also heterogeneous in others aspects, such as geographic, social and economic dimensions.However, are these structural factors the only reasons for the variance in policy outputs?This paper's main argument disagrees.The hypothesis to be tested is that the political system's dynamics at a sub-national level under the new democratic Era impact governmental performance.
The basis is that those who make decisions on public policy, e.g.elected officials and parties, respond to incentives and constraints, as electoral rules or the level of political participation (CLEARY, 2007;IMMERGUT, 2006;PUTNAM, 1999).In other words, the understanding of a policy's determinants comes from the choices' intentionality.Policy outputs are consequences of the decision making process in which stakeholders are affected not only by structural factors, but also by democratic political dynamics.
The inquiry employs a large-N cross sub-national quantitative analysis based on electoral, partisan, socioeconomic and geographic data collected from almost every Brazilian municipality.Synthetic indicators, formulated from 2009 nationwide surveys on public education, health, housing and welfare policies, are our dependent variables -local government performance.
Besides this introduction, the article is organized in five other sections.Next, the institutional and political transformations embedded in the democratization process are discussed, considering their importance to understanding the current framework of social policy implementation.In the section three, the descriptions of the indexes of local government performance are presented.The fourth section begins with some theoretical background followed by the variables used in the statistical models.Finally, empirical results and conclusions are debated in the last two sections.

Democratization and institutional changes in Brazil
After twenty-one years of military regime, in the mid 80's Brazil began to experience a transition to democracy.Although some civil and political rights, such as the freedom of association, right to assemble and to vote were restored in the beginning of this decade, the country was governed by a politician not elected by the people's vote until 1990.The federal Constitution of 1988 came to intensify the democratization process, reestablishing social rights and new institutional arrangements for public policies.
There is consensus that the Brazilian welfare system before the Constitution was essentially corporatist, political and financially centralized in the federal government (DRAIBE, 1990;MEDEIROS, 2001;IPEA, 2009).Social policies were very restricted and fragmented.In other words, the targeted population was restricted, mostly, to urban workers and their benefits, and services were provided by a dispersed range of organizations.
During Constitutional debates, nonetheless, pressures for a proactive state in the social area and for a fair division of responsibilities and revenues among government levels were evident.Social movements and interest groups lobbied congressmen for the following demands: inclusion of population sectors not covered by welfare or other social policy; permanent and transparent mechanisms of policy financing; equal rights between urban and rural workers; citizens' participation in the policymaking processes and; above all, fiscal and administrative decentralization.Consequently, the decisions changed for good public policy and states' responsibilities in Brazil.By law, health, education, housing and welfare programs became compulsory governmental with citizen's rights to all in need.
Universalistic principles came along with a remaining corporativist basis (ESPING-ANDERSEN, 1991;CARVALHO, 2002) in the new social policy framework that included mandatory governmental expenditures in public health and education, welfare benefits without previous contributions and labor rights, such as a national minimum wage, public pension system and equivalent treatment for rural workers.Therefore, the formal assumptions were formulated to accomplish widespread and equal access to public policy, in order to move towards a universalistic model of a welfare state.
Regarding policy management, local politicians transformed the social sector as a result of intensive lobbying.Their demands involved the restructuring of federative relations, local communities' empowerment, governmental efficiency and efficacy and also an opening process to citizens' participation in policymaking (SOUZA, 2004).
Consequently, political, fiscal and administrative decentralization were reinforced (ARRETCHE, 2000;MONTERO;SAMUELS, 2004).The Constitution established the federal state as an immutable clause and all municipalities became autonomous entities, an innovative feature in federative nations.Every governmental level -federal, state, Federal District and municipal -have executive and legislative branches elected by people for a four-year period, except for federal senators that have eight-year terms in office.In addition, all government levels share a certain degree of administrative autonomy and a range of responsibilities, mostly concerning fiscal and social policy.
Because of these characteristics, Brazil is one of the most decentralized countries in the world (SAMUELS, 2004;ABRUCIO, 2005).Nevertheless, there is a dilemma in the country's governmental relationships that alternate between cooperative and competitive federalism.The latter consists of a division of responsibilities in order to preserve autonomy and competition in certain policy areas.On the other hand, the former means that all governmental authorities have the same range of obligations in the interest of a joint policymaking process (ELAZAR, 1987).Definitely, the Brazilian Constitutional decisions became closer to the cooperative model, since it states that important social policy implementation, such as welfare, education, health and housing, shall be common responsibilities among local, state and federal administrations.
Despite this ideal purpose difficulties in systematically adopting these formal requirements in many policies have prevailed.Most of them lack institutional mechanisms of intergovernmental coordination and cooperation.Hence, policy implementations have been heterogeneous processes so far that may vary accordingly to the issue's salience in the agenda, specific framework, division of responsibilities and financial resources.
Moreover, there are three aspects to which the literature converges to a certain degree of consensus.First, local governments' pivotal role is due to their proximity to people's preferences and Brazil's continental dimensions.Secondly, federal administration is a protagonist player, mainly because of its financial and regulatory powers.Finally, as well as in other successful decentralization processes, in Brazil the process is highly dependent on the implementation framework formulated to simultaneously constrain and motivate sub-national officials to cooperate.This kind of strategy is grounded in agreements about governmental duties and technical and financial support from federal administration (ARRETCHE, 2004;KAUFMAN;NELSON, 2004).
The most important social policies in Brazil are the ones with the same formal and informal rules for every municipality.The legal and operational frameworks, broadly established by the 1988 Constitution and, above all, implemented by the federal governmental, are supposed to embrace impersonal and technical procedures.For instance, municipal expenditures in education and health policies are legally established as federal intergovernmental transfers.Nevertheless, these frameworks do not explain all policy outputs, since local government performance is far from homogeneous.This highlights how crucial it is to investigate the determinants of local politicians' behavior.

