The Effect of Pre-Electoral Party Coordination on Vote Choice: Evidence from the Italian Regional Elections

Abstract

Although it is theoretically relevant, the effect of pre-electoral coalitions on vote choice remains a largely unexplored area in the field of party strategy. The article addresses this gap by focusing on the Italian regional elections, where the electoral rules allow parties to run alone and, at the same time, to present pre-electoral coalitions on the ballot paper. This setting allows us to investigate under what conditions citizens vote for their preferred party and the coalition that this party indicated to coalesce with. The results suggest that ideological congruence and the size of the parties entering a pre-electoral agreement are decisive factors. Findings also reveal that the time elapsed since the establishment of a coalition has no effect on vote choice.

Keywords

pre-electoral coalitions party coordination strategic voting candidates Italian regional elections

Introduction

The incentives to coordinate for both parties and voters depend on the institutional environment in which they operate. Under majoritarian rules, voters coordinate by deserting those parties unlikely to obtain any seats and thus to avoid ‘wasting’ their vote. Small parties are discouraged from participating in majoritarian elections since they are unlikely to obtain any seats. When it comes to proportional systems, these can be designed to guarantee representation for all but the tiniest of parties, to a large extent obviating strategic incentives (Cox, 1997). The result is that strategic coordination by both voters and parties is typically higher in disproportional systems.

Political parties can coordinate in several ways. They can compete with a joint manifesto, support common candidates or announce that they want to form a governing coalition after the elections (Powell, 2000). One of the most common coordination mechanisms is the formation of pre-electoral coalitions. Despite the fact that pre-electoral coalitions are very common and can significantly influence both election and policy outcomes (Golder, 2006), there has been little theoretical or empirical research on pre-electoral coalitions. As of today, the majority of the literature focuses almost exclusively on government coalitions that form after an election (Laver, 1998; Martin and Stevenson, 2001); only more recently has attention been given to the formation of pre-electoral coalitions (see Golder, 2005, 2006).

However, while we know today about the conditions of pre- and post-electoral coalitions’ formation, the effect of pre-electoral coalitions on vote choice remains a largely unexplored area. In the words of Thomas Gschwend and Marc Hooghe (2008), ‘at present, there is no research available on the question of how voters respond to the formation of pre-electoral coalitions’ (p. 557).

Determining under what conditions citizens vote for their preferred party and the coalition that this party indicated to coalesce with is not an easy task. The main difficulty in examining this research question is that voters only express one vote and the assessment of the impact of pre-electoral agreements can only be a counterfactual one (Gschwend and Hooghe, 2008). Thus, unless high-quality survey data are available to reveal voters’ motivations, such assessment is not possible. This is the reason why the only available work on the impact of pre-electoral coalitions on vote choice uses a laboratory experiment. The experimental study conducted in Belgium asked 1255 students to cast two votes: one vote on a ballot with individual parties and another vote on a ballot with coalitions. Gschwend and Hooghe (2008: 556) found that the decision whether to follow the coalition or not can be traced back to four considerations: dislike of the coalition partner, ideological congruence between coalition partners, size of the initially preferred party, and being attracted to a specific high-profile candidate.

This article employs the Italian regional elections to study the effect of pre-electoral coalitions on the election outcome in order to fill this gap in the literature. Italians vote for their regional councils by casting two votes on the same electoral ballot. One vote is for a party on the proportional tier of the ballot paper and another vote is for the future president of the region on the majoritarian tier. Parties often coordinate by running as independent entities on the proportional tier but presenting common candidates and symbols on the majoritarian tier; for this reason, the majoritarian vote is usually a vote for a pre-electoral coalition. Examining vote patterns in the Italian case then offers an opportunity to observe under what conditions voters follow their party into the coalition. At the same time, it enables us to control for the effect of other voting determinants, primarily candidate-specific factors.

This study is important not only because it aims to explore a neglected area in the field of party strategy but also because it offers insights into voting behaviour under an expanding set of electoral rules, mixed systems and also into the widespread practice of presenting voters with pre-electoral coalitions in many of these systems. In addition, this study can make a contribution to the understanding of strategic voting more generally. Studies of strategic voting specify a certain chain of reasoning the voter employs before casting a vote for a party in a way that would seem to be counter-intuitive to a preference for a party. I analyse this counter-intuitive behaviour further by including the effect of formal pre-electoral cartels on vote choice. Specifically, the uniqueness of the Italian case allows us to assess voters’ strategic reactions to the formation of pre-electoral cartels for the first time.

This article is structured as follows. It first provides a general theory and develops the hypotheses to be tested. Then it introduces the case under study and it presents the data and methods. Subsequently, the article describes the results based on the use of aggregate data and extends the analysis to the individual level using surveys. Finally, the article discusses the results and their implications.

Theoretical Arguments

Of 292 elections in 20 countries, Sona Golder (2005, 2006) found that 44% had pre-electoral coalitions, and one-quarter of the governments were formed by a pre-electoral coalition. Recent works suggest that voters are not only aware of pre-electoral pacts (e.g. Bowler et al., 2010) but also that the coalition preferences of voters have an impact on the outcome of the government formation process after the elections (Debus and Müller, 2013). Yet we do not know how voters respond to the formation of pre-electoral coalitions. The only study that has examined this question is the experimental study conducted by Gschwend and Hooghe (2008). My aim is to evaluate hypotheses about the effect of pre-electoral coalitions on voting behaviour using real-world data.

I note first that when party elites decide to form a pre-electoral coalition, voters subsequently have two options: they could follow their preferred party’s elites and stay within the coalition, or voters could desert the pre-electoral coalition to which their party belongs and cast their vote for some other party or coalition. There are two different reasons to expect that many voters will follow the lead of their party elites. First, it is to be expected that those who identify with a party or strongly prefer a party above all the others will be more likely to ‘simply’ follow their party into the coalition. Second, if voters are instrumentally motivated and looking for ways to optimise the effect of their vote (Cox, 1997), it seems likely that they will stay within the coalition. Gschwend and Hooghe (2008: 560) explain that instrumentally motivated voters will indeed understand the considerable benefits of joining a pre-electoral coalition.

