Do political parties matter for turnout? Number of parties, electoral rules and local elections in Brazil and Bolivia

Transcription

1 University of Colorado From the SelectedWorks of Carew E Boulding March, 2013 Do political parties matter for turnout? Number of parties, electoral rules and local elections in Brazil and Bolivia Carew E Boulding, University of Colorado Boulder David Brown, University of Colorado Available at:

2 Do Political Parties Matter for Turnout? Number of Parties, Electoral Rules, and Local Elections in Brazil and Bolivia Abstract Does the number of political parties influence voter turnout in developing democracies? Some scholars argue large party systems facilitate matching voter preferences with a specific party, increasing turnout. Others argue multi-party systems produce too many alternatives, decreasing turnout. In developing democracies, there is debate over whether these institutions matter at all. We argue that party systems do matter for turnout in developing countries, but the relationship between turnout and the number of political parties is conditional on the electoral formula. Under proportional representation systems, large numbers of parties increase turnout. Under winner take all systems, large numbers of parties depress turnout. Since electoral rules also influence the number of parties, we use an innovative sub-national research design, taking advantage of local variation in the number of parties that is largely unrelated to the electoral system. Specifically, we test these relationships by analyzing turnout data at the municipal level in Brazil and Bolivia, countries with very different electoral rules. Overall, we find evidence that party systems influence turnout, but in different ways depending on the election rules. Keywords Vote Turnout; Party Systems; Electoral Systems; Brazil; Bolivia 1

3 1. INTRODUCTION The act of voting is fundamental to democracy. Surprisingly little consensus exists, however, on what best explains the decision to vote. Most agree that some combination of institutional, socio-economic and election-specific factors are important, but a significant number of questions remain about the conditions under which these factors matter the most, particularly in developing countries. This paper focuses on one particular question: how does the number of political parties affect voter turnout in developing democracies? Some argue that a large number of parties allows voters to select candidates/parties that embody views close to their own, increasing turnout. Others argue additional parties flood the electoral process with too much information, making it difficult for voters to choose between options, thereby depressing turnout (Downs 1957; Jackman 1987). To further complicate matters, large numbers of political parties at the national level are often associated with proportional representation (PR) and indirect selection of the executive through parliamentary procedures. Conflating the two makes it difficult to determine whether PR or the number of parties is affecting voter turnout. Under PR rules, the threshold for representation tends to be lower for individual parties, increasing the sense that votes matter, therefore boosting turnout. However, others argue that the coalition formation between small parties that often accompany PR elections with large number of parties obscure the choices for voters and drive turnout down (Downs 1957). We argue that these theoretical contradictions in the literature can be reconciled with attention to an important contextual variable: the electoral formula. Under electoral formulas where relatively few votes are wasted (PR voting rules, for example), an increase in the number of parties should increase turnout: people can support politicians close to their ideal points 2

4 without throwing their votes away. Under electoral rules where a relatively large number of votes are wasted (first past the post rules, for example), an increase in the number of parties should have the opposite effect: although increasing choice, a large number of parties lowers the probability that any single one will win office, discouraging voter turnout. It is difficult to determine the relative influence of parties and electoral systems in crossnational studies because country-level electoral rules have a clear effect on the number of parties. We propose a unique research design to establish whether the number of parties increases or decreases turnout. We collected data on municipal level elections in Brazil and Bolivia during two very similar rounds of elections in both countries. The design allows us to do two things: 1) vary the number of parties participating in an election while holding constant the electoral system; 2) explicitly test whether the number of parties and their impact on turnout is affected by the difference in the electoral formula employed in the Bolivian and Brazilian mayoral elections. As expected, we find a strong positive relationship between the number of parties and turnout in Bolivia, but a strong negative relationship between the number of parties and turnout in Brazil. These findings are interesting for the debate over party systems and voter turnout, but they also speak to a larger debate over the importance of political institutions in developing democracies. Many scholars register skepticism concerning electoral rules and their importance in contexts that witness significant problems like corruption, electoral access, and clientelist power brokering. Our results suggest that political institutions hold significant consequences. Voters and political parties respond to rational incentives even in the context of relatively new democracies like Bolivia and Brazil. In Latin America, large numbers of political parties have been seen as a pathology of weak and fractionalized party systems. We argue, however, that weak political parties are not the whole story: voter participation hinges on the value voters and 3

