Introduction

The decision by political parties to participate in a multiparty coalition presidential primary involves significant strategizing. In presidential democracies with two-party systems, regardless of how the party selects its presidential candidate, there will likely be just one rival in the general election. But in democracies with an institutionalized party system and stable multiparty coalitions, parties can choose to run as a single party in the general election or as a part of a multiparty pre-electoral coalition. Parties can also decide to participate in multiparty coalition primaries to select their presidential candidates.

For parties, joining a multiparty coalition presidential primary implies downside risks and potential upside gains. There is the risk of losing the primary to a candidate from another party and, thus, not having a candidate in the first-round vote-which presumably denies its legislative candidates a presidential coattail vote effect. There is also the potential upside gain that, by having its candidate win the presidential primary, the party will have an edge, affording it stronger electoral support in the first-round vote, which will benefit down-the-ballot legislative candidates from the coattail vote effect of the party’s presidential candidate. In the runoff election, in case the party’s candidate makes it that far, there are additional benefits, as it is easier to bring together likeminded parties and voters when the multiparty coalition is formed before the election than when coalition formation takes place between the first- and second-round votes.

We use the case of Chile, a presidential democracy with an institutionalized multiparty system, stable coalitions, and, since an electoral reform in 2012, optional simultaneous presidential primaries organized by the national electoral authority. We assess whether the turnout in the coalition primary impacts the vote share received by the primary winner in the first-round election and by the candidate supported by the coalition in the runoff vote in all the elections held since the alternative of optional primaries has existed. In 2013, two multiparty coalitions held competitive presidential primaries. In 2017 and 2021, two of the three multiparty coalitions held presidential primaries. Thus, out of eight possible multiparty coalition presidential primaries, there were six instances held in those three election cycles. In three cases, the coalition primary winner made it to the runoff election. In all six cases, the coalitions that held the primary made it to the runoff or endorsed another candidate in the runoff election. Using data from Chile’s 345 municipalities, we evaluate whether turnout in presidential primaries impacts the vote share in the general election received by the presidential candidate who won the primary and the vote share in the runoff election received by the candidate supported by the respective coalition. Altogether, we have 2,760 datapoints of turnout at the municipal level in multiparty coalition presidential primaries, first-round vote, and runoff vote in three election cycles (2013-2021) in Chile.

In what follows, we review the reasons why parties hold primaries and the potential upside gains and downside risks associated with holding multiparty coalition primaries. Then, we discuss the determinants of turnout and vote choice and the potential effect of turnout in primaries on the electoral support for a party in the general election. We postulate two hypotheses. After discussing our methodology, we present the case of Chile between 2013 and 2021 and then describe the results of our models and the implications of our findings for the debate on the pros and cons of holding presidential primaries in multiparty democracies.

1. Pros and Cons of Holding Presidential Primaries

The set of rules for selecting candidates is part of the institutional design of democracies (^{Alcántara Sáez 2002}; ^{Freidenberg 2003}; ^{Hazan and Rahat 2010}; ^{Siavelis and Morgenstern 2008}; Rahat and Hazan 2001). In democracies with institutionalized party systems, candidate nomination has historically been a prerogative of political parties. Increasingly, in recent decades, in a push to make themselves more accountable to voters and reduce political disaffection, many parties have opted to hold primaries to select their candidates, thus renouncing their historical prerogative (Hazan and Rahat 2010).

Parties tend to hold primaries to unify political factions and reduce internal conflicts (^{Kemahlioglu, Weitz-Shapiro, and Hirano 2009}). They have primaries when “a dissenting faction threatens to pursue its favored policy outside of the party” and “to pull an existing faction into the party if each is sufficiently close ideologically” (^{Hortala-Vallve and Mueller 2015}, 290), as well as to select more competitive candidates (^{Adams and Merrill III 2008}). Similarly, parties are more likely to hold primaries when they are in the opposition or after losing an election (^{Cross and Blais 2012}).