Yet several conditions exist under which voters may deviate from these baseline predictions. The first hypothesis is based on the influence that candidates might have. Several scholars have proven that voters’ evaluations of the personalities of candidates play an increasingly important role in determining electoral preferences (Shugart et al., 2005). Thus, the hypothesis is that when voters prefer a candidate who is nominated by a different party or coalition, they face serious cross-pressures to desert the pre-electoral coalition to which their party belongs in order to vote for that preferred candidate. Conversely, when the preferred candidate belongs to the coalition endorsed by the preferred party, this might serve as an incentive to remain loyal to the coalition.

There are at least three factors related to pre-electoral coalition features that may have an impact on voting. The most straightforward is ideological similarity: voters are not likely to vote for the sponsored coalition of the party they otherwise support if they expect that there will be too many policy concessions to make. Golder (2006) shows that the successful formation of a pre-electoral coalition is largely dependent upon the ideological distance between the coalition partners. Gschwend and Hooghe (2008) show that a similar logic is relevant for voters too. A result echoed by Marc Debus and Jochen Müller (2014), who find that voters favour coalition governments with a low degree of internal programmatic heterogeneity. Consequently, the more congruent the ideological positions of the coalition partners, the more likely it is that supporters of the constituent parties will support the coalition at the polls.

The size of the coalition partners is, however, a crucial element. While small parties’ supporters should understand the considerable benefits of joining a pre-electoral coalition, they are also those expected to defect more easily. In fact, it can be assumed that the larger coalition partners will, to a large extent, be able to impose their views on the junior coalition partners (see also Martin and Vanberg, 2003). This means that junior coalition partners will face a difficult task in incorporating their views into the joint platform, and for supporters of these small partners, there are thus fewer reasons to feel represented by the manifesto of the cartel (Gschwend and Hooghe, 2008). The defection of small parties’ supporters can come about for other reasons as well (e.g. protest vote); in any case, the expectation is that supporters of junior coalition partners may defect more easily than supporters of the major coalition parties. This argument finds support in the experimental study by Gschwend and Hooghe (2008).

It can be argued that time also plays a role in determining the responses of voters. The formation of a pre-electoral coalition implies that voters, to some extent, are required to re-adjust their mental map of the political space according to clues provided by the party elites (see also Kabashima and Reed, 2000). The likelihood of voters following their party into a coalition should increase as the time elapsed since the establishment of the pre-electoral coalition increases. Gschwend and Hooghe (2008) found that loyalty towards the cartel increases over years despite this being a small effect.

The Italian Case

The Italian regional electoral system is a rather complex mixed system (D’Alimonte, 2005: 273; Di Virgilio, 2005). In countries using mixed rules, citizens have two votes: one for a party under proportional rules and one for a candidate under majoritarian rules. The two votes may be cast on the same electoral ballot, as is the case in Germany and New Zealand; in other electoral systems, such as the Japanese and Scottish cases, voters are presented with two separate ballot papers, one with a list of parties and one with a list of candidates. Today, coordination efforts of parties on the majoritarian tier are increasingly common. Since the 2003 election in Japan, for instance, parties frequently enter into formal pre-electoral coalitions by supporting common candidates (Burden, 2009). In Germany, the smaller parties Free Democratic Party (FDP) and Greens usually commit themselves in advance to one of the larger parties instructing voters on how to ‘split’ their vote in order to favour a specific coalition government outcome (Pappi and Thurner, 2002).

Maurice Duverger (1954) identified two distinct logics, both of which we see in operation under mixed rules (e.g. Ferrara and Herron, 2005). First, elite-level coordination on the majoritarian tier aims at conserving resources to expend on candidates who can actually win. These coordination efforts are intended to anticipate the second strategic coordination discussed by Duverger, the voter-level coordination. On one hand, voters who prefer small parties have strong incentives to split their vote by supporting a bigger party’s candidate, thus avoiding wasting their vote on a candidate who has no chance of getting elected (Bawn, 1999). On the other hand, voters who prefer big parties face similar strategic incentives to split by supporting a smaller party on the proportional tier to help it cross the electoral threshold of entering parliament and thus favouring a specific coalition government (Gschwend, 2007).

Under the Italian regional electoral system introduced in 1995,¹ voters have two votes: they choose by proportional rules a party list at the provincial level, and with the second vote they choose by plurality a list at the regional level connected to a candidate for the presidency of the region. The proportional vote is cast using open lists with a threshold of 3%, unless the party is connected to a presidential candidate who obtains more than 5% of the vote, in which case the threshold for the single party is only 1.5% (also known as the ‘double-threshold’ mechanism) (Di Virgilio, 2002). Thus, 80% of the seats are assigned on the proportional tier. In addition, the winning regional list will receive a bonus of 20% (or more) – enough to ensure a majority of at least 55% of the seats in the regional council.²

Since the introduction of the new electoral rules, parties have increasingly coordinated by presenting common symbols and forming pre-electoral coalitions on the majoritarian tier, while still running as independent entities on the proportional tier (Di Virgilio, 2007). The electoral ballot usually displays two large pre-electoral coalitions, the centre-left and the centre-right, each endorsed by one of the two largest Italian parties plus several junior coalition allies (The Appendix provides an example of the ballot paper). The two largest parties today are as follows: on the left, Partito Democratico (PD), up until 2007 named Democratici di Sinistra (DS) and before 1998 named Partito Democratico della Sinistra (PDS); on the right, Il Popolo della Libertà (PdL) formed in 2009 by the fusion between Forza Italia (FI) and Alleanza Nazionale (AN). Besides the two largest pre-electoral coalitions, there are several smaller ones generally endorsed by smaller parties. Party systems and pre-electoral coalitions differ from one region to another, while they are the same across districts of the same region; only the composition of the pre-electoral coalitions, for example, the number of coalition partners, may slightly change from one district to another within the same region.