5 parties place on each individual ballot. Electoral laws designed to maximize each individual vote are crucial to fostering democratic participation. 2. PREVIOUS WORK Political institutions are important determinants of cross-national variation in turnout among established democracies (Powell 1980; Jackman 1987). A growing number of scholars argue institutions matter for turnout in developing countries (Fornos, Power, and Garand 2004; Kostadinova and Power 2007; Dettrey and Schwindt-Bayer 2009; Lehoucq and Wall 2004; Perez-Linan 2001; Kuenzi and Lambright 2007). Although some agreement exists on the important role institutional rules play in determining turnout (e.g., compulsory voting rules tend to increase voter turnout), there is much less agreement on multipartism, electoral rules and voter turnout. For example, scholars debate whether winner-take-all (SMD) or proportional representation (PR) schemes generate higher levels of participation. Most find PR promotes participation (Powell 1980; Andre Blais and Carty 1990; Franklin 2004; Jackman and Miller 1995; Radcliff and Davis 2000; Norris 2002). However, the mechanisms driving the result are less understood (Brockington 2004; Blais 2006). Although higher rates of turnout are associated with PR, one of its central features (relatively large party systems) seems to decrease voter turnout. In one of the foundational studies on multipartyism and turnout, Robert Jackman (1987) found a strong, negative relationship between the number of parties and turnout among 19 industrialized democracies. Jackman explained the result by relying heavily on Anthony Downs insights into multiparty systems (Jackman 1987, p. 408). Since there is a high correlation between multipartyism and coalition government, voters may be discouraged from participating 4

6 when the actual formation of government is only indirectly under their control (Downs 1957). For Downs, it was entirely possible that a voter would be so befuddled that he stays home on election day (Downs p. 152). Preliminary evidence indicates that coalitions are not the culprit: systems that produce single-party majority governments do not have higher rates of turnout (Blais and Carty 1990; Blais and Dobrzynska 1998). Individual-level data collected in the industrialized democracies also demonstrates that strong partisans those who would lose most when their parties enter unwieldy coalitions feel just as efficacious when voting in the context of coalitions (Brockington 2004). And in presidential elections where multipartism in the race does not translate into multipartism in the office, because only one candidate wins the presence of many candidates still makes it more difficult for voters to discern the policy implications of their vote (Dettrey and Schwindt-Bayer 2009, 1327). Nevertheless, even with these additional studies, a recent reviewer and significant contributor to the field writes The bottom line is that we have a poor understanding of the relationship between the number of parties and turnout (Blais 2006, 118). Although the bulk of empirical research on turnout (most of which has studied advanced democracies) finds a positive relationship between PR and higher turnout, this trend is less evident in Latin America than it is in Europe (Blais and Aarts 2006). This relationship, however, is not the primary focus of our paper. Rather, we are interested in the relationship between the number of parties and turnout. Although PR systems often have a greater number of parties, very few studies have investigated how the number of parties influences turnout. Jackman s (1987) study is an exception in that it does directly address this question finding that higher numbers of parties are associated with lower turnout. Others, like Powell (1982), find that other features 5

7 of political parties such as linkages to social groups matter for turnout, but the number of parties is not central. Our contribution, then, is to explore how the number of parties on the ballot affects voter turnout under different electoral formulas. This focus addresses the puzzling finding that PR systems tend to have higher turnout than majoritarian electoral systems yet the number of parties does not seem to increase turnout (despite being a prominent feature associated with PR systems). This puzzle and the question of how number of parties shapes turnout has been identified by experts on the voter turnout literature (Blais 2006; Blais and Aarts 2006), and is important for understanding the causal links between electoral formula, party systems, and turnout. Rather than focusing on turnout at the national level in PR versus majoritarian systems, we focus on the relationship between number of parties and turnout in local elections. Overall, the relationship between number of parties and turnout is much less established in the existing literature than the relationship between PR and turnout. This is partly because fewer studies have directly looked at this (again, Jackman 1987 is an exception). More importantly, in cross-national studies of turnout it is very difficult to separate out the causal effect of the electoral formula from the causal effect of number of parties since we know electoral formulas strongly influence the number of parties competing (Duverger 1954). The subnational comparison of local elections in Brazil and Bolivia that we employ helps solve this problem. Although the political party system in each country is clearly related to the electoral rules, variation in the number of parties in local elections cannot be the result of the electoral formula since all local elections follow the same electoral formula. Although Brazil tends to have higher overall turnout than Bolivia, we exploit variation in the number of parties in local elections that cannot be attributed to the electoral formula (which is constant in each 6