Measuring governmental performance
Given the 1988 Constitutional outcomes, this analysis is focused on management and implementation of the most developed social policies in the country: education, health, housing and welfare.I begin with a brief description of the method used to measure local government performance, its description, followed by an overview of each social policy.
How can local government performance, our dependent variable, be measured?Performance in public affairs is an extremely complex subject that may involve a vast range of approaches.In order to analyze local politicians' behavior and test the impact of the political system on them, I choose to examine policy outputs of education, health, housing and welfare instead of their outcomes.The former more precisely reflects local government decisions and actions, whereas policy outcomes are more influenced by structural variables and by other programs from different governmental levels, known as the marble cake problem (SHARKANSKY, 1967).
Policy outputs are complicated to define and manage as well, since they can involve a large set of indicators.It is worth mentioning the important literature about the determinants of public service performance conducted by relevant scholars (MEIER; O 'TOOLE, 2002;BOYNE, 2004).Differently from this paper's approach, they primarily investigated the impact of management strategies on organizational performance, which is, normally, measured by the quantity and quality of outputs, efficiency, effectiveness and equity.In that sense, I formulated synthetic indexes that combine several indicators of two policy-output dimensions: institutional capacity and service provision.These indexes are based on 2009 nationwide local government surveys (MUNIC/IBGE) and datasets from the Ministries of Education, Health and Social Development.Because these datasets are composed of a broad list of variables, the research employs principal component analysis (PCA) to aggregate and reduce the data to just four indexes for each dimension.The appendix describes how these synthetic indexes are formulated, including their thematic indicators that basically encompass a set of indicators that expose important aspects of both dimensions.Considering the great population and size differences among cities in Brazil, the synthetic indexes Pedro Luiz Costa Cavalcante were elaborated after separating the cities into seven population ranges.After being normalized, the indexes vary from 0 to 1001 .
The institutional capacity indexes intend to depict an overview of the local government framework of each social policy.They include indicators of sectors' human resources, departments' organization, legislation and management tools, management resources and intergovernmental relations.The second performance measurement, service provision indexes, is based on different indicators of each social policy, because their features and frameworks are highly distinctive.In sum, these indexes demonstrate the variety and scope of social services provided by municipalities.
It is worth mentioning that in both cases, despite the fact that most of the primary data were collected between 2008 and 2009, it is impossible, for instance, to know precisely when a specific management tool was established or a social service started being provided.Because of this, the research's basic assumption is that policy implementation is not a product of a single term of office, but a result of incremental process.
Regarding the education policy, many important changes since 1988 have generated an entirely new administrative and financial framework.Local governments have become responsible for basic education, from kindergarten through junior high school, and legally forced to invest 25% of their budget on it (FRANZESE; ABRUCIO, 2009).In 2007, a federative fund (Fundeb) was approved in an effort to increase policy resources, consequently, to improve the quality of educational systems.Furthermore, federal transfers to sub-national governments are based on technical requirements, such as number of child enrollment.Although education access has been universalized, its quality is far from the developed country stage and is also characterized by inequalities in many perspectives.
The education indexes distribution by region, presented visually in Figure 1, demonstrates their intense variance either intra and inter regions in both dimensions.The institutional capacity's indexes have a heterogeneous pattern with a standard deviation around 17 in almost all regions, which can be noticed by the symmetries of the distributions around the median value (box's darker line).The Midwest has the highest average (55) and the highest standard deviation (17.5).In terms of service provision, there is a clear difference among intra-regional dispersion.It is evident that the poorest regions, the North (23) and Northeast (21), have index averages lower than the national (35).www.revistaoes.ufba.br

