Abstract
Do parties’ experiences in subnational elections predict when parties enter national competition and compete for the presidency? Building upon the party nationalization literature, we argue that a party’s presence in elections across subnational units and its subsequent performance in these elections are determining factors for whether it enters the presidential race. To conduct our analysis, we have assembled an original dataset on parties’ presence and performance in subnational elections and presidential entry in 17 countries in Europe and Latin America from 1990 to 2013. We find that a party’s presence and performance in subnational elections are significant predictors of its decision to run for president, even when the party ran in the previous election and when the elections are concurrent. These findings have important implications for understanding how subnational elections relate to national party systems and democratic representation, more generally.
Introduction
In Ecuador, a new political party called the Living Well Movement (Movimiento Vivir Bien) recently began contesting elections in the province of Imbabura (La Hora, 2013). The party grew in popularity due to its critique of President Correa’s ‘good living’ (or ‘buen vivir’) environmental policy, which has promoted – rather than hindered – mining efforts in one of the world’s most unique ecosystems: the Intag Valley, located in the heart of the province (CODELCO out of Intag, 2015). Ultimately, the party, which represents poor subsistence farmers, ran in three mayoral contests in February 2014, and won two of them, including that of Cotacachi, Ecuador’s third largest city. Subnational elections are often the entry point to politics for new parties such as Living Well, which have specific local interests. However, with both their electoral success and their message, Living Well garnered the attention of national parties as well as that of the national media (El Norte EC, 2013). Currently, Ecuador does not have a nationwide party focused exclusively on ‘green’ issues (Eisenstadt and West [no date]) but, with its success, Living Well is a contender to fill that vacuum. Can we expect parties like Living Well to also participate in national politics? More generally, to what extent does the experience of a party in subnational races predict its decision to enter national elections?
Investigating how geographic subunits structure national political competition has captured the attention of scholars at least since Schattschneider (1960) recognized the lure of local politics for members of the US Congress. Although there is a growing body of literature on party system nationalization, few investigations have focused on how nationalization occurs specifically in presidential systems (Amorim Neto and Cox, 1997; Golder, 2006; Hicken and Stoll, 2008; Morgenstern et al., 2009), and even fewer have focused on the influence of subnational elections in the nationalization process (Harbers, 2010; Jones, 1995; Moreno, 2003; Samuels, 2003). Furthermore, the studies that do investigate the influence of subnational races examine how they affect national legislative politics rather than presidential contests. Yet, both in theory and in practice, subnational politics have the potential to be significant predictors of parties’ presidential strategies. This article fills this gap in the literature by explicitly theorizing and analyzing the connection between parties’ presence and performance in subnational elections and their decision to enter the presidential election.
To investigate the connection between subnational elections and the presidential race, we have assembled an original dataset that includes electoral results for all parties in the highest level of subnational races (governorships or subnational legislative bodies) for 17 countries in Europe and Latin America from 1990 to 2013. We then identify whether those parties ran their own candidates for president in the concurrent or subsequent presidential election. By focusing on subnational elections at the level of the individual party, we are able to explicitly model how parties’ subnational electoral presence and performance predict their approach to one of the most significant national-level elections – the presidency.
Subnational presence, performance, and presidential elections
There are strong theoretical rationales for why subnational elections should predict parties’ decisions to enter the presidential race. A party’s subnational presence is the extent to which it contests subnational elections within a country. Some political parties have high presence in subnational elections, competing for governorships or regional legislatures in nearly every subnational unit. However, as the Living Well example demonstrates, regional concerns can drive both candidate and voter behavior in subnational races. Candidates may form their own regionally-focused party to address local concerns that have been neglected by national parties. This is often in response to voters who express preferences that are dominated by such local concerns. Because subnational offices are the most appropriate venue for affecting local policy, in addition to being the most accessible for regional parties to win, we would expect regional parties to be especially likely to emerge to contest subnational elections. Empirically, this expectation is demonstrated in our sample (see Figure 1).
Distribution of subnational presence.