8 country). This research design allows us to explore the effect of number of parties on the ballot in a context where the number of parties is not directly the result of the electoral formula. 3. THEORY We argue that the number of parties in an election has varying effects on voter turnout depending on the electoral formula. In other words, the impact the number of parties has on turnout is conditional on the electoral formula. 1 It is intuitively appealing that a greater number of parties should encourage voter turnout. When more parties field candidates in an election, voters can choose parties closer to their ideal point than would be possible in a two-party system. However, as reported above, very little empirical support exists for this hypothesis. Instead, there seems to be a robust negative relationship between the number of parties and voter turnout in cross-national comparisons. The critical dimension, we argue, is the value both parties and voters place on each vote. Additional alternatives additional parties will produce little change in turnout when votes are wasted. A greater number of parties under conditions of proportional representation means that voters may get to vote closer to their ideal point and have a good shot at their party getting a seat. Both are necessary conditions for increasing turnout. Under rules that favor larger parties, the appeal of voting for the preferred party is counteracted by the low probability that the vote will have an impact on the election. We argue that the balance between vote choice and strategic concerns about how much that vote is likely to matter influences the behavior of voters and political parties. For voters, this can be an individual calculation to choose a second choice candidate who is more likely to win. Party tactics to mobilize support also matter. Under conditions where the value of each vote is 1 (Cox 1997) raises the importance of electoral systems in structuring the strategic incentives for political parties to mobilize voters. 7

9 relatively high, parties have strong incentives to mobilize every voter possible. Under conditions where there is a clear threshold beyond which additional votes are wasted, there is little incentive for parties to mobilize voters beyond the necessary minimum. 2 More importantly, these incentives vary with the electoral system. In a proportional system, parties have an incentive to mobilize core supporters and fringe supporters, since both might result in more seats. In a majoritarian system, parties need only mobilize up to a certain threshold. Elections vary on many dimensions that can affect the value of each individual vote (district magnitude, overall levels of participation, or restrictions on the franchise, for example), but we focus on the clear distinction between proportional and majoritarian voting. Although there may be variation in the value of each vote in different elections, it is clear that on average individuals votes count more in proportional elections than in majoritarian elections. Under conditions of proportional representation in local elections, then, we expect that the number of political parties in the contest will be positively associated with voter turnout. Since many parties can win representation on the council, the threshold to representation is relatively low, meaning that many parties can run and reasonably gain enough votes to get a seat. In this environment, larger numbers of parties allow voters to support candidates closer to their ideal point. In majoritarian municipal elections, we expect the opposite effect. H1: Under conditions where relatively few votes are wasted, an increase in the number of political parties tends to increase voter turnout. 2 This incentive structure has been shown to change the way political parties target goods in Indian elections, depending on whether an electoral context is two-party or multiparty (Chhibber and Nooruddin 2004). Where only two parties are competing, they tend to provide goods that reach more broadly. Where more parties are competing and only a plurality is needed to win, goods are targeted more narrowly. 8

10 H2: Under conditions where relatively large numbers of votes are wasted, an increase in the number of political parties tends to decrease voter turnout. 4. RESEARCH DESIGN Instead of examining aggregate cross-national data, we examine voter turnout at the municipal level in Bolivia and Brazil. This is an interesting comparison as both countries share a number of institutional features that matter for voter turnout, but have quite different electoral formulas. Both have compulsory voting, non-concurrent elections, and multiparty elections at the local level. Both transitioned to democracy around the same time, following military dictatorships (Bolivia in 1982, Brazil in 1985). With these local level data we can test whether the number of parties affects turnout within Brazil and within Bolivia in addition to determining whether electoral differences between the two countries affect the relationship between the number of parties and turnout. 3 In Bolivia, local elections are much like local versions of parliamentary elections. Voters choose between party lists, seats on the city council are allocated proportionally by party, and the party with the plurality of votes selects the mayor from their list. According to Article 94 of the Bolivian Electoral Code (available at from the National Electoral Court website (CNE 2009)) council members are elected for five year terms following a system of proportional representation based on the number of votes each political party, citizen s group, indigenous group, front, or alliance receives. The mayoral candidate for each party is the candidate at the top 3 Unlike local elections in the United States, local elections in both Brazil and Bolivia follow consistent electoral rules across each country, making it unlikely that variation in the number of parties at the local level is a function of local variation in election rules. 9

11 of the party list. If one party wins an absolute majority of votes, their candidate becomes the mayor. If no party wins an absolute majority, the newly elected council votes between the candidates from the top two parties. If the vote ties, the party with the plurality of votes wins the mayor s seat. The number of city council members is determined by the population of the municipality. Municipalities with up to 25,000 people have a five-member council, with 2 additional council members for each additional 25,000 people. The municipal governments of the department capitals have eleven council members. Brazilian mayors, in contrast, are directly elected in a first-past-the-post system. Starting in 2000, mayors could be re-elected for a second term in office. In towns with more than 200,000 inhabitants, if no candidate wins 50 percent of the vote, there is a run-off election between the top two vote getters. Along with voting for mayor, Brazilians elect vereadores (city councilmen and women) using a proportional representational system. The size of the council (the municipal legislature) is based on population (the number of legislatures varies from 9-55) (Sakurai 2008 p. 303). The proportional system does encourage more parties to participate in city council elections relative to mayoral contests. The essential difference, however, between Brazilian mayoral elections and their Bolivian counterparts is that the mayor is directly elected by the population, in Bolivia the mayor is chosen indirectly. Under these conditions, the number of political parties has a very different effect on voter turnout. As the number of parties increases, smaller parties face a decreasing chance of reaching the winning threshold. Unlike PR systems, there is no compensation of winning seats on the council; the candidate either wins or loses the election. Consequently, an increase in the number of parties in a local election of this type reduces voter turnout, as voters realize the chances of their favored candidate winning the election is small. 10

