Assessing congruence between citizens and parties in Swiss referenda

Assessing congruence between citizens and parties in Swiss referenda Kathrin Kissau* (Kathrin.Kissau@fors.unil.ch) Jan Rosset* (Jan.Rosset@fors.unil.ch) *Swiss Foundation for Research in Social Sciences FORS c/o University of Lausanne Paper prepared for the Annual Meeting of the Swiss Political Science Association, University of Geneva, January 7 th and 8 th 2010 Abstract The Swiss semi-direct democratic political system comprises both representative and direct democratic institutions, in both of which parties play a key role. In this paper, we analyse the circumstances under which citizens express the same position as their preferred party in other words their representative in parliament in the occasion of a popular vote. We find that both contextual and individual characteristics mediate this relation which can also be termed as the congruence between partisans and the party. It notably seems that citizens vote more in line with their party in the case of popular initiatives compared to referenda and Swiss voters on the left of the political spectrum tend to agree less with their preferred parties. Additionally, the new pattern of party coalitions in which the SVP stands against all other large parties seems to polarize citizens voting behaviour bringing about more party-partisan congruence. 1

1 Introduction On November 29th 2009 the Minaret initiative to forbid construction of Minarets in Switzerland was approved by 57.5 percent of the voters. This yes not only came as a surprise for the majority of political analysts and the mainstream media, but also for the Federal Council, the coalition of SP, GPS, FDP and CVP as well as for the major trade associations, who had all recommended the initiative to be dismissed. In contrast, the main political supporter 1 of the initiative, the SVP, was able to celebrate a sweeping triumph as it had been able to mobilize not only its own partisans but other voters as well achieving a very high turnout (above 50 percent) and support for its initiative that was almost twice as high as its normal share of votes. This result came about because many voters had not followed their preferred party s or the Federal Council s voting recommendation. This particular event shows the complicated relation between on the one hand representative and on the other direct democratic features of the Swiss political system. Indeed, on this occasion the representative institutions were contradicted by the popular vote which resulted in a different decision as if it had been for the parliament to decide to adopt or reject the Minaret bill. This example shows that under the Swiss semi-direct democratic system representatives and the population who elected them do not necessarily agree. However, this contrariness remains rather rare - probably because of the mutual dependence of representative institutions and direct democratic procedures. Indeed, Swiss political actors are presented with the challenge to assume dual roles: To come to terms with their requirements and possibilities in representative as well as direct democratic procedures. Alongside to their activities in parliament, especially parties have grown to be a central part of direct democratic votes, by coining campaigns, issuing voting recommendations or even by initiating votes. This role of the parties is one of the reasons the differences between the Swiss semi-direct democratic system and a purely representative one should not be overestimated. The difference between the two types of democracy is all the more limited because [...] direct democracy is also likely to be mediated by political organizations and their representatives (Kriesi 2005: 7). In fact, both representative and direct democratic institutions in Switzerland have to be seen as overlapping. Referenda and initiatives complement the work of the parliament. And at the same time, key actors of representative institution also actively take part in direct democratic processes. The Minaret vote has, however, demonstrated that direct democratic and representative institutions not always go hand in hand, and party or federal council recommendations are not always followed by the sovereign. 1 The initiative came from the Egerkingen Committee which was mainly composed of SVP members. This party was also the only sizeable organisation to back it during the campaign. 2

As these popular votes are carried out often in Switzerland, the interaction between citizens and political elites is increased in contrast to representative systems. There the policy preferences of the citizens can only be deduced indirectly through public opinion surveys, whereas popular votes enable the immediate analysis of the issue congruence 2 between political elites and voters. We thus advance previous work on political congruence, particularly on congruence between citizens and parties by focussing on popular votes in Switzerland. By analyzing under what circumstances citizens follow their preferred party s recommendations in popular votes - and are thus congruent - we shed light on their relationship, more specifically the influence of parties and the party loyalty of voters in direct democratic procedures. In this paper we will first give a short introduction on direct democratic procedures in Switzerland, especially on the position of parties within this semi-direct democratic system. In a second step we will discuss the rich literature on opinion formation in popular votes by putting an emphasis on the particular aspects that might be relevant in explaining congruent voting. Next we will develop a model that includes both contextual and individual factors to predict the odds that citizens vote in line with the party s recommendation. We will describe the results and discuss what can be concluded from these for congruence, also in representative systems. 2 Background: Direct democratic procedures in Switzerland Switzerland is the country where direct democratic institutions are most developed at the national level. To complement the representative legislative and executive bodies which deal with most political matters, there is a series of possibilities for civil society actors and citizens to hold the parliament accountable on specific issues and to participate directly in agenda setting and political decisions. 2.1 Popular votes The Swiss constitution allows four different types of popular votes. Firstly, as in most other countries with direct democratic institutions, some parliamentary decisions that directly affect the constitution or imply some important institutional changes are automatically subject to referendum: this is the so called compulsory referendum. The votes for the approval of changes in the constitution and the membership in international organisations fall under this category. This type of referendum has most chances of success in the popular vote. Secondly, there is the optional referendum. It takes place when at least 50 000 citizens have signed a petition within a period of 100 days after the decision of introducing a new 2 We define congruence as the proximity of the political preferences of a citizen and the political elite. 3

