Political Competition and Local Social Spending: Evidence from Brazil

Transcription

1 University of Colorado From the SelectedWorks of Carew E Boulding June, 2014 Political Competition and Local Social Spending: Evidence from Brazil Carew E Boulding, University of Colorado Boulder David Brown, University of Colorado Available at:

2 Political Competition and Local Social Spending: Evidence from Brazil Carew Boulding David Brown Abstract Electoral theories of democracy imply electoral competition insures accountability. Using data on local elections, socio-economic factors, and municipal budgets from more than 5000 municipalities in Brazil for the years 1996, 2000, and 2004, we find municipalities with more competitive elections allocate less to social spending compared to municipalities with little political competition. We argue that previous theory on political competition and public goods obscures the critical role that financial resources play in shaping the dynamics of social spending and political competition. Municipalities with small budgets lack the resources necessary to engineer convincing electoral victories. Where resources are negligible, voter turnout is low, and incumbents rarely win reelection. Incumbent parties in municipalities with large financial resources win big. Armed with adequate resources, incumbent parties mobilize voters and win by large margins. This new argument and evidence reconciles contradictory findings in the existing literature on competition and public goods.

3 Introduction What is the relationship between political competition and social spending? Theories of democracy argue electoral competition promotes accountability and responsiveness. Close elections can compel politicians to perform better in the hopes of keeping their job. Under-performing politicians, the argument goes, lose elections; Close elections serve as motivation. Since governments often make important decisions about who gets what, political competition influences spending on education, health care, and housing. Cross-national empirical studies (Brown and Hunter 1999; Brown and Hunter 2004; Avelino, Brown, and Hunter 2005; Lake and Baum 2001) suggest more democratic read competitive politics are positively associated with the provision of public goods. Studies that focus on political competition, however, offer much less evidence of a strong, positive relationship (Cleary 2007; Stonecash 1987; Boyne 1998). Local governments offer a unique opportunity to examine political competition and social spending by holding national-level variables constant. Our sub-national analysis explores these issues across municipalities in Brazil. We first ask whether spending is higher in more competitive municipalities. Our results surprise: Municipalities with more competitive elections spend less than municipalities with little electoral competition. A robust finding, it holds across elections years, rich and poor municipalities, highly educated and less educated municipalities, and different measures of both political competition and social spending. This finding is puzzling given the strong conventional wisdom that closer elections ought to make politicians more eager to provide benefits to constituents. To make sense of this puzzle, we focus on the role of budget resources. Budgetary resources constrain spending decisions on education, health care, and housing, in a way that deeply influences the nature of political competition. The simple logic connecting competition to spending obscures the critical role 2

4 resources play in shaping how politicians spend and build political support. We offer an alternative explanation: political competition is a function of resources. Politicians constrained by small budgets lack the resources necessary to build political support. Politicians face close elections when they have relatively little to spend; They can t afford to mobilize support. Without much to spend, voter turnout is low and incumbents find it difficult to win. Our focus on resources reconciles two seemingly contradictory empirical patterns: wealthy municipalities experience less political competition yet higher voter turnout. Endowed with significant resources, incumbents win by large margins and mobilize a relatively large number of voters. In the next section we discuss the existing research on political competition and government responsiveness. We then make our argument and lay out our research design. Competitive Elections and Government Responsiveness The logic that competitive elections force politicians to address constituent demands follows a simple and compelling logic. Threatened with removal from office, politicians facing competitive elections must perform. There is evidence at the national level implying political competition works this way: democracies tend to spend more on social spending (Brown and Hunter 1999) more on education (Brown and Hunter 2004) and they tend to provide more public goods (Lake and Baum 2001). The shift to multiparty elections in African countries generates more education spending (Stasavage 2005) and legislative competition directs more resources toward primary education at the state level in Mexico (Hecock 2006). Some argue, however, the relationship between political competition and greater government responsiveness is not that straightforward particularly in the developing world. The number of parties (Chhibber and Nooruddin 2004), how well-informed the electorate is (Besley and Burgess 2002; Besley and Burgess 2001) and the presence of patron-client lines (Kitschelt and Wilkinson 3