12 5. DATA AND MODEL We use OLS with errors clustered at the state/province level to estimate the impact socioeconomic variables, political competition, and the number of parties have on voting turnout in Brazil and Bolivia. 4 The turnout variables were collected from the 2000 and 2004 municipal elections in Brazil and from 1999 and 2004 in Bolivia. In both cases, the turnout variable represents the number of votes cast (whether valid, blank, or nullified) as a percentage of the eligible population. As can be observed from Table 1, turnout is higher in Brazil (the mean in Brazil is 88 percent compared to Bolivia s 62). There is considerable variance in turnout even when we average across states in Brazil and provinces in Bolivia. In Brazil, the states with the lowest turnout averages are situated in the Amazon (Acre 82.5%, Amazonas 80.6%, and Rondonia 82.5%) while the highest are found in the more populated areas in the South and Southeast (Rio de Janeiro 89.4%, São Paulo 88.3%, and Rio Grande do Sul 93.2%). In Bolivia, the highest averages were in La Paz (64.2%), the capitol, and Oruro (71.7%), a municipality a few hours away. The lowest averages were found in more rural and sparsely populated areas, both in the Amazon, (Beni, 58.7%), and in the Andean highlands (Potosi, 59.7%). Because we are interested in the number of parties, we chose to use the raw number of parties competing in each election as our main independent variable. We decided to use the number of parties rather than the effective number of parties (ENP) since ENP entails combining the number of parties with the vote share. By combining vote share with the number of parties, it 4 Data for Brazil was obtained from the following website Electoral data from Bolivia was obtained from Other Bolivian data was obtained from 11

13 would be difficult to ascertain whether the party structure of local elections and its impact on turnout was being driven purely by the number of choices voters had or the level of competition in the race. We prefer the number of parties as a measure because it is a simple and direct measure of the choices that voters are confronted with on a ballot. We believe this measure best captures the variation we are interested in. We do, however, recognize that many studies prefer the use of the effective number of parties as a measure. As a robustness check, we also estimated the models for 2004 using the effective number of parties 5 in place of number of parties and find very similar results in both sign and significance. The effective number of parties and the raw number of parties for both countries in 2004 are also highly correlated (.85), giving us further confidence that our measure is appropriate. Our analysis consists of estimating regressions for all the municipalities within each country separately (Models 1 and 2), Finally, since Brazilian states and Bolivian provinces feature distinct political processes even when compared within each country, we clustered errors by state/province. We know that a number of socio-economic variables are important determinants of turnout, both in cross-national comparisons and in comparisons of local elections. We include these as control variables. In particular, turnout is higher among more educated citizens (Wolfinger and Rosenstone 1980; Blais 2000). Turnout is also higher in districts with smaller populations (Powell 1982; Blais and Carty 1990; Blais and Dobrzynska 1998), although it is less clear how population size or district size affects turnout in local elections (Blais 2006; Oliver 2000). The context of the election itself is also important. Very close electoral competitions tend to drive higher turnout, as people interpret their vote as 5 We calculate the effective number of parties using vote share according to the formula proposed by Laakso and Taagepera (1979). 12

14 potentially having a greater impact on the outcome of the election (Blais 2000, 60). 6 For the competition variable we simply took the margin of victory (the difference between the winning vote getter and the second place finisher). Given both countries history with democratization, we included the number of votes for leftist parties since their strategies typically tend to emphasize mobilization relative to more conservative parties. Essentially, we wanted to be able to distinguish between increased turnout that was due to the number of parties and an increased turnout that was primarily the result of mobilization from one side of the political spectrum. 7 We used the following model to explain the variance in turnout at the local level among both Brazilian and Bolivian municipalities in addition to specifically testing whether political competition and the number of parties holds different consequences for turnout in Brazil and Bolivia: 1) Percentage of Eligible Voters Casting Ballots = a + b 1 (Previous Turnout ) + b 2 (Income per capita) + b 3 (Population) + b 4 ( Rural Population) + b 5 (Literacy) + b 6 (Votes for the Left) + b 6 (Margin of Victory) + b 7 (Number of Parties) + e. In order to compare differences between Bolivia and Brazil, we also estimate fully interactive models that register the differences between Bolivia and Brazil (Model 3) and the differences between the border region and the municipalities elsewhere in both countries (Model 4). In the fully interactive models every independent variable is interacted with a Brazil/Bolivia dummy or a dummy that indicates whether the state or province lies on the border between Bolivia and Brazil. The following is the fully interactive model: 6 Cox (1988) makes the case that margin of victory can be spuriously correlated with turnout and recommends using raw vote margin when comparing districts of equal population. In our case, the population size varies widely between municipalities and we believe that percentage margin is more appropriate. 7 Although votes for the left variable was statistically significant in the basic model (Model 1), its inclusion or exclusion in the model had very little impact on the other estimates. 13