Primaries reduce the discretionary power of parties to select candidates (^{Cutright and Rossi 1958}; ^{Ranney 1968}, 1972). Nevertheless, when there is low turnout in the primaries, the power of the party elites is not transferred to the general public but to interest groups that can mobilize voters or resources to influence the primary results (^{Gallo 2005}; ^{Kenney and Rice 1985}). Since primaries attract a lower turnout than general elections, the candidates who prevail in primaries can be more distant from the median voter than candidates nominated by the party leadership; thus, they might be potentially less electable (^{Brady, Han, and Pope 2007}, 80; Ranney 1968). In addition, primary conflicts among party leaders might harm the electoral performance of the primary winner in the general election (^{Lengle, Owen, and Sonner 1995}; ^{Polsby 1983}). Primary campaigns might also have negative consequences as they might deepen intra-party divisions and conflicts (^{Buquet and Chasquetti 2008}; ^{Ansolabehere, Iyengar, and Simon 1999}). In fact, while, in some cases, primaries promote intraparty competition and often more options to voters, in other cases, they hinder competition and allow elites to reduce the options for voters (Buquet and Gallo 2022).

A potential disadvantage of holding a primary is that its winner is the candidate who best reflects the views of the primary voters, not the candidate who can best reflect the views of the general public (^{Aldrich 1995}; ^{Ranney 1968}). However, since primary voters are also concerned with winning the general election, they might take into account a candidate’s electability when voting in the primaries (^{Abramowitz 1989}; ^{Geer 1988}). Moreover, candidates are also interested in winning the primary and the general election; thus, they are also concerned about the general election median voter when campaigning for their party primaries (^{Brady, Han, and Pope 2007}, 80). Still, because both types of candidates have different priorities, candidates in primaries must decide whether they will target only primary voters or the voters who will likely participate in the general election later in the year (^{Brady, Han, and Pope 2007}). As candidates need to protect their reputation and avoid flip-flopping, they normally stand away in the primaries from taking positions that are too radical and cost them voters in the general election (^{Brady, Han, and Pope 2007}, 82). When incumbents can run for re-election, they might also face challengers in the primaries who adopt more radical positions, which underlines the ongoing problem of having to face two different electorates-the general public and the likely primary voters (^{Brady, Han, and Pope 2007}, 82).

Holding primaries alters the convergence to median voter dynamics associated with a two-person race. In party primaries, candidates have incentives to deviate from the national median voter and seek the support of the primary median voter. As a result, in the primary campaign, candidates adopt positions that diverge from the median voter in the general election. When a party primary is highly competitive and turnout levels are expected to be low-either because other primaries are held on the same day or because the party usually commands limited popular support-the incentives for candidates to deviate from the median voter are higher. In turn, when a competitive primary is held by a more popular party, or there are no other primaries on the same day-and all voters are allowed to participate in the primary-a higher expected turnout might induce candidates to deviate less from the median voter. The dynamics of a primary campaign might also motivate parties to use negative campaigning to drive moderate voters away while targeting specific support bases (^{Ansolabehere and Iyengar 1995}).

A two-person primary with a higher turnout should produce winners whose positions are closer to the median voter than two multi-candidate primaries held by parties on the opposite end of the political spectrum. Thus, when there is a multi-person primary, a plurality winner might still be far away from the median voter. Since the dynamics of two-party systems are different from those of multiparty systems-even when those parties might end up forming multiparty coalitions or when runoff provisions turn the second round into a two-person race-the findings reported for how primaries in two-party systems (as in the United States) impact the first-round vote for the primary winner should not be expected to be the same as in multiparty systems. In short, holding a primary does not always need to have the same effect on the vote share in the first-round election for the primary winner.

2. Presidential Primaries in Latin American Democracies

Since the mid-1960s, holding Democratic and Republican presidential primaries in the United States has become a benchmark for other presidential democracies. Primaries were not widely held in Latin American democracies for the first decades after the transitions to democracy in the late 1980s (^{Colomer 2002}), although there were cases of primaries held at the local, regional, and, occasionally, the national level (^{Alcántara Sáez 2002}; ^{Freidenberg 2003}; ^{Gallo 2005}). In some countries, such as Argentina, presidential primaries became mandatory in the first decade of the twenty-first century-although parties have managed to turn those primaries into an expensive non-competitive confirmation process for the parties’ presidential nominees (^{Buquet and Gallo 2022}). In other countries, primaries are organized by each party or coalition outside the institutional legal electoral process, like Paraguay or Chile before 2013.

