Partisan alignment and public policies provision in Brazil

This research examines whether local governments politically aligned with the president’s party offer more public policies. The delivery of public services is one of the central activities of governments. Since the political parties control the executive branch in both federal and local government, it is reasonable to expect that their partisan and electoral interests will influence policy implementation. This study analyzes the coverage of primary healthcare as an indicator of policy delivery. Given the strong sharing of responsibilities among federated entities, I argue that the federal government implements public policies strategically, increasing the supply of services in partisan-aligned municipalities. To empirically test this relationship, I estimate the causal effect of alignment using a regression-discontinuity design for close elections. The results indicate that partisan-aligned municipalities have, on average, 3% higher coverage of primary healthcare than others governed by opposition parties. In a city with 10,000 inhabitants, for example, this would mean 300 more people receiving healthcare.


INTRODUCTION
Delivering on public policy is one of the main activities of democratic governments.As a means for voters to assess the performance of governments (Fearon, 1999;Manin, Przeworski, & Stokes, 1999), it becomes an important source of political legitimacy (Pierson, 1995).With decentralizing reforms taking place from the 1980s onwards, the provision of public services has gradually become an activity administered by local governments (Falleti, 2005(Falleti, , 2010)).One of the main reasons for this phenomenon is the association between decentralization and stronger and more efficient social protection systems (Schneider, 2003;Sellers & Lidström, 2007).
As a result, the volume of empirical work on the local provision of public policies is on the increase, with a considerable portion highlighting the large variation that exists at the subnational level (Giraudy & Pribble, 2020;Niedzwiecki, 2018;Wampler, Sugiyama, & Touchton, 2019).The principal explanations for these results involve institutional design (Rogers, 2021), the extent of state capacity (Ziblatt, 2008) and the reproduction of regional economic inequalities (Rodrigues-Silveira, 2019).
In an attempt to minimize the negative effects of decentralization, several countries have invested powers in the central government to coordinate and supervise the activities of local governments in the implementation of policies (Obinger, Leibfried, & Castles, 2005).This structure has increased the interdependence between levels of government, giving central-local relations great importance for the development of social policies (Arretche, 2010;Pierson, 1995).
Studies that consider this aspect as an explanatory variable have for the most part analyzed the institutional mechanisms that guide intergovernmental relations, such as constitutional design and financial resource transfer systems (Abrucio, 2005;Almeida, 2005;Arretche, 2002a).Equally important, but relatively less explored, are institutions' actors and their interests (Pierson, 1995).Thus, I include myself in the debate that discusses the role of the main actors in democratic governments, namely, political parties (Keman, 2006).More specifically, I analyze the effect that partisan alignment between Federal and local government has on the delivery of health care services in municipalities.
As parties compete for votes at both the national and local levels, I expect to observe political direction in the implementation of public policies with a view to gaining electoral advantage over political opponents.I argue that the parties in the Federal Government implement policies strategically, increasing the offer of services in municipalities that are aligned to the president's party.
Considering the case of Brazil, one of the strongest federal systems in the world (Stepan, 1999), I empirically investigate this relationship through a quasi-experimental design: regression-discontinuity design for close elections.The results indicate that political direction in implementation does indeed exist.When a municipality is aligned, there is a 3% increase in primary health care coverage.
The article as it follows on from this introduction (the first section) is structured as follows: in the second section, I outline the debates about parties, public policy, and implementation.In the third section, I discuss the implementation of public policies.In the fourth section, I present my argument and hypothesis.In the fifth section, I explain the selection of the case and the Brazilian institutional context.In the sixth section, I present my research design and methodological choices.In the seventh section I present the results and a discussion of them.And in the eighth section, I make my final remarks.