15 2) Percentage of Eligible Voters Casting Ballots = a + b 1 (Previous Turnout ) + b 2 (Income per capita) + b 3 (Population) + b 4 ( Rural Population) + b 5 (Literacy) + b 6 (Votes for the Left) + b 7 (Margin of Victory) + b 8 (Number of Parties) + b 9 (Country Dummy) + b b 16 (Independent Variables*Country Dummy) + e. As an additional test of our argument that the number of parties has different effects under different electoral rules because of the structure of incentives we also compare municipalities with close elections to those with little or no political competition. Electoral contests where there is more political competition are likely to attract the attention of both political parties and voters. If our hypothesis that calculations to vote and to mobilize voters depends on the value of each vote, we would expect to observe that the number of parties and its impact on turnout in Bolivia (where votes also count towards producing a city council) does not vary between competitive and uncompetitive races. In Brazil, however, where plurality reigns, we would expect the number of parties to have a very different relationship with turnout depending on the competitiveness of the race. 5. RESULTS We start by estimating the effect of the number of parties on voter turnout in separate regressions for the 310 municipalities in Bolivia (see Model 1 in Table 2), and 5290 municipalities in Brazil (Model 2 in Table 2). In Bolivia, where local elections are run according to proportional representation rules, voter turnout increases as the number of parties increases. In Brazil, where local elections are run according to majoritarian rules, voter turnout decreases as the number of parties rises. Other variables have a similar effect on turnout in both countries: the previous elections turnout is a strong positive predictor of turnout in the current election; turnout is higher in smaller municipalities and in more educated ones. And there are some other interesting differences: poorer municipalities had higher turnout in Bolivia, but Income has no 14

16 significant effect in Brazil, perhaps reflecting the electoral strength of the pro-poor party of Evo Morales in the 2004 Bolivian elections. Votes for the left also influence turnout differently. In Bolivia, where the MAS did very well in this election, votes for the left increase turnout. In Brazil, the effect is the opposite. To test the significance of these differences, we estimated a fully interactive model which allows both the slopes and intercepts to vary between Brazil and Bolivia for each independent variable. In the fully interactive model, Model (3), we find the relationship between the number of parties and turnout in Bolivia is statistically significant. As expected from the separate regressions in Models 1 and 2, Income, votes for the Left, and the number of parties all show significantly different relationships in Bolivia and Brazil. To gain a better understanding of the substantive significance of the relationship between the number of parties and voting turnout in the two countries, we generated predicted values for each country as the number of parties ranges from its minimum value to its maximum, using the fully interactive model and holding all other variables constant at their mean values (see Figure 1). As the lines on the graph indicate, the relationship between number of parties and voter turnout is much stronger in Bolivia than in Brazil, in addition to working in the opposite direction. In Bolivia, an increase in the number of parties is associated with a fairly substantial change in turnout. In Brazil, the slope is negative and relatively flat. According to Figure 1, at roughly one standard deviation below the mean (4 parties), Bolivia observes turnout in municipal elections of roughly 60 percent of the eligible population. At the mean, there is an increase of roughly four percentage points. Finally, at one standard deviation above the mean, roughly 69 percent of the Bolivian population goes to the polls: a nine percentage-point increase from one standard deviation below the mean. As the confidence intervals indicate, the errors around the 15