The adoption of presidential primaries in Latin American democracies has been “touted as moves toward openness and internal party democracy” (^{Carey and Polga-Hecimovich 2006}, 530). However, since parties are aware of the potential negative externalities of primaries, there is concern that holding primaries might be “an electoral handicap” (^{Carey and Polga-Hecimovich 2006}, 530). Presidential primaries have been criticized because they produce weak presidential candidates who might struggle to appeal beyond the party base (^{Chasquetti and Buquet 2004}). A study on Uruguayan primaries warns that primary winners tend to be more extremists; however, the rules of the primary can minimize the chances of selecting candidates who are too polarizing and can foster the election of competitive candidates (^{Altman 2013}). It also matters whether primaries are mandatory for all parties or optional (Altman 2013; Buquet and Gallo 2022; ^{Cruz 2021}). A study on optional primaries in Mexico concluded that rules matter: “Weaker organizations combined with low entry requirements for aspiring precandidates resulted in primaries that were costly for the parties’ general election efforts” (^{Bruhn 2010}, 28). Nevertheless, the study also warned that primaries might induce more negative campaigning (Bruhn 2010, 29). Another study found that primary candidates with closer ties to the party apparatus do better due to their ability to mobilize voters in the optional primaries (^{Ascencio 2021}). Previous research has also focused on what accounts for the decision by parties to hold primaries (^{De Luca, Jones, and Tula 2002}) and on the impact of incumbency status on how competitive primaries are (^{Skigin 2022}).

The concerns over holding primaries in Latin America mirror those raised by the literature in the United States, but so do the pros of replacing a party-centered mechanism to nominate candidates with a process that allows for popular participation (^{Carey and Polga-Hecimovich 2006}). A key difference between the US and Latin American primary processes is that, in the US two-party system, both parties hold presidential primaries; on the contrary, in those Latin American countries where primaries are optional or held outside the formal institutional setting, not all parties that comprise the often unstable and deinstitutionalized party systems have presidential primaries. When each party in the system holds primaries, negative externalities can be minimized as all parties are exposed to them (^{Buquet and Gallo 2022}). However, when parties can opt in or out of holding primaries, both the positive and negative externalities are unevenly distributed (^{Carey and Polga-Hecimovich 2006}, 531).

In Latin America, the potential centrifugal incentives of primaries that would drive candidates away from the median voter might be exaggerated, as that concern is more relevant in two-party systems and institutionalized party systems where voters assign ideological positions to the different parties. When multiparty systems are not institutionalized, parties do not have clear ideological positions, and voters will have difficulties identifying which parties or candidates are closer to the median voter, regardless of whether the party holds competitive primaries. In presidential democracies with runoff provisions and multiparty systems, convergence to the median voter might not be a priority in the first-round vote. In those cases, spatial models are not predictors of the type of competition that takes place among the several presidential candidates in the first round (^{Carey and Polga-Hecimovich 2006}, 532). Moreover “primaries may be general election assets rather than liabilities” (^{Carey and Polga-Hecimovich 2006}, 533). Primary voters might be better at identifying more competitive candidates than party insiders, and general election voters might value candidates elected in primaries more than those selected by party elites (^{Carey and Polga-Hecimovich 2006}, 542). The positive externalities of primaries can be summarized as follows: “If voters value openness, transparency, and internal party democracy in the selection of presidential candidates, and if primaries deliver those qualities in larger measure than other candidate selection procedures, then primaries should also provide a stamp of legitimacy that is an asset to candidates relative to rivals selected by other procedures” (^{Carey and Polga-Hecimovich 2006}, 534).

3. The Effect of Primaries and Other Determinants of Vote Choice

There are three dominant schools that account for electoral behavior and political preferences (^{Bartels 2011}). The Columbia School underlines the importance of social cleavages as the base for the structure of the party system and determinant of electoral preferences and vote choice. Based on the work of ^{Lipset and Rokkan (1967)}, the Columbia School predicts stable electoral preferences that are well-aligned with institutionalized political parties. The Michigan School, or the school of political socialization, builds on the work of ^{Campbell et al. (1960)} and focuses on the impact of socialization processes on electoral preferences and political behavior. Finally, the Rochester School, or the economic vote theory school, places more importance on short-term economic perception determinants to account for the voters’ decision to punish or reward incumbents (^{Duch and Stevenson 2008}).