law or the modification of a law has been taken by the parliament. These optional referenda are less successful. Between 1874 and 2005 roughly 50 percent of the laws that were challenged passed the popular vote (see Lutz 2006). These two forms of referenda correspond to an increased accountability of elites who are subject to citizens control not only in the legislature long-term perspective but on each specific issue. Thirdly, citizens can also initiate new laws, modify or revoke present ones. It is the institution of the popular initiative that guarantees this possibility. To submit a text to the vote of citizens, the committee that initiates it must collect 100 000 signatures. Then - and this is when the fourth instrument of direct democracy comes in play - the parliament has the possibility to propose an alternative to this popular initiative. This is called a counter-project as it often has the goal to oppose a popular initiative by submitting a more moderate proposal than the one put forward by the initiative committee. When a counter-proposal is formulated, then the committee has the choice to maintain or withdraw its initiative. All together less than 10 percent of the popular initiatives were approved at the polls (see Lutz 2006). Despite these extensive rights of the citizens, representative institutions are not powerless in direct democratic procedures. In particular, the Federal Council (executive) usually formulates voting recommendations. And in parallel to the direct democratic process, the parliament also votes. This happens not only in the obvious case of referenda but also for popular initiatives. Alongside the Federal Council, parties as well as trade unions and other professional organisations issue voting recommendations 3. In the past 50 years the recommendations of the political elite have most often been followed by the population 4. However, the stability of this phenomenon is decreasing (see Sciarini and Trechsel 1996). Citizens have rated the recommendations of the government to be most important for their own decision, followed by their favoured party s recommendation (see Trechsel and Sciarini 1998: 116). 2.2 Parties in Switzerland The Swiss political landscape is characterised by a relatively large number of parties, but especially four parties occupy a central role. Each of these has representatives in the Federal Council: the Swiss people s party (SVP, populist, national conservative), the Social Democratic Party (SPS, centre-left), the Free Democratic Party (FDP, liberal) and the Christian Democratic Party (CVP, centrist). Although their combined share of votes remained rather constant during the second part of the 20 th century (Ladner 2001), the 3 This of course is not compulsory. However in an overwhelming majority of cases parties indeed issue voting recommendations. 4 For the years 1951-2009, only 23% of compulsory and 35% of optional referenda have been rejected the population. Citizens followed the elites even more in the case of popular initiative, only 7,5% of which have been accepted (source : Federal Statistical Office). 4