5 2007) all influence the extent to which politicians respond to competition. The evidence directly linking higher levels of competition to spending is mixed. For example, studies of state and local political competition in the United States find only a weak link between party competition and welfare policies (Dawson and Robinson 1963; Holbrook and Van Dunk 1993; Stonecash 1987). Similar studies in the UK also find little evidence that competition counts (Boyne 1998; Alt 1971; Karran 1982; Hoggart 1985). In the developing world, the results are also inconsistent. Hecock (2006) finds Mexican states with higher electoral competition spend more on education, but Cleary (2007) finds no relationship between competition and public goods in Mexican municipalities. Cleary also finds that competition does not have indirect effects on literacy rates or poverty. Instead, political participation influences public goods provision. What explains these divergent results? Perhaps conditions in elections specific to developing countries explain the difference. The lack of information, a tendency to vote along ethnic lines, and the absence of credible political challengers can all render competitive elections useless (Keefer and Khemani 2005). Keefer (2007) finds that young democracies are corrupt, fail to uphold the rule of law, and exhibit lower levels of bureaucratic quality. Because competitors can t make credible promises in this context, politicians prefer targeting transfers instead of providing public goods like secondary education. Competition may not work when poverty and inequality are pervasive. Although some argue competition breaks down clientelist ties (Fox 1994), others remain much more skeptical. Some posit competition reinforces clientelism. For example, Kitschelt and Wilkinson argue political parties make more effort to build principal-agent linkages of accountability whenever competitiveness is intense. Whether this competitiveness translates into more clientelistic or more programmatic responsiveness, however, is contingent upon levels of development (Kitschelt and Wilkinson 2007, 28). K&W argue competition encourages parties to spend more on programmatic goods in wealthy 4

6 societies. Competition, they argue, encourages parties to spend more on clientelistic policies in poor societies: Especially among poor countries, competition enhances clientelism. Because competition intensifies ethnocultural mobilization (Wilkinson 2004), and ethnic groups can promote clientelism, politicians will move to employ every imaginable strategy of attracting constituencies, subject to a general budget constraint (Kitschelt and Wilkinson 2007, 32). Interestingly, even if political competition strengthens patron-client ties, politicians may still provide public goods. There is increasing evidence that clientelism does not preclude the satisfactory delivery of public goods. Kitschelt and Wilkinson state there are in fact contexts in which clientelism seems to deliver rather satisfactory outcomes in terms of economic growth and distributional equity (Kitschelt and Wilkinson 2007, 329). Keefer and Vlaicu (2008, 371) present a model of electoral competition in which candidates have two costly means to make themselves credible: spending resources to communicate directly with voters and exploiting pre-existing patronclient network. They demonstrate that in low-credibility states, political appeals to patron-client networks may be welfare enhancing, but in the long run, they delay political development by discouraging direct appeals to voters that are essential for credible mass-based political parties. As the discussion above indicates, there are a number of different reasons why elections often do not function as intended. We offer an explanation that subsumes most of these reasons while reconciling two contradictory findings. First, we recognize the importance of economic development issues by focusing on the budgetary constraints inherent in many politicians decisions. Second, our theory reconciles some of the inconsistent findings on the relationship between democratic political competition and provision of public goods, especially social spending. By focusing on the financial constraints small budgets create, we gain a much fuller understanding of the links between competition, social spending and citizen participation. In the next section, we detail how the size of budgets affects both the ability of politicians to mobilize support and the level 5

7 of competition we observe. Budgetary Resources and Competition The relationship between electoral competition and social spending has been difficult to establish because a third factor strongly influences them both: the overall size of the budget. In very poor municipalities, budgets constrain government spending. In these municipalities, fiscal limits are so severe that no amount of electoral competition can overcome them. Without resources, politicians with strong incentives to increase social spending simply lack the money to do so. In these cases, social spending is comparatively lower and political competition is higher than in wealthier municipalities, but not because electoral competition fails to motivate politicians. Instead, without sufficient resources at their disposal, incumbents are less likely to ward off challengers. In municipalities with relatively large budgets, incumbent mayors have the means to target resources toward supporters or potential supporters. Where sufficient resources exist, social spending is higher, and incumbents regularly defeat their opponents. 1 Before continuing, it is important to acknowledge there are theoretical reasons to suspect relatively large budgetary resources would actually lead to more political competition: the more money at stake (the bigger the budget), a larger number of politicians are compelled to enter the race, increasing competition. While theoretically this explanation captures an important strategic component in the relationship between social spending, the size of the budget, and the decision to run for office, the empirical implication that follows is a positive relationship between the size of 1 There is a well-developed literature on the incumbency advantage in the United States, which has demonstrated that U.S. legislative incumbents have a strong advantage in elections over challenge. Reasons for this advantage include the ability to provide favors, gain visibility and raise money for campaigning (Erikson 1971), redistricting (Cox and Katz 1999), the seniority system, which gives voters an incentive to favor incumbents (McKelvey and Riezman 1992), the scare-off effect of high quality candidates choosing not to run against incumbents (Cox and Katz 1996). Cox and Morgenstern argue that the increase in the incumbency advantage is due to larger budgets (1993). 6