Vote choice determinants also influence electoral participation, but turnout is impacted by other determinants as well. Electoral participation is a collective phenomenon, but the decision to vote in an election is made at the individual level (^{Franklin 2004}). More people might be convinced to turn to vote when those persons believe that others will also vote (^{Ferejohn and Fiorina 1974}); however, the determinants for why people participate in elections can also be associated with socio-demographic indicators and event-specific conditions and events (^{Brady, Verba, and Schlozman 1995}; Franklin 2004). In fact, “the factors most strongly associated with the likelihood that someone will vote are their age (young people are less likely to turn out), their education […] and the extent to which they are embedded in social structures (people who are members of churches, unions, and other organizations are more likely to vote; loners are less likely to do so)” (Franklin 2004, 16).

Studies on electoral participation in Latin America usually focus on institutional and circumstantial factors that impact electoral participation (^{Fornos, Power, and Garand 2004}; ^{Pérez-Liñán 2001}). Socio-demographic indicators-like education, age, and income-are more significant predictors of turnout than subjective motivations-like trust, interest in politics, or political efficacy-and civic mobilization networks play a more relevant role than clientelist networks or religious congregations (^{Carreras and Castañeda-Angarita 2014}, 1098).

Participation and vote choice are also endogenous to the nature of the election. More competitive elections attract more voters. The intensity of the competition tends to increase when the economic conditions are less favorable (^{Franklin 2004}, 111) and when voters are dissatisfied with the government and want to punish the incumbent (^{Ferejohn and Fiorina 1974}). Finally, electoral participation is also associated with institutional incentives (^{Lijphart 1997}; ^{Powell Jr. 1986}). Mandatory voting, automatic registration, or easier access to registration and scheduling elections on holidays and during good weather seasons promote higher turnout (Lijphart 1997). In turn, registration requirements, short voting hours, and a longer travel time required for people to vote negatively impact turnout (Lijphart 1997).

Turnout in primaries is influenced by the same variables that account for turnout in general elections, like the socioeconomic characteristics of voters, legal restrictions on participation, campaign spending, and the political context (^{Kenney and Rice 1985}, 101). Since the determinants of turnout are a constant for all parties, the decisions by some parties to opt in and others to opt out cannot be associated with institutional rules. Parties will look at the base of likely voters and decide whether they are in a better position to mobilize their base and attract other voters than their potential rivals. Thus, the decision by a party to participate in a primary is more likely explained by its capacity to mobilize voters and the tradeoff expectations associated with winning or losing the primaries.

A primary election allows parties to mobilize their base and build electoral support. Presumably, when parties mobilize voters in a primary, those voters will also be more likely to participate in the general election and vote for the party’s candidate. But when primaries are part of the institutional design of the electoral process, there might be institutional incentives for parties to participate in the process-like free television time or additional public funding for campaigns. As mobilization networks are central to explaining turnout (^{Carreras and Castañeda-Angarita 2014}), when parties participate in primaries, they can develop tools to mobilize their electoral base. We would expect that parties participating in a coalition primary have already used or will use the primaries to develop a stronger set of mobilization strategies.

The ability of political parties to mobilize voters impacts the vote share received by the party candidate in the primary (^{Ascencio 2021}). It would follow that the turnout in a primary should positively impact the turnout for the primary winner in the general election. This association is justified by the assumption that the parties that mobilize voters in a multiparty primary will also mobilize them in the general election even if their candidate does not win the primary. When parties lose the primary, they are required to formally support the winning candidate of the multiparty coalition. In fact, the losing parties become minority partners in the multiparty coalition. Indeed, those parties might not be as enthusiastic about the coalition presidential candidate; however, it is safe to assume that they still prefer, by a long shot, that their coalition wins the presidential election. Thus, we propose a hypothesis associating turnout in the coalition primaries with the vote share for the coalition candidate in the general election.

H1: Turnout in a multiparty coalition presidential primary is positively associated with the coalition’s presidential candidate’s vote share in the first-round and, when applicable, the runoff election.