However, elections can be considered to function as ‘economies of scale’ for political parties – the more elections in which a party becomes involved, even if they are not at the same time, the less costly each election is because parties will have already built up the necessary resources and organizational strength to run a campaign. Given the nationwide scale of the race, presidential elections tend to require organizational presence across the country. In addition, presidential candidates benefit from having regional bases of support among constituents who can contribute valuable resources to the campaign – time, money, and, most importantly, votes. A regional party that contests subnational elections in fewer than a quarter of a country’s subnational units likely does not have the organizational support or resources to finance a nationwide campaign required to contest the presidency. However, as a party’s subnational presence increases, so too should its organizational capacity, which suggests that a more national presence is a good predictor that a party will contest the presidency. This leads to our first hypothesis:
H1: Parties that have a larger presence in subnational races – i.e. run for office in more of their country’s subnational units – will be more likely to run a candidate for president.
It is not, moreover, merely participating in subnational elections that may predict when parties enter presidential elections. Subnational performance, or the average vote share that a party receives in elections for governor or regional legislatures, should also affect a party’s decision to enter the presidential race. Similar to subnational presence, greater subnational performance indicates that a party has the organizational capacity and voter support needed to launch a national campaign for the presidency. However, there is another mechanism at work that connects subnational performance with entry in presidential contests, and that is the introduction of influential new members to the political party. Indeed, research on individual parties has illustrated how strong performance in subnational elections may help predict a party’s national political strategy. For example, Samuels (2004) shows that the Brazilian Workers’ Party’s success in state and municipal elections introduced more moderates to the party, which ultimately led the party to transition from being a radical niche party to a mainstream political competitor that eventually won the presidency. Other scholars have attributed the national-level success of ‘New Left’ parties in Latin America to their achievement in subnational governments. In particular, control of municipal governments gave left parties an opportunity to solidify their organization and support bases, gain experience, and establish reputations for administrative competence that was crucial for their ability to launch and win presidential contests across the region (Chávez and Goldfrank, 2004; Levitsky and Roberts, 2012: 9).
In sum, greater subnational performance means that there are new politicians joining the party label to compete for governorships and regional legislatures and that they are enjoying some success. Such success indicates that these politicians are likely pragmatic – able to adjust their political strategy to gain votes – and are also likely to be more moderate – or at least appeal to the median voter within their particular region. This influx of relatively powerful new party members is likely to influence the party’s national electoral strategy towards increasing the party’s office-holding opportunities and towards engaging in political competition for higher-level offices. Given these possible influences, we argue that, like subnational presence, subnational performance is also a good predictor of when parties enter the presidential race. This leads to our second hypothesis:
H2: Parties that have greater performance in subnational races – i.e. stronger bases of support and greater electoral success across subnational units – will be more likely to run a candidate for president.
Our argument focuses explicitly on how greater experience in subnational elections is a predictor of parties’ decisions to enter the presidential race. We argue that subnational elections are strong predictors of presidential entry even when parties compete for the presidency for other reasons – either to gain office or earn credibility as coalition-partners and policy advocates (West and Spoon, 2013) – and then supplement their presidential campaign by competing for subnational office. Parties are more likely to treat subnational elections as complements to the presidential race under two conditions: 1) the party already has strong organizational capacity across the subnational units; and 2) the party believes that it can benefit from expending valuable resources to compete for subnational office. The first condition is met if the party competed for the presidency in the previous election; it would then have the nationwide capacity to launch a presidential campaign regardless of its participation in subnational elections. The second condition exists primarily if the presidential election is concurrent with subnational elections. In concurrent elections, entering subnational races does not necessarily require the launching of separate campaigns as much as it requires adding to the existing national campaign. Furthermore, given the power of the coattails effect of presidential elections in concurrent election cycles, parties can expect to be more successful in subnational races when they are simultaneously competing for the presidency (Golder, 2006; Shugart 1995). Therefore, even under these conditions, parties’ subnational experience still serves as a strong predictor of their presidential entry decisions. 1 Our third hypothesis is thus:
H3: Parties that have greater performance and presence in subnational races will be more likely to run a candidate for president even when elections are concurrent and the party previously contested the presidency.