THE INFLUENCE OF POLITICAL PARTIES ON PUBLIC POLICIES
The relationship between political parties and public policies is one of the most central characteristics of democratic systems (Stokes, 1999).One of the classic assumptions set forth in the literature is that political parties formulate such policies in order to win elections (Downs, 1999), and that there are two main reasons for this.Firstly, public administration is under the direct control of political parties (Hjern & Hull, 1982).Therefore, government and party decisions are closely linked (Strøm & Müller, 1999).Secondly, both voters and candidates are interested in public policy, since it is an important component of electoral competition (Wittman, 1983).
As rational agents, the behavior of competitive political parties can be divided into three categories: vote-seeking, office-seeking, and policy-seeking.According the first model, the central objective of the parties is the maximization of votes in an election.According to the second, it is to increase its control over political offices.According to the third, it is to maximize its influence on public policies (Strøm, 1990).Although analytically separate, votes have no intrinsic value for parties, being merely a means, in democratic societies, to be appointed to office and influence policies.In turn, controlling positions yields benefits to parties, such as extracting income and increasing power and prestige, since policies are public goods for all actors in the system, even for those that do not participate in government (Budge & Laver, 1986;Strøm, 1990).
These three dimensions are not mutually exclusive, and, in practice, no party will pursue only one of them (Wolinetz, 2002).Thus, political parties act as organizations with multiple objectives: valuing votes, valuing offices, and valuing policy (Schmidt, 1996).The importance of each varies according to their central motivation.In this article, I assume that the main interest of political parties is to maximize votes to control public office.To this end, policy is valued instrumentally, with parties choosing electorally profitable public policies, which consequently maximize their chances of winning elections (Strøm & Müller, 1999).
In contrast to the convergence school, which argued that parties were not important to explain the results of public policies, so-called 'politics matters' studies have started to argue that this relationship does indeed exist and, in fact, matters (Imbeau, Pétry, & Lamari, 2001).Also known as "partisan theory, " this set of theories is based on the assumption that politics is a market in which politicians and governments deliver policy in exchange for political support.Thus, differences in the party composition of governments affect the adoption of policies, expenditure on them and outcomes (Schmidt, 1996).Studies have highlighted the effects of the ideological composition of governments on macroeconomic results (Alvarez, Garrett, & Lange, 1993;Hibbs, 1977Hibbs, , 1992)), as well as expenditure and coverage of social policies (Allan & Scruggs, 2004;Castles, 2009;Huber & Stephens, 2001, 2012;Pierson, 1996).
At the same time, an extensive literature points to the relationship between partisan alignment and maximization of local well-being (Golden & Min, 2013).Instead of treating this favoritism as a benevolent action on the part of the governing parties, their behavior is rather associated with an opportunistic strategy that is aimed at extracting income, gaining voters, maximizing chances of re-election and building political alliances (Arulampalam, Desgupta, Dhillon, & Dutta, 2009;Brollo & Nannicini, 2012).
The relationship that structures this behavior is simple.On the supply side, political authorities have electoral incentives that permeate decisions on the provision of public services that usually favor their allies (Bueno, 2017).On the demand side, voters see the delivery of more services as a sign of the local incumbent's skill and competence, which increases their chances of voting for that candidate in subsequent elections (Bracco, Lockwood, Porcelli, & Redoano, 2015).This combination makes central government parties strategically allocate resources to party-aligned subnational governments, given that greater performance at the local level increases the electoral chances of both the local and national incumbents (Arulampalam et al., 2009;Bueno, 2017;Fouirnaies & Mutlu-Eren, 2015).
In general, public policies provide a means for evaluating governments, since the level of wellbeing perceived by the electorate is a way of making inferences about incumbents (Fearon, 1999).This mechanism, known as retrospective voting, works as a means of vertical accountability (Berry & Howell, 2007).The evidence indicates that voters reward parties that make good governments (Healy & Lenz, 2017;Kriner & Reeves, 2012;Lewis-beck & Stegmaier, 2000;Peltzman, 1987).In short, public policy lies at the heart of party competition (Stokes, 1999).