Italian voters have two options. They can follow their party into the coalition by casting a straight vote or they can split the vote by supporting a different coalition than the one that their party indicated to coalesce with. As the electoral rules allow for the formation of pre-electoral coalitions, only one party for each coalition will present its own candidate, and thus, several parties’ supporters will be ‘frustrated’ (Benoit et al., 2006) because their candidate is not running. Given that a voter prefers a party nominating its own candidate and another voter prefers a party with a candidate of an allying coalition, a straight-ticket vote does not have an identical meaning to both voters. As explained below, the presence on the ballot paper of the parties’ candidate should and will be empirically controlled for. Voters can also abstain in one part of the ballot by supporting only a party or a coalition/candidate.³

Dependent Variables

At the aggregate level, there are two ways of examining under what conditions pre-electoral coalitions are successful; both need to take into account that under the Italian regional electoral rules, the vote for a pre-electoral coalition is also a vote for a candidate. The first way is by measuring the ‘coalition vote gap’ calculated as the vote difference between the votes gained by the coalition/candidate and the parties endorsing that coalition/candidate at the district level. The vote gap is a measure of how much the coalition/candidate outperforms the linked parties. Positive values indicate that the coalition/candidate did better than the parties, whereas a negative gap means that the parties outperformed the coalition/candidate.

The vote gap measure is problematic. First, these differences mask cross-voting among the two tiers of the electoral ballot, a problem known as the ‘ecological fallacy’ (Robinson, 1950). Second, official electoral results only report the total vote for the coalition/candidate and the vote for each party. Consequently, it is only possible to measure the vote gap at the coalition level and no information is available at the party level. Due to the limitations of the vote gap measure, this article also considers estimates of the vote patterns obtained using Rosen et al.’s (2001) method.

This method represents an extension to the multi-party settings of King’s (1997) ecological inference method which is best applied to situations with two parties and two candidates. Rosen et al.’s method uses Bayesian statistics to estimate intra- and inter-coalition flows of voting employing aggregate electoral results. The lowest level of aggregation for which electoral data are available for the Italian regional elections is the precinct level (Italian comune). In order to ‘validate’ the method, several tests have been carried out and presented in the Supplementary Material of this article. There I also discuss possible limitations of the data. As the efforts of validating the method appear convincing, we are able to proceed with some confidence into exploring the covariates of split-ticket voting using this method.

In summary, the two measures, vote gap and split-ticket voting, capture different phenomena. In the words of Roberto D’Alimonte (2001), it is possible to separate the ‘gluing’ effect, when the coalition/candidate is able to retain voters on the majoritarian tier (lower split-ticket voting), and the ‘magnetic’ effect, when the coalition/candidate is able to attract more votes than the linked parties (higher vote gap). Both measures, the vote gap and ecological inference estimates, are obtained using official electoral results available from the Italian Ministry of Interior website (www.elezionistorico.interno.it). The unit of analysis for the vote gap is coalitions at the provincial district level. For split-ticket voting, the unit of analysis is parties at the provincial district level. The analysis covers the four elections (1995–2010) held in the ‘ordinary’ regions under mixed rules.

An extension of the aggregate-level analysis is carried out in the last section of the article using surveys. To this end, the analysis of split-ticket voting is replicated using a post-election telephone study conducted by Ipsos following the regional election in 2010. The availability of individual-level data enables us to expand some of the results gathered through the use of aggregate voting figures alone by having access to voters’ preferences. (The Supplementary Information provides descriptive statistics of all variables employed in the analysis.)

Hypotheses and Operationalisation

The theoretical overview listed several reasons that could explain whether party supporters follow their initially preferred party into the coalition and cast a straight ticket or whether they defect by casting a split-ticket vote. First, in the case of voters who identify with a party or who strongly prefer a party above all the others, parties function as the principal voting cue (Campbell and Miller, 1957). For this reason, it makes sense to hypothesise that those voters will be more likely to ‘simply’ follow their party into the coalition, thus casting a straight vote. The effect of a strong commitment to a party can be investigated at the individual level as well as at the aggregate level. In the first case, I use the question, ‘Do you usually think of yourself as close to any political party?’ and expect that identifiers will be more likely to cast a straight vote than non-identifiers (Party identification hypothesis).

The effect of partisanship can also be investigated at the aggregate level allowing room for contextual voting patterns. In the Italian context, party choice has been historically linked to specific areas of the country. Centre-left parties have been traditionally very strong in the central regions of the country. The same holds true for the centre-right parties in the northern and southern regions, albeit to a lesser extent (Agnew, 2002; Bellucci and Segatti, 2011). Thus, one expects lower levels of vote gaps and lower levels of split-ticket voting for specific political parties in those regions where they have been historically strong (Party stronghold hypothesis). The stronghold variable takes a value of 1 for a specific coalition when it has never lost an election across the four elections under investigation in this article.⁴

In the Italian context, a straight vote is also consistent with strategic voting. Strategic voters are those who do not vote in line with their preferences since they are primarily concerned with expectations about how parties (or candidates) will perform at the elections (Fisher, 2004). The classical formulation of strategic voting under mixed systems relates to what is often called the ‘wasted-vote’ hypothesis: voters are more likely to deviate from a straight vote when their party’s candidate has no chance of winning on the majoritarian tier (Bawn, 1999). A more recent strand of research has examined another strategic hypothesis, related to split-ticket voting, often defined as the ‘coalition-insurance’ hypothesis: if voters’ most preferred coalition has a junior partner who is in danger of falling below the national threshold of representation, voters might be motivated to employ a coalition-insurance strategy (Gschwend, 2007).