8 local coffers and political competition. We find the opposite. Our argument is similar to one put forth in a very different context by Nooruddin and Chhibber (2008). To explain vote volatility in Indian elections, they argue financial resources play a critical role in determining when incumbents can build political support. Fiscally endowed governments consolidate electoral support while fiscally constrained politicians are more open to political challengers. We take this argument a step further to address directly the relationship between resources, social spending and political competition. From this argument, we derive several empirical implications. First, under conditions of wide variation in local fiscal resources, we expect a negative relationship between political competition and social spending again, not because competition reduces spending, but because governments with resources use them to build political support, reducing political competition in subsequent elections. Local governments with limited resources are unable to use this strategy, not because competition is missing, but because the resources are missing. In these cases, incumbents have fewer tools to build support. Consequently, we should observe both higher political competition in subsequent elections, and lower overall social spending when local governments lack resources. This hypothesized relationship between competition, resources, and social spending helps explain the weak empirical support for claims linking political competition to political performance. Although budgetary resources are often controlled for in the models mentioned above, the empirical tests are often limited to a single cross-section, which obscures the relationship we posit. In sum, we expect a negative relationship between political competition and social spending. 2 2 It is important to note that in focusing on social spending we are intentionally avoiding the contentious battle over whether spending in health, education and housing are public or private goods. Empirically, it is quite difficult to distinguish between public expenditures that are targeted narrowly and those that are targeted more broadly (a school, for example, may be a broad public good, but the contract to build the school clearly benefits specific interests). For the sake of simplicity, we assume that most politicians pursue a combination of strategies, including rewarding 7

9 Second, resources influence whether voters are interested in electoral contests. Resources also influence incentives to contest elections as well as decisions about mobilization strategies. Over time we expect resource-constrained governments to experience more competitive elections, but not because voters want to replace ineffective mayors. A number of mechanisms could be at play here. First, voters could become dissatisfied with poor government services, a situation created by a strong budgetary constraint. Second, decreasing budgetary resources in specific government programs reduces the amount of patronage at stake in an election, thereby reducing participation. Finally, larger coffers could provide incumbent politicians with the means to mobilize the masses for rallies and for voting on election day. Following this logic, we expect voter turnout to be higher in municipalities with relatively large budgets. In financially constrained municipalities, the lack of resources dampens turnout both because the relative stakes of the election are low and because incumbents do not have resources to mobilize voters. When politicians have access to resources, incumbent politicians can mobilize political support, and voters are more likely to care: they see elections as a meaningful space to settle conflicts of redistribution. There are reasons, however, to expect that there could be a negative association between resources and turnout intensive competition for relatively few resources. However, even if fewer available resources actually generates more interest and competition, the inability of incumbent politicians to secure support through government spending could still dampen turnout and competition in the aggregate. Finally, if larger budgets allow incumbent politicians to build political support, generating key supporters, and broader efforts to bolster public opinion (Diaz-Cayeros 2008). Health, education, and housing spending in different contexts can be either narrowly or broadly focused. Our argument does not rely on assumptions about the true nature of public spending. As Kitshelt and Wilkinson convincingly argue, politicians can build support both through patron-client linkages and through programmatic ones. Critically, both strategies require financial resources. Politicians with relatively more to spend simply have more freedom to adopt different strategies than those who are not equally endowed. 8

10 higher voter turnout and wider margins of victory over time, we expect larger budgets will be associated with party electoral success. Conversely, smaller budgets should be associated with incumbent defeat. If, as we expect, the availability of financial resources is a critical constraint on politicians ability to build electoral support, then having more resources should have an observable and significant effect on how well incumbents do in elections. Taken together with the test of the effect of resources on voter turnout, this would be convincing evidence that budgets are often the determining factor in the relationship between political competition and social spending. Research Design: Budget Resources, Municipal Elections and Social Spending in Brazil Motivated by the disconnect between the strong theoretical argument that electoral competition drives government provision of public goods and the surprising lack of consistent empirical evidence for this relationship within democracies at the local level, we set out to explore our hypotheses in the context of Brazilian municipal elections. We first focus on the relationship between size of the municipal budget, political competition in mayoral elections, and social spending on health care, education and housing. Our interest in social spending stems from a line of research that uses social spending as a proxy for government efforts towards public goods provision (Brown and Hunter 2004; Brown and Hunter 1999; Lake and Baum 2001; Hecock 2006). Since spending on housing, health, and education does not capture all spending on public goods, we provide evidence in Appendix A that shows our results are not dependent on focusing exclusively on spending in these areas. 3 We demonstrate that in Brazil greater political competition (closer elections) tend to be associated with less social spending compared with municipalities where elections are won by large margins. We then turn our attention to the hypothesized causal relationships at work, exploring how 3 The same relationship that holds for social spending also holds for total municipal spending as well (Appendix A). 9