When a coalition holds a primary and the winning candidate either drops out of the race before the first-round vote or fails to make it to the runoff election, the coalition has a difficult decision to make. Typically, parties nominate an alternative candidate for the first-round vote, and when their candidate does not make it to the runoff, they endorse a candidate who does. In the case of multiparty coalitions, the coalition might split as some of its parties support one candidate, and others support the other candidate or abstain in the runoff. Since the coalition’s primary winner has not made it to the runoff, the support transfer from the party to either of the two candidates in the runoff might not be as clear-cut. Still, in an institutionalized party system with stable coalitions, the parties comprising the coalition tend to act in blocs, and coalitions normally endorse candidates for the runoff election. Moreover, as the electorate in such systems is normally aligned along a left-right scale, voters will support the runoff candidate that most closely represents their views. In that sense, runoff elections are more likely to reflect political divides or social cleavages than first-round elections. In the first round, voters might have more than one option that is ideologically close to their own views. In the runoff election, campaigns are usually more polarized; thus, voters are more likely to align along the same lines as their likeminded parties.

That would lead us to expect that the bulk of support for the coalition candidate that lost in the first-round vote will go to the runoff presidential candidate supported by the coalition. Thus, we would expect that the parties that comprise that coalition will mobilize voters to help their runoff candidate-even if that candidate did not participate in the primaries. As primary voters are already more likely to hold strong political views, they will more easily align behind the candidate supported by the parties these voters identify with and have voted for in the primaries. As a result, we would expect that turnout in presidential primaries would be positively associated with the vote share received by the candidate supported by that coalition in the runoff. Our second hypothesis formally states the following:

H2: When a multiparty coalition candidate fails to make it to the runoff election, the vote share received by the candidate supported by that coalition in the runoff is positively associated with the turnout in multiparty coalition primaries.

4. Presidential Primaries in Chile’s Multiparty Democracy

After the return of democracy in 1990 and until the 2017 elections, the political party system in Chile was structured around two large competitive multiparty coalitions: the right-wing Alianza and the left-wing Concertación (^{Sehnbruch and Siavelis 2014}). The political system was defined as uprooted but stable in the first decades of post-authoritarian democracy (^{Luna and Altman 2011}), with some authors questioning the strength of the party system (Luna and Mardones 2010). In 2017, partially as a result of an electoral reform that increased the district magnitude for legislative elections (^{Gamboa and Morales 2016}), the number of coalitions increased. The 2019 social uprising was seen as evidencing a crisis of democratic representation and manifesting the lack of societal ties among institutionalized parties (^{Heiss 2021}; ^{Morales Quiroga 2020}). The debate on the strength of the party system and the alleged popular discontent with democratic representation in Chile has been a source of debate for decades (^{Castiglioni and Kaltwasser 2016}; ^{Lechner 1998}; ^{Moulián 1997}). In the past 20 years, several texts suggested adopting institutional reforms-including primaries-to strengthen the link between parties and voters (Luna et al. 2012).

Optional presidential primaries were first adopted in an electoral reform introduced in 2012. However, the history of presidential primaries in the current democratic period dates to the 1990s in Chile. In 1993, the right-wing multiparty coalition Alianza held a convention comprised of party delegates to nominate its presidential candidate. The left-wing Concertación coalition held primaries that required prior registration for party militants and sympathizers. In 1999, Concertación organized open presidential primaries, which proved a success as 1.2 million people-19 % of those who voted later in the general election-participated (^{Navia and Rojas Morales 2008}). In 2005, presidential primaries in the ruling Concertación were canceled when one candidate dropped out, and Michelle Bachelet, from the Socialist Party (PS), automatically became the coalition candidate. In 2009, the ruling Concertación held limited presidential primaries that confirmed former President Frei (PDC) as the Concertación candidate.

An electoral reform established optional presidential primaries in 2012 (Law #20640 enacted on December 6, 2012). The national Electoral Service (Servel) organizes the primaries for those parties and coalitions that choose to hold them 30 days before the registration deadline for the upcoming presidential and legislative election. Parties that participate in the primaries can legally raise funds according to the finance campaign legislation, have access to public campaign finance, and can use free television time in the weeks before the primaries. Primaries are binding, which means that participating parties cannot abandon the multiparty coalition after the primary and that the primary winner cannot be replaced unless they withdraw from the race.