The unique features of the Italian regional electoral system provide voters no incentive to split strategically following either one of the two mentioned ‘strategies’ (i.e. the wasted-vote or coalition-insurance hypothesis). The double-threshold mechanism previously mentioned reduces strategic incentives to a minimum; even if voters prefer a different coalition/candidate than the one endorsed by the party they voted for, it makes little sense for them to split because this would reduce the opportunity of their party to gain seats (Di Giovine and Pizzetti, 1996). Put simply, via the double-threshold mechanism, the more votes there are for the allying coalition/candidate, the higher the chances that the preferred party will need just 1.5% of the vote rather than 3% to receive seats. Thus, based on expectations of the election outcome, voters should cast a straight vote even when they prefer a different coalition/candidate.

When measuring strategic voting at the individual level, and based on the idea that highly educated and sophisticated voters are more likely to ‘respond’ to strategic incentives, it makes sense to expect that these voters (via education or political interest) will be more likely to cast a straight vote than their least aware counterparts (Voter sophistication hypothesis). The literature on strategic voting has also considered aggregate district-level features supposedly correlated with a strategic reaction to the electoral rules (e.g. Niemi et al.,1992). Specifically for mixed systems, Robert Moser and Ethan Scheiner (2005) suggest that the relationship between straight vote and competitiveness of the district race should help differentiate between strategic and personal voting for candidates. In the Italian context, the mechanism is as follows.

When the district race is competitive because the vote difference between the two top-ranked coalitions/candidates is low, voters of the parties linked to the two top-ranked coalitions/candidates should cast a straight vote even when they prefer a different candidate. That is, since the two top-ranked coalitions/candidates are at risk of losing the majoritarian election by just a few votes, supporters of the constituent parties should vote for them casting a straight ticket. On the other hand, more split-ticket voting in competitive races means that the higher the risk for the coalition/candidate to lose the election, the more likely people are to switch and vote for another coalition/candidate. This result would make no sense from a strategic point of view, and it is more easily interpreted considering, for instance, the presence of a strong candidate on the electoral ballot (which, when endorsed by another pre-electoral coalition, decreases the likelihood of a straight vote). Of course this non-strategic straight vote can come about for other reasons, such as protest voting; despite this, however, the effect of candidate-specific factors is arguably the most important. For lowest ranked coalitions/candidates, the matter is more complicated: if a more competitive race should encourage a lower degree of straight votes in order to help the victory of the least-worst option among the two top-ranked coalitions/candidates, at the same time, less straight votes can be entirely consistent with personal voting (Moser and Scheiner, 2005: 263) (District competitiveness hypothesis).⁵

To sum up, when people cast a straight vote, this behaviour is consistent with both a party-centred choice and strategic voting. If the vote is party-centred, then we should observe variables such as party identification and party stronghold decreasing the vote gap and increasing straight-ticket voting. Alternatively, if strategic voting exists, voter sophistication at the individual-level as well as district-level variables should have an effect on vote choice. Specifically, they should decrease the vote gap and increase straight-ticket voting.

Patterns of lower vote gaps and higher straight-ticket voting can also be consistent with candidate-centred voting. To assess the impact of candidates, in addition to the use of the district-level indicators discussed, this article also gives attention to their key features, such as incumbency and being a nationally visible politician. The expectation is that high-profile candidates (via their personalities, media visibility and so forth) are more likely to increase the coalition/candidate vote share, and decrease split-ticket voting, than other candidates (Incumbent and party leader hypotheses).

When discussing the effect of candidates, it is important to consider the actual presence of the party’s candidate on the ballot paper. Each coalition is formed by one candidate and at least one party, but in most cases there are several parties endorsing one candidate. I defined as ‘frustrated’ the voters who prefer a party that is endorsing another party’s candidate on the majoritarian tier of the ballot paper. It is reasonable to expect that frustrated voters will be more likely to split their vote than voters who are non-frustrated. At the district level, the presence of the candidate on the majoritarian tier is controlled for by building a variable which takes a value of 1 when the party did not run its own candidate and 0 otherwise (Frustrated party hypothesis). When using survey data, a variable is built to control whether the party supported by the voter on the proportional tier ran its own candidate or not (Frustrated voter hypothesis). Employing the same logic, an increase in ticket-splitting and lower vote gaps is expected where there are fewer candidates; thus, I control for the number of candidates running on the majoritarian ballot. Overall, I expect candidate-centred voting to increase the vote gap and decrease split-ticket voting.

Based on the fact that a vote for a candidate is de facto a vote for a coalition, higher vote gaps and lower split-ticket voting can also come about due to the specificity of pre-electoral coalition agreements. To distinguish the specific effect of pre-electoral coalition agreements, I include indicators of the quality of such agreements (ideological congruence and coalition experience) which should decrease the vote gap and split-ticket voting.

First, ideological similarity: the more congruent the ideological positions of the coalition partners, the more likely it is that supporters of the constituent parties will support the coalition at the polls. Considering the lack of information on the policy positions of parties at the Italian regional level, ideological similarity is empirically controlled using a variable which takes a value of 1 when the pre-electoral coalition does not include extremist parties and 0 otherwise. In addition, I consider whether or not the coalition underwent controversial discussion before the elections (Policy-congruent coalition hypothesis). For instance, with regard to the 2010 election, in those regions where the Unione di Centro, a centre-right party, joined the centre-left coalition provoking a split of other left-wing coalition allies, the centre-left coalition has been identified as a non-congruent coalition. Second, the size of the coalition is considered: supporters of junior coalition partners may defect more easily than voters of the major coalition parties (Small party hypothesis). Third, time: the likelihood of voters following their party into a coalition should increase as the time elapsed since the establishment of the pre-electoral coalition increases (Coalition experience hypothesis). I built a variable which takes a value of 1 when the parties form the same coalition as in the previous election-year and 0 otherwise. Ideological congruence and coalition experience should decrease both the vote gap and split-ticket voting.⁶

Findings

Figure 1 shows values of the coalition vote gap on the left and split-ticket on the right. As mentioned, the vote gap is measured as the percentage difference between the votes gained by the coalition/candidate and those obtained by the linked party/parties at the provincial district level (on the total of each coalition vote). Figure 1 shows that the values of the vote gap are often below 5% and that these values are rarely negative: this means that in the majority of cases, coalitions/candidates outperform the parties linked to them. There is a striking difference between the two big coalitions and the smaller ones where the latter tend to obtain larger vote gaps. On the right, Figure 1 shows percentages of split voting by party (on the total of the party vote) at the provincial district level as estimated by Rosen et al.’s method. The figure indicates that values of split-ticket voting are much higher for smaller coalitions when compared to the two big coalitions.