11 the size of the municipal budget affects voter turnout and incumbent party electoral success. Examining competition, budget resources, and social spending at the local level represents an ideal test since budgetary decisions--especially in education, health, and housing--are made at the local level. Brazilian municipalities, in particular, are useful in this regard. First, unlike congressional districts, which in some cases cut across populations in odd ways, municipalities tend to reflect a consistent population grouping (analogous to counties in the United States). These political units offer a large number of comparable cases (over 5,000), that share the same national level political context while exhibiting wide variation on the variables of interest: political competition, budgetary resources, and social spending. Because the Brazilian government has made local level data available from a wide variety of sources, we are also able to control for many important contextual features of Brazilian municipalities. Second, the responsibility to provide certain public goods, including primary education, housing, and health care, belongs to the municipal governments in Brazil since the decentralization reforms embedded in the 1988 constitution. Since the constituent assembly that drafted Brazil s constitution in 1988 was influenced heavily by politicians who had served at the local level, an attempt was made to transfer resources and responsibility from the federal government to the state and local levels (Souza 1997). This is particularly important since only a small percentage of revenue in the smaller towns actually comes from local sources in 2001 the estimate was 7% of the resources allocated to small towns and cities came from their own revenues (Souza 2002; p. 8). Most local revenue is generated by a tax on services (Imposto Sobre Serviços; ISS) which is levied on businesses, the self-employed, and groups of professionals working together (hospitals, clinics, law offices, etc.). Municipalities are free to vary the service tax between 2% and 10%. Local governments also administer a property tax on urban real estate (Imposto Sobre Propriedade Predial e Territorial Urbana: IPTU); Municipalities are free to implement different formulas when calculating 10

12 these taxes. Local governments receive transfers from a state government tax on goods and services (Imposto Sobre Circulação de Mercadorias e Prestação de Serviços: ICMS), equaling 75% of the state taxes collected in their area. The remaining 25% is redistributed back to municipalities according to state law which often stipulates criteria for municipalities to perform in a number of different ways e.g., education and environmental protection (Souza 2002; p. 12). In terms of federal transfers, the bulk comes from an income tax (Imposto de Renda: IR) and a tax on industrial products (Imposto sobre Produtos Industrializados: IPI). These two federal taxes make up what is called the Fundo de Participação Municipal (Municipal Participation Fund, FPM) which transfers 10% to state capitals, and 86.4% to smaller municipalities. The remaining 3.6% is transferred to municipalities with populations greater than 156,216 inhabitants (Souza 2002; p. 12). The FPM is the most important transfer from the federal government to the local level which averages 40% of local revenue (Brollo et. al, 2009). The size of each municipality s share of the FPM is calculated by the Federal Court of Accounting (Tribunal de Contas da União, TCU). Although these federal transfers are required by law to conform to a strict algorithm based on population calculated by the TCU, there is some divergence (Brollo et. al, 2009). There is also some preliminary evidence that in very tightly contested municipalities, mayors who are associated with the opposition receive less than their fair share in the two years running up to an election (Brollo and Nannicini 2011). To account for this possibility, we include a variable that records whether the mayor is from the president s party. Table 2 summarizes the kinds of local taxes collected and the constitutional transfers to the municipalities. Tables 1 & 2 about here In sum, the framers of the constitution sought to give sub-national levels of government the freedom to make important fiscal decisions. Although constitutional amendments have stipulated 11

13 municipalities must spend a minimum level of revenue on education and health over the years, state and municipal governments are free to establish the maximum. There is an increasing amount of room for local governments to maneuver especially since primary education and primary health care are constitutionally under the purview of municipal government. Third, municipal governments include both an executive (mayor) and a legislative body (city council) with candidates that compete in local elections, providing a microcosm of similar state-level institutions but with a much greater number of cases and a great deal more variation on levels of competitiveness and share of the budget devoted to social spending. Finally, election rules are constant across municipalities in Brazil (unlike local elections in the United States), making local elections more comparable. Data on municipal election returns, local government spending, and demographic controls are available for more than 5,000 municipalities for three municipal election cycles (1996, 2000, and 2004) allowing for both cross sectional and time series analysis and a much higher number of observations that equivalent cross-national studies could offer. 4 Data and Variables To measure social spending comparatively across municipalities, we use the amount of the municipal budget devoted to social spending (defined as spending on education, health or housing), per capita. We chose to use social spending rather than other measures of public goods provision such as water or sewerage provision because social spending is somewhat more responsive to decisions made by the local government. Although there may be problems in how spending translates into outcomes, for our purposes we are interested in the first step over which politicians exercise control choosing to allocate resources towards social issues or towards other budget items. Because of the skewed nature of the variable s distribution, we use the logged form in the models. 4 For a discussion of the advantages of sub-national quantitative analysis see Snyder (2001). 12