Swiss people s party has seen an important rise in its electorate since the 1990 s, while the FDP and CVP experienced a slow decline. Additionally, the importance of the Green Party (GPS, ecologist) has continuously increased since the late 1980 s. Since the year 2000 its share of votes ranged between 5 and 8 percent. Altogether, these five parties represented more than 85 percent of votes and more than 90 percent of parliamentary seats in the 1999-2003 and 2003-2007 legislatures. One of the specificities of Swiss politics is that the party system at the national level does not match the party system in each of the cantons, where some of the important parties at the national level can have a marginal share of votes (e.g. the CVP in some protestant cantons). Because of the importance of cantonal level politics and their specificities, parties at the national level have rather weak organisations and function as the overarching organisations of cantonal parties. The formation of these parties and the development of direct democracy in Switzerland are inextricably wedded to one another (see Ladner and Brändle 1999: 284) and parties have thus evolved and adjusted to the circumstances of direct democratic institutions. It is even argued that in contrast to other countries, parties in Switzerland have developed not as much as the result of electoral processes but rather as a result of popular votes (Hug 1994: 85). When trying to assess the relationship between citizens and parties in Switzerland, specifically their congruence in popular votes, it helps to have a look at the challenges and advantages that a party s setting within a semi-direct democracy has: While generally political parties seem to be losing importance, especially in terms of interest aggregation or political agenda setting, parties in direct democracies are additionally challenged by the emergence of other political actors on the political arena, such as social movements or interest groups (see Ladner and Brändle 1999: 286). Foremost single-issue groups have an advantage when trying to influence the popular votes and mobilize citizens, as they stand for clear-cut positions in contrast to broad party programs. Furthermore, elections, where parties usually dominate the field, are of less importance in Switzerland, as weighty issues can be separately voted on in referenda (see Neidhardt 1992: 31). These popular votes also impose additional work for parties and require extra financial resources for campaign and volunteers. At the same time these direct democratic requirements can weaken parties by making their internal heterogeneity more visible. In Switzerland the process of deciding a party s overall position on a subject issue to a popular vote is often a source of conflict which leaves cantonal parties taking a different stand than the national one (see Ladner and Brändle 1999: 286). On the other side it is also argued that parties frequently adapt to direct democratic procedures and even benefit from them (see Smith 2005). Direct democracy can thus promote the versatility of parties, as they constantly have to deal with new issues, 5

respond to social change or take up new political ideas (see Ladner and Brändle 1999: 287). Popular votes also give parties the chance to test or present new party programs or ideas (see Nef 1988: 178) and also encourage parties to remain at a relatively high level of political activity. Initiatives, which are often launched by parties themselves or organizations close to them, are thus one possibility to mobilize their voters, to restore their political participation (see Bowler/Donovan 2006: 650). As parties also play an important role in the campaign proceeding the vote by framing the public debate and giving voting recommendations, they seem to have adopted popular votes as one possibility to influence the decision making process outside of parliament (see Budge 2005). Ladner and Brändle also point out that initiatives and referendums enable governmental parties to act independently of the parties they share power with and this can help to break the chains of consociationalism (1999: 288). Added together Kriesi comes to the conclusion that under conditions of contemporary party democracies, direct democratic procedures are likely to be guided and controlled by political parties and related political organizations (2005: 7). While this speaks for the influence of parties in Switzerland, the recent Minaret vote showed that the relationship between parties and partisans is versatile and the strength of parties to sway their partisans is not steady. Rather, other influences are also present which determine the voting decision of the citizens. 3 Theory: Opinion Formation in Popular Votes Given the historically deeply embedded doubts about the capacity of simple citizens to take informed and reasonable decisions on political matters (see Schumpeter 1943), students of direct democracy have been interested in the vote choice of citizens. In that literature two broad approaches can be distinguished. On the one hand, research has focused on the outcomes of direct democratic votes and the way they are consistent with the elites depending on the degree of agreement within them (see Papadopoulos 1996; Sciarini and Trechsel 1996) and on the other hand, research has concentrated on the opinion formation process at the individual level. This second axis of research has been very fruitful, with a wealth studies, gathered in an edited book, documenting the role played by different individual characteristics in that process (see Kriesi 1993). Building on the findings of their predecessors and integrating new concepts from the literature on cognitive heuristics and on elites influence on opinion formation (see Zaller 1992), newer studies have developed ever finer explanatory models in which citizens competence and different decision making strategies interact (Sciarini and Marquis 2000; Kriesi 2005; Marquis 2006). In the perspective of these studies, different cognitive strategies can be pursued to reach a decision. They imply either following a certain cue (it can be the party position, the 6