Data and methods
To test our hypotheses on how subnational elections predict presidential entry strategies, we compare party behavior in 17 countries in Europe and Latin America, of which 7 are in Europe and 10 are in Latin America. 2 All of these countries have a directly elected president. The European countries are semi-presidential (as defined by Elgie (1999)) and the Latin American countries are presidential. Our original dataset covers the period 1990–2013 and includes 57 subnational elections. The subnational units are those that have been identified as having a significant level of both shared and self-rule in the Regional Authority Index (RAI) (Hooghe et al., 2008) for the countries included and by our own assessment based on the RAI criteria (for the Latin American cases), as well as whether they held subnational elections. These subnational units vary from municipalities (e.g. Finland), to counties (e.g. Romania and Croatia), departments (e.g. Paraguay, Peru, and Uruguay), regions (e.g. France), and states (e.g. Austria, Brazil, and Venezuela). If a country has subnational executive elections, we included those results, and if the country only has elections to a subnational legislative body we included those. Of the 57 elections, 34 are for subnational executives and 23 are for subnational legislative bodies. Our unit of analysis is the political party in the subnational election. Our dataset includes all parties that ran in more than one subnational unit, as well as parties that only ran in one subnational unit, but won the election in that unit. This results in a total of 1036 party observations.
Similar to West and Spoon (2013), our dependent variable is a binary variable indicating whether a party runs a candidate in the concurrent or subsequent presidential election, coded 1 if so and 0 if not. Because we are focused on how participation in subnational elections influences presidential entry, parties that do not compete in subnational elections are not included in the dataset. In the coding of our variable, we follow Spoon and West (2015) and define running for president as cases where parties nominate one of their own party leaders as a presidential candidate. We did not consider parties joining a pre-electoral coalition to support the candidate of another party as running a candidate for president. 3 Of the 1036 party observations in our dataset, 296 (28.6%) ran in a presidential election.
Our key independent variables relate to the party’s presence and performance in the subnational elections either preceding or at the same time as the presidential election. First, we account for the presence of the party in subnational elections by including a variable that measures the percentage of subnational units in which the party ran. Although the mean value for presence is 37.6%, nearly half of our observations are less than or equal to 20%, and one-fifth of our observations are greater than or equal to 80%. Thus, nearly 70% of the parties in our dataset can either be considered to be regional parties, which have concentrated their subnational efforts, or large nationalized parties that are contesting almost all subnational elections. The distribution of the presence variable is illustrated in Figure 1. Second, we measure the party’s performance by using its average vote share in each subnational election across all subnational units. 4 The mean value of performance is 6.5%. To test our third hypothesis – that a party’s subnational presence and performance are predictors of running for president regardless of election timing or running in the previous presidential election – we include several interaction terms. First, we interact both the presence and performance variables with a dichotomous variable for whether the subnational and presidential elections were concurrent. When subnational elections are not all on the same date (e.g. in Argentina and Austria), we code it as concurrent if a majority of the elections are on the same date as the presidential election and as non-concurrent if the majority of elections are not. Second, we interact both variables with the lagged dependent variable, which indicates whether a party ran a candidate for president in the previous presidential election.
We also incorporate several control variables in our analysis. First, we include a variable that measures the percentage of subnational elections won by a political party. Second, we include a dichotomous variable that accounts for whether the party ran in the most populous subnational entity. 5 We expect winning more subnational offices and competing in the most populous unit to positively influence a party’s likelihood of running a candidate for president. Third, we control for the type of subnational election with an executive variable, which is coded as 1 for an executive election and 0 for a legislative election. We posit that running in a subnational executive election will provide parties with the executive experience necessary to then successfully contest the presidency. Among the countries in our dataset, we have eight countries that have executive elections and six that have legislative elections. Fourth, we include a dichotomous variable for whether a state is federal. Following the argument that when parties gain access to local office, especially when it is relatively powerful, they may be less likely to expend resources in national elections (Chhibber and Kollman, 1998, 2004; Harbers, 2010; Samuels, 2003), we posit that parties may be less likely to run presidential candidates in federal systems. 6 Fifth, we utilize a dichotomous variable to account for the electoral rules used in the presidential election. Following Amorim Neto and Cox (1997), Jones (1999, 2004), and Golder (2006), we argue that parties are less likely to run presidential candidates in plurality systems. Sixth, to account for the consolidation of democratic institutions, we include a democracy control. We posit that, when the party system is more democratic, parties will be more likely to enter the presidential race, given that competition is more fair and open. To measure democracy, we use the country’s Freedom House (2015) scores for the year of the subnational election. We add the political rights and civil liberties scores and then take the inverse, so that higher values indicate more democratic countries. Finally, we include a time component by controlling for whether the party ran in the previous presidential election by lagging the dependent variable. See Appendix 1 for the descriptive statistics of all of the variables included in our analysis.
