37º Encontro Anual da ANPOCS; ST 04 - Comportamento Político Presidential Coattails in Coalitional Presidentialism

37º Encontro Anual da ANPOCS; ST 04 - Comportamento Político Presidential Coattails in Coalitional Presidentialism André Borges (Universidade de Brasília) Mathieu Turgeon (Universidade de Brasília) 1

Past research on presidential elections in the U.S. and elsewhere have shown that candidates for Congress or state legislature sharing a party affiliation with a popular presidential or gubernatorial candidate generally benefit from the latter s success. The comparative literature has argued that presidential elections shape the legislative party system whenever there is a coattails effect in that the presidential candidate pulls with him other candidates from the same party label competing at lower level elections. One of the central claims of research on presidential coattails is that the direct election of the chief executive is likely to have a reductive effect on party fragmentation and help nationalize elections under certain conditions are met. Precisely, the impact of the presidential race on the party system should be highest when elections for the executive and legislative branches of government are held concurrently or in close temporal proximity to one another (Ferenjohn and Calvert 1984, Shugart and Carey 1992, Jones 1994, Shugart 1995, Golder 2006, Hickens and Stoll 2011). Although there is plenty of evidence on the systemic effects of presidential coattails, especially in what concerns party system fragmentation, much less is known about the mechanisms that induce voters, political parties and candidates to establish a link between executive and legislative elections. There is also a lack of research and theorization on the effect of presidential elections on legislative parties under distinct electoral and party rules. In this paper, we intend to fill part of this gap by exploring presidential coattails in a multiparty context, in supplement of most of the work done on two-party systems. Specifically, we examine coattails effects in coalitional presidentialism. By coalitional presidentialism, we mean a model of presidential governance in which presidents must seek the support of at least another party through the division of cabinet positions to obtain a majority in Congress (Chasquetti 2001, Cheibub, Przeworski and Saiegh 2004, Colomer and Negretto 2005, Amorim Neto 2006). This type of multiparty presidential government is frequently observed in several Latin American countries and deserves closer examination. The paper focuses on electoral instead of governing coalitions in presidentialism. Despite the fact that the governing and electoral arenas are related in presidential systems, the specificity of coalition dynamics in presidentialism suggests it is necessary to make an analytical distinction between these two types of coalitions. In parliamentary systems, government coalitions are formed after the election and are binding, as government survival depends on parties support. By contrast, party coalitions in 2

presidential systems often take place before the election and they are not binding. Obtaining support of various parties during the presidential election does not guarantee legislative support in the post-election period in the hypothesis of a candidate being elected. By the same reasoning, a party may support the government in Congress during a president s term and yet decide not to join the electoral coalition in the following election. This implies that governing coalitions sometimes differ markedly from electoral coalitions in presidential systems (Mainwaring and Shugart 1997). The argument developed in this paper applies mainly to presidential regimes in which presidential candidates face strong incentives to form a multiparty electoral coalition to secure the presidency. Even though a governing coalition may be formed regardless of the fact that the elected president previously formed an electoral coalition, the comparative evidence suggests that electoral coalitions are not uncommon in multiparty presidential regimes (Chasquetti 2001). This is not surprising considering that in a fragmented party system even the largest parties may be unable to mobilize alone the majority required to elect the president. The functioning of multiparty coalitions in the governing and electoral arenas has profound implications for the interaction between presidential and legislative races. First, because the presidency is the major electoral prize in presidential systems, legislative candidates have an incentive to organize their campaigns around their party s presidential candidate. Second, voters also recognize the overwhelming importance of the presidency relative to other political offices and, for that reason, they typically rely on the party of their presidential candidate as an information shortcut to help them decide how to vote in legislative elections(shugart, et al. 1992, Shugart 1995, Golder 2006). These two factors reinforce the connection between presidential and legislative races. But, what is the impact on presidential coattails? We argue that presidential coattails in coalitional presidentialism should benefit mostly the electoral coalition parties, and this, frequently at the expense of the president's party. In presidential elections, candidates must mobilize and win the vote of a majority of the national electorate. Parties filing presidential candidates thus have strong incentives to adopt a vote-maximizing strategy, even if this implies weakening the parties' performance in other electoral contests (Samuels 2002, Samuels and Shugart 2010). In multiparty settings, the trade-off between strengthening the "executive" or the "legislative" party branch is all the more evident, because parties must construct broad 3