We include the three electoral cycles since the adoption of legal optional presidential primaries. In 2013, the two dominant multiparty coalitions, Concertación (renamed New Majority that year) and the ruling Alianza (renamed Let’s Go Chile), held presidential primaries. In the Let’s Go Chile primaries, Andrés Allamand, a former legislator and minister from National Renewal (RN), faced Pablo Longueira, a former legislator and minister from the Independent Democratic Union (UDI). Longueira won with 51.4 % and a 6.0 % turnout among registered voters. In the opposition New Majority coalition primaries, former president Michelle Bachelet (PS) (2006-2010) easily defeated three rivals with 73.1 % of the votes. Turnout in those primaries was 15.8 % of registered voters (^{Castiglioni 2014}).

In 2017, the ruling New Majority nominated Senator Alejandro Guillier, a left-wing independent, as its presidential candidate without holding primaries. The opposition Let’s Go Chile coalition held a three-person primary. Former President Sebastián Piñera won with 58.4 % of the votes. An emerging left-wing coalition to the left of New Majority, Broad Front, also held a two-person presidential primary. News anchorwoman Beatriz Sánchez, who had no prior political experience, won with 67.6 % and became the Broad Front candidate.

In 2021, the ruling Let’s Go Chile coalition and Broad Front held primaries, but New Majority did not. In the Let’s Go Chile primaries, right-wing independent Sebastián Sichel defeated the candidates of the three established parties. In the Broad Front primaries, former student leader and legislator Gabriel Boric defeated the frontrunner candidate, Mayor Daniel Jadue, from the Communist Party (PC), the largest party in the coalition. Boric came in second (25.8 %) in the first-round vote, while Sichel received a disappointing 12.7 % in the first round. In the runoff election, Boric won by defeating right-wing independent José Antonio Kast-who did not participate in the Let’s Go Chile primaries.

Table 1 shows the presidential primary results for the 2013, 2017, and 2021 elections, turnout rates, and the share of registered voters that participated in the primaries. Turnout ranged from highs of 22.2 % in 2013 and 20.9 % in 2021 to a low of 13.4 % in 2013. The level of competitiveness varied in the six primaries held in those three election cycles. The most competitive race was in the Let’s Go Chile coalition in 2013, while the least competitive race was in the 2013 New Majority coalition when Bachelet received an overwhelming 73.1 % of the votes (and about half of all votes cast in the primaries that year).

5. Methodology

We assess the effect of turnout in primaries on the vote share received by the coalition candidate in the first-round and runoff presidential elections in Chile. We use the three elections (2013, 2017, and 2021) after the introduction of optional and centrally organized individual party and multiparty coalition primaries. Our unit of analysis is the municipality, given that this allows us to assess variance across municipalities in each electoral cycle and between electoral cycles. Since our objective is to evaluate the effect of turnout in the general election, we use a subnational unit of analysis to have variance within each electoral cycle. We have built a dataset with information from the 345 municipalities in the country on presidential primary election results, first-round and runoff presidential election results, and socio-demographic indicators. The dataset includes turnout rates in the primaries, first-round vote, and runoff election (as a percentage of registered voters). In 2013, two multi-party coalitions held primaries-the center-right Let’s Go Chile (Chile Vamos) and the center-left New Majority. In 2017, Let’s Go Chile and an emerging left-wing coalition, Broad Front (Frente Amplio), held presidential primaries, but not New Majority (Nueva Mayoría). In 2021, presidential primaries were again held by Let’s Go Chile and Broad Front, but not by the parties that previously comprised New Majority.

As we assess the effect of turnout in primaries on the vote in the general election, we have data for six primaries in 345 municipalities (2,070 observations). Our dependent variable takes two values: the vote share received by the coalition candidate in the first-round election and the vote share received by the coalition candidate in the runoff election (or, if the coalition candidate did not make it to the runoff, by the candidate supported by that coalition in the runoff). Since voting in the primaries and general election was optional, and all eligible voters were automatically registered in those three cycles, for comparison purposes, we assess the proportions over the number of registered voters rather than over those who turned out to vote. After all, we are interested in the effect of turnout in the primaries over the vote share for the coalition candidate in the general election.