14 We measure political competition as the vote margin between the winning party and the second largest party in the mayoral election. For ease of interpretation, we subtract the percentage point difference between the winning candidate and the second place finisher from one, giving us a measure that increases as elections become tighter (more competitive). To compare municipal budget resources we record the size of the budget in per capita terms. Because of the distribution of this variable, we use the logged form. As an alternate measure, we also use budget expenditures per capita (logged) to capture the level of government spending, not just the available budget. We also control for a number of other variables that might affect social spending, voter turnout, or incumbent party success. First, we control for political factors that tend to be associated with higher levels of social spending, particularly whether the mayor of the town is a member of the PT, the leftist political party that campaigned on greater social spending as one of its key issues. Including the PT variable in the equation also allows us to determine whether being a member of the president s party matters. Second, we control for a number of demographic factors that might influence the demand for social spending (income, population size, urban-rural population, education, and percent of the population over the age of 15). Voter turnout is measured as the percentage of the eligible population that voted in the municipal election. The electoral success of incumbent political parties is measured with a dichotomous variable, indicating whether the incumbent party won the election. All the voting data are available from the Instituto de Pesquisa Econômica Aplicada (IPEA 2010). Given the cross-sectional nature of the data and the continuous nature of the dependent variable, for the majority of the models we use OLS with robust standard errors Although we employ cross-sectional models, we capitalize on the annual spending data we have for each Brazilian municipality for 1996 through

15 The Puzzling Relationship between Competition and Spending in Brazil Our analysis begins with the finding that more competitive elections at the local level in Brazil are associated with less social spending. Given the strong conventional wisdom that competition ought to compel politicians to provide better government services out of fear of losing office, Table 3 reports surprising estimates. Models 1 and 2 demonstrate the negative relationship between political competition and social spending in two different time periods. More competitive elections in 2000 were associated with less social spending per capita (logged) in 2001 (Model 1). Similarly, more competitive elections in 2004 were associated with less social spending per capita in 2004 (Model 2). In both models the sign on the coefficient is negative and significant, indicating that as competition grows, the amount spent on health, education, and housing actually decreases. < Insert Table 3 About Here > The negative association between political competition and social spending is strong and robust. Whether measured in the year after the election or by the difference between average spending levels, the result is the same: increased competition is associated with less social spending. In terms of its substantive impact, the coefficient indicates that for a one unit increase e.g.,.79 to.80 in political competition, the amount spent per capita on health, education, and housing decreases.24 percent. Holding all of the control variables at their means, the predicted value of social spending when competition is one standard deviation below the mean is 311 Reais per capita (95% interval: 308, 316). At one standard deviation above the mean, the predicted value is 290 Reais per capita (95% interval: 287, 295). To put that figure in context, it translates into a 21 unit (the Real) decrease in per capita spending, a seven-percentage point drop in the mean level of social spending per capita. 5 5 The estimates reported above are stable: we subjected them to a number of tests to gage their stability. To make sure the results were not driven by swings in annual spending, we calculated the 14

16 We also test the stability of our results with respect to different operationalizations of spending and budget resources. Appendix A reports the results from three different regressions, using the amount spent on social programs as a percentage of the total budget (Model 1), the size of the entire budget expressed in per capita terms (Model 2), and the revenues collected by or transferred to the municipality (Model 3). The models are run to test whether competition in the 2004 municipal elections is associated with spending in the year that follows (2005). Although the coefficient on political competition is not statistically significant when using the share of the budget devoted to social spending (Model 1), it is statistically significant in the regressions using total expenditures (Model 2) and revenue (Model 3). Overall, it appears there is a strong empirical pattern between political competition and spending on health, education, and housing, the question is why? In the next section we discuss a number of alternative explanations for the strong negative relationship between local political competition and social spending that we find in Brazil. We also explore the secondary hypotheses of our theory, including the effect of budgetary resources on electoral success for incumbents and on voter turnout. Do Brazilian Voters Want Less Spending? A number of explanations exist to account for the negative relationship between political competition and social spending in addition to the one we offer that resources affect both mean levels of spending for and We checked to see whether competition in 2000 had any association with the difference between the two averages for those years (see Model 3 in Table 3). We found political competition had a negative association with the difference in spending between the second period and the first We also took the mean level of spending between 1996 and 2005 and regressed it on the mean level of competition between three different elections (1996, 2000, and 2004); The mean level of political competition was strongly correlated with the mean level of social spending. Finally, to account for the important differences between Brazilian states--remember the models so far include regional dummy variables- -we re-estimated the regression for the 2004 election using state fixed effects. Including state fixed effects had no substantive impact on the results. 15

