This is a guest post by André Borges and Ryan Lloyd based on their recent article in Electoral Studies
The literature on presidential coattails has, until now, focused mainly on the role played by presidential elections in shaping national legislative races. Comparative research has demonstrated that in the presence of a sufficiently low number of candidates running for president, concurrent presidential and lower chamber elections deflate the national party system. Presidentialism, however, is often associated with federal institutions, which should complicate party aggregation by introducing the issue of vertical integration.
In a recent paper (Borges and Lloyd, 2016), we argue that the coattails effect may operate not only “horizontally,” by shaping national legislative elections, but also “vertically,” by shaping elections held at lower levels of government. All else being equal, concurrent national (presidential) and subnational (gubernatorial) elections will foster coordination because parties and voters are aware that the presidency is the most important electoral prize in a presidential regime. No candidate for subnational executive office receives as much media attention or as many campaign contributions as the top presidential contenders do. Voters also recognize the overwhelming importance of the presidency in comparison to other political offices in presidential regimes, and they typically pay more attention to presidential candidates than those running for other offices (Golder, 2006).
Our central claim is that the congruence between national and subnational elections increases when elections are temporally proximate and the effective number of presidential candidates is sufficiently low. On the one hand, parties running for president have strong incentives to coordinate strategies between national and subnational electoral arenas because they are required to mobilize a national majority of the vote to win, which in turn requires coordination with local candidates. On the other hand, parties that lack viable presidential candidates will respond strategically to the deflation of the presidential party system by coalescing around one of the major presidential contenders because supporting third candidates (or not participating in the presidential election) may cost them votes in subnational races.
As coordination efforts are repeated over time and national party divisions are successfully reproduced at the subnational level, voters should respond accordingly and make congruent choices in national and regional elections. This is especially true because party coordination provides an external cue for voters in subnational elections. Given that voting is an information- and time-intensive activity, voters are likely to rely on national policies and national party dynamics as a cognitive heuristic for making decisions about subnational elections (Rodden and Wibbels, 2011). Our second hypothesis is ,therefore, that a presidential coattails effect should exist at the individual-voter level when coordination is effective and leads to vertical party linkage.
We evaluate these hypotheses using district-level data from Brazilian gubernatorial and presidential elections from 1945 to 2010. We complement our time-series cross-sectional (TSCS) analysis by running a series of logit regressions on survey data in order to assess the effects of presidential coattails on Brazilian gubernatorial elections. Our logit regressions use surveys from two electoral periods that were characterized by distinct levels of presidential party fragmentation: 2002 and 2010.
Brazil is an ideal case study for analyzing the effects of presidential elections and federalism on party linkage between levels of government because rules governing presidential and gubernatorial elections were changed relatively recently. From 1945 to 1962, presidential and gubernatorial elections concurred on only a few occasions, and not in all states. Lower-chamber elections concurred at the same time as presidential elections in 1945 and 1950, but not for the two elections immediately afterwards (1955 and 1960). In contrast, all elections to national and state-level posts have occurred concurrently since 1994, thereby greatly increasing the stakes of the presidential race. Given that the major traits of Brazil’s political system (presidentialism, federalism, electoral system and legislation on political parties) have mostly remained constant across these two periods, Brazil’s case allows us to test our first hypothesis with a quasi-experimental design.
Our empirical findings indicate that concurrent elections have a negative effect on dissimilarity as long as the effective number of presidential candidates is sufficiently low. Party system incongruence does decrease when presidential and gubernatorial elections concur, but this effect disappears as fragmentation of the presidential vote at the district level surpasses 2.6.
These results are fully independent from subnational dynamics. Previous work on Brazil claimed that reverse coattails exert a substantial impact on the presidential vote, as presidential candidates depend on the support of state party leaders and their political machines (Samuels, 2003). If this hypothesis were correct, we would expect to see low dissimilarity whenever the main contenders in the presidential race count on the endorsement of subnational party organizations. To control for such effects, we created a dummy variable that indicates whether or not the incumbent governor’s party was a member of either one of the two largest coalitions disputing the presidential race. As a proxy for incumbent parties’ strength at the state level, we included a measure of terms completed in gubernatorial office. Overall, although subnational party dynamics does have an impact on dissimilarity – the presence of a coalition incumbent governor does decrease dissimilarity, especially for mean levels of continuity in office – this effect pales in comparison with effect of concurrent races.