Table 1 shows results for six multivariate models. In the vote gap analysis (Models 1–3), the dependent variable is the coalition vote gap. The total number of observations for each year of election is obtained by multiplying the number of coalitions contesting each provincial district by the total number of provincial districts. Models 1–3 for the vote gap are estimated using Ordinary Least Squares (OLS). As the spread of the residuals is somewhat wider towards the middle right of the graph than at the left, suggesting some heteroscedasticity, a robust option for estimating the standard errors using the Huber–White sandwich estimators is included. In addition, the dataset contains data on coalitions from 89 districts in 15 regions, so it is possible that the scores within each district may not be independent, and this could lead to residuals that are not independent within districts; consequently, I treat errors as clustered by district within regions.

Table 1.

Explaining Coalition-Level Vote Gap and Party-Level Split-Ticket Voting, Pooled Data (1995–2010).

	Dependent variable: COALITION vote gap (OLS)			Dependent variable: PARTY split-ticket vote (WLS)
	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
Incumbent candidate	−0.03***		−0.04**	−0.05**		−0.01
Incumbent candidate	(0.01)		(0.01)	(0.02)		(0.02)
Party leader candidate	0.09***		0.10***	−0.09***		−0.10***
Party leader candidate	(0.01)		(0.01)	(0.03)		(0.02)
First place	−0.26***		−0.25***	−0.60***		−0.60***
First place	(0.01)		(0.02)	(0.02)		(0.03)
Second place	−0.24***		−0.22***	−0.51***		−0.56***
Second place	(0.01)		(0.01)	(0.04)		(0.03)
First × District competitiveness	0.03		0.03	−0.23*		−0.05
First × District competitiveness	(0.05)		(0.05)	(0.09)		(0.09)
Second × District competitiveness	0.00		0.01	−0.33*		−0.25*
Second × District competitiveness	(0.05)		(0.04)	(0.13)		(0.11)
Lower × District competitiveness	0.14		0.13	−0.00		0.03
Lower × District competitiveness	(0.14)		(0.14)	(0.08)		(0.08)
Policy-congruent coalition		−0.14***	0.00		−0.29***	−0.04*
Policy-congruent coalition		(0.01)	(0.01)		(0.03)	(0.02)
Coalition experience		−0.15***	−0.05**		−0.32***	−0.02
Coalition experience		(0.02)	(0.01)		(0.03)	(0.02)
Frustrated party					0.09***	0.06***
Frustrated party					(0.02)	(0.01)
Small party					0.21***	0.14***
Small party					(0.03)	(0.02)
Party stronghold			0.03			−0.06**
Party stronghold			(0.02)			(0.02)
Number of SMD candidates	0.01*	0.02***	0.01*	0.02*	0.05***	0.02**
Number of SMD candidates	(0.00)	(0.00)	(0.00)	(0.01)	(0.01)	(0.01)
Constant	0.28***	0.19***	0.27***	0.59***	0.24**	0.57***
Constant	(0.02)	(0.02)	(0.02)	(0.06)	(0.08)	(0.05)
Observations	1412	1412	1412	4149	4149	4149
Adjusted R²	0.30	0.22	0.31	0.62	0.45	0.66
AIC	−601.55	−456.87	−606.22	−468.43	1022.62	−950.50

SMD: single-member district; AIC: Akaike information criterion.

Source: Own elaboration from data available at www.elezionistorico.interno.it

Entries are results from Ordinary Least Squares (OLS) and Weighted Least Squares (WLS) analyses.

All models include dummy variables for each year of election (not shown).

Robust standard errors in parentheses: *p < 0.05; **p < 0.01; ***p < 0.001.

When it comes to split-ticket voting, the dependent variable measures the percentage of split voting by party on the total of the party vote at the provincial district level as estimated by Rosen et al.’s method. Because in this case the dependent variable is constituted by estimates, I employ Weighted Least Squares (WLS) regression. The use of the WLS is recommended by the methodological literature (Adolph et al., 2003) and commonly used for regression models using ecological estimates as dependent variable (e.g. Burden and Kimball, 1998). In the WLS analysis, observations (estimates) are weighted by the inverse of the estimates’ standard error as provided by Rosen et al.’s method, thus giving greater weight to observations with more precise estimates. Models 4–6 are estimated using robust errors again with provincial districts being clustered in regions. Finally, because the models in Table 1 employ pooled data (1995–2010), four dummy variables, one for each year of election, are included (but not shown in the table). For all models presented in Table 1, substantive results are unchanged if two-sided tobit models are estimated to account for a dependent variable censored at 0 and 1 or when using beta regression to account for a dependent variable measured as a proportion. Adding a random effect instead of a Huber–White correction does not change substantive conclusions.

In the hypotheses section, I listed several arguments which refer broadly to candidates, coalitions and parties. Starting with candidates, Model 1 for the vote gap and Model 4 for the split-ticket voting, respectively, show the effect of candidate-specific factors. In line with the incumbent and party leader hypothesis, we find that an incumbent candidate has the effect of decreasing both the vote gap and split-ticket voting, whereas running a party leader increases the vote gap and reduces the amount of split-ticket voting. If anything, one expects the competitiveness of the district race in Italy to have a positive effect on the vote gap and a negative effect on split-ticket voting in the case of the two top-ranked candidates. These results would be in line with personal voting for candidates rather than strategic considerations. Table 1 demonstrates that this is indeed what we find but only for split-ticket voting, whereas all the district variables are insignificant in the vote gap model.