Elections and local government performance in Brazil
The circumstances are not different for health policy.During the 90's, the federal government conducted an intense decentralization process in order to implement the constitutional principles of universal and free health services for the entire population.The Unified Health System (SUS) framework is grounded in regionalized and hierarchical resource transfers to sub-national administrations that also have counterpart funding (ARRETCHE, 2002).Although far from ideal, there are important improvements in health policy compared to past decades.SUS has included a significant portion of the population in the public system, expanded the range of services offered and transformed the centralized model to a truly federative arrangement.As a result, there has been considerable growth of health care units, from about 29,000 in 1985 to 77,000 in 2005, mostly at the municipal level (IPEA, 2009).
Nevertheless, the health policy implementation is clearly asymmetric among local governments.The composite index distributions; outlined in figure 2, confirm the heterogeneity in both institutional capacity and service provision, although the first dimension presents a more homogenous distribution among regions.In this case, the expected lower performance between North/Northeast and the other regions is not noticeable.In both cases, the highest averages were in the Northeast and Midwest, whereas the South has the worst performance.One possible explanation may be the higher demand for public health services in the poorest regions of the country, considering that in South and Southeast municipalities the private sector has a greater representation.

Figure 2 -Institutional capacity and service provision indexeshealth, by region.
Source: The author.
In a subsequent transformation process, welfare was characterized by private sector dominance and residual actions.Even though the 1988 Constitution introduced new scope and objectives, particularly conceiving of social welfare as a right and formally part of the Social Security system in Brazil, its implementation was poorly organized, without effective coordination among all governmental levels.The area undertook its national plan only in 2004 with a complete new framework based on public services and income transfers focused on the poorest.The recently created Unified Welfare System (SUAS) is quite similar to SUS, in particular, because it established decentralization and intergovernmental partnerships as guidelines.Accordingly, the new governance arrangement aims to turn local government into a pivotal player.Nevertheless, differently from health and education, the federal government Pedro Luiz Costa Cavalcante concentrates on budgeting and regulation.In fact, the welfare institutionalization process is still ongoing, which reflects on how local government performance is described.
Focusing on social welfare's outputs, figure 3 confirms the perceived disparity among municipalities.From a regional perspective, it is evident how the index distributions are more intense in the service provision dimension; while the institutional capacity disparities among the regions are quite slighter.In both dimensions, the North's municipalities have the worst performance (76 and 49) and Midwest cities stand out positively (81 and 58).Finally, Brazilian public housing policy is the most financially centralized and less regulated among these four sectors.The decentralized process was not successful, partly because of the lack of regulation in the Federal Constitution that did not define management responsibilities among governmental levels, leading to an uncoordinated implementation process (ARRETCHE, 2000).Only in 1995 did the National Urban Development Policy propose a more decentralized framework for federal expenditures.In 2004, the National Housing Plan was approved and since then it represents the main guideline for this policy implementation.Formally, among other principles, it highlights the need for joint action articulated at the three levels of government, although its actual results are not well-evaluated so far (MINISTRY OF CITIES, 2010).
Given this context, it is reasonable to understand the greatest disparity among social policy indexes, which is reflected by the highest standard deviations, approximately 22 and 18 in institutional capacity and service provision indexes.They are really large, especially compared to the housing indexes' averages, around 25 and 34, respectively.As shown in figure 4, unlike the other sectors, the housing implementation variability is also revealed by data asymmetry, i.e., the median lines within the box are not equidistant from the ends, which can be related to the low degree of institutionalization of this policy.Regarding the averages, Southern municipalities (40) in the first dimension and the cities from the Midwest (40) present the best performance in the provision of housing services.Source: The author.