To determine the probability of a party entering the presidential race, we use a logistic model and cluster the standard errors by country–year. We use this technique because the party-level observations in each country for a given subnational election are not independent.
Results
Table 1 presents the results from our logistic regression analysis. Model 1 demonstrates support for our first two hypotheses. A party’s presence across geographic subnational units (H1) and its performance subnational elections (H2) are significant predictors of its likelihood of running a presidential candidate. Thus, parties that are more nationalized, i.e. run in more subnational races, and that perform better in subnational elections are also more likely to enter the presidential race.
Performance and presence in subnational elections.
p < 0.01, **p < 0.05, *p < 0.10. †Results are from logistic regressions with clustered robust standard errors.
To show the substantive effect of our results, we simulate predicted probabilities. Figure 2 plots the probability of entering the presidential race across the range of presence (Figure 2(a)) and the range of performance (Figure 2(b)). The dashed lines are the 95% confidence intervals and the tick marks on the bottom of the figures indicate the distribution of the presence and performance variables. Figure 2(a) demonstrates a strong relationship between subnational presence and presidential entry across the distribution of presence. The more subnational races a party contests, the more likely it is to run a presidential candidate. Holding all variables at their means, a party has a 22.5% probability of entering the presidential race. For a more regional party that only contests 10% of subnational races, the probability of entry drops to 12.7%. Conversely, for a party that contests 80% of subnational races, the probability of entry is 45.3%. Figure 2(b) shows that there is only a statistically significant relationship for performance when parties on average earn below about 30% vote share in subnational elections. Thus, as a party’s subnational performance increases to 30%, so does its probability of running a candidate. Above 30%, the relationship is no longer statistically significant. Increasing average vote share to one standard deviation above the mean – 17.7% – the likelihood of entry increases to 27.0%. All probabilities are significant at p = .05. Thus, these findings show that presence, or the degree to which a party is nationalized, is a more robust predictor of a party’s probability of running for president across the full distribution of the variable.
Effect of subnational electoral presence and performance on presidential entry. (a) Presence; (b) performance.
Returning to the example of Ecuador’s Living Well Movement, our results suggest that for the party to contest the presidency, it must first expand beyond the borders of Imbabura. In February 2014, the party ran in municipal elections in only 1 of Ecuador’s 24 provinces (4.2%). Should it choose to continue to contest provincial elections in the next two years, based on our analysis, it has around a 17% likelihood of fielding a candidate for president if it focuses only on the single province of Imbabura. Running in four additional provinces (20.8% of provinces) increases the likelihood of entering the race to around 20%. Based on our results, the party will need to contest roughly 75% of provincial elections (18 provinces) to increase its probability of running a presidential candidate to about 40%.
In Hypothesis 3, we argued that a party’s subnational presence and performance are strong predictors of running for president regardless of election timing or running in the previous presidential election. The results of Models 2 and 3 support this argument. In Model 2, although the coefficients for the interaction terms between presence and performance and concurrent elections are negative, they are not statistically significant. Similarly, in Model 3, the coefficients for the interaction terms between presence and performance and previously running a presidential candidate are negative, but they are also not statistically significant. These findings suggest that a party’s subnational electoral choices do reflect its decision to enter the presidential race, and that neither timing of the election nor previously entering the presidential race appear to condition this predictive relationship.
Several of our control variables also have a significant relationship with the likelihood of running a presidential candidate. First, contesting the election in the most populous subnational district has a positive and significant relationship with presidential entry. Given the media attention and notoriety that both parties and candidates gain by contesting subnational races in a country’s most populous district, this finding may be attributed to the fact that subnational elections help parties build significant resources that can be readily transferred to the presidential campaign. Second, running for executive office at the subnational level also positively relates to presidential entry. Competing for governorships may give parties the experience, credibility, and notoriety necessary for constructing a bid for the higher-level executive office of the presidency. Third, consistent with West and Spoon (2013) Now added to the bibliography, we find that electing the president through a plurality system increases a party’s likelihood of running. Finally, as expected, running a presidential candidate in the previous election also strongly predicts entry.