Our logit analysis of survey data on the 2002 and 2010 elections was supportive of our second hypothesis. Multilevel electoral coordination between parties does indeed seem to be reinforced and reflected in individual-level data, as we find evidence that presidential evaluations have significant effects on the probability that one will vote for that the gubernatorial candidate of the presidential candidate’s coalition. In other words, presidential coattails voting exists at the level of the individual voter in gubernatorial elections.
To ensure that our estimates did not suffer from simultaneity bias because of the possible effects of a reverse coattails effect, we ran several tests. First, we used a bivariate probit model with the same control variables as our normal model, specifying it with the presidential and gubernatorial votes as our joint dependent variables. This specification allowed us to account for a possible correlation between the presidential and gubernatorial votes by not assuming that errors in the two equations were uncorrelated. Even accounting for this potential correlation, the presidential evaluation had a strong, significant effect on the gubernatorial vote for both the Workers’ Party (PT) and the PSDB (Brazilian Social Democratic Party).
Second, we ran a logit model using our individual survey data with votes in second-round gubernatorial elections in 2002 and 2010 as our dependent variable. The advantage of this latter specification is that it allows for the inclusion of controls for voters’ preferences in the gubernatorial election (the first-round vote for governor) that are not simultaneous with the (second-round) vote for president. In the presence of reverse coattails, presidential evaluations would be strongly correlated with the first-round gubernatorial vote and would therefore contribute little to the explanation of the second-round vote for both president and governor. This, however, is not the case: our models show that the first-round vote and presidential evaluations both have significant effects, suggesting that the coattails effect existed even with these controls.
Interestingly, we found that the coattails effect was stronger for PT candidates than for PSDB candidates, which is consistent with our hypothesis. Because the Workers’ Party coordinated its national and regional strategies more effectively, this induced greater congruence in voters’ choices. The PT’s presidential candidate, for instance, faced no internal resistance in either 2002 or 2010, whereas the PSDB dealt with internal leadership disputes in both years. It is therefore plausible that fractiousness in the PSDB led to less effective coordination in 2002 and 2010 in comparison to the PT. Furthermore, these differences cannot be attributed to higher levels of partisanship among PT supporters, as both models control for party identification.
The article presents important contributions for two distinct literatures. First, we develop a novel set of hypotheses building on the literature on vote congruence and second-order elections, that had previously focused almost solely on parliamentary countries. We demonstrate that multilevel electoral coordination in presidential systems has some important peculiarities that had not yet been incorporated into theoretical models. Second, we contribute for research on presidential elections and party systems, by incorporating issues of vertical party linkage and multilevel electoral coordination into the analysis.
Our empirical findings indicate that the choice of electoral rules for electing presidents and governors is key for building effective federal institutions, as long as it may have a relevant impact on the degree of party integration. When parties and party systems are poorly integrated policy coordination across levels of government will be harder to achieve. Although we do not claim that concurrent elections have produced an integrated, nationalized party system in Brazil, as dissimilarity has remained high in the recent democratic period, we believe that, in the absence of vertical simultaneity of elections, the Brazilian party system would likely be much more volatile and unstable. Furthermore, because Brazil is a least-likely case in which extreme multiparty system, decentralized party and electoral institutions, and low levels of party institutionalization all conspire against effective coordination, we expect such effects to be stronger in other, more favorable settings. In any case, our findings suggest that no account of party system formation in multilevel presidential systems will be adequate without an analysis of coordination across national and subnational electoral arenas and related coattail effects.
Borges, André, and Ryan Lloyd. 2016. “Presidential coattails and electoral coordination in multilevel elections: Comparative lessons from Brazil.” Electoral Studies no. 46:104-114.
Golder, Matt. 2006. “Presidential Coattails and Legislative Fragmentation.” American Journal of Political Science no. 50 (1):34-48.
Rodden, Jonathan, and Erik Wibbels. 2011. “Dual accountability and the nationalization of party competition: Evidence from four federations.” Party Politics no. 17 (5):629-653. doi: 10.1177/1354068810376182.
Samuels, David. 2003. Ambition, federalism, and legislative politics in Brazil. Cambridge: Cambridge University Press.