Preferences for Political Coalitions in Spain

Spanish political parties opt whether or not to form alliances with other parties for a number of reasons, one of which is the coalition preferences of their voters. This work explores to what extent party preferences and ideological proximity on the two main dimensions of political competition in Spain affect voters' coalition preferences. Using survey data from 2009, we find that voters' coalition preferences are greatly affected by ideological proximity, especially on the left–right axis, whereas decentralisation preferences are much more salient for the supporters of regionalist parties. However, this general pattern varies depending on the coalition being assessed.

opposition benches. Second, they evaluate the consequences of political alliances for intra-party cohesion. And, last but not least, they assess their vote-seeking objectives in future national and regional elections.
These claims are in line with Strom's (1984;1990) consideration of minority governments as rational solutions, under certain conditions. In the case of Spain, opposition parties are often able to exercise influence over policy while at the same time avoiding the attrition resulting from participating in government. Similarly, Müller and Strom (1999) argue that parties frequently forgo immediate office-seeking goals in exchange for future vote-seeking benefits. For all these reasons, Spanish singleparty minority cabinets appear to be a rational outcome in the government formation process.
However, it is important to note that the vote-seeking argument has an implicit builtin assumption regarding the reaction of voters to inter-party agreements. If Spanish parties are willing to remain in opposition without tasting the sweetness of government, it is because they anticipate, among other factors, the response of voters. If parties trade office for votes, this must be because voters may react badly to specific coalitional strategies. Although voters seem to be significant actors in this story, their reactions to actual or potential coalitions, and the reasons underlying them, are not at all clear.
In this work we seek to investigate this issue in detail. We will first examine the general influence of party preferences on coalition preferences. That is, we will analyse whether having voted for a certain party affects what type of government Spanish voters prefer and which specific coalitions they support or oppose. Also, we will consider whether there are further elements that help voters make up their minds about these preferences. Second, we will investigate the role of ideological proximity on the two traditional axes of party competition in Spain, specifically on voters' preferences for coalitions. We will also focus in on various groups of party supporters in order to evaluate which ideological dimension weighs most heavily in their building of coalition preferences. That is, in a multidimensional space of political competition, potential coalitions can be distant on one axis but close on another, and that scenario requires voters to decide which dimension is most salient for them to support, or to oppose, a given inter-party agreement.

Coalition Preferences: What Do We Know?
Parliamentary systems are prevalent all over Western Europe. Elections in these systems can produce a clear winner who controls an absolute majority of seats and who is able to form a single-party majority government. However, most often various parties need to ally with each other in order to reach the necessary 50 per cent seat share threshold to be able to pass laws and make policy. These deals can take the form of coalition governments or policy-specific agreements between opposition parties and the minority government in office.
Although these kinds of inter-party pacts are extremely frequent in multi-party parliamentary systems, we still know relatively little regarding what voters think about them. Clearly, parties accept or reject coalitions for reasons that go beyond the reaction of the electorate yet, they are also concerned about what their voters think about them. It is therefore surprising that '[w]orks on coalition formation and survival have traditionally focused on tactical and policy considerations at the elite level, and frequently neglected the role of voters' (Jou 2010). As a result, the very notion of voters' coalition preferences has received scant attention in previous political science research.
In recent years various studies have emphasised the role of coalition preferences in predicting voting behaviour. According to Kedar (2005) voters can, and often do, exhibit outcome-oriented behaviour by sacrificing their vote for the party that is closest to their own position and opting for another party in order to make the formation of their preferred coalition more likely. Evidence of this has been provided for the independent role of coalition preferences on voters' behaviour for countries such as Austria (Meffert & Gschwend 2010), Belgium (Gschwend & Hooghe 2008), and Israel (Bargsted & Kedar 2009;Blais et al. 2006), among others. However, the dependent variable for these studies has still been party choice, rather than coalition preferences.
Additionally, Jou (2010) has focused on how policy preferences of voters constrain parties' coalition choices and affect their subsequent electoral performance in New Zealand and Japan. Nonetheless, this author does not deal with voters' coalition preferences as such, and the main aim of his study is to understand parties' coalitional strategies and not voters' coalition tastes.