Conclusions
Do the ‘races at the bottom’ predict parties’ presidential entry strategies? Through our analysis of the relationship between subnational and presidential elections, we find that a party’s decision of whether to run a candidate in the presidential election is predicted by its behavior at the subnational level. Specifically, we find that a party’s subnational electoral presence across districts strongly predicts its probability of entering the presidential race. Performance across subnational units also predicts the party’s likelihood of presidential entry, but its predictive power is limited to smaller parties. Importantly, neither the timing of the election nor contesting the previous presidential election condition how a party’s subnational presence or electoral success predicts its decision to run a presidential candidate. Thus, our findings lend support to our argument that it is the degree of a party’s nationalization that predicts whether it will enter the presidential race.
We do, however, acknowledge that through our analysis, we are only able to determine that party nationalization predicts presidential entry. We suspect that the features of subnational electoral participation that we highlight above – e.g. national organizational presence and the entrance of influential new party members – may cause parties like Ecuador’s Living Well, which originate at the subnational level, to enter the presidential race. However, because entry in the presidential race is not randomly assigned and we are unable to control for all confounding causal factors (such as party ideology, leadership appeal, etc.), we cannot clearly establish a causal relationship between subnational presence and performance and presidential strategy. Future research should seek to establish the causal relationship between subnational electoral behavior and running presidential candidates by using an approach similar to that of Dinas and Foos (2013), who use the discontinuities generated by the 5% electoral threshold rule in Germany to determine how subnational elections affect performance in national elections.
Unpacking how parties incorporate subnational politics into their national electoral strategies has several implications for understanding representative democracy in presidential systems. First, our project is a first step toward demonstrating the national political consequences of parties’ subnational decisions and performance. Though extant research has shown that subnational elections influence national legislative party systems (Moreno, 2003; Samuels, 2003), we demonstrate that subnational elections also help shape presidential elections by predicting when subnationally-competitive and successful parties enter the presidential race. Using an original cross-national dataset, we show that when parties gain experience at the subnational level, they are also more likely to participate in the presidential race, which also corresponds with an increased size of the national party system. Second, our project shows how regionally focused parties may influence national politics. By predicting when small parties, like Ecuador’s Living Well, enter the presidential race, we find that contesting subnational elections may provide marginalized populations and minority groups with greater opportunity in national politics, in addition to enhancing their chances to influence regional politics.
Footnotes
Appendix 1
Descriptive statistics.
| n | Mean | SD | Minimum | Maximum | |
|---|---|---|---|---|---|
| Presidential candidate | 1036 | 0.286 | 0.452 | 0 | 1 |
| Presidential candidate (t − 1) | 1036 | 0.198 | 0.399 | 0 | 1 |
| Presence | 1036 | 0.376 | 0.343 | 0.002 | 1 |
| Performance | 1036 | 0.065 | 0.112 | 0.00 | 0.896 |
| Won | 1036 | 0.054 | 0.144 | 0 | 0.964 |
| Populous | 1036 | 0.516 | 0.500 | 0 | 1 |
| Concurrent | 1036 | 0.193 | 0.395 | 0 | 1 |
| Governor | 1036 | 0.550 | 0.498 | 0 | 1 |
| Federal | 1036 | 0.331 | 0.471 | 0 | 1 |
| Plurality | 1036 | 0.149 | 0.356 | 0 | 1 |
| Democracy | 1036 | 11.35 | 1.851 | 6 | 14 |
Acknowledgements
We thank Eric Keels, Tatyana Kelman Kisin, Amalia Pulido-Gomez, and Rebecca Sanders for their excellent research assistance and the two Research & Politics reviewers and Catherine de Vries for their most helpful suggestions. All errors remain our own.
Authors’ Note
The author order reflects the principal of rotation. Both authors have contributed equally to all work.
Declaration of conflicting interest
None declared.
Funding
This work was partially supported by a University of North Texas Research Opportunity Grant (grant number GA 9383).