17 competition and spending. Perhaps in more competitive elections, politicians are compelled to behave in a more fiscally responsible way, limiting the amount local government spends on health, housing, and education. When elections are not competitive, politicians are free to spend on these programs to benefit interests that supported them in their campaigns. Another explanation treats spending on health, education, and housing as a club good politicians can use to target specific interests who have pledged their support. The larger this group becomes and as levels of competition decrease the more goods politicians must provide, increasing the level of spending on social programs. In either case, political competition forces politicians to be more responsible with public revenues. If social spending was seen as wasteful, and only politicians with large levels of electoral support could raise levels of spending, we should expect to see a negative relationship between deficit spending and political competition. When political competition is high, politicians should be less willing to spend more than they take in. To test that proposition, we examined whether political competition in 2004 was associated with deficit spending in To calculate the extent to which municipalities are spending more than they are taking in, we took the overall amount spent in the municipal budget and divided it by revenue. Table 3 reports the results of various regressions that help us explore a number of different causal mechanisms. Models 3 and 4 in Table 3 indicate that competition is not strongly associated with deficit spending. Although political competition in 2000 is associated with deficit spending in 2001, there is no relationship between competition in 2004 and deficits in In other words, there is little evidence of a relationship between political competition and the extent to which politicians are willing to spend more than they are generating in revenues. This indicates, therefore, that there is at least some preliminary evidence that truly competitive races are not compelling politicians to spend less. It seems there is no Tea Party in Brazil. Instead, we believe that the critical factor in shaping both competition and social spending is 16

18 the availability of budget resources. Models 5 and 6 (Table 3) estimate social spending per capita (logged) as a function of political competition, including the size of the budget as an independent variable (measured as the logged budget per capita). In both models the size of the budget is a very strong predictor of social spending. Note also that when including the budget on the right-hand side of the equation, the significance of the competition variable decreases dramatically in one (model 6) with the sign actually flipping in the other (model 5). Voter Turnout If, in fact, larger budgets actually enable politicians to mobilize their constituents to maintain their seats, we would expect the size of the budget to be positively related to turnout. In other words, if resources are necessary to generate large winning margins, resources should be positively associated with turnout. To test that proposition, we regressed turnout in the 2004 election on three different measures that record the size of the budget: size of the budget (logged), size of the budget per capita (logged), and the change in the budget from the period to (see Table 4). In every instance, the size of the budget had a very strong and positive association with turnout. Models 7 and 8 show the relationship between the size of the budget and turnout in 1999 and 2004, respectively. Models 9 and 10 show the effect of budget expenditures on turnout, again for 1999 and And finally, model 11, shows the effect of the change in social spending before and after 2000 on turnout. In every model across two election years and three different specifications of the budget variable larger budgets are associated with higher turnout. When there s something to fight over--a big budget--it appears Brazilians are more likely to participate. < Table 4 about here > Budget Resources and the Incumbency Advantage A key part of our explanation for how political competition and social spending are connected is that the availability of resources limits the extent to which local politicians can spend their way to 17

19 an incumbency advantage. Here, we test that part of the argument directly, exploring the relationship between the budget and the incumbency advantage. We expect that as the number of resources available to the incumbent party increases, their ability to thwart the efforts of the opposition increases. This line of thought suggests that as budgets increase in size, incumbent parties should win and the opposition will be increasingly marginalized. In other words, when the resources available to politicians increase, the degree of political competition should decrease. A look back at the data does, in fact, seem to indicate that competition is more likely when the resources in question are relatively low. In municipalities with big budgets, the degree of political competition tends to be low. < Table 5 about here > To examine the veracity of these claims we examine whether party incumbency (whether the same party stayed in power from one race to the next) was associated with the overall size of the budget. We find that the size of the municipal budget the year before an election is positively correlated with the probability of a party retaining mayoral office in both 2000 and 2004 (Models 12 and 13). The same relationship holds using budget expenditures per capita rather than the overall size of the budget (Models 14 and 15). Similarly, increases in social spending are also associated with incumbent victory (Model 16). 6 Overall, these results present a remarkably robust image of the effect of budgets on 6 As the regressions in Appendix B show, the results are not simply capturing the differences between the big metropolitan areas in the country and those of the small rural villages or a relationship that only exists at high levels of political competition. Instead, incumbent parties with big budgets are more likely to win in municipalities above and below the mean of political competition (Models 4 and 5). Likewise, financial resources make incumbent party victory more likely in both large and small municipalities alike (Models 6 and 7) 18