Hence, the more basic question of what determines coalition preferences in the first place (that is, taking them as the dependent variable) has seldom been addressed. But do voters have coalition preferences? Are voters prepared to evaluate abstract objects such as coalitions? In a spatial scenario, voters place themselves in different dimensions. From a proximity perspective, they should rank their preferences with regard to parties and coalitions: those closest to them in the highest positions, those further away, last. This, however, assumes that voters are able to identify the location of the objects to be ranked. For parties, the problem is not so serious. Parties are at the centre of political life. They are the objects to which the public and media pay attention. Voters receive information about their positions and platforms in campaigns, but also during the legislative term. For coalitions, though, the situation is more ambiguous. Except in very specific cases, coalitions are hypothetical constructs with positions and characteristics that are not immediately retrievable by voters. Voters' views on coalitions may merely be derived from the views of their individual members. However, there may be other factors at play that help us explain why some coalitions are preferred over others.
The study of the predictors of coalition preferences dates back to Norpoth (1980), who analysed the preferences of the (West) German public with respect to the coalitions formed by party leaders during the 1961 -76 period. Almost twenty years later, Thurner and Pappi (1999) also chose the German case and addressed the question of whether or not coalition preferences mirrored party and candidate preferences. However, they did so only in passing, as their main goal was to analyse voters' choice at election polls, rather than coalition preferences themselves.
Very few other studies have enquired about the explanatory factors behind voters' preferences for coalitions. In one, Meffert, Gschwend, and Schütze (2009) found that the party preferences of Austrian voters precede their coalition preferences, because a longer time is spent answering survey questions related to the latter. Nonetheless, they also claimed that coalition preferences are more than just weighted averages of party preferences. They concluded that party preferences are indeed the most important predictors of preferences for pacts, but they only explain a limited amount of the variance. Candidate evaluations also seem to play a significant role whereas, quite surprisingly, ideological and policy-based explanations largely fail to provide any further explanation. They conclude by arguing for the necessity of future research, which the present work in part seeks to address (Meffert, Gschwend & Schütze 2009, p. 14): The considerable unexplained variance rather suggests that other factors must play a role as well that were not captured in the analysis. But most important, it implies that coalition preferences are to a considerable degree independent of party preferences and a unique and necessary predictor of vote decisions. The challenge to future research is clear. Not only have these findings to be replicated for other countries with multiparty systems, but better explanations about the sources and origins of coalition preferences are necessary.
An interesting scenario to do that is precisely one in which no formal coalition governments have formed, such as in Spain. It is true that if voters' evaluation of abstract constructs such as coalitions is difficult in most party systems, it should be even more complex in a country like Spain. 1 However, as previously mentioned, Spanish voters are extremely familiar with inter-party agreements and also with coalition cabinets at regional levels. On the other hand, the absence of coalition governments could also be viewed as an advantage, as voters will not be biased in favour of certain arrangements that they have (or have not) experienced before, versus others. In any case, it is intriguing to ask why these governments have not formed at national level. Knowing more about the attitudes of voters towards coalitions could provide us with clues to better understand Spanish (non-)coalitions. This is the key aim of this paper.

Coalition Preferences in Spain: What Should We Expect?
The political space in Spain is traditionally characterised by the presence of two dimensions of electoral competition. To the traditional left -right dimension one has to add a second axis regarding the territorial organisation of the state: more centralised vs. more decentralised. The existence of heterogeneous feelings of national belonging in the different territories of the country shapes Spaniards' preferences towards decentralisation and the Spanish political system as a whole (Moral 1998;Fernández-Albertos 2002). Obviously, voters and parties place themselves on more dimensions, since there are issues that fall beyond the scope of either of the two. Nevertheless these two better reflect the Spanish political space and can also be considered to serve as a heuristic for many other issues.
The placement of the Spanish political parties in this two-dimensional space can be identified in a number of ways, one being how voters view them. Figure 1 shows the positions of the PSOE, PP, IU, and the most significant regionalist parties in these two political axes, ranging from zero to ten, according to Spanish voters in 2009. 2 Although the ideological variation across Spanish regionalist parties is worth noting, we only plot the placement of the regionalist parties CiU, ERC, and PNV for presentational purposes. 3 The circles represent the averages of parties' attributed placements, which seem to be consistent with conventional wisdom, at least from the perspective of the resulting rank-orderings. The IU is the leftist state-wide party, followed by the regionalist ERC, the centre-left PSOE, and then the conservative PP, with the two largest regionalist parties (CiU and PNV) more or less in between the latter two. On the axis relating to the territorial organisation of the state, the PP appears as the most pro-centralisation, followed by the PSOE and IU. The regionalist parties obviously emerge as the most pro-decentralisation.