17 predicted values are fairly small (the grey-shaded areas surrounding the two lines). The dramatic difference between the slope of the Brazilian and Bolivian regression lines holds an important implication worth mentioning. Given the sub-national nature of the data, we can say that there is a direct correlation between the number of parties and turnout in Bolivia, but this is highly contingent on the electoral system. Unless we distinguish between the different contexts under which these conditions operate, the relationship between the number of parties and turnout will remain obscured. It is also clear from Figure 1 that overall turnout is higher in Brazil than it in Bolivia. In fact, average turnout across all municipalities in 2004 in Brazil is 88 percent; in Bolivia in the same year it is 62 percent. The conventional wisdom is that turnout is usually higher in PR systems such as Bolivia than in majoritarian systems such as Brazil, so this difference deserves some explanation. First, it is important to point out our research is aimed at explaining variation at the local level in both countries precisely because that internal variation cannot be attributed to the national level electoral rules or party system. It is more than likely, however, that the overall differences between the two countries are related to the national electoral rules and party system. Brazil s national elections unlike the local elections are held using open-list proportional representation. Bolivia s national elections run according to a mixed system in which some representatives are chosen proportionally and others are chosen through plurality rules. The difference in national averages, then, may partly be due to differences in national electoral formulas, which are more proportional in Brazil. Figure 2 illustrates the partial regression plots by country, giving another image of these differences. Another possibility is that other features of both the electorate and the rules associated with voting affect turnout. Although voting is mandatory in both countries, Brazil has a much 16

18 more developed bureaucracy for enforcing such laws than Bolivia does. Bolivia overall is a much less developed country, with a higher percentage of the population living in extreme poverty, living in rural areas, and with much lower rates of literacy and education (all of which predict lower turnout rates). Sparsely populated and remote areas are associated with lower levels of turnout in both countries. However, a much larger percentage of the Brazilian population lives in urban areas (58%) relative to Bolivia (19%). Moreover, in an effort to combat clientelism, all vehicle traffic is prohibited in Bolivia on Election Day to prevent political parties from offering rides to polling places in exchange for votes. As a result, most Bolivians in rural areas have to walk a long way to their polling station, an obstacle that may reasonably depress turnout as well. Finally, to test whether electoral difference between Brazil and Bolivia is the key factor, we created a new country called Bralivia, comprised of states and provinces adjacent to the Brazilian-Bolivian border. We then estimated a fully interactive model that allows the slopes and intercepts to vary for each independent variable according to whether the case is along the border ( Bralivia ) or somewhere else in Brazil or Bolivia. By doing so, we are mixing Brazilian cases with Bolivian cases. If national differences are important, dividing municipalities this way should produce different estimates. If the interaction between the dummy variable and number of parties in this equation (Model 4) resembles the estimates in Model (3), it would suggest our hypothesis about the importance of electoral systems is wrong. If the distinction between Brazil and Bolivia is important and institutional differences exist that identify important processes between the two countries (indirect vs. direct mayoral elections), we would expect to find no important difference between the slopes on the number of parties with this model. That is, in fact, what we find in Model (4). When the interaction does not distinguish 17

19 between two different electoral regimes but instead mixes them, we find that the coefficients for the interaction terms of political competition and the number of parties cannot be distinguished from zero at any acceptable level of confidence. This analysis of the border municipalities suggests that the number of parties increases turnout in Bolivia but decreases turnout in Brazil is not just a function of Bolivia being a poor, largely rural country and Brazil being a much richer more developed country. Even the most similar municipalities in Bolivia and Brazil exhibit clear differences in how the number of parties affects turnout. To make sure our results are not driven by regional disparities within each country, in addition to clustering the errors by state/province, we took the basic regressions for each country separately and included state/province dummies. In those regressions we found the same strong positive correlation between number of parties and turnout in Bolivia while finding a negative and statistically insignificant result for Brazil. Finally, as a final check of the result s stability, we re-estimated both the basic Brazilian and Bolivian models (models 1 & 2; Table 2) with incumbency measures. We found that the incumbent party winning in Brazil had a positive and significant impact on turnout, but it had no impact on the number of parties variable (the variable of interest). In Bolivia, the incumbent variable was not significant and its inclusion had no impact on the positive relationship between the number of parties and turnout Political Competition Thus far we have demonstrated that the number of parties has very a different effect on voter turnout in Bolivian municipalities and Brazilian municipalities. We have argued that 8 For Brazil, we used a 0/1 variable that indicated whether the same party that had won in 2000 won again in For Bolivia, we used a 0/1 variable that indicated whether an incumbent ran in the race in Our use of different incumbency measures in this instance was purely due to data availability. 18