For H1, the independent variable is the turnout rate in the coalition primaries. We estimate the turnout rate based on the number of registered voters in each municipality. We measure turnout as the proportion of votes received by all coalition candidates in each primary. As null and blank votes are reported for the entire primary and not for each coalition, we do not consider those votes as part of the turnout for each coalition. We expect a higher turnout in the respective multi-coalition primaries to positively impact the vote share received by the primary winners in the first-round election.

For H2, the independent variable is whether the coalition candidate made it to the runoff election. This indicator takes the value of 0 when the coalition candidate did not make it to the runoff election. We expect that when the coalition candidate did not make it, the impact of the turnout in the primaries would be lower than when the coalition candidate made it to the runoff.

We use controls for the number of candidates in the coalition presidential primaries and the primary vote share of the coalition primary winner. We expect a higher value to positively impact the candidate’s vote share in the presidential election. We also include controls for the margin of victory of the coalition winner; that is, we estimate the proportion of votes received by the winner with respect to the runner-up. We also control for the coalition vote in the previous presidential election. For the case of Broad Front-a coalition that first ran in 2017-we use the combined vote share for all left-wing candidates (excluding Bachelet) in the 2013 election as a proxy for the Broad Front vote share in the previous election. Finally, we also use controls for the poverty level and the proportion of the population that lives in rural areas in the municipality. We finally include controls for the election cycle, using 2013 and our reference category. Table 2 shows the descriptive statistics for the dependent, independent, and control variables used.

6. Results

We estimated ordinary least squares (OLS) models where the dependent variable is the vote share for the coalition candidate in the first-round (Models 1 and 4 in Table 3) and runoff (Models 2, 3, 5, and 6 in Table 3) presidential elections. In Models 1-3, we include as a control variable the vote share received by the coalition in the previous presidential election. In Models 3 and 6, we include an interaction effect between the presence of the coalition candidate in the runoff and the impact of the coalition primaries turnout on the vote share for the coalition candidate in the runoff.

H1 associates the participation rate in the coalition primaries with the vote share received by the coalition presidential candidate in the first-round vote and in the runoff election. All the models show a positive and significant effect. A higher turnout in the primaries is associated with a higher vote share for that coalition candidate in the first-round presidential vote. Notice that the effect is stronger for the runoff than for the first-round coalition candidate vote. Since there are only two candidates in the runoff, their individual vote share is always higher than the vote share of candidates in the first-round vote. Still, the positive association indicates that in all the municipalities where the coalition had a strong turnout in the primaries, the vote share for the coalition candidate in the general election is higher.

The models that include the vote share for the coalition in the previous presidential election (Models 1-3) and that exclude it (Models 4-6) show similar coefficient magnitudes. This suggests that the vote share in the previous election for the coalition presidential candidate is not driving the association between the primary turnout rate and the vote share for the coalition candidate in the general election. In fact, the effect of the vote share for the coalition candidate in the previous election on the vote share in the current election is marginally negative. This points to the fact that the turnout in the coalition primaries is a better predictor of the vote share for the coalition presidential candidate than the vote share received by the coalition candidate in the previous election.

H2 associates the vote share in the coalition primaries with the vote share for the coalition candidate in the runoff election, distinguishing between elections where the coalition candidate made it to the runoff and where the coalition candidate did not make it to the runoff (and the coalition endorsed a different candidate). Among the six coalition primaries included in the study, the coalition candidate that won the primaries did not make it to the runoff on two occasions (Broad Front in 2017 and the right-wing coalition in 2021).

The models show two things. First, the magnitude of the effect of turnout in the primaries is stronger for the runoff vote than for the first-round vote. As we discussed, the individual vote share for the two candidates in the runoff is typically higher than the vote share received by each of the multiple candidates in the first-round vote. Second, the impact of having the coalition candidate in the runoff is positive on the vote share for the coalition candidate in the runoff. Nevertheless, the interaction variable of having the candidate in the runoff and the effect of turnout in the primaries is negative on the vote share for the coalition candidate. That means that when the coalition candidate does not make it to the runoff, the turnout level in the primaries has a stronger positive impact on the vote share of the coalition candidate than when the coalition candidate makes it to the runoff. As runoff elections are more polarizing, voters are more likely to align along ideological lines. Given that runoff elections are usually seen as a choice between the lesser of two evils for many voters, ideologically mobilized voters support the candidate they dislike least.