IMPLEMENTATION OF PUBLIC POLICIES
Within the public policy cycle, implementation is the stage in which governments mobilize effort, knowledge, and resources to transform decisions into practice (Howlett, Ramesh, & Perl, 2020).This definition is not unique in the literature but reflects the complexity of implementing a policy characterized by dynamic processes that involve different groups of actors in constant negotiation, bargaining, communication, and conflict (Schofield, 2001).
Implementation is both a political and an administrative process, and it is very difficult to separate them.Normatively, in democratic systems, control over public policy should be exercised by agents whose power derives from their responsiveness to voters (Matland, 1995).Thus, institutions and political actors promote results in terms of public policies (Hjern & Hull, 1982).
During the 1970s and 1980s, studies on the subject were organized mainly on the basis of two theoretical currents, 'top-down' and 'bottom-up.' The first sees policy makers as the central actors, so that implementing agents have clear information and well-defined objectives.As a criticism of the rigidity of this process, the second stream emphasizes the bureaucrats who deliver the services and their interactions with the target audience (Matland, 1995).
Despite being useful for understanding that the two perspectives play an important role, the subsequent debates have suggested an attempt at synthesis, through the mapping of implementation structures (Winter, 2003b).Thus, the most recent research in the area is characterized by the analysis of performance and its causes and consequences based on several theoretical matrices, as well as evaluation methodologies and standards (Winter, 2003a).
The literature has identified several factors that influence the implementation of public policies, such as the characteristics of the program, the availability of financial resources, the organizational environment of the policy, power relations, the competence of the implementing bureaucracy, the coordination between the actors and intergovernmental relations (Hasenfeld & Brock, 1991;Hill & Hupe, 2002;Hjern & Hull, 1982;Howlett et al., 2020;Matland, 1995).
Furthermore, recent Brazilian literature has been pointing to the importance of state capacities as mechanisms of coercion, the existence of laws and the bureaucratic and administrative body as influential factors in the results of public policies (Marenco, 2017;Pires & Gomide, 2016).The literature shows evidence that high capacities are associated with more product delivery and superior fiscal, educational and human development performance (Coelho, Guth, & Loureiro, 2020;Grin, Nascimento, Abrucio, & Fernandes, 2018;Pires & Gomide, 2016;Segatto, Euclydes, & Abrucio, 2021).
Despite not theoretically using this analytical lens, since it would require investigations beyond the scope of this work, I focus the explanation on one of the aspects that are also part of this debate: intergovernmental relations (Grin, 2016;Macedo & Ferreira, 2020).Three reasons justify this choice.Firstly, multi-organizational implementation is practically the rule, especially when dealing with complex problems and political systems with shared authority (O'Toole, 1986).Secondly, different levels of government are often involved in implementing the same policy, which demands greater coordination capacity (Howlett et al., 2020).Thirdly, the interdependence between governments affects how local implementing agents receive financial and informational resources from the center, with the latter acting to control policy execution (Van Meter & Van Horn, 1975).
The empirical analyzes on implementation had a very strong focus on case studies, listing problems, barriers, and implementation failures (Barrett, 2004;Winter, 2003b), which generated an analytical bias (Lima & D' Ascenzi, 2013).Besides it being unrealistic to believe that a program will be implemented entirely according to the design and means foreseen by its formulators (Arretche, 2001), little information about the performance of such policies can be obtained based on this choice (Winter, 2003b).
To overcome this problem, the literature suggests two actions.The first, focused on the empirical strategy, is to reduce the number of variables to those considered critical, increase the number of cases, and introduce control elements to make the base similar and comparable (Goggin, 1986).The second, focused on the object, is to evaluate the policy outputs, i.e., the services being delivered to the population (Rossi, Lipsey, & Freeman, 2004;Winter, 2003b).Therefore, implementation research is more informative according to the evaluation and with parsimony (Winter, 2003b).

ARGUMENT AND HYPOTHESIS
The centrality of public policies in the process of electoral competition makes them a valuable arsenal of political legitimacy (Pierson, 1995).Therefore, with an eye on the potential electoral benefits that parties can obtain from its execution, central government authorities tend to favor political allies in subnational governments (Bueno, 2017).The motivation for this is relatively simple: the performance of parties at the local level influences national election results, acting as a reverse coattail effect (Ventura, 2021).
Based on this, I argue that the parties in the Federal Government implement public policies in a strategic way, increasing the offer of services in municipalities that are aligned to the president's party.By politically directing policy implementation, the incumbent party seeks to improve its evaluation by the electorate, which will increase its performance in subsequent elections.Therefore, the hypothesis of this article is: • H: partisan alignment between president and mayors increases the delivery of public policies.