20 differences in the electoral system create varying incentives for political parties and voters, which explain the difference. In Bolivia, even small parties with no chance at winning the mayor s seat can gain a place on the city council, giving incentives to small parties to mobilize their voters, and for voters to bother voting. In Brazil, while small Brazilian parties may gain from placing a candidate in the municipal legislative body, the winner-take-all nature of mayoral races means that small political parties have little to gain from mobilizing voters, and voters have little to gain from casting their ballots. As a further test of this argument, we explore several additional implications. First, if our argument is true, we should also observe differences between Brazil and Bolivia in terms of how the closeness of the election affects the relationship between number of parties and voter turnout because of the difference in the value of each vote. In Brazil we should see a stronger negative relationship between number of parties and voter turnout in municipalities with low levels of political competition. That is, the greater the probability of electoral loss (measured by the size of the margin between the two largest political parties), the more likely it is that political parties conserve their resources by not attempting to mobilize voters, and the more likely that voters will decide to stay home rather than vote for a lost cause. Second, in Bolivia, we expect there to be little difference between competitive and uncompetitive races, since even under conditions of low competition, smaller parties can gain some degree of representation or a role in selecting the next mayor. To test these secondary hypotheses, we estimate our main model with an interaction for political competition. If there was a larger than 3% difference between the two largest parties, the race was considered uncompetitive, if there was a smaller than 3% margin, the race was 19

21 considered competitive. 9 Table 3 shows the results for Bolivia (model 1) and Brazil (model 2). For easier interpretation, we have also graphed the predicted values of voter turnout against the number of political parties in competitive and uncompetitive municipalities for Brazil (see Figure 3) and for Bolivia (see Figure 4). As expected, the negative relationship between number of parties and turnout in Brazil is significantly stronger in uncompetitive races. In Bolivia, on the other hand, there is very little difference between competitive and uncompetitive races: voter turnout increases with the number of political parties in both types of municipalities. 7. CONCLUSIONS This paper explores the influence of the number of parties in an election on voter turnout in two very different local electoral systems. We argue that the effect of the number of parties on turnout must be understood in the context of the electoral rules. Where voters have a wider range of choices combined with a good chance their favored party will win some representation, we expect an increase in turnout. However, where voters have a wider range of choices, but only one party can win, we expect a decrease in turnout. Our test of these hypotheses in the context of municipal elections in Bolivia and Brazil yields promising results. We find that under electoral systems that resemble parliamentary systems (party list voting, proportional representation on the city council, and the mayor chosen by the plurality party) an increase in the number of parties increases voter turnout. However, under an electoral system that resembles presidential systems (with direct plurality voting for the mayor) an increase in the number of parties decreases voter turnout. 9 We use a 3 percent threshold for competitiveness (Nardulli 2005), but our results are robust to different thresholds (see Appendix A). 20

22 In Bolivia, the implication is that a greater number of parties at the local level has a positive effect on turnout. We can see evidence of this in the four-point boost in turnout between the 2000 and 2004 municipal elections that followed a change in electoral rules to allow many more local political groups to organize and run in local elections. Since 2004, however, the MAS party has consolidated power at all levels of politics, and the majority of the older establishment political parties have collapsed. If this trend continues, and it results in fewer parties running in local elections, we may see a decline in voter turnout. In Brazil, the implication is that a greater number of parties at the local level has no effect on turnout. A great deal has been made of Brazil s party system and the impact it has on representation, accountability, and democracy itself (Mainwaring 1995). These results indicate that at the local level, the nature of the party system (as measured by the number of parties), has no perceptible impact on turnout. This project contributes to several different debates. First, it continues in the tradition of seeking to understand variation in voter turnout and offers an explanation for the previously inconclusive findings on the relationship between number of parties and voter turnout. Since, as we argue, the number of parties can have opposite effects under different electoral rules, it makes sense that aggregate studies have been contradictory. Second, we add to the growing body of literature that seeks to explain turnout in developing countries, moving away from the bias of testing these relationships primarily in the context of Europe and North America. Third, this project helps us understand the dynamics of municipal level elections as being responsive to similar institutional conditions that govern national elections. Finally, we hope to have isolated an important dimension upon which important electoral institutions are contingent in their impact on political participation. The evidence presented above suggests that the incentives voters and 21

23 parties face in terms of the values they place on each individual vote alters that impact that the number of parties can have on participation. It is less clear how the incentives created at the local level by electoral formula and the number of parties are related to national level elections and political parties. We note that overall higher levels of turnout in Brazil may be related to electoral rules at the national level (as well as things like level of development and enforcement of compulsory voting laws). Our research design, however, allows us to hold constant those inter-country effects. By estimating models at the sub-national level in each country, we can hold that variable constant and examine the relationship between number of parties and turnout. While the differences that exist between Brazil and Bolivia could lead to a shift in the intercept, since the local elections are nonconcurrent with the national elections, we can think of no obvious reasons why the national electoral system would influence the functional relationship between the number of parties and turnout. At the very least, since local elections are non-concurrent with national elections, the national election s impact on turnout should be equally muted in both countries. Understanding how local and national factors interact to shape political participation is, however, a promising agenda for future research. 22