Policy determinants in Brazil
After all, what are the causes of policy outputs?It is, definitely, not a simple question.Since the 50's political scientists, economists and policy analysts have studied the impacts of socioeconomic and political variables on governmental performance.Hence, this subfield of comparative politics has changed the analytical strategy from thick description to Large N studies.In doing so, it has experienced a substantial progress in both theoretical and methodological basis, including an evident improvement in statistical techniques (BLOMQUIST, 1999).
Recently, Imbeau, Pétry and Lamari (2001) reviewed the subfield literature in order to elaborate a taxonomy of policy determinants.On one hand, advocates of the 'convergence' school argue that political and institutional differences do not matter in explaining a policy's outputs, since industrialized societies have become increasingly homogenous.As a result, they face the same social and economic problems; because of this governments implement similar kinds of solutions.On the opposite side, some scholars support that politics do matter, arguing that economic and social factors cannot be neglected, but political variables are relevant in explaining governmental outputs as well.
Overall, the literature's main goal, particularly in political science, is to identify and measure the effects of political systems, in their many aspects, on public policy in order to test core assumptions of the electoral democracy theory.In the accountability approach, democracy supposedly tends to produce governments controlled by and accountable to the voters in which political parties act as central players in an open, fair and regular electoral process.The belief is that electoral results and partisan rules, even informal ones, operate as democratic incentives and constraints to better politician performance (MANIN; PRZEWORSKI; STOKES, 1999).So, elections are the fundamental mechanism of accountability that allows citizens to assess their incumbents.The voter behavior varies between the retrospective to prospective voting logics.The first voters grade politicians' or political parties' past performance in office, rewarding them with re-election or punishing them with defeat at the polls (FIORINA, 1981).