CASE SELECTION AND INSTITUTIONAL CONTEXT
The case chosen to empirically study the relationship proposed here is Brazil, considered one of the strongest federalisms in the world (Stepan, 1999).The Federal Constitution of 1988 established common competences for the Union, states, and municipalities in several areas of social policies, including health, education, social assistance and culture, among others (Almeida, 2005).At the same time, it has increased sources of revenue for subnational governments without clear spending responsibilities (Eaton & Dickovick, 2004).The combination of growth in spending and little discipline accentuated the fiscal crisis and allowed central government elites to justify federal regulatory measures as a matter of urgency for the Brazilian State (Arretche, 2009).
The reforms in federal relations, which took place mainly from 1995 onwards, continued the sectoral decentralization process, but linked it to the strengthening of control exercised by the federal level (Melo, 2005).As a result, federal laws began to regulate the powers of taxation, spending and the implementation of public policies of states and municipalities (Arretche, 2009).Such reforms did not encounter obstacles, since the new rules presented an incentive structure that made them attractive to subnational governments (Arretche, 2002a).
This "recentralization" process resulted in an increase in federal coordination mechanisms, constitutionalization of intergovernmental transfers and standardization of guidelines for states and municipalities (Abrucio, 2005;Arretche, 2004).Thus, the Federal Government monopolized policy decision-making, while policy-making, i.e., the responsibility for executing policy, became the responsibility of subnational governments (Arretche, 2010).
In addition, I chose to analyze a public health policy, more specifically, primary health care in Brazilian municipalities.Three reasons justify this choice.Firstly, health has been, over the years, one of the areas in which the country has invested the most, especially when considering the federal budget (Vieira, Servo, Benevides, Piola, & Orair, 2020).Secondly, it is the most universal public policy, as it can be -and is -used by citizens of all age groups.Thirdly, it is an important measure of well-being (Touchton, Sugiyama, & Wampler, 2017), as it has an impact on human development, school results, and individuals' incomes (Banerjee & Duflo, 2011).

RESEARCH DESIGN
To operationalize the offer of public policies, I selected the primary health care coverage variable, available on the Ministry of Health's Primary Care Information and Management portal.The coverage rate is calculated based on the percentage of the municipality's population covered by Family Health Strategy (Estratégia Saúde da Família) teams and equivalent traditional basic health care teams in relation to the population estimate. 1n order to estimate the causal effect of partisan alignment, I adopted a regression-discontinuity design for close elections (Lee, Moretti, & Butler, 2004).This technique has been used by studies that investigate alignment as a type of treatment (Brollo & Nannicini, 2012;Bueno, 2017;Meireles, 2019).Here, the comparison is made between municipalities that elected a mayor aligned with the Federal Government (which forms the treatment group) and others that elected political opponents (the control group) by a very small margin of votes.In such cases, we can assume that the election was decided by chance and that the formation of the treatment or the control group is the result of a distribution as if it were random.
Following the procedures of previous works (Brollo & Nannicini, 2012;Bueno, 2017;Meireles, 2019), I only selected municipalities that, in the previous local election, had two or three mayoral candidates, one of which was affiliated with the same party as the president.Elections in which there was a tie between the first and second places were discarded, since, in these cases, the oldest candidate is elected.In cases where there was a second round, only this election was used.From a total of 45,806 available observations, the final sample totals 4,401 cases in which one of the candidates is aligned with the President of the Republic.The time frame 2009-2017 was chosen due to the availability and comparability of the dependent variable and covariates.
The estimate was obtained based on the model proposed by Calonico, Cattaneo, and Titiunik (2014), which combines local linear regression with automatic selection of observations whose winning margins are small enough to be considered random.This procedure applies a greater weight to observations that are very close to the cutoff point, and less weight as the margin of victory increases.
For the design to be valid, it must meet the assumption of continuity around the cutoff point, that is, in this case, that candidates cannot determine their own votes in close elections.The available evidence indicates that this assumption is met in the Brazilian case (Eggers, Fowler, Hainmueller, Hall, & Snyder, 2015).For my dataset, I performed the density test proposed by McCrary (2008) to confirm proper use of the design.In addition, I used observational and placebo sensitivity tests for alternative cutoffs.All procedures mentioned are set out in Graphs 3, 4, and 5 of the Appendix.
Finally, to ensure that the observations close to the cutoff point are, on average, similar in terms of possible confounding variables (Dunning, 2012), I performed the balancing of covariates using the T test to compare means.The results, presented in Graph 6, in the Appendix, indicate that observations very close to the cutoff point (in the ranges -0.1 to 0.1 and -0.5 to 0.5) have statistically similar means (p > 0. 05) on five characteristics: demographic (resident population of the municipality -Source: Brazilian Institute of Geography and Statistics [IBGE]), fiscal (total constitutional transfers and own IPTU collection -Sources: National Treasury and National Treasury Secretariat [STN]) and administrative (total municipal servants and proportion of statutory servants -Source: IBGE).