In strongholds for the left or right, the turnout in the primaries is positively associated with the vote for the respective candidate in the runoff. For example, in 2021, when right-wing independent José Antonio Kast-who did not participate in the Let’s Go Chile right-wing coalition primaries-made it to the runoff and Let’s Go Chile threw its support behind him in the runoff after its candidate had failed to make it to the runoff, he did better in those municipalities with high turnout in the Let’s Go Chile primaries was higher than in municipalities where the turnout for the Let’s Go Chile primaries was lower. Similarly, wherever the Broad Front candidate in 2017, Beatriz Sánchez, did well, the left-wing candidate in the runoff, New Majority’s Alejandro Guillier, had strong support as well. In the municipalities with higher participation in the Broad Front primaries in 2017, Guillier was more likely to do better in the runoff election than in municipalities in those where the Broad Front primaries attracted fewer voters.

The control variables behave as expected, with idiosyncratic elements playing a role in each election cycle and with a positive impact of poverty levels and rural population on turnout in elections, a factor regularly mentioned in studies on turnout in Chile (^{Contreras, Joignant, and Morales 2016}).

Figure 1 shows the predicted probabilities estimated from the results in Table 3. Figures 1A and 1B show a positive impact of turnout in the primaries on the vote share for the coalition candidate in the general election, but the slope of the curve is more pronounced for the runoff vote (1B) than for the first-round vote (1A). This is probably because, in the runoff, when there are only two candidates, voters have fewer choices to support a candidate close to their own views. Since primaries attract more ideological voters, there is a stronger effect of turnout in the primaries in the runoff election than in the first-round vote.

In turn, Figure 1C, based on Model 3, shows the predicted probability for the vote share received by the coalition candidate in the runoff, distinguishing when the primary winner made it to the runoff and when that did not happen. The slope of the line is steeper when the primary winner was not in the runoff, but the difference is only significant at lower values of primary turnout rates. To clarify this effect, let’s return to the 2021 runoff election. Because its coalition candidate, Sebastián Sichel, lost in the first-round vote, Let’s Go Chile supported independent right-winger José Antonio Kast in the runoff. The vote share for Kast benefited more from a marginally higher turnout in the Let’s Go Chile primaries in municipalities with low turnout rates in primaries than in those where the turnout in the primaries was higher. This is again likely associated with turnout in the primaries being more ideological; therefore, in the runoff election, when the choices are more limited, ideological voters go for the candidate closest to their views.

Source: Own elaboration with data from Servicio Electoral.

Figure 1 Predicted probabilities for the effect of turnout in primaries on the coalition candidate’s vote share in the presidential election 

Conclusion

Prior studies have identified a mobilization impact of party presidential primaries on the vote share in the general election for the primary winner, but also negative externalities associated with the primary winner being more distant from the median voter in the general election. As the devil is in the details, electoral rules and particularities of party system institutionalization (including the number of parties and the presence of multiparty coalitions) make it difficult to identify patterns in the impact of primary turnout on the electoral performance of the primary winner in the general election.

We use the case of Chile, a presidential democracy with an institutionalized party system with stable multiparty coalitions and simultaneous optional presidential primaries. Turnout in the multiparty coalition primaries is positively associated with the vote share for the primary winner in the first-round and runoff elections. However, the association between turnout in the coalition primaries and the runoff election is stronger than the association between turnout and the first-round vote. We show that in a presidential democracy with an institutionalized party system and stable multiparty coalitions, holding multiparty coalition presidential primaries is positively associated with a higher vote share for that coalition in the presidential election.

Electoral rules matter. When an institutionalized party system is structured around stable multiparty coalitions, the turnout in coalition primaries has a higher impact on the vote share for the coalition candidate in the runoff than in the first-round vote. Even when the coalition primary winner does not make it to the runoff, turnout in the coalition primaries positively impacts the vote share in the runoff for the candidate supported by that coalition. In institutionalized party systems, multiparty coalition primaries might have some potentially negative consequences but also a positive impact on the vote share for the coalition presidential candidate in the general election.