Moving to the effect of pre-electoral agreements, Model 2 and Model 5 show that a policy-congruent coalition and coalition experience are both likely to decrease the vote gap and they reduce defection. Because the analysis of split-ticket voting is at the party level, Model 5 includes controls for the size of the party entering the coalition as well as whether or not the party runs its own candidate. The results suggest that junior coalition partners are characterised by higher percentages of split-ticket vote than large coalition partners. Similarly, parties which do not present their own candidate are characterised by higher levels of split-ticket voting. Running complete models increases the explanatory power as the higher value of the Adjusted R² suggests. In complete models, the size and the effect of candidate-specific variables remain about the same, whereas coalition- and district-specific factors have a lower effect. Concerning the party stronghold hypothesis, Table 1 shows that long-term voting predictors are important for explaining loyalty to the coalition: where parties are historically strong, we find higher vote gaps and lower levels of split-ticket voting, despite the results being statistically significant only for the latter. Finally, if anything, one expects more ticket-splitting and less vote gap in regions where there are fewer candidates/coalitions than where there are more. The results indicate, however, that a larger number of candidates slightly increase both the vote gap and split-ticket voting.

Overall, the findings suggest that pre-electoral coalition features are important for explaining both the vote gap and split-ticket voting: they represent the gluing effect which helps parties retain their voters on the majoritarian tier. On the other hand, the features of the candidates relate more to the magnetic effect which helps to explain whether or not the candidates will outperform the parties linked to them.

An Analysis at the Individual Level

This section uses surveys to extend the evidence found using aggregate data by including individual-level determinants. It is also possible to check whether similar patterns are to be found when switching the level of observation. Although a direct comparison between aggregate- and individual-level data should be done with caution, as the aggregate-level analysis uses pooled data and the surveys only cover the 2010 election, the use of individual-level data, using the same independent variables while changing the level of investigation from the party to the individual, can still provide a test for the findings presented so far.

The dependent variable in this section is dichotomous: whether or not the voter casts a split-ticket vote. Consequently, logit regression models are used. Models 1–3 in Table 2 present the results using only the voting predictors used in the aggregate analysis; Model 4 uses only individual-level predictors, while Model 5 presents the results obtained combining individual- and aggregate-level variables. I estimate random effects logit models to account for the fact that we are considering both individual- and aggregate-level indicators.

Table 2.

Explaining Split-Ticket Vote at the Individual Level, 2010 Election: Random Effects Logit Coefficients.

	Dependent variable: straight vs split-ticket vote
	Model 1	Model 2	Model 3	Model 4	Model 5
Incumbent candidate	0.07		0.01		0.06
Incumbent candidate	(0.23)		(0.31)		(0.31)
Party leader candidate	0.38		−0.13		−0.13
Party leader candidate	(0.37)		(0.39)		(0.39)
First place	−3.48***		−3.04***		−3.08***
First place	(0.25)		(0.29)		(0.29)
Second place	−3.02***		−2.82***		−2.81***
Second place	(0.21)		(0.22)		(0.22)
First × District competitiveness	−0.46		−0.31		−0.41
First × District competitiveness	(1.33)		(1.35)		(1.35)
Second × District competitiveness	1.42		0.16		0.06
Second × District competitiveness	(1.31)		(1.36)		(1.36)
Lower × District competitiveness	5.03**		4.00*		4.25**
Lower × District competitiveness	(1.64)		(1.63)		(1.64)
Policy-congruent coalition		−1.72***	−0.91***		−0.87***
Policy-congruent coalition		(0.27)	(0.25)		(0.25)
Coalition experience		−0.09	−0.10		−0.14
Coalition experience		(0.22)	(0.26)		(0.26)
Party stronghold		−0.88***	−0.61**		−0.64**
Party stronghold		(0.19)	(0.22)		(0.23)
Frustrated voter		−0.24	−0.13		−0.16
Frustrated voter		(0.13)	(0.16)		(0.16)
Small party		0.73***	0.72***		0.78***
Small party		(0.19)	(0.20)		(0.20)
Number of SMD candidates	0.15	0.10	0.17		0.16
Number of SMD candidates	(0.10)	(0.09)	(0.10)		(0.10)
Party identification				−0.55***	−0.66***
Party identification				(0.14)	(0.15)
Education				−0.34*	−0.46**
Education				(0.13)	(0.14)
Political interest				−0.42***	−0.37***
Political interest				(0.10)	(0.11)
Constant	−0.65	−2.15***	−0.45	−1.90***	0.03
Constant	(0.45)	(0.44)	(0.49)	(0.10)	(0.50)
Observations	5038	5038	5038	5038	5038
Groups	74	74	74	74	74
LL	−1414.41	−1531.33	−1392.15	−1561.79	−1368.90
Rho	0.21	0.21	0.21	0.04	0.21
Rho	(0.04)	(0.06)	(0.04)	(0.01)	(0.04)

SMD: single-member district; LL: log-likelihood.

Source: Ipsos opinion poll regional elections 2010.

Entries are results from random effects logit models. Standard errors in parenthesis: *p < 0.05; **p < 0.01; ***p < 0.001. The models include only respondents who cast both votes. Sample respondents who cast both a party and a candidate vote are as follows (with regional population as for 2013 in parenthesis): 478 respondents in Piemonte (population: 4,436,798), 1034 in Lombardia (9,973,397), 550 in Veneto (4,926,818), 163 in Liguria (1,591,939), 497 in Emilia-Romagna (4,446,354), 376 in Toscana (3,750,511), 70 in Umbria (896,742), 642 in Lazio (5,870,451), 564 in Campania (5,869,965), 366 in Puglia (4,090,266), 73 in Basilicata (578,391) and 225 in Calabria (1,980,533).