coalitions to succeed in securing the highest electoral prize, that is, the presidency. To forge a winning coalition, parties of presidential candidates negotiate support for their candidate by supporting, in turn, candidates from other parties in other electoral contests including, most notably, national legislative and gubernatorial elections.. This electoral strategy to win the presidency has a high cost for parties filing presidential candidates. In this paper, we examine how this strategy affects presidential coattails in coalitional presidentialism. The summary of our argument is that in coalitional presidentialism, parties filing presidential candidates have strong incentives to make concessions to electoral coalition parties in lower level elections, resulting in a diffuse coattails effect where coalition partners are most benefitted. We analyze and test the hypothesis that presidential coattails in coalitional presidentialism is most benefecial to the electoral coalition parties than the president's own party in lower level elections by examining municipal level data for Brazilian presidential and legislative elections in the years 1998, 2006 and 2010. In all these years, the incumbent party and/or president was victorious in the presidential contest. Yet, as we show later in this section, the good showing of the incumbent candidate did not necessarily bring substantial gains to the president's party in lower chamber national elections. To the contrary, the findings show that the president's electoral coalition parties benefited more from presidential coattails than the president's own party in lower chamber national elections. In the following sections, we provide a discussion of coattails effects in presidential system with a particular attention given to their effects in coalitional presidentialism. We then discuss the case of Brazil and present evidence that presidential coattails have distinct effect in coalitional presidentialism. The paper concludes with brief remarks about the implications of these findings. Presidential Coattails Effects in Coalitional Presidentialism Traditional arguments about presidential coattails rest on the assumption that regardless of the electoral rules employed to allocate legislative seats and independent of the existence or not of coalitions disputing the presidency, the coattails effect produces the greatest benefit for the president s party (Shugart and Carey, 1992; Hickens and Stoll, 2011). We challenge the claim that the coattails effect is independent of the existence or not of electoral coalitions in presidential contests. To the contrary, we argue that the 4

coattails effect should work differently in coalitional presidentialism. More precisely, presidential coattails should benefit most coalition partners because the imperatives of coalition-making and sustaining force presidential candidates and their co-partisans to make concessions to allied parties in lower-level elections in exchange for their support in the presidential race. In the end, presidential coattails should benefit mostly coalition partners. Below we review the relevant literature on presidential coattails and develop hypotheses about their effect in coalitional presidentialism. Presidential coattails have two main determinants. First, in a presidential system the national executive is the major electoral prize. The singularity and importance of the presidential election implies that presidential candidates receive the lion's share of campaign finance and national media attention. Legislative candidates have an incentive to organize their campaigns around their party's presidential candidate in the hope of benefiting from his or her organizational, financial and media advantages. Second, voters also recognize the overwhelming importance of the presidency relative to other political offices in presidential regimes, and they typically pay more attention to the presidential race. They rely on the party of their presidential candidate as an information shortcut to help them decide how to vote in legislative elections (Golder, 2006). Because voters know that the president will need legislative support, they have an incentive to vote for the legislative ticket affiliated with their preferred presidential candidate. Recognizing the centrality of the presidential election in voters choices, legislative candidates have further incentives to coordinate their own campaigns with their party's presidential party (or coalition).(shugart 1995, Samuels 2002, Golder 2006) Research on the impact of presidential elections on party systems argues that the direct election of the national executive promotes, under certain conditions, less fragmented party systems. The dispute for the presidency can be understood as an election for a nationwide district with a single seat (Samuels, 2002). Under plurality rule, Duverger s law should prevail, deflating the number of effective presidential candidates. When congressional elections are concurrent with presidential elections, one should observe stronger coattails and hence, a lower number of effective legislative parties (Shugart, et al. 1992, Golder 2006). That is, concurrent (or close) presidential and legislative elections entail voters to support more strongly presidential candidates' parties, as opposed to those not affiliated with any of the presidential front-runners, and 5

encourage, in turn, legislative candidates to coalesce around the leading presidential candidates (Hickens, et al. 2011). The reductive or deflationary effect of concurrent (or quasi-concurrent) presidential and legislative elections should be strongest in countries that employ PR rules to allocated legislative seats (Golder, 2006). Shugart and Carey (1992) suggest that the presidential coattails effect under the combination of plurality and PR rules for executive and legislative elections, respectively, weakens even more multipartyism. But, when PR electoral rules are combined with runoff elections, the deflationary effect of the presidential contest on the number of presidential candidates and parties does not hold. Runoff systems usually produce a higher effective number of presidential candidates because the incentives for strategic voting are weaker than in plurality systems. In contrast to plurality elections, parties may have an incentive to enter the presidential race even when the likelihood of winning the presidency is very low. This is because they know that a good performance in the first round can leave them in a strong bargaining position to obtain concessions from one of the two frontrunners in the second round in exchange for their support. Specifically, parties will enter the presidential race with the primary aim of increasing their legislative representation, regardless of their chances of winning the presidency (Shugart, et al. 1992, Jones 1994, Shugart 1995, Golder 2006). Research evidence reveals indeed that there is greater legislative fragmentation in countries using runoffs for presidential elections than in countries using plurality rule (Golder, 2006; Jones, 1994). When there are two large parties that obtain most of the national vote in both legislative and presidential elections, it is reasonable to assume that the parties of the main presidential contenders benefits most from coattails, and this mechanism, in turn, results in lower fragmentation. The parties that mobilize the lion s share of the national vote should increase their representation in the lower chamber at the expense of the other parties that do not compete in the presidential election or do so with weaker presidential candidates. However, this is not necessarily true in multiparty, coalitional presidentialism, mainly due to the impact of the separation of powers on party strategy and due to the dynamics of coalition making and maintaining In a presidential system the parties with actual chances of winning have strong incentives to privilege the presidential contest to the detriment of the legislative election. Besides, the majoritarian nature of the presidential election means that 6