Given the positions of Spanish political parties on the two main dimensions of political competition in the country, we can easily imagine the placement of the coalitions that could credibly form in Spain. 4 As coalitions can be considered to be the sum of their parts, we should logically expect the location of the potential coalitions in this two-dimensional space to be consistent with that of their member parties. Hence, the portrayal of the political space of Spanish (credible) coalitions on each axis would look something like in Figure 1 (coalitions in triangles).
These mean locations are simply the result of averaging the placement attributed by the interviewees to each potential partner. This provides a rough assessment of the spatial location of coalitions on the two main axes of political competition in Spain, Given this spatial illustration, and also that the only possibility for a single-party government to form would be an office run by either the PSOE or the PP, we can now present what we expect about the coalition preferences of Spanish voters. First, regarding preferences for the generic type of government, we could say the following: IU voters would (1) prefer a single party not to win an absolute majority of seats in parliament (namely, they would like to avoid a single-party majority government of the PSOE or the PP), but (2) it is not so clear whether they would prefer a coalition or a single-party minority government in generic terms. The latter would clearly depend on the specific coalition being formed. Therefore, generically, IU voters are not expected to have any clear preference for a coalition government.
PSOE voters would prefer a single party to win an absolute majority of seats in parliament only if the winning party was the PSOE, but not if it was the PP. Hence, (1) PSOE voters' generic preferences for single-party majority governments as opposed to single-party minority or coalition offices have no clear direction. (2) As for the dilemma between the latter two, the answer would again depend on who would be the leading party. If it was the PSOE, then a single-party minority government would be preferred, as it would leave the party in the best position to negotiate with the other parties, depending on the specific policy. Conversely, though, any PP-led coalition would probably be better than a single-party PP government from the PSOE voters' view.
For PP voters, the situation should be very similar: (1) no clear preference regarding a single-party majority vs. a minority or a coalition government, and (2) no clear preference regarding what type of government should form if no single-party obtained an absolute majority of seats, whether it be a multi-party or a single-party cabinet. Just as with PSOE voters, therefore, the preferences for the type of government are too context specific to derive a clear-cut expectation.
For the voters of regionalist parties it is clear that they would prefer no party (i.e. the PSOE or PP) to win an absolute majority of seats, since that would give their parties no opportunity to influence policy-making at all. Regarding the second choice between a coalition and a single-party minority government, regionalist voters would prefer one or the other depending on the specific coalition. The CiU and PNV may benefit more from a PP or a PSOE minority government rather than a PSOE-IU coalition or a PSOE -PP one, while ERC voters would more clearly prefer any coalition except an agreement between the two main parties.
Any further pattern emerging in the empirical analyses may respond to which coalitions voters believe are more likely to form (which would make it clearer whether a coalition or a single-party minority government is preferable) or to other more generic tastes for or against specific types of governments which are also unobservable. For instance, a single-party minority government could be preferred because of their tendency to offer median policy outcomes in each dimension, or less preferred because of their greater instability. None of this, though, can be captured by the data we have available.
Beyond generic preferences, this paper will also look at which specific coalitions Spanish voters prefer. First, simply looking at the party choice of voters, we can expect the preferred coalitions (among the menu of credible coalitions in Figure 1) to be as shown in Table 1. 5 The expectations summarised in the table are derived from a proximity perspective, where one assumes that voters will opt for the alternative closest to their own spatial placement. For the doubtful cases (IU and ERC voters when PP is the formateur), the two specified alternatives are almost at an equal distance to the placement of their own party. For instance, ERC voters may prefer the average PP -REG coalition from the point of view of the decentralisation dimension, but a PP -PSOE coalition would be closer to ERC's position on the left -right axis. Obviously, the specific choice would depend on which dimension was more salient for each voter. We will address the dimension saliency issue below.
It is worth mentioning that these are 'on average' expectations. It is clear that not all voters of a given party hold the same ideological position, and it is neither the case that all of them would place each coalition at the same point. We have thus derived the expectations of voters' coalition preferences from the attributed placement of their respective parties and the average position of the coalitions. Nonetheless, to evaluate the importance of ideological proximity on the coalition preferences of Spanish voters, we shall also take an individual-level perspective.
More specifically, we first expect that the farther a given voter places a specific coalition from themself, the less likely they are to prefer that coalition. This should be true for the two main axes of political competition. However, an open question would be (the proximity to) which of the two dimensions matters more for coalition preferences. As noted, both the traditional left-right dimension and that regarding the territorial organisation of the state are clearly prevalent in political debate. However, some have found that while the traditional left -right dimension matters for all Spaniards when it comes to the time to cast their ballot at the polls-regardless of their partisanship-the decentralisation dimension is only relevant for regionalist partisans (Aguilar & Sánchez-Cuenca 2008). One could easily apply this rationale to the sphere of coalition preferences and hypothesise that while proximity on the left -right axis is going to matter for all Spanish voters, proximity on the decentralisation axis will only be relevant for the voters of regionalist parties. A more conservative approach would simply expect that for all voters but regionalists the left-right dimension will be more salient than the decentralisation one in their assessment of coalition alternatives, while for regionalists the two dimensions will be equally important. All these are Spain-specific hypotheses that can nonetheless be contextualised into a more general theoretical framework. We have first argued that ceteris paribus voters will generically prefer the type of government that is more likely to deliver policies closer to their preferred ones. Second, we have hypothesised that voters' coalition preferences will be guided by ideological proximity (especially) on the axes that are more salient for the parties they vote for. These arguments can easily apply to broader settings where more than one dimension of political competition is present and/or where there are significant single-issue parties. The scope of this paper, however, is limited to the case of Spain.

Data, Variables, and Methods
The empirical analysis is based on data from the Centro de Investigaciones Socioló gicas survey (2799), which took place in the spring of 2009 and contains various items measuring Spanish public opinion with respect to coalition preferences (for more details about this survey see also the introduction of this special issue). In this empirical study there are mainly two dependent variables. The first relates to the preferences of Spanish voters for the generic type of government, while the second refers to preferences for coalitions between specific parties. For the analysis of the former, we draw on one of the survey questions that asks whether the respondent prefers a party winning an absolute majority of seats or not and, if the latter, whether the preferred alternative would be a single-party minority government or a coalition cabinet. As a result, we run two separate probit regressions: one taking the dichotomous choice between singleparty absolute majority and not, and another considering the dichotomy between a coalition and a single-party minority government.
The second dependent variable of this study is voters' preferences for specific coalitions. Its operationalisation is based on two questions. The first asks the following: 'If general elections were to be held tomorrow and they were won by the PSOE without an absolute majority, which of the following alternatives would you prefer?' The choice set is composed of: (1) PSOE governing in coalition with IU; (2) PSOE governing in coalition with Unió n Progreso y Democracia (UPyD); (3) PSOE governing in coalition with regionalist parties; (4) PSOE governing in coalition with PP; (5) PSOE governing alone, followed by the conventional don't know, don't answer options. The same applies for the second question, with the PSOE substituted by the PP, although the question does not offer the respondent the bizarre possibility of a PP -IU coalition. 6 It is true that the question explicitly asks for governmental coalitions. Although there have been no coalition cabinets in the national arena, one might rather safely extend the logic of preferences for governmental coalitions and consider preferences for legislative coalitions-in Laver and Schofield (1990) terms-or inter-party agreements more generally. That is, we can interpret support for a given coalition as support for parties giving priority to pacts, deals, and agreements in legislative policy-making with one another, rather than with other parties. Given the number of minority governments that Spain has had, these sorts of agreements have been very prevalent.
For the analysis of the type of government generic preferences, the main independent variables are the standard ideological self-placement on a zero to ten left-right scale (Left-Right), and self-placement on the zero to ten axis that measures preferences for decentralisation (Preferences for Decentralisation). To differentiate among specific party voters, we have also included the variable Vote Recall on the 9 March 2008 elections, in the form of various groups of dummies depending on the particular econometric model. Additionally, we control for the experience of voters regarding different types of governments by incorporating the Share (of legislatures) with Single-Party Majority Governments in the Autonomous Community, the Share of Single-Party Minority Cabinets, and the Share of Coalitions. We include the first-mentioned in the analysis of the generic preference for single-party absolute majorities, whereas we incorporate the latter two in the analysis of the choice between a coalition and a single-party minority.
For the analysis of specific coalition preferences, we include the Vote Recall variables to identify which are the preferred coalitions for each group of voters. In the final analyses, the main independent variables measuring ideological proximity are Distance to Alternative (Left-Right) and Distance to Alternative (Preferences for Decentralisation), referring to the absolute difference between the self-placement of the individual and the placement they attribute to the coalition on the left -right and decentralisation dimension, respectively (for the exact absolute proximity formula and for different spatial models of issue voting see Queralt's article in this special issue). Since we do not have data on how voters locate coalitions spatially, we follow the same approach as graphically illustrated above and consider the placement of coalitions as a nonweighted average of the values attributed to the parties forming the coalition. 7 In order to evaluate whether the effect of proximity on each dimension varies across individuals with different partisan interests, we also interact the Distance to Alternative variables with PID (Party X), which is a dummy that captures whether the interviewee sympathises with party X. Given the alternative-specific nature of some of these variables, we use a conditional logistic regression as a statistical technique.
Finally, in order to isolate the effect of our main explanatory variables, several additional controls are introduced in the analysis of both generic and specific coalition preferences. These are standard socio-demographic variables such as Gender (Female), Age, and Level of Education dummies (none, primary, secondary, vocational training, tertiary (#3 years degree), tertiary (4 years degree or higher). Although all analyses are run controlling for these variables, the related estimates are not displayed in the tables for the sake of presentational simplicity. Table 2 presents the estimates regarding the type of government generically preferred by Spaniards. 8 The dependent variable in columns 1 -4 refers to preferences for (one) or against (zero) single-party absolute majorities in parliament. Ideology seems to exert the strongest effect, left-wing voters being more inclined to dislike single-party majority governments, compared with minority and coalition governments. Similarly, those who are more pro-decentralisation tend to prefer the latter to the former. Additionally, voters who live in Autonomous Communities with a higher prevalence of single-party majority governments tend to prefer, on average, absolute majorities. Regarding different groups of voters, their preferences tend to be in line with expectations, IU, ERC, and PNV voters being generally unfavourable to single-party absolute majorities. PP voters, on the other hand, prefer the latter, while for PSOE voters the direction of the preference is not clear. Rather surprisingly, CiU voters are not clearly against single-party absolute majorities. Although it is clear that at the nationwide level a single-party absolute majority would leave this regionalist party aside, it is also true that CiU has traditionally formed single-party governments in the regional arena, and that may explain the emerging pattern here.

Results
Columns 5-8 present the estimates for the dependent variable coding voters' generic preferences for coalitions (one) relative to single-party minority governments (zero). 9 Interestingly, ideological self-placement does not seem to matter in this particular dilemma, while party preferences only have explanatory power in a couple of cases. IU voters tend to prefer coalitions rather than single-party governments, while CiU voters again show a preference for single-party governments. Also, we find that the more the Autonomous Community of the voter has been under the rule of single-party minority cabinets, the less likely it is that they support these governments and the more likely they will prefer coalitions instead. As expected, we find less clear patterns for this dilemma due to the uncertainty regarding hypothetical coalition formation processes. Table 3 presents the conditional logit estimates to evaluate the effect of party choices on the preferences for PSOE-and PP-led coalitions (columns 1-3 and 4 -5, respectively). Regarding the former, the first thing we see is that party choices predominately influence coalition preferences in the expected directions. IU voters tend to prefer a PSOE-IU coalition, while regionalist party voters largely prefer a PSOE-REG coalition, and their most disliked alternative would be a grand coalition between PSOE and PP (especially for PNV and ERC voters). PP voters seem to reject any alternative that does not include their party in government (i.e. they have a clear preference for a PSOE -PP coalition and are against PSOE-IU and PSOE -REG). Interestingly, PSOE voters tend to favour a PSOE -IU coalition and dislike the other two alternatives. As for PP-led coalitions, an agreement between the PP and the regionalists is preferred over a PP minority government by all voters except, precisely, PP voters. On the other hand, a grand coalition led by the PP would be preferred by PSOE voters but not by those of the PP, while the voters of the regionalist PNV show an overt opposition to this coalition.
Concerning the ideological proximity arguments, we also ran conditional logit analyses and summarised the results in eight two-way plots (illustrated in Figure 2). These graphs show the influence of ideological proximity (distance to alternative), on each axis, on the conditional probability of the choice of the preferred coalition. 10 It is clear that the two main independent variables play an important role in voters'  coalition preferences. In the two leftmost plots we can see that the farther they place the parties forming the coalition from their self-placement in the ideological space, the less likely is that alternative to be preferred both on the left -right and decentralisation dimensions and for both PSOE-and PP-led coalitions. It is worth mentioning, though, that coalition preferences seem to be more sensitive to distance on the left -right axis than on the decentralisation dimension. This is in line with the expectations stated in the third section of this paper, as well as the previous literature. The downward slopes of the solid lines (left -right dimension) are steeper than the dashed ones (decentralisation dimension), although this is far clearer for the preferences about the coalitions led by the PSOE. However, this general pattern varies notably if we examine different partisan attachments in the remaining six plots. It is clear that PP sympathisers tend to place more emphasis on the left-right dimension, rather than decentralisation, to choose their preferred coalition. That is, they seem to accept coalitions that are distant from themselves on this dimension, but are much more sensitive to deviations on the leftright axis. In sharp contrast, regionalists seem to care little about the left -right dimension, whereas decentralisation preferences are much more salient for them when building their coalition priorities. Interestingly, IU partisans follow the general pattern in the case of PSOE-led coalitions, but when the PP is the formateur their coalition preferences seem to respond more to the decentralisation dimension than the leftright one. This probably has to do with their general aversion to a grand coalition of PP and PSOE relative to an agreement between PP and regionalist parties. Finally, those identified with PSOE follow the general trend in their preferences for PP-led coalitions, but their choice of coalition partners when their party is the formateur seems to respond more or less equally to both spatial dimensions.
It is perhaps a bit surprising that in a couple of rare cases the distance effects appear to change sign. This counter-intuitive pattern emerges for IU and regionalist partisans in the case of PP-led coalitions. The most plausible interpretation may be that in these cases the decentralisation dimension completely supersedes the left -right one, so that voters ignore the latter to the extent that it even appears to have an effect in the opposite direction of what one would expect. If this is the case, the identified patterns might be better viewed as spurious relationships created by the dominance of one dimension over the other, rather than an odd substantive finding on ideological proximity.

Concluding Remarks
This work has aimed to study voters' preferences for coalitions using the particular case of Spain after the 2008 elections. It has explored the coalition preferences of Spanish voters by concentrating on the effect of party preferences and ideological proximity on the two traditional dimensions of political competition in the country for different groups of partisans. We have first shown that Spanish voters' generic preferences for different types of cabinets are not only influenced by their party preferences, but also by their ideological placement on the left -right and decentralisation axes, as well as by the experience they have with different types of cabinets in their Autonomous Community. The further to the right, the more likely the voter prefers a single-party majority government (even after controlling for vote choice), and the more pro-decentralisation, the less likely. On the other hand, the preference between coalitions and single-party minorities seems to respond more to idiosyncratic factors that are less related to ideological and party preferences.
In our second set of analyses we have found that party preferences significantly determine specific coalition preferences (namely, support or opposition to governmental agreements between specific parties), but ideological proximity to the coalition is perhaps even more important. Our empirical results have revealed that voters are more sensitive to proximity on the left-right dimension than on the decentralisation axis for their coalition preferences. However, regionalist partisans are a notable exception, as they are much more willing to accept a distant coalition on the left -right axis as long as it is located closer to their more salient territorial dimension. In conclusion, we believe this study has taken a step forward towards a greater understanding of the determinants of voters' coalition preferences and towards providing better explanations about their sources and origins.
[5] The expected preferences for single-party minority governments of PSOE and PP voters assume that this type of cabinet leaves their party more room to pursue its desired policies, closing deals with this or that partner depending on the specific policy and circumstances.
[6] The alternatives with UPyD as a partner are dropped from the analyses. See note 4.
[7] These averages are clearly a very rough account of what would be the ideological position of each coalition. Any weighted average (e.g. by electoral or seat share), though, would create problems in differentiating between coalitions given the much larger size of the two main parties with respect to the others. In any case, from a rank-ordering perspective, the unweighted averages make sense.
[8] All analyses are corrected using a sampling weight for the territorial representativeness of the survey, which denotes the inverse of the probability that the observation is included because of the sampling design.
[9] This dependent variable is based on a question that is only asked to those who in the first place oppose single-party absolute majorities. As a consequence, the estimates in columns 5 to 8 of Table 2 may suffer from a sample selection bias. Accordingly, we ran a Heckman probit in which the selection equation determined the preference for non-absolute majorities and the outcome equation was the preference for coalitions vs. single-party minority cabinets. Nonetheless, the post-estimation Wald tests never allowed the rejection of the null hypothesis of independence of equations. This is why we opted for the presentation of two separate equations for each model. Heckman probits are available from the author upon request.
[10] Local polynomial smooth plots with confidence intervals. The results displayed in traditional estimates' tables are available from the author upon request.