position of the government or simply that of preferring stability over an unknown change) or elaborating a more complex strategy that involves a systematic processing of arguments (e.g. rational choice, to maximize individual benefits (see Linder 2001: 10). These studies thus suggest that the decision process is best conceived as a complex development in which both contextual and individual characteristics play a role and influence one another. For example, politically more competent citizens react differently to a situation in which there is a high level of agreement within the elites than less competent (Sciarini and Marquis 2000). Below, we address some of the main influences of this decision making process in the situation of a popular vote. These influences can be found on various levels. Thus, we first discuss the contextual characteristics (vote, elite) and then those of the individual citizen that play a role for the opinion formation. Hereby, we combine both approaches to vote choice research addressed above, though we do not raise the claim to develop an encompassing model. Rather, we shift the focus away from vote choice itself to congruent voting. To explain under which circumstances citizens are more likely to vote in congruence with their party and when not, we assess what influences have generally been identified on vote choice itself. We argue that these contextual and individual characteristics can also be used to predict congruent voting of Swiss citizens. Generally, partisanship is known to be a significant predictor of voting behaviour (see Butler and Stokes 1969; Campbell et al. 1960). Equally, in popular votes party supporters are expected to use the recommendations of their preferred party as a cue on the utility of either alternative. This is referred to as the likeability heuristic which implies that voters infer their own position depending on the position of the group they like or dislike (see Hobolt 2006: 628). Thus, voting recommendations of parties are most frequently applied as a predictor for the individual s preference in a popular vote. We take a different perspective by viewing the relationship between voting choice and party recommendation (congruence) to gain insight into the interrelation of direct democratic procedures and representative institutions, specifically between citizens popular voting behaviour and parties recommendations in Switzerland s semi-direct democracy. 3.1 Contextual characteristics Issues. A number of studies have shown that some policy issues result in higher turnout and influence the way the campaign is waged (e.g. morality issues or those referring to international policy) which itself can affect vote choice (see Mooney 2001: 8). Gaps between citizens and elites preferences were also found with regard to fiscal issues, as citizens seem to care more about fiscal discipline than the elite (see Feld and Kirchgässer 2001: 365). Linder explains these observed differences by pointing out that depending on the issue to be decided, people's political behaviour is determined by 7

traditional social ties, by self-interest, and by politically mediated values that seek to generate solidarity or altruism (Linder 2001: 10). We will test the importance of the issue voted on by analyzing whether congruent voting is more likely to take place when social policy or ecology is voted on (RQ 1a). The effect of the issue on the citizen s choice also seems to be mediated by his or her position within the political spectrum. Kriesi observed this in the case of conservative voters who tended to agree more with the government on agricultural policies and significantly less in relation to institutional change and foreign policy (see 2005). Taking a slightly different approach, we will test whether the effect of an issue on congruence is different for partisans of a party which has put particular emphasis on this issue. We will thus look at the obvious case of congruence of voters of the Green party with their party when they vote on ecological issues, expecting a higher degree of congruence on these specific matters (RQ 1b). Types of votes. Various studies have analyzed voting behaviour in view of the different types of popular votes (initiatives, facultative referendum, obligatory referendum, counter-project). Trechsel and Sciarini found that political parties recommendations seem to be more influential in the case of obligatory referendums and initiatives than in the case of optional referendums (see 1998: 120). We will test these results and investigate if the type of votes (initiative, facultative referendum, obligatory referendum) influences the odds of voting in line with the party recommendation (RQ 2). Elite consensus. A part from the issues and the type of vote, the consensus among the elite also influences the success of popular votes as Trechsel and Sciarini revealed: The higher the parliamentary consensus the higher is the probability of governmental success at the polls (1998: 119). Also, it is conceivable that not only the degree of agreement within the elite but also the way the elites are split has an influence on the voter s decision. Thus, we will investigate which form of coalition within the five central political parties increases or decreases the likelihood for congruent voting (RQ 3a). Beyond consensus among political elites as a whole, the degree to which parties internally agree on a certain issue might also influence the vote choice. In the Swiss party system cantonal sections sometimes overtly take different stands than the party at the national level, which we argue might have invoke a similar impression as disagreement within parliament has. We will therefore test if congruence is more likely if the preferred party is internally homogenous (RQ 3b). However, these general observations on elite consensus effects are mediated by other individual factors. Zaller for example differentiated the effects of elite consensus or conflicts, contending that when there is a high consensus among elites then the more politically informed and engaged agree more with government (mainstream effect). In contrast, when there is conflict within the elite then the more informed and engaged 8