presidential candidates must appeal to the median voter, regardless of the position of their own parties in the ideological spectrum (Samuels 2002). From this, one can infer that, even if voters cast votes on a partisan basis and support the parties of their preferred presidential coalition in the presidential election, it is not certain that they should most often prefer the president s party rather than some other coalition party in the legislative contest. If the main presidential contenders campaign on a policy agenda that represents their coalition rather than their own party as it is often the case, voters may support party A in the legislative election because they identify with it in terms of ideology or policy issues, and vote for a coalition that includes party A in the presidential election but which is led by another party. The key point here is that in multiparty, coalitional presidentialism, it is often the case that no political party, including that of the leading presidential contender, have neither a majority of the legislative seats, nor a party organization capable of mobilizing alone a majority of the national vote. This implies that forming a strong and broad party coalition is strategic to win the presidency. Given the imperatives of coalition-making, the parties competing for the presidency cannot afford alienating their potential allies by running a legislative campaign devised to steal votes from other parties. That is, coalitional presidentialism exacerbates the trade-off between the objectives of winning the presidency and maximizing the legislative vote of the party, as long as the forging of a victorious coalition in the presidential election will require making concessions to allied parties in the lower level elections, including the legislative contest. In many instances, coalition partners will help the party running for the presidency to reach ideological niches and regions where it had no appeal or lacked a strong local organization previous to the election. To secure the support of coalition partners, a presidential candidate must campaign on behalf of the whole coalition and take care that his co-partisans do not to invade the electoral markets where the allies are strong in legislative elections. The point we want to make is that there is a crucial difference in the dynamics of presidential coattails in two-party and multiparty settings with the presence of coalitions. In coalitional presidentialism, some of the parties that are allied in the presidential election are adversaries in the legislative contest. In a two-party, majoritarian system such as the U.S., there are two parties that compete against each other in both presidential and legislative elections. In other words, in coalitional 7

presidentialism politicians form different teams to dispute presidential and legislative elections, whereas in two-party presidentialism there are only two teams disputing both elections. In a two-party system the strength of the association between the presidential and the legislative performance of parties depends mostly on whether electoral rules employed in legislative elections promote nationalizing or localizing effects. Electoral systems that create an incentive for a personal vote may promote ticket splitting, as long as legislators are able to establish a connection with voters that is independent from the national reputation of their party. On the other hand, in a political system with nationalized and centralized parties, with electoral rules that leave little room for localoriented campaign strategies, choices tend to be purely partisan and motivated by national policy (Shugart 1995, Chhibber, Kollman and ebrary Inc. 2004). In a majority, bipartisan system there would be no ticket-splitting, as legislative candidates would have strong incentives to make campaign on the same national policies advocated by their party s presidential candidate. In coalitional presidentialism, voters are faced with three choices in legislative elections in that they may cast a vote for the party of their preferred presidential candidate, they can support another member of the coalition they voted for in presidential elections or they may vote for a party that is not in the coalition. Even if electoral rules are nationalizing, there is no reason to expect voters to prefer the first choice to the second. Because a presidential candidate leading a broad coalition cannot be expected to privilege his own party s agenda and, also, he/she has strong incentives to adopt a vote-maximizing strategy to appeal to voters of the other parties that are part of the coalition, it is reasonable to expect voting choices to follow a coalition logic. For the reasons explained above, even if voters and candidates are mainly concerned about national policies and issues, the coattails effect should benefit the whole coalition, and not only the president s party, a situation we define as a diffuse coattails effect. Brazil Coalitional Presidentialism Brazil is a good test case for our theoretical argument for two main reasons. First, it is a widely studied case of coalitional presidentialism, in which multiparty coalitions play a fundamental role in the governing and electoral arena (references). Second, the country's electoral system and federal institutions have led analysts to conclude that 8