Pedro Luiz Costa Cavalcante
Alternatively, prospective voting stresses that voters base their choice on expectations about future politicians' performance (MARAVALL, 2009).Obviously, these accountability mechanisms are conditioned by a minimalist definition of democracy in which election is not sufficient but rather a necessary requirement.
In the Brazilian literature, despite studies about the determinants of health policy (MARQUES; ARRETCHE, 2002), welfare program adoption (COELHO, 2010) and social expenditures in municipalities (RIBEIRO, 2005), the area of study is still undeveloped.Paradoxically, the political, institutional and administrative characteristics provide excellent conditions to undertake this type of comparative inquiry.Analyzing policy determinants in Brazilian municipalities generates some advantages.A large number of observations facilitate the findings' generalization (KING; KEOHANE; VERBA, 1994) and the focus on sub-national units provides a greater variation of the phenomena to be explained and controls for cultural and historic aspects as well (SNYDER, 2001).
In order to contribute to this subfield, the main goal of this inquiry is to test the hypothesis that not only structural factors cause the variance on policy outputs, but also that political system dynamics have influenced municipal performance in Brazil.As mentioned before, this comparative policy subfield is grounded in the assumption that stakeholder intentionality has, in many ways, an impact on policy outputs.Hence, it is fundamental to test how political variables, products of the new democratic Era, affect local governments' ability to perform well.
The first factor that should influence policy outputs is electoral competition that has been used as independent variables from the beginning of the subfield (KEY JUNIOR, 1951) through to more current inquiries (CLEARY, 2007).Under many perspectives, the core premise is that electoral competition functions as an accountability mechanism, in some sense of how threatened and worried the incumbent party should feel should feel about losing the next election.There are many ways to measure how competitive the elections are; I chose effective number of parties' index, the most common one that has an updated version formulated by Grigorii Golosov (2010): Where, N p : electoral competitiveness; S 1 : the largest component of the number of votes or the number of seats received by a party; S i : the number of votes or the number of seats received by the i-th party; x: the smallest component of the number of votes or the number of seats received by a party.
Ideology is also a common variable in most of the policy determinant analyses (IMBEAU; PÉTRY; LAMARI, 2001).The hypothesis borrows Anthony Downs (1957) view that leftist parties tend to have a more proactive position concerning governmental intervention in the economy whereas right-wing parties are supporters of less state involvement and the free market.Hence, it is expected that the more leftist a mayor's party is, the better his or her performance will be in social policy.To measure ideology I use a left-right estimation of the political parties in Brazil.The estimates, elaborated by Power and Zucco Junior (2009), are based on politicians' opinions about government intervention in the economy.The party scores vary from 0 (extreme left) to 10 (extreme right).I use the average from the last three mayors' terms of office.
The third and four factors intend to test how political relationships between different levels of government affect a municipality`s capacity to perform.Therefore, the variables governor and federal alignments aim to check whether party differences among intergovernmental politicians facilitate or prevent cooperation in policy implementation, as the literature predicts (COX; MCCUBBINS, 1986;ARRETCHE;RODDEN, 2004).The first means that the mayor and the state governor are in the same party.On the other hand, federal alignment means that the mayor's party is part of the Presidential coalition.Both are measured in years from 1997 to 2008 (0 to 12 years).
Another important aspect in policy analysis is the influence of the legislative branch.Scholars have demonstrated how formal and informal institutions embedded in the Legislative-Executive relations matter to the decision and policy-making processes in Brazil (FIGUEIREDO; LIMONGI, 2006) and worldwide (STEIN et al., 2006).Despite the literature focus on the national level, this dynamic is relevant to study sub-national government as well, especially in Brazil where local representatives are elected and have a extensive range of duties and responsibilities, such as law approval, budgeting and governmental oversight, among others.This institutional arrangement entails a setting in which the local city council (Câmara dos Vereadores) may act as a veto player, borrowing George Tsebelis (2002) terminology.Consequently, I expect that the level of legislative control, measured by the percentage of local representatives that are members of the same party or winning coalition of the city's mayor, can influence his or her performance.
All these previous political variables include data from three municipal elections (1996, 2000 and 2004) and their subsequent mandates.As mentioned earlier, the assumption is that policy implementation is a result of an incremental process.
The last political variable that should be linked to policy outputs involves the theory of social capital.Robert Putnam (1999) argues that citizen involvement in public decisions tends to positively affect governmental performance.Thus, the level of citizen participation openness is a social capital proxy that measures how the municipality's social policies committees are organized and running.This variable follows the same technique employed on the dependent variables indexes, principal component analysis (PCA).The synthetic index is based on data of the existence, features and effectiveness of a municipality's committees, also separated by population range, all detailed in the paper's appendix.
Apparently, city wealth might affect policy performance, because local governments should be able to provide better services with a larger budget (GRAY, 1976).Therefore, per capita income is used not only as an independent variable, but also to condition the effects of electoral competition and ideology.In wealthier municipalities, because of their economic dynamism, these two variables would be more influential on social policy performance.Finally, considering the country's demographic diversity, the model also controls for municipalities of North and Northeast regions, population and urbanization rate.The latter is usually used as an important component of economic development that tends to expand governmental responsibilities.Because of the highly skewed distributions of the income and population variables, logarithmic transformations were used to make their distribution more approximately normal.Table 1 presents the list of independent variables -political and control ones -their sources and expected signs in terms of positive or negative relations to the dependent variables and descriptive statistics.Using these set of variables to explain the local performance variety in the two dimensions selected, the research employs multiple regression analysis, more precisely, ordinary least square (OLS) method on a cross-section of data (WOOLDRIDGE, 2006).The research covers a majority of Brazilian municipalities, over 5500 units of analysis.The following equations are estimated: Where, Y: synthetic indexes of administrative capacity and service provision (education, health, housing and welfare); β 0 : Y-intercept; β pol, β int and β con : coefficients of political, interaction and control variables; X pol : matrix of political variables; X wealth : per capita income; X con : matrix of control variables.