RESULTS AND DISCUSSION
Observing the evolution of the average percentage of primary health care coverage over time, in Graph 1, some differences are clear.First, the average coverage in Brazilian municipalities went from just over 85% in 2009 to an average of 91% in 2017.Despite the upward trend, there was a decline between 2016 and 2017.When disaggregated by partisan alignment with the president, the data indicate that coverage in non-aligned municipalities follows, in an almost parallel fashion, the joint trend.

GRAPH 1 EVOLUTION OF THE AVERAGE PERCENTAGE OF PRIMARY HEALTH CARE COVERAGE
Graph 1 Evolution of the average percentage of primary health care coverage Source: Elaborated by the author.
Despite being relatively smaller over the time series, coverage in party-aligned municipalities presents a curious pattern.The slope of the line is considerably greater, i.e., the expansion of the health offer in these locations happened more quickly.It is interesting to note that, in the 2015-2016 period, both the national average and that of non-aligned municipalities remained practically stable, while that of aligned municipalities followed a strong growth trend until it surpassed the two straight lines.
These descriptive data should be viewed with caution, as the observations are very unevenly numbered: 5,468 aligned (11.94% of the total sample) and 40,338 (88.06%).
From the territorial point of view, as can be seen in Map 1, virtually all regions show great variation in the percentages of primary health care coverage in 2009.In the last year of the sample, the expansion of assistance meant a reduction in inequalities between municipalities.All regions now show darker colors on the map, which indicates Source: Elaborated by the author.
Despite being relatively smaller over the time series, coverage in party-aligned municipalities presents a curious pattern.The slope of the line is considerably greater, i.e., the expansion of the health offer in these locations happened more quickly.It is interesting to note that, in the 2015-2016 period, both the national average and that of non-aligned municipalities remained practically stable, while that of aligned municipalities followed a strong growth trend until it surpassed the two straight lines.These descriptive data should be viewed with caution, as the observations are very unevenly numbered: 5,468 aligned (11.94% of the total sample) and 40,338 (88.06%).
From the territorial point of view, as can be seen in Map 1, virtually all regions show great variation in the percentages of primary health care coverage in 2009.In the last year of the sample, the expansion of assistance meant a reduction in inequalities between municipalities.All regions now show darker colors on the map, which indicates greater coverage of primary health care.Graph 1 and Map 1 above provide a better understanding of the process of expansion of basic health services in Brazil.Despite the evolution observed over the years, there is still uneven coverage among municipalities.Is it possible, then, that partisan alignment between presidents and mayors results in a greater supply of health services?The answer is yes.
Source: Elaborated by the author.
Graph 1 and Map 1 above provide a better understanding of the process of expansion of basic health services in Brazil.Despite the evolution observed over the years, there is still uneven coverage among municipalities.Is it possible, then, that partisan alignment between presidents and mayors results in a greater supply of health services?The answer is yes.In Graph 2, I present the result of the RDD.The local estimate demonstrates a causal effect of, on average, 3%.That is, municipalities aligned with the president have around 3% more of the population covered by basic health care.Thinking about a municipality with 10,000 inhabitants, for example, this would mean 300 more people being adequately served by health teams.Table 1 shows the estimates of the conventional, bias correction and robust models, the last of which is, by default, the main model.All estimates show the existence of a statistically significant causal effect (p < 0.05).