presidential coattails are either absent or irrelevant due to gubernatorial coattails (Samuels 2003, Ames, Baker and Renno 2009). We present evidence that that there is a coattails effect, though it follows a coalition logic. This is so despite institutional rules that conspire against the linking of presidential and legislative elections, which suggests that the diffuse coattails effect analyzed here may be even stronger in institutional settings less conducive to split-ticket voting. Brazil Electoral Rules and their impact on political behavior Brazil adopts a PR, open-list electoral system in which the states function as multimember districts with magnitudes ranging from 8 to 70. As a consequence of permissive electoral rules party fragmentation has increased since the first multiparty elections held in 1990, reaching a total of ten effective parties in 2010. Open-list PR creates incentives for individualistic strategies of legislative campaigning and a localist orientation of parliamentarians. Candidates to the federal chamber have often succeeded in cultivating a personal vote independent from their party's reputation. Several surveys applied in the Chamber of Deputies indicate that for the majority of incumbent parliamentarians success in the electoral arena depends mostly on their personal efforts to obtain funds and mobilize voters, rather than on the party's structure and organization. Incumbent candidates' ability to bring pork to their electoral bailiwicks has been shown to be an important predictor of reelection, even though other factors, such as presidential coattails have a significant impact on reelection rates as well (Ames 2001, Pereira and Rennó 2001, 2007). Survey research on voters behavior in legislative elections in two Brazilian cities has demonstrated that most citizens prefer candidates preoccupied with local issues, rather than national ones. However, a substantial percentage of voters - around one-third - make their choices based on candidates' ideology or policy positions (Ames, Baker and Rennó 2008). These results are congruent with the diverse set of strategies of electoral mobilization adopted by candidates to the lower chamber. Whereas some candidates develop strategies based on the targeting of pork-barrel to territorially delimited constituencies, others prefer to make campaign on their party's programme or by taking position on issues or policies - such as the environment - that allow them to mobilize constituencies dispersed across a state's territory (Carvalho 2003). Levels of partisanship among voters remain low notwithstanding the gradual institutionalization of the national party system in the recent democratic period. The 9

leftist Worker's Party (PT) is currently the only Brazilian political organization that achieves significant levels of party identification among voters: a 2007 survey indicated that 20% of Brazilian voters were PT sympathizers (Samuels 2008). By comparison, the Brazilian Social Democratic Party (PSDB) reached only 6% of party identification in that same year, even though the party governed Brazil during eight years before the PT came to power (1995-2002). Differences in levels of partisanship reflect, to a great extent, the distinctive organizational structure of national parties. The PT and a few other left-wing organizations most notably, the Brazilian Communist Party (PC do B) emerged as externally mobilized parties, with a strong national leadership and a coherent party program, but they are an exception to the general rule of weak party organization. On the other hand, the major national parties have distinctive ideological profiles and behave accordingly in Congress, which suggests that ideology does play a role in national legislative races (Figueiredo and Limongi 1999, Power 2008, Hagopian, Gervasoni and Moraes 2009). Arguably, ideological mobilization and the reliance on pork and local networks are not necessarily mutually excluding strategies, as different types of voters may respond differently to each of these political incentives (Kitschelt and Wilkinson 2007). Since the return to democracy, Brazilian presidents have been elected by majorityrunoff. The frontrunner candidate must obtain over 50% of the national vote in the first round to win the presidency; otherwise, the two leading candidates dispute a runoff election. Despite theoretical expectations and empirical evidence on the inflationary effect of majority-runoff on party fragmentation (Shugart, et al. 1992, Golder 2006), the number of effective candidates has varied from 2 to 3 in the recent democratic period, with the exception of the first post-transition election for president held in 1989. This indicates that the least competitive parties have preferred to join a coalition instead of launching a candidate with dim chances of wining the presidency. [Figure 1 about here] Two major parties - the left-wing PT and the centre PSDB - have polarized presidential elections, obtaining the bulk of the national vote and electing all presidents since 1994. The PSDB has led a centre-right coalition formed most often by Brazil's largest right-wing party the DEM (Democrats) and other smaller conservative organizations. The PSDB's emergence as a major player in national politics coincided 10