20 incumbent victory. Over two elections and different specifications of the budget variable, incumbent political parties have a higher probability of winning elections in wealthier municipalities. Conclusion This paper offers a new way of thinking about the relationship between political competition and social spending. The direct relationship between competition and social spending (as well as other indicators of accountability, such as public goods provision and social services) has been difficult to establish. Despite compelling evidence that democracies outperform non-democracies, evidence for political competition as the driving mechanism for public spending is weak at best. In our study, we find a strong negative relationship highly competitive local elections in Brazil tend to be associated with less social spending. To explain both the contradictions in the existing literature and our puzzling empirical finding, we offer an argument that places budgetary resources front and center. The size of a budget is a good predictor of both the level of political competition in local elections and patterns of social spending. There are several important implications of this research. First, it is a strong reminder of the constraints fiscal resources place on a government not just in terms of what government can provide in the way of social services, but also for the dynamics of how democracy works. Politicians with substantial budgets can build political support through both targeted and public spending, leading to less political competition over time as incumbent parties consolidate support. Where budgets are constrained, however, elections appear more competitive, but not necessarily as a sign of vibrant democracy as a sign of chronic dissatisfaction with local government and local government services. Second, these findings suggest a careful reconsideration of how measures of political competition are used in political analysis. The concept political competition measured as the margin between parties, or the share of votes or seats for a certain party, or even the effective 19

21 number of parties is commonly used as a proxy for quality of democracy. In some cases this might be justified if there is a clear difference between competitive and uncompetitive elections and evidence that the lack of competition is an indicator of fraud, for example. But in many cases the dynamics at play are less straightforward. We often assume that close elections represent democracy working well, but they might also indicate chronic dissatisfaction. Likewise, a single party winning repeatedly flies in the face of many definitions of democracy, but could either indicate low levels of competition, or a very successful popular leader. 20

22 References Alt, James E Some Social and Political Correlates of County Borough Expenditures. British Journal of Political Science 1 (01): doi: /s Avelino, George, David S. Brown, and Wendy Hunter The Effects of Capital Mobility, Trade Openness, and Democracy on Social Spending in Latin America, American Journal of Political Science 49 (3) (July): Besley, Timothy, and Robin Burgess Political agency, government responsiveness and the role of the media. European Economic Review 45 (4-6) (May): doi: /s (01) The Political Economy of Government Responsiveness: Theory and Evidence from India*. Quarterly Journal of Economics 117 (4) (November 1): doi: / Boyne, George A Party Competition and Local Spending Decisions. British Journal of Political Science 28 (1) (January 1): Brollo, Fernanda, Tommaso Nannicini, Roberto Perotti, and Guido Tabellini. (2009) Federal Transfers, Corruption, and Political Selection: Evidence from Brazil Unpublished Manuscript (Bocconi University) Brollo, Fernanda and Tommaso Nannicini Tying Your Enemy s Hands in Close Races: The Politics of Federal Transfers in Brazil. Institute for the Study of Labor (IZA), IZA Discussion Paper No Brown, David S., and Wendy Hunter Democracy and Social Spending in Latin America, The American Political Science Review 93 (4) (December): doi: / Democracy and Human Capital Formation: Education Spending in Latin America, Comparative Political Studies 37 (7): Chhibber, Pradeep, and Irfan Nooruddin Do Party Systems Count?: The Number of Parties and Government Performance in the Indian States. Comparative Political Studies 37 (2) (March 1): doi: / Cleary, Matthew R Electoral Competition, Participation, and Government Responsiveness in Mexico. American Journal of Political Science 51 (2) (April): doi: / Cox, Gary W., and Jonathan N. Katz Why Did the Incumbency Advantage in U.S. House Elections Grow? American Journal of Political Science 40 (2) (May 1): doi: / The Reapportionment Revolution and Bias in U. S. Congressional Elections. American Journal of Political Science 43 (3) (July 1): doi: /

23 Cox, Gary W., and Scott Morgenstern The Increasing Advantage of Incumbency in the U. S. States. Legislative Studies Quarterly 18 (4) (November 1): doi: / Dawson, Richard E., and James A. Robinson Inter-Party Competition, Economic Variables, and Welfare Policies in the American States. The Journal of Politics 25 (02): doi: / Diaz-Cayeros, Alberto Electoral Risk and Redistributive Politics in Mexico and the United States. Studies in Comparative International Development 43(2): Erikson, Robert S The Advantage of Incumbency in Congressional Elections. Polity 3 (3) (April 1): doi: / Fox, Jonathan The Difficult Transition from Clientelism to Citizenship: Lessons from Mexico. World Politics 46 (2) (January): Hecock, R. Douglas Electoral Competition, Globalization, and Subnational Education Spending in Mexico, American Journal of Political Science 50 (4) (October): Hoggart, Keith Political party control and the sale of local authority dwellings Environment and Planning C: Government and Policy 3 (4): Holbrook, Thomas M., and Emily Van Dunk Electoral Competition in the American States. The American Political Science Review 87 (4) (December 1): doi: / Karran, Terence Borough Politics and County Government Administrative Styles in the Old Structure. Policy & Politics 10 (July): doi: / Keefer, Philip and Tuti Khemani Public Expenditures and the Poor. The World Bank Research Observer, vol. 20. No. 1 (Spring). Keefer, Philip, and Razvan Vlaicu Democracy, Credibility, and Clientelism. J Law Econ Organ 24 (2) (October 1): doi: /jleo/ewm054. Kitschelt, Herbert, and Steven Wilkinson Patrons, clients, and policies. Cambridge University Press. Lake, David, and Matthew Baum The Invisible Hand of Democracy: Political Control and the Provision of Public Services. August 1. McKelvey, Richard D., and Raymond Riezman Seniority in Legislatures. The American Political Science Review 86 (4) (December 1): doi: / Moreno-Jaimes, Carlos Do Competitive Elections Produce Better-Quality Governments?: Evidence From Mexican Municipalities, Latin American Research Review 42 (2):