The findings are broadly consistent across surveys and aggregate data. First and foremost, the effect of candidate features, such as incumbency, is again negative, but it is very small and fails to show significance across all models. With regard to district competitiveness, we again find results in line with personal voting for candidates rather than strategic considerations. Concerning coalition-specific effects, we can confirm that a policy-congruent coalition is likely to reduce defection; however, the variable ‘coalition experience’ is now not statistically significant. Model 4 shows results including only individual-level indicators. The findings suggest that voters with higher levels of political interest are less likely to split as one would have expected from the analysis of the electoral system rules. The model also tells us that having a party attachment decreases the probability of splitting the ticket. Model 5 in Table 2 brings individual- and aggregate-level indicators together. Whereas the broader conclusions do not change for each of the coefficients, the power of the model is now at its highest, suggesting that the simultaneous account of both factors is indeed an important one.

Concluding Remarks

One of the basic and most-often used mechanisms to coordinate entry into the electoral market is the formation of pre-electoral coalitions (Golder, 2006). Although it is theoretically relevant, the effect of pre-electoral coalitions on vote choice remains a largely unexplored area in the field of party strategy. This article attempts to fill this gap in the literature employing the Italian regional elections. The electoral rules for the Italian regional elections allow parties to run alone, and, at the same time, to present pre-electoral coalitions on the electoral ballot. Voters can express two votes: one for a party and one for a candidate representing a pre-electoral coalition. This peculiarity offers an opportunity to test under what conditions supporters follow their initially preferred party into the coalition as well as more conventional vote choice explanations concerning primarily parties and candidates.

First and foremost, as was the case for the experimental study conducted by Gschwend and Hooghe (2008), the overwhelming majority of voters vote for their preferred party and the coalition that this party indicated to coalesce with. This behaviour may come about as a consequence of the fact that the vote is party-centred or strategic. With regard to parties, the analysis showed parties are very important to explain vote choice and that voters are paying primary attention to party factors in deciding how to vote. Conversely, none of the variables used to control for strategic voting point to the conclusion that a straight or split-ticket vote is a consequence of strategic coordination by voters.

The article also inquired into what causes people to split their vote. When it comes to candidates, as voting on the majoritarian tier enables Italians to choose the future president of their regions, a strong effect was anticipated. Instead, the results indicate that sometimes candidates exert a magnetic effect attracting votes beyond those of the linked parties, but they are not able to perform a gluing effect stopping voters from splitting to other coalitions. These results nicely explain why Gschwend and Hooghe (2008) find a weak effect of candidates in their experimental study.

The article also controlled for several factors correlated with pre-electoral coalitions. The analysis demonstrates that the more congruent the ideological positions of the coalition partners, the more likely it is that supporters of the constituent parties will support the coalition at the polls. The size of the coalition partners is, however, a crucial element and supporters of small parties are much more likely to desert their party’s coalition. It can therefore be confirmed that when negotiating a pre-electoral coalition, party elites have every reason to be responsive to the wishes of the small coalition party’s supporters because they are the first who desert the pre-electoral coalition. Conversely, there is little evidence to support the conclusion that the time elapsed since the establishment of the pre-electoral coalition is an important factor in determining split-ticket voting.

These results have implications for analysing voting behaviour and coalition politics. If voters know that coalition governments are likely to form, they are inclined to develop preferences regarding potential coalitions (Hobolt and Karp, 2010) and their vote is subject to these preferences and expectations. For coalition politics, the results confirm recent findings in Germany (Debus and Müller, 2014) that parties are still the decisive actors in the government formation process, yet they are restricted in their choices by the coalition preferences of the electorate which they need to take into account. The lesson of this study for Italian electoral behaviour and beyond is that future research should look into the effect of pre-electoral agreements on vote choice, a subject of political science that cries out for research.

Footnotes

Appendix

Acknowledgements

The author would like to thank Kenneth Benoit, Lorenzo De Sio, Michael Marsh, Markus Wagner and the anonymous reviewers for helpful comments on earlier versions of this article which was presented at the 2011 meeting of the Italian Political Science Association (SISP). All errors are mine.

Funding

The author gratefully acknowledges the financial support offered by the Irish Research Council (IRC) during her doctoral studies at Trinity College Dublin, Ireland. Calculations were performed on the Lonsdale cluster maintained by the Trinity Centre for High Performance Computing. This cluster was funded through grants from Science Foundation Ireland.

Supplementary Information

Additional supplementary information may be found with the online version of this article.

S1: The Italian regional electoral law.

S2: Discussion of existing ecological estimation methods.

S3: Descriptive statistics of dependent and independent variables.

Notes

Author Biography

Carolina Plescia is Assistant Professor in the Department of Methods in the Social Sciences at the University of Vienna. She obtained a PhD in Political Science from Trinity College Dublin, Ireland, in November 2013. Her PhD thesis was awarded the 2014 European Consortium for Political Research (ECPR) Jean Blondel Prize for the best thesis in Politics. Her main research interests include comparative electoral behaviour, coalition governments, representation and research methodologies.

References

Adolph

King

Herron

. (2003) A Consensus on Second-Stage Analyses in Ecological Inference Models. Political Analysis 11 (1): 86–94.

Agnew

(2002) Place and Politics in Modern Italy. Chicago, IL: University of Chicago Press.

Bawn

(1999) Voter responses to Electoral Complexity: Ticket Splitting, Rational Voters and Representation in the Federal Republic of Germany. British Journal of Political Science 29 (3): 487–505.

Bellucci

Segatti

(2011) Votare in Italia. Dall’appartenenza alla scelta. Bologna: Il Mulino.

Benoit

Giannetti

Laver

(2006) Voter Strategies with Restricted Choice Menus. British Journal of Political Science 36 (3): 459–485.

Bowler

Karp

Donovan

(2010) Strategic Coalition Voting: Evidence from New Zealand. Electoral Studies 29 (3): 350–357.