with the election of then Minister of Finance, Fernando Henrique Cardoso, to the presidency in 1994. The PT has become the leader of the Brazilian left bloc following the decay of the Democratic Labour Party (PDT) in national elections (Melo 2006)). The Brazilian Communist Party (PC do B) and the Brazilian Socialist Party (PSB) have supported all PT presidential candidates since 1994, with the exception of Lula's candidacy in 2002 1. Since Lula's victory in the 2002 presidential election, the PT has become increasingly pragmatic in what concerns electoral strategy, allying to right-wing parties such as the PR (Republican Party) in an effort to attract moderate voters. Presidential and national legislative elections have been held concurrently since 1994. All gubernatorial and subnational legislative elections have also coincided with the presidential contest, with exception of the 1990 election, when all legislative elections (national and subnational) coincided with the election of state governors. More recently, with the approval of the reelection law in 1998, presidents gained an important mechanism to enable the construction and consolidation of electoral coalitions. Because a popular president is a natural candidate, with strong advantages vis-à-vis other potential competitors - presidents may take advantage of their powers over policy and patronage to mobilize a personal following and secure party politicians' support and loyalty -, he is less likely to be challenged by co-partisans or by allied parties, which in turn contributes for coalition unity. All else being equal, concurrent elections and the possibility of reelection should strengthen presidential coattails, increasing the impact of the dispute for the national executive on the electoral fortunes of candidates to the lower chamber. One of the few studies on presidential coattails in Brazilian elections, by Pereira and Rennó (2007), found that a deputy's alignment to the president increased likelihood of reelection to the Chamber of Deputies in 1998, when president Cardoso disputed a second term. The role of federalism and subnational alliances in Brazil Under Brazil s federal arrangements, subnational electoral alliances play a key role in national coalition-making. Because subnational party organization and electoral strategy matters for parties performance in national legislative elections, independent 1 In 2002 the PSB decided to dispute the presidency against the PT. The PSB candidate, Anthony Garotinho, ranked third in the first round, with roughly 18% of the national vote. 11

of presidential coattails, a viable candidacy to the presidency cannot avoid engaging in a complex game of intra-coalition coordination at the state level. Following the 1988 Constitution, governors and mayors were empowered with substantial policy responsibilities and financial resources. Governors emerged as national power brokers, as they took advantage of fiscal decentralization to strengthen regional party machines, loosely controlled by national party leaders. State Executives have the power to implement public policies in areas such as health, education and infra-structure, they exert control over the nomination of thousands of bureaucratic posts and over the allocation of investments within state s territory. More recently, following the stabilization of the economy and a series of reforms of fiscal federalism in the 1990s and 2000s, state governors have lost significant powers to tax, spend and borrow (Almeida, 2005; Arretche, 2007). Despite the federal government s successful moves to recentralize policy authority and resources, state governments remain important actors in national electoral and party processes. Electoral rules reinforce governors influence in national politics as several important decisions relative to elections and party organization such as party primaries to select candidates to all relevant national posts, with the exception of the presidency are taken at the state level. Given sub-national politicians control over patronage and policy, and their influence in party organizations at the national and subnational levels, candidates to the federal chamber have strong incentives to associate their campaigns with the gubernatorial race. (Abrucio, 1998; Samuels, 2003). Not different from other federal countries, gubernatorial coattails play a significant role in national legislative elections. This is not to say that presidential coattails do not matter. In Brazilian presidentialism, the national executive is an extremely powerful institution, whose capacity to formulate macroeconomic and social policies, distribute budget resources and nominate thousands of political appointees (Amorim Neto, 2007) greatly affects the career prospects of subnational officials. Local and state politicians have strong incentives to seek presidents support, which implies that national and sub-national elections are interdependent. Recent work by Limongi and Cortez (2010) reveals that gubernatorial elections have become increasingly nationalized as the main presidential contenders, the PT and the PSDB and their most loyal coalition partners have greatly increased their share of the national vote since the early 1990s. Among the largest parties in gubernatorial elections, the PSB and the PFL have adopted a durable alliance with the PT and the 12

PSDB respectively, and these four parties, with few exceptions, have competed for the state executive as two clearly delimited blocs. Between 1990 and 2010 the share of the national vote held by the PT-PSB and the PSDB-PFL blocs in gubernatorial elections increased from 32.5% to 65%. These four parties elected 21 out of 27 governors in 2010, as compared to only nine in 1990 (Limongi and Cortez, 2010, p. 32). The PMDB is the only party not consistently aligned with one of the two presidential blocs that has remained competitive in the subntional arena. The party elected five governors and obtained 18.2% of the national vote in 2010. Different from the PSB and the PFL, the PMDB has adopted an erratic strategy in what concerns national electoral coalitions: in 2002 it supported the PSDB, and in 2010, the PT. The party decided to join president Lula s cabinet in 2004, even though it had integrated the electoral coalition of José Serra, who was defeated by Lula in the runoff elections 2. Because gubernatorial elections matter for the national legislative dispute, the PT and the PSDB have sought to subordinate subnational electoral strategies to the imperatives of coalition making at the national level. Table 1 below shows the number of times the PSDB and the PT participated of gubernatorial elections either by fielding a candidate or by supporting another party, and the number of gubernatorial candidacies, from 1994 to 2010. [Table 1 about here] With the exception of the PT in 2002, in all the other cases the number of participations in gubernatorial elections was greater than the number of gubernatorial candidacies. Note that the difference between the two columns equals the number of times the PT and the PSDB decided to support an allied party in gubernatorial elections. In 2010, for instance, the PT participated of gubernatorial elections in all 27 states, but it led a gubernatorial coalition in only 10 states. In the remaining 17 cases, with a single exception, the PT supported one of the parties that integrated the electoral coalition 2 The PMDB disputed presidential elections in 1989 and 1994, with disappointing results. In 1998 the party leaders decided not to support a presidential candidate to obtain maximum autonomy to adapt electoral coalition strategies to local context. In 2006, the party supported Lula's re-election but did not join the coalition. Also, the PMDB has been much less keen to subordinate its subnational electoral strategies to national coalition pacts as compared to the PFL and the PSB. 13