Bandwidth type mserd
Source: Elaborated by the author.
In Graph 2, I present the result of the RDD.The local estimate demonstrates a causal effect of, on average, 3%.That is, municipalities aligned with the president have around 3% more of the population covered by basic health care.Thinking about a municipality with 10,000 inhabitants, for example, this would mean 300 more people being adequately served by health teams.Table 1 shows the estimates of the conventional, bias correction and robust models, the last of which is, by default, the main model.All estimates show the existence of a statistically significant causal effect (p < 0.05).The direct link between the Federal Government and municipalities is the result of the process of municipalization of primary health care through strong federal regulation (Vazquez, 2014).As local governments joined the Unified Health System (SUS), they became more dependent on the Federal Government, both from a fiscal point of view and in terms of operational guidelines (Viana & Machado, 2009).Therefore, the transfer of responsibility for basic health management to municipalities reorganized relations between federative entities (Arretche, 2002b(Arretche, , 2002a)).
The observed alignment effect is produced through two mechanisms, characterized as policy instruments, which are the means available to directly affect the nature, type, quantity and distribution of goods and services (Howlett, 2000).These are classified into three types: regulatory, economic, and informational.Economic mechanisms comprise the distribution or withdrawal of resources, such as cash transfers or benefits, but may also be of a non-monetary nature.Informational mechanisms are the dissemination of knowledge, recommendations, and guidelines, with the aim of influencing other actors (Vedung, 1998).Since the rules are the same for all subnational governments, the parties will manipulate the distribution of the last two.
With regard to economic matters, the central government increases the volume of discretionary resources for the aligned municipalities (Bracco et al., 2015;Brollo & Nannicini, 2012;Bugarin & Marciniuk, 2017;Fouirnaies & Mutlu-Eren, 2015).Given that these resources are associated with the provision of public goods and services (Bueno, 2017), it is natural to expect that this increase in the municipal budget, in addition to constitutional transfers, will contribute to the expansion of supply.
With regard to the provision of information, parties facilitate coordination and the flow of information between levels of government (Bolleyer, 2011).This happens because of the sharing of structures (Thorlakson, 2009).National parties, therefore, establish operational guidelines for their parties at the local level (Van Houten, 2009), delivering information regarding the production of public policies (Saglie, 2019).Thus, the parties make resources available and guide their use so that mayors aligned with the Federal Government offer more services.

FINAL CONSIDERATIONS
In this work, I have analyzed the relationship between partisan alignment and the provision of public policies.In so doing, I sought to bring evidence to the substantive results of party-political activity, which is relatively little studied in the literature.The delivery of public services, such as the coverage of basic health care as explored here, is the final point in a chain of decisions taken by elected public agents and which affect the daily lives of a large part of the population.
The distribution of goods and services based on party and electoral incentives to which these actors respond has practical consequences: the production of inequalities in implementation, since the available resources are limited, since allocating in a way that favors some groups means harming others.Evidently, I do not believe that there is unrestricted control over its distribution, since there are legally defined rules for this.However, the literature cites considerable evidence of political direction where there is room to exercise discretion.
Even with a voluminous set of data and robust techniques, it is incumbent on me to mention some of the limitations of this study.Firstly, the fact that the study considered a period of less than a decade may restrict its explanatory power.Secondly, regression-discontinuity designs offer only a local explanation.That is, it is possible to make inferences only for observations close to the cutoff point, which does not allow for investigation into and understanding of what supply patterns exist in municipalities where the aligned candidate won by a large margin of victory.Cases like these are important to understand the behavior of individuals in government who are not constrained by strong electoral competition.
It is necessary to answer several questions so that a more expressive set of empirical evidence can be brought to bear on the relationship between party politics and the implementation of public policies.Investigating causal mechanisms more systematically is one of the first steps.Although much is said about intergovernmental coordination, little is known about how this happens in the daily activities of governments.Who are the most relevant actors?How often do they interact?What is the role of political parties, bureaucracy, and the relationship between them?These are just some of the points needed to theoretically and empirically develop this theme.
Coverage in Brazilian territory in 2009 and 2017Source: Elaborated by the author.

Table 1
Model estimates