Results
The following table shows the estimated coefficients and standard errors in parentheses.First of all, it is important to mention that all eight equations indicate F greater than the critical F-values.It confirms the paper's hypothesis that when socioeconomic variables are controlled, political system variables have significant independent impacts on local government performance.However, these effects vary extensively, in particular, among the social policy areas.These findings are reasonable and coincide with the literature on policy determinants (BLOMQUIST, 1999).
Moreover, I observed similarities between both policy implementation dimensions, but also different patterns.Generally, the coefficients of determination are low, except for the education service provision.In this case, the control variables' estimates, such as income, urbanization and region, are very high, which can explain the elevated adjusted R 2 .Despite the little explanatory power of the models, there are some interesting results.
Contrary to expectation, one of the most used political variables in the literature, electoral competition, does not reveal a consistent significant impact overall.In fact, it showed negative effects on welfare service provision and positive institutional capacity of housing policy.In most cases, electoral threat doesn't seem to matter in how local governments organize and implement these policies.These results are intriguing, since theoretically the level of rivalry in electoral processes is supposed to produce better policy performance (CLEARY, 2007).These findings deserve further investigation.
Governor and federal alignments show little direct effects.Partisan alliance with the governor has negative impact only on welfare service provision, but with a weak effect.The proximity between mayor party and Presidential coalition show statistical significance just in education service provision, nevertheless, its substantive effect is also weak.On one hand, the empirical results contradict some of the literature findings (COX; MCCUBBINS, 1986;ARRETCHE;RODDEN, 2004), on the other; they confirm the perception that decentralization processes were predominant based on technical grounds instead of being partisan oriented.To put it another way, it is reasonable to infer that the bulk of legal and operational frameworks formulated after the Constitution of 1988 have been followed throughout the social policy implementation processes.continued from previous page www.revistaoes.ufba.brSource: The author.