assembled by presidential candidate Dilma Roussef 3. Because Roussef assembled a large coalition that included nine parties, four of which controlled governorships throughout the country, the PT had to sacrifice its gubernatorial ambitions to support coalition allies. What are the implications of subnational coalition dynamics for presidential coattails? Given our argument on the logic of coattails in multiparty presidentialism we expect intra-coalition coordination in gubernatorial elections to favour allied parties at the expense of the leading coalition parties. Once the PT or the PSDB decide to support an allied party rather than launching a gubernatorial candidacy, this implies that presidential candidates will have to campaign on behalf of a subnational ticket led by another party. Also, by supporting allied parties where they count on a strong local organization or a competitive gubernatorial candidate or both, presidential candidates allow these organizations to secure and even strengthen their local strongholds, yet at the same time reducing the prospects of electoral growth of their co-partisans in legislative elections in these districts. This is especially so because in Brazilian federalism candidates to the federal chamber depend on subnational party organization and on subnational alliances to connect national and subnational races. Local partisans provide not only resources and access to local networks, but they also provide important informational shortcuts to voters, facilitating the translation of national politics into local context. Given the integration of national and subnational party alliances, those parties that are competitive in gubernatorial elections in some states but lack the national organization required to win the presidency may benefit from coattails by strategically linking their gubernatorial and national legislative candidacies to one of the presidential frontrunners. The end result of the optimum strategies pursued by presidential candidates and their parties, and coalition parties, is a diffuse coattails effect that should benefit all the coalition and not only or mostly the leading party. Data To evaluate the hypothesis that presidential coattails are most beneficial for coalition partners in lower level legislative elections than for the president's own party in 3 In the state of Roraima, the PT supported Neudo Campos of the PP (Popular Party). The PP did not join Roussef s electoral coalition, though it did integrate the legislative coalition led by the PT in the Chamber of Deputies. 14

coalitional presidentialism, we gathered data from the three Brazilian presidential and lower chamber elections where the incumbent presidential candidate was successfully reelected. These are the 1998, 2006 and 2010 elections. In 1988, Fernando Henrique Cardoso (PSDB) was successfully reelected in the first round. In 2006, Lula (PT) reelected himself and in 2010, Lula's candidate, Dilma Roussef, also managed to keep the presidency in the hands of the PT for a third consecutive term. The electoral coalitions for all three contests were different each time. In the 1998 Fernando Henrique Cardoso reelection, the PSDB built an electoral coalition with four other parties (PFL, PTB, PPB and PSD). In 2006, Lula's presidential bid benefitted from the electoral support of three coalition partners (PSB, PRB and PC do B). 4 And finally, in 2010, the PT built a coalition with nine other parties (PC do B, PDT, PSB, PR, PMDB, PRB, PTC, PTN and PSC) to help Dilma Roussef elect herself as Brazil's first women president 5. Our interest lies in examining how much the presidential vote for the incumbent presidential candidate matters for the president's own party and that of its coalition partners in lower-level elections. To evaluate this presidential coattails effects, we disaggregated the electoral results for both presidential and lower chamber elections at the municipal level. Brazil has over 5,500 municipalities spread across its five regions: North, Northeast, Midwest, Southeast and South. In addition, we gathered other election results, also at the municipal level, for gubernatorial contests because governors also exert coattails effects of their own. Information about the party of the mayors during those elections was also tallied. Election results for all three years were collected from the IPEA 6 (Instituto de Pesquisa Econômica Aplicada), the CEM 7 (Centro de Estudo da Metrópole) and the TSE (Tribunal Superior Eleitoral) 8. For each election, we grouped the electoral results for all electoral coalition partners (lower chamber and gubernatorial contests) and stacked the data for the three election years. This leaves us with lower chamber and gubernatorial elections results for 4 In some parts of Brazil, the PMDB also supported Lula, but its support was not officially backed from the national party. The PMDB remained divided and some important regional sections campaigned against Lula and the PT. Even though the PSB did not formally join the PT's coalition as well, the party maintained its strategy of aligning with the PT and its presidential candidate in most states, reason why we considered it part of the coalition. 5 We excluded from our analysis the coalition parties that received a very small percentage of the national vote (below 2%). These are the PRB, the PSD, the PTC, the PTN and the PSC. 6 www.ipeadata.gov.br 7 http://www.fflch.usp.br/centrodametropole/ 8 http://www.tse.jus.br/eleicoes/estatisticas 15