Burden

Kimball

(1998) A New Approach to the Study of Ticket Splitting. American Political Science Review 92 (3): 533–544.

Burden

(2009) Candidate-Driven Ticket Splitting in the 2000 Japanese Elections. Electoral Studies 28 (1): 33–40.

Campbell

Miller

(1957) The Motivational Basis of Straight and Split Ticket Voting. American Political Science Review 51 (2): 293–312.

10.

Cox

(1997) Making Votes Count: Strategic Coordination in the World’s Electoral Systems. Cambridge: Cambridge University Press.

11.

D’Alimonte

(2001) Mixed Electoral Rules Partisan Realignment and Party System Change in Italy. In: Shugart

Wattenberg

(eds) Mixed-Member Electoral Systems: The Best of Both Worlds. Oxford: Oxford University Press, pp.323–350.

12.

D’Alimonte

(2005) Italy: A Case of Fragmented Bipolarism. In: Gallagher

Mitchell

(eds) The Politics of Electoral Systems. Oxford: Oxford University Press, pp.253–276.

13.

D’Alimonte

Bartolini

(1998) How to Lose a Majority: The Competition in Single-Member Districts. European Journal of Political Research 34 (1): 63–106.

14.

Debus

Müller

(2013) Do Voters’ Coalition Preferences Affect Government Formation? West European Politics 36 (5): 1007–1028.

15.

Debus

Müller

(2014) Expected Utility or Learned Familiarity? The Formation of Voters’ Coalition Preferences. Electoral Studies 34 (1): 54–67.

16.

Di Giovine

Pizzetti

(1996) Osservazioni sulla nuova legge elettorale per i Consigli regionali. Le Regioni 1: 411–423.

17.

Di Virgilio

(2002) L’offerta elettorale: la politica delle alleanze si istituzionalizza. In: D’Alimonte

Bartolini

(eds) Maggioritario finalmente? La transizione elettorale 1994–2001. Bologna: Il Mulino, pp.79–129.

18.

Di Virgilio

(2005) The Italian Regional Elections of April 2005: Does the Triumph of the Union Signal the End of the Berlusconi Era? South European Society and Politics 10 (3): 477–490.

19.

Di Virgilio

(2007) La costruzione dell’offerta politica: progettazione nazionale, realizzazione regionale. In: Chiaramonte

Tarli-Barbieri

(eds) Riforme istituzionali e rappresentanza politica nelle Regioni italiane. Bologna: Il Mulino, pp.117–144.

20.

Duverger

(1954) Political Parties. New York: Wiley.

21.

Ferrara

Herron

(2005) Going It Alone? Strategic Entry under Mixed Electoral Rules. American Journal of Political Science 49 (1): 16–31.

22.

Fisher

(2004) Definition and Measurement of Tactical Voting: The Role of Rational Choice. British Journal of Political Science 34 (1): 152–166.

23.

Golder

(2005) Pre-Electoral Coalitions in Comparative Perspective: A Test of Existing Hypotheses. Electoral Studies 24 (4): 643–664.

24.

Golder

(2006) Pre-Electoral Coalition Formation in Parliamentary Democracies. British Journal of Political Science 36 (2): 193–212.

25.

Gschwend

(2007) Ticket-Splitting and Strategic Voting under Mixed Electoral Rules: Evidence from Germany. European Journal of Political Research 46 (1): 1–23.

26.

Gschwend

Hooghe

(2008) Should I Stay or Should I Go? An Experimental Study on Voter Responses to Pre-Electoral Coalitions. European Journal of Political Research 47 (5): 556–577.

27.

Hobolt

Karp

(2010) Voters and Coalition Governments. Electoral Studies 29 (3): 299–307.

28.

Kabashima

Reed

(2000) Voter Reactions to Strange Bedfellows: The Japanese Voter Faces a Kaleidoscope of Changing Coalitions. Japanese Journal of Political Science 1 (2): 229–248.

29.

King

(1997) A Solution to the Ecological Inference Problem: Reconstructing Individual Behavior from Aggregate Data. Princeton, NJ: Princeton University Press.

30.

Laver

(1998) Models of Government Formation. Annual Review of Political Science 1 (1): 1–25.

31.

Martin

Stevenson

(2001) Government Formation in Parliamentary Democracies. American Journal of Political Science 45 (1): 33–50.

32.

Martin

Vanberg

(2003) Wasting Time? The Impact of Ideology and Size on Delay in Coalition Formation. British Journal of Political Science 33 (2): 323–332.

33.

Moser

Scheiner

(2005) Strategic Ticket Splitting and the Personal Vote in Mixed-Member Electoral Systems. Legislative Studies Quarterly 30 (2): 259–276.

34.

Niemi

Written

Franklin

(1992) Constituency Characteristics, Individual Characteristics and Tactical Voting in the 1987 British General Election. British Journal of Political Science 22 (2): 229–240.

35.

Pappi

Thurner

(2002) Electoral Behaviour in a Two-Vote System: Incentives for Ticket Splitting in German Bundestag Elections. European Journal of Political Research 41 (2): 207–232.

36.

Powell

(2000) Elections as Instruments of Democracy: Majoritarian and Proportional Visions. New Haven, CT: Yale University Press.

37.

Renwick

Hanretty

Hine

(2009) Partisan Self-Interest and Electoral Reform: The New Italian Electoral Law of 2005. Electoral Studies 28 (3): 437–447.

38.

Robinson

(1950) Ecological Correlations and the Behavior of Individuals. American Sociological Review 15 (3): 351–367.

39.

Rosen

Jiang

King

. (2001) Bayesian and Frequentist Inference for Ecological Inference: The RXC Case. Statistica Neerlandica 55 (2): 134–156.

40.

Shugart

Valdini

Suominen

(2005) Looking for Locals: Voter Information Demands and Personal Vote-Earning Attributes of Legislators under Proportional Representation. American Journal of Political Science 49 (2): 437–449.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.48 MB