24 Nooruddin, Irfan and Pradeep Chhibber Unstable Politics: Fiscal Space and Electoral Volatility in the Indian States. Comparative Political Studies. 41 (8): Snyder, Richard Scaling Down: The Subnational Comparative Method. Studies in Comparative International Development (SCID) 36 (1) (March 6): doi: /bf Souza, Celina Constitutional engineering in Brazil: the politics of federalism and decentralization. Houndmills and London: Macmillan; New York: St. Martin s Press. Souza, Celina Brazil s System of Local Government, Local Finance and Intergovernmental Relations. Paper is part of the EngKaR Research Project 8070: Building Municiapl Capacity for Finance and Budgeting. It has been commissioned by the International Development Department of the School of Public Policy, University of Birmingham (UK). Stonecash, Jeffrey M INTER PARTY COMPETITION, POLITICAL DIALOGUE, AND PUBLIC POLICY: A CRITICAL REVIEW. Policy Studies Journal 16 (2) (December 1): doi: /j tb00778.x. Wantchekon, Leonard Why do resource dependent countries have authoritarian governments? Journal of African Finance and Economic Development. Vol. 5, no.2: pp

25 24

26 Table 1: Summary Statistics Variable mean sd min max Difference in Average level of Spending , Social Spending per capita (Reais) Social Spending as a % of the Budget * Budget per capita (Reais) Average level of Competition Income per Capita in 2000 (logged) Population in 2000 (logged) Percent of Pop. In Rural Areas Literacy Rate Turnout in Effective Number of Parties % Households with TV and Electricity PT Mayor in Population older than *value above 1 represents a few cases where federal transfers pushed the amount spent on social services over the total budgetary outlay. 25

27 Table 2: Local Taxes and Transfers from Federal and State Levels Local Tax Federal Transfer State Transfer Service Tax (ISS) 22.5% of income tax 25% of value-added tax (ICMS) Urban property tax (IPTU) 50% of rural property tax 50% of motor vehicle registration tax (IPVA) Frontage Tax Property transfers (ITBI) 25% of the tax on industrial products (IPI) 70% of the tax o financial operations on gold (IPF/gold) Source: Souza

28 Political Competition in 2000 Table 3: Relationship between Political Competitions and Spending Social Social Social Spending Deficit Deficit Spending Spending Social Spending 2005 (1) (2) (3) (4) (5) (6) *** * *** (0.0407) (0.0141) (0.0146) Political Competition in *** (0.0439) (0.0105) (0.0202) Turnout in *** *** (0.0960) (0.0334) (0.0393) Turnout in *** *** * (0.1211) (0.0302) (0.0602) Size of Budget (logged) *** *** (0.0078) (0.0097) ENP *** ** (0.0138) (0.0042) (0.0053) ENP ** ** (0.0117) (0.0031) (0.0058) Mayor is PT in *** (0.0278) (0.0094) (0.0094) Mayor is PT in (0.0201) (0.0047) (0.0095) Income per capita in *** *** ** *** (0.0287) (0.0288) (0.0084) (0.0067) (0.0125) (0.0128) Pop. in 2000 (Logged) *** *** *** *** *** (0.0076) (0.0076) (0.0020) (0.0017) (0.0060) (0.0080) Rural Pop (% of total) * ** *** *** (0.0326) (0.0312) (0.0105) (0.0080) (0.0132) (0.0161) Literacy Rate in *** *** ** ** (0.0013) (0.0011) (0.0003) (0.0003) (0.0007) (0.0005) % Homes with Electricity and TV Number of observations *** *** *** (0.0006) (0.0005) (0.0002) (0.0001) (0.0003) (0.0002) 4,711 4,014 4,712 4,014 4,711 4,014 Adjusted R note: *** p<0.01, ** p<0.05, * p<0.1 *All regression use robust standard errors. **Regional dummy variables were used in each regression but are not reported above. 27