both the president's party and his or her electoral coalition partners at the municipal level for three election cycles. Thus our analysis is not dependent upon the specificities of each election. Rather, it focuses on the dynamics of presidential coattails effects in coalitional presidentialism independent of parties. Results As a first test to evaluate the hypothesis that presidential coattails effects are most beneficial for electoral coalition partners in coalitional presidentialism, we present in Table 2 correlations between the president's vote and that of her party and her coalition partners in lower chamber elections. The findings are quite telling. Presidential votes correlated a lot more with lower chamber votes for coalition partners than with the president's own party. Specifically, presidential votes correlates only very weakly with the president's party lower chamber votes (.04). It correlates nearly six times more with lower chamber votes of coalition partners (.22). A year-by-year analysis reveals that presidential votes always correlate more strongly with lower chamber votes for coalition partners. [Table 2 about here] The results presented in Table 2 are supportive of the hypothesis that presidential coattails effects are most beneficial for electoral coalition partners. Below, we present are more robust test of this hypothesis by proposing a multivariate analysis of both the president's and her electoral coalition partners lower chamber votes. The first and second independent variables are common for both equations (i.e., that for the president's party and that for her electoral coalition partners). The first independent variable is the president's vote to account for presidential coattails. Our expectation is that this variable will show a positive sign for both equations, but it should exert a stronger effect on the president's electoral coalition partners' lower chamber votes. The second independent variable is a measure of the local economy's dependence on government expenditures. This variable measures the contribution of public administration to the municipality's GDP. According to Zucco (2008), local economies that rely heavily on government activities generally tend to show stronger support for the government incumbent candidates. We expect this variable to positively affect both the president's and her electoral coalition partners lower chamber vote. 16

The model includes three other independent variables, each measured specifically for president's party and her electoral coalition partners. The first of these three variables is the party's (or coalition of parties) lower chamber electoral votes it received in the previous election. The purpose of this variable is to measure the party's (or coalition of parties') past electoral strength. We expect that the president's party (or her electoral coalition partners) will do well where they did well in the previous election. For the 1998 election, for example, we explain the PSDB's lower chamber vote by the vote it received in 1994 for those same elections. As for the coalition of parties, this represents the sum of the lower chamber votes received by the PSDB's electoral coalition partners, that is, the PFL, PTB, and PPB. The same is done for the other two elections. The model also includes a variable that measures the gubernatorial coattails. It has been shown elsewhere that gubernatorial elections exert coattails effects of their own (Samuels, 2003). Thus we include the votes received by the gubernatorial candidates from the president's party (electoral coalition partners') to explain the president's (electoral coalition partners') lower chamber votes. As for presidential coattails, we expect this variable to positively affect both the president's (electoral coalition partners') lower chamber vote. Finally, we include dummy variables to identify municipalities controlled by the president's party or her electoral coalition partners. Mayoral elections take place every four years in between federal elections. We believe that the president's party or her electoral coalition partners benefit from the presence of mayors from their party or coalition of parties. Thus we expect a positive sign for this variable too. Table 3 presents the multivariate regression analysis. The equations were estimated by ordinary least squares. The first column shows the results for the president's party equation and the second column that for the electoral coalition parties. First, we note that the all variables show statistical significance at the.001 level. Also, all variables but the Government Dependence variable show the expected sign. More importantly for present purposes, we find that the Presidential Vote exert a much stronger effect on the electoral coalition parties than on the president's party, even after controlling for many other important determinants of lower chamber votes. The size of the effect of Presidential Vote on the electoral coalition parties is a little more than twice the effect it has on the president's own party. As the president's vote increase by 17

1%, that of the electoral coalition parties lower chamber vote increases by.16% while it only increases by.08% for the president's party. This is strong evidence that presidential coattails matter most for electoral coalition parties in coalitional presidentialism. [Table 3 about here] It is also worth noting that gubernatorial candidates also exert coattails effects of their own. The coattails effects of gubernatorial candidates are stronger than that of the president for both the president's party and her electoral coalition partners. It is also distinctively stronger for electoral coalition parties. As governors' vote from the electoral coalition parties increases by 1%, than of those electoral coalition parties lower chamber vote increases by.24%. On the other hand, when the governors' vote from the president's party increases by 1% that of the president's party lower chamber vote increases by.15%. The president's party and her electoral coalition previous votes are also strong predicators of their current respective lower chamber votes. The effect on the president's and her coalition partners is.51 and.46, respectively. Finally, both the president's party and her electoral coalition partners benefit from controlling the mayoral posts. The effect of both lower chamber votes is similar, providing each with about 8% more vote in those municipalities. Finally, and curiously, municipalities that are more dependent on federal government funds are less like to support the president's party of her electoral coalition partners. Conclusion Comparative research on presidential systems and elections has yet to deal adequately with the distinctive impact of presidential coattails on parties and party systems in multiparty settings. The theoretical argument and hypotheses developed in this article represent an initial effort to understand coattails dynamics in coalitional presidentialism. Following Samuels (2002) and Samuels and Shugart (2010), we start with the assumption that the separation of powers creates strong vote-maximizing incentives, as competition for the highest political prize, the presidency, forces parties to mobilize a national majority of the vote. Given these vote-seeking incentives, presidential candidates have incentives to adopt broader (or less restrictive) policy positions relative to their parties, even if this implies harming their co-partisans electoral prospects in the legislative race. The trade-off between strengthening the "presidential" or the "legislative" party branch is especially acute in coalitional presidentialism, because 18