24 References Blais, André, and Kees Aarts Electoral Systems and Turnout. Acta Politica 41(2): Blais, and Dobrzynska Turnout in electoral democracies. European Journal of Political Research 33, no. 2 (March 1): doi: /a: Blais, Andre To Vote or Not to Vote?: The Merits and Limits of Rational Choice Theory. University of Pittsburgh Press. Blais, Andre, and R.K. Carty Does proportional representation foster voter turnout? European Journal of Political Research 18, no. 2 (March): doi:doi: /j tb00227.x. Blais, André What Affects Voter Turnout. Annual Review of Political Science 9: Brockington, David The Paradox of Proportional Representation: The Effect of Party Systems and Coalitions on Individuals' Electoral Participation. Political Studies 52, no. 3: Chhibber, Pradeep, and Irfan Nooruddin Do Party Systems Count?: The Number of Parties and Government Performance in the Indian States. Comparative Political Studies 37, no. 2 (March 1): doi: / CNE Código electoral. Normas Electorales: Corte Nacional Electoral. Cox, Gary W Closeness and Turnout: a Methodological Note. The Journal of Politics 50(03): Cox, Gary W Making votes count. Cambridge University Press. Cox, Gary W., Frances M. Rosenbluth, and Michael F. Thies Mobilization, Social Networks, and Turnout: Evidence from Japan. World Politics 50(03): Dettrey, Bryan J., and Leslie A. Schwindt-Bayer Voter Turnout in Presidential Democracies. Comparative Political Studies 42, no. 10 (October 1): doi: / Downs, Anthony An Economic Theory of Political Action in a Democracy. The Journal of Political Economy 65, no. 2 (April): doi: / Duverger, Maurice Political Parties: Their Organization and Activity in the Modern State. Taylor & Francis. Fornos, Carolina A., Timothy J. Power, and James C. Garand Explaining Voter Turnout in Latin America, 1980 to Comparative Political Studies 37, no. 8 (October 1): Franklin, Mark N Voter turnout and the dynamics of electoral competition in established democracies since Cambridge University Press. Grofman, Bernard, and Peter Selb Turnout and the (effective) Number of Parties at the National and District Levels: A Puzzle-Solving Approach. Party Politics 17(1): Jackman, Robert W Political Institutions and Voter Turnout in the Industrial Democracies. The American Political Science Review 81, no. 2 (June): Jackman, Robert W., and Ross A. Miller Voter Turnout in the Industrial Democracies during the 1980s. Comparative Political Studies 27, no. 4: doi: / Kostadinova, Tatiana, and Timothy J. Power Does Democratization Depress 23

25 Participation? Voter Turnout in the Latin American and Eastern European Transitional Democracies. Political Research Quarterly 60, no. 3 (September 1): doi: / Kuenzi, Michelle, and Gina M. S. Lambright Voter Turnout in Africa's Multiparty Regimes. Comparative Political Studies 40, no. 6 (June 1): doi: / Laakso, Markku, and Rein Taagepera The "Effective" Number of Parties: A Measure with Applicaiton to West Europe. Comparative Political Studies 12, no. 1: 3. Lehoucq, Fabrice, and David L. Wall Explaining voter turnout rates in new democracies: Guatemala. Electoral Studies 23, no. 3 (September): doi: /s (03) Mainwaring, Scott Brazil: weak parties, feckless democracy. in Mainwaring, S., and TR Scully eds. Building Democratic Institutions: Party Systems in Latin America. Stanford University Press. Norris, Pippa Democratic Phoenix: Reinventing Political Activism. Cambridge University Press, September 2. Oliver, JE City Size and civic involvement in metropolitan America. American Political Science Review 94: Perez-Linan, Anibal Neoinstitutional accounts of voter turnout: moving beyond industrial democracies. Electoral Studies 20, no. 2 (June): doi: /s (00) Powell, G. Bingham Voting Turnout in thirty democracies. In Electoral Participation, ed. R. Rose, Sage Publications Contemporary Democracies: Participation, Stability, and Violence. Harvard University Press. Radcliff, Benjamin, and Patricia Davis Labor Organization and Electoral Participation in Industrial Democracies. American Journal of Political Science 44, no. 1 (January): Sakurai, S. N., & Menezes-Filho, N. A. (2008). Fiscal policy and reelection in Brazilian municipalities. Public Choice, 137(1-2), doi: /s Wolfinger, Raymond E., and Steven J. Rosenstone Who Votes. Yale University Press. 24

26 Table 1: Descriptive Statistics for Brazil and Bolivia Variable N mean sd min max Turnout Tunout Income Per Capita (in US $) Population Rural Population (% of total) Literacy (% of pop above 15) Left Support Margin of Victory Number of Parties Growth in Parties Bolivian Cases Turnout Tunout Income Per Capita (in US $) Population Rural Population (% of total) Literacy (% of pop above 15) Left Support Margin of Victory Number of Parties Growth in Parties