Pedro Luiz Costa Cavalcante
The Executive-Legislative relations' estimates are significant in three cases with coefficients relatively high compared to other political variables.Legislative control seems to result in better performance in education services and both dimensions regarding housing policy.Furthermore, partisan ideology coefficients presented in every policy areas and dimensions have negative signs.Some of them are significant which helps to support the expected influence of ideology on social policy implementation, e.g., the more leftist the local government is the more interventionist its performance in social policy tends to be (IMBEAU; PÉTRY; LAMARI, 2001).For instance, holding all other factors constant, in health institutional capacity, a reduction of one point in the ideology measurement reflects an increase of 1.15 in the municipality performance index, whereas the same variation in the ideology indicator tends to increase 1.2 in the welfare service provision index.
The degree to which citizens are able to participate in the policymaking process appears statistically significant in all policies.The impacts are positive and quite robust in many cases.For example, an increase of 10% in the citizen participation openness index generates an increase of 13 points in welfare service provision or education institutional capacity scores, ceteris paribus.These consistent effects confirm theoretical expectations and, consequently, Robert Putnam's hypothesis (1999) that higher social participation leads to better governmental performance.
A municipality's wealth impacts policy performance in five of the eight cases analyzed, as expected (GRAY, 1976).In some policy areas, such as housing and health, a higher level of city income per capita generates robust effects on performance indexes, holding all other factors constant.Regarding the interactive effects, although some of the political variables conditionalized by municipality income per capita presented statistical significance, most of them demonstrated coefficients substantial in a practical sense.The exception was one case in which a higher electoral competition decreases its influence on housing institutional capacity indexes as city income grows.
As projected, local performance is affected positively by urbanization, except for housing institutional capacity.So, in general, local governments respond in urbanized cities with a more structured and superior range of services in most of the social policies analyzed.In the same way, population seems to positively impact mayor performance.Concerning a city's geographic location, the dummy North-Northeast's results indicate a different pattern of effects.The empirical evidences are not sufficient to confirm the common knowledge that these regions would demonstrate worse performance based on their population's socioeconomic conditions.The results show that municipality performance varies substantially depending on the dimensions and policies observed.
Finally, an interesting finding is that politics impacts are indifferent regarding the degree of policy institutionalization.An institutionalized policy corresponds to a sector in which its framework is grounded in a set of rules that shapes actors and organizational actions and behaviors (PIERSON, 1995;NORTH, 1990).In Brazil, education and health are historically more regulated and structured than welfare and housing (FRANZESE; ABRUCIO, 2009).For instance, while the former have mandatory expenditures for all levels of government and intergovernmental transfers established by federal law, the latter have not reached this level of financing.Initially, we expected a minor influence of politics on more regulated policies; however, the empirical results are quite similar.

Final remarks
Brazil is a successful case of a democratization process that, in a relatively short period of time, has reestablished political and civil rights and built a comprehensive and innovative social policy system.The new policy governance is grounded on constitutional rights, complex intergovernmental relationships and citizen participation.Because of these institutional transformations, analyses of democracy's impact on policymaking and its outputs are essential.In this context, this paper achieved its goal by featuring implications both in methodological and substantive aspects.www.revistaoes.ufba.br

Elections and local government performance in Brazil
First, the research overcomes the many variables and small N problem of the decentralization case studies.The methodological strategy is able to minimize this dilemma by increasing its ability to generalize inferences.Besides, formulating original and comprehensive performance measurements that describe how municipalities organize and provide public services contributes to analyze a complex issue: the different levels of policy implementation.This operational technique may be an alternative to deal with the intrinsic complexity of studying policy outputs.
Regarding the substantive aspect, the empirical results demonstrate that political factors cannot be ignored in this kind of policy analysis.Notwithstanding, some important variables do not present a direct effect on government performance, such as electoral competition.Likewise, surprisingly, the effects of the interactions between income and political variables were generally not relevant.On the other hand, ideology and citizens participation results support the argument in favor of the actual influence of democratization on policymaking frameworks and outputs.Moreover, the research confirms previous findings that politics effects vary among policy sectors, although there weren't obvious patterns.
Lastly, it is evident in all cases that environmental aspects hold an exploratory position even more consistent than political system factors.However, it doesn't refute the hypothesis that politics variables do matter to how politicians make decisions and implement policy under the new Brazilian democratic Era.In doing so, this paper contributes to expand our understanding of political system impact on policy outputs, which is extremely important not only for academic purposes but also as decision making support for stakeholders.
In this sense, the paper's findings provide an interesting range of research questions that could be explored by case studies or Qualitative Comparative Analysis (QCA) strategies.In terms of content, issues such as the role of estate government in social policy or the determinants of policy outcomes are, among others, appropriate for a future research agenda.

Figure 1 -
Figure 1 -Institutional capacity and service provision indexeseducation, by region.

Figure 3 -
Figure 3 -Institutional capacity and service provision indexeswelfare, by region.Source: The author.

Figure 4 -
Figure 4 -Institutional capacity and service provision indexeshousing, by region.

Table 1 -Independent variables.
Source: The author.
Source: The author.