presidential candidates are often required to forge a broad party coalition to obtain a national majority of the vote. This implies that presidential candidates must make concessions to allied parties in lower level elections, including national legislative races, in exchange for their support in the presidential race. Evidence from Brazilian presidential and national legislative elections confirms our central claim that the requirements of coalition-making and maintaining produce a diffuse coattails effect that benefits mostly coalition parties at the expense of presidential candidates' parties. Presidential coattails estimated for coalition parties only are up to two times larger than the purely "partisan" coattails effect estimated for the president's party. This result holds even in the presence of control variables that include parties' performance in gubernatorial elections and a measure of parties' performance in the previous legislative election. We believe that this is so because presidential candidates have powerful incentives to campaign on behalf of the whole coalition, even if this implies sacrificing their co-partisans ambitions in the legislative race. Future research should explore further the broader implications of the diffuse coattails effect for multiparty presidential systems. There are good reasons to believe that presidential coattails in coalitional presidentialism should maintain or increase party fragmentation in the lower chamber. That is, different from traditional arguments on presidential coattails and the party system, the theoretical argument and empirical evidence presented in this article indicate that the coattails effect, even in the presence of concurrent elections and a low number of effective presidential candidates, may contribute for the survival and growth of smaller parties. Given that coalitional presidentialism is now the modal model of governance in Latin America these sorts of questions surely deserve greater attention from comparativists. 19

References Ames, B. (2001), The Deadlock of Democracy in Brazil, Michigan: The University of Michigan Press: Ann Harbor. Ames, B., Baker, A., and Rennó, L. (2008), "The Quality of Elections in Brazil: Policy, Performance, Pageantry or Pork?," in Democratic Brazil Revisited, eds. P. Kingstone and T. J. Power, Pittsburgh, PA: University of Pittsburgh Press. Ames, B., Baker, A., and Renno, L. R. (2009), "Split-Ticket Voting as the Rule: Voters and Permanent Divided Government in Brazil," Electoral Studies, 28, 8-20. Amorim Neto, O. (2006), "The Presidential Calculus: Executive Policy Making and Cabinet Formation in the Americas," Comparative Political Studies, 39, 415-440. Carvalho, N. R. d. (2003), E No Início Eram as Bases: Geografia Política Do Voto E Comportamento Legislativo No Brasil, Rio de Janeiro: Revan. Chasquetti, D. (2001), "Democracia, Multipartidismo Y Coaliciones En América Latina: Evaluando La Difícil Combinación," in Tipos De Presidencialismos Y Coaliciones Políticas En América Latina, ed. J. Lanzaro, Buenos Aires: CLACSO, pp. 319-357. Cheibub, J. A., Przeworski, A., and Saiegh, S. M. (2004), "Government Coalitions and Legislative Success under Presidentialism and Parliamentarism," British Journal of Poltical Studies, 34, 565-587. Chhibber, P. K., Kollman, K., and ebrary Inc. (2004), "The Formation of National Party Systems Federalism and Party Competition in Canada, Great Britain, India, and the United States." Colomer, J. M., and Negretto, G. L. (2005), "Can Presidentialism Work Like Parliamentarism?," Government & Opposition, 40, 60-89. Ferenjohn, J. A., and Calvert, R. (1984), "Presidential Coattails in Historical Perspective," American Journal of Political Science, 28, 127-146. Figueiredo, A., and Limongi, F. (1999), Executivo E Legislativo Na Nova Ordem Constitucional, Rio de Janeiro: Editoral FGV. Golder, M. (2006), "Presidential Coattails and Legislative Fragmentation," American Journal of Political Science, 50, 34-48. Hagopian, F., Gervasoni, C., and Moraes, J. A. (2009), "From Patronage to Program the Emergence of Party-Oriented Legislators in Brazil," Comparative Political Studies, 42, 360-391. Hickens, A., and Stoll, H. (2011), "Presidents and Parties: How Presidential Elections Shape Coordination in Legislative Elections," Comparative Political Studies, 44, 854-883. 20