citizens will have higher degrees of divergence depending on their party preference (polarisation effect) (see Zaller 1992). We will therefore assess how party homogeneity interacts with political competence of the individual to affect congruent voting behaviour (RQ 3c). 3.2 Individual characteristics Party. Vote choice of citizens also appears to differ depending on which party or which type of party they identify with. Trechsel and Sciarini for example revealed that party arguments of the governmental parties influenced voters differently than arguments of the extra-governmental opposition (see 1998). Supporters of populist parties also seem to be less receptive to their parties recommendations (see Bowler et al. 2003). Thus, depending on the party s position on the left-right spectrum and on its position within the government, citizen s relationship to their preferred party and thus their choice of voting in line with it might vary. We will investigate whether the chance for congruence is increased or decreased depending on which party a citizen favours (RQ 4a). Also we test if these differences between the parties vary depending on the topic voted on (RQ 4b). Competence of voters. The choice when voting has also been found to differ between politically competent and incompetent citizens. For instance voters tend to say no more often than yes ( Neinsager effect ). This effect is most prominent for the less informed citizens (see Passy 1993). Also, those unsure about their preferred party s position tend to vote against their party more frequently (see Kobi 1993). Here, Marcinkowski points out the complexity of this process by arguing that the information provided in a campaign or the way the issue is framed during the campaign can also influence whether or not a citizen feels competent about a topic or not, and if he or she will then rely on their own predisposition or take the information provided or the recommendations expressed into account (see 2007:105). We will investigate whether high or low political competence enhances congruence between the vote of the citizen and the party recommendation (RQ 5a). Additionally, we test whether the Neinsager effect also affects the odds for voting congruently or not (RQ 5b). Salience of the issue. A number of studies have come to the conclusion that it is not so much the issue of the proposal itself as the salience 5 of this topic that influence citizens voting decision, with salient issues bringing about more congruence (see Gilens et al. 2001: 14). Congruence is thus expected more for politically important issues such as domestic ones (see Monroe 1998). Though salience is usually measured by analyzing campaigns, the media or parliamentary activity on the issue, recent studies have pointed 5 Salient issues are those that the public (Franklin and Wlezien 1997) or a person (see Giger 2009) cares about and thus has a meaningful opinion. 9

out that salience could also be viewed as an individual characteristic (see Giger 2009). We will therefore test whether congruence is more likely for issues that are salient for the general public (RQ 6a) or if it is more individual saliency that affects vote choice in line with the party (RQ 6b). Left/Right attribution. There is reason to assume as well that the decision process might be different depending on the position of the citizen within the left-right spectrum. Donovan and Karp observed such differences between citizens most peripheral to politics on the far right of the ideological spectrum and those on the left. Left-wing party supporters apparently participate more in popular votes and vote in line with their party s recommendations, though this effect seems to be diminishing (see Donovan and Karp 2006: 684; Lutz 2005). Differences in popular voting behaviour between citizens on the left and right of the political spectrum were also revealed by Trechsel and Sciarini who found that in the case of popular initiatives right wing partisans tend to follow government recommendations while left wing/socialist partisans tend to oppose (see 1998: 118). We will test whether citizens on the left or right of the political spectrum are more likely to vote in line with their party in popular votes (RQ 7a). Additionally, we will investigate if congruence is more probable if the individual and his or her favoured party are ideologically close in terms of their position on the left-right scale (RQ 7b). Socio-demography. Voting decisions do not seem to be determined to a large degree by socio-demographic factors (see Hobolt 2006: 634). However, in some cases men and women at least seem to use different cognitive strategies to come to a conclusion depending on the stimulus (e.g. issue of the vote) (see Nai and Lloren 2009). Additionally, three geographically embedded cleavages have traditionally structured the vote of Swiss citizens: the religious protestant-catholic divide, the so-called Roesti- Graben a barrier between on the one hand German speaking and on the other French speaking cantons and the urban rural divide, with citizens from cantons with a large urban agglomeration often voting differently than those from more rural areas (see Seitz 2009). In our model we will control for the rural or urban residence of the respondent as well as for the language spoken in the canton were they live. Although not expecting large effects, we test whether the cleavages that have an influence on the direction of the vote itself, also play a role in following the preferred party recommendation or not (RQ 8). 4 Method 4.1 Data To analyze under what circumstances Swiss citizens follow their preferred party s recommendations in popular votes we make use of the VOXIT Data, which is a 10

standardised dataset that includes all the VOX surveys since 1977. In these surveys ca. 1 000 eligible voters are contacted within two or three weeks after the popular vote. The focus of these surveys is the interest, motivation, and competence of the citizens in matters related to voting and politics in general. The central topics covered in the interviews include: general political opinions and habits, political and social affinities, degree of understanding of the items put to vote as well as various aspects related to the decision on how to vote on these issues. The sample we analyse comprises 73 votes that took place during the 1999-2003 and 2003-2007 legislatures (see Table 1). These popular votes had an average participation rate of 45.6%. Table 1: Types of popular votes between 2000 and 2007 2 1 2 11 1 7 7 5 5 1 1 2 1 2 2 3 3 1 5 4 4 1 1 1 2000 2001 2002 2003 2004 2005 2006 2007 compulsory referendum optional referendum popular initiative counter project Source: authors own illustration based on Swissvotes data, N = 73 In our analysis we focus on the four large parties that have representatives in the Federal Council, namely the Swiss people s party (SVP), the Social Democratic Party (SPS), the Free Democratic Party (FDP) and the Christian Democratic Party (CVP) as well as the Green party (GPS). Altogether, these five parties represented more than 85% of votes and more than 90% of parliamentary seats during the period studied. As respondents who actually voted and those who did not relate differently to the actual voting decision, we only included cases in our analysis where respondents actually voted and indicated to be close to one of the five major parties (N=19 051). 11

4.2 Measures Dependent variable. The congruence between citizens and parties in popular votes was measured by comparing the voter s decision with his or her preferred party s recommendation on each vote. If the vote was in line with the party we defined this as congruent (coded as 1 ). If the voter decided differently than the party, the relationship was defined as non congruent (coded as 0 ). In this case it did not matter whether the party had recommended a yes vote and the citizen had voted no or vice versa (see Table 2). Independent variables. A number of variables were included in our estimations that referred to the above discussed contextual and individual characteristics of the voting decision. To assess the influence of the issue voted on, we classified the 73 votes included in our analysis. Doing this we identified two major and distinct issues: Social policy, which 19.3% of the votes could be attributed to, and ecology which was the topic of 15.1% of the votes. For the type of vote the categories described above (see chapter 2.1) were used. Thus, we constructed four dummy variables: popular initiative (35 votes), optional referenda (24 votes), compulsory referenda (10 votes) and counter-proposal (4 votes). In the analysis, the initiative was the reference category. For the elite consensus we classified the votes into different coalition situations. The most common divide between the five largest parties was - in concordance with the literature focusing on earlier periods (see Hug 1994) - that of a left-right cleavage when the SVP, FDP and CVP took the same stands and opposed a coalition formed of the SPS and GPS (36 votes). But it also happened on 6 occasions that the CVP, the most centrist of the right parties, joined forces with the parties situated on the left of the political spectrum. The most frequent divide among right parties was that between the SVP on the one hand and the two more moderate parties on the other which took the same stand as the left parties on these occasions (12 votes). There were also issues (10 in total) on which all the large parties agreed and formulated the same vote recommendations to their voters. The rest of our sample comprises more unusual cases in which for instance the Green Party stood alone against all others (2 cases) or the SPS, FDP and CVP allied together (2 cases), which we classified as other coalitions. We created dummy variables for these five situations, using the right coalition as a reference category. We measured the party homogeneity by determining the existence of an intraparty opposition regarding the party s stand on an issue voted on. We set up an ordinal scale ranging from 0 to 3 and classified parties in each vote depending on how many of the up to 26 cantonal parties opposed the stand of the national party. When 10 and more 12

opposed cantonal parties we coded 0 (2.7%), if 5 to 9 cantons differed we coded 1 (6.1%), if only 1 to 4 cantonal parties recommended to vote differently than the national party we coded 2 (24.0%) and for complete homogeneity we coded 3 (67.3%). For party differences we created dummy variables for each of the five major parties (CVP, SVP, FDP, SPS, GPS) included in this study. 13.8% of those sampled were partisans of the CVP, 25.4% of the SVP, 20.2% of the FDP, 35.0% of the SPS and 5.7% of the GPS. The CVP was used as a reference category. For political competence of the individuals we constructed an index from -2 (not very competent) to 2 (very competent) from four questions that were asked in the VOX survey: How often are you asked for your opinion regarding a decision on a political topic? How much are you interested in politics? Generally, do you find it difficult or easy to get an idea of the consequences of a Yes or No vote for yourself? What bills were voted on in the last popular vote? For voting no we recoded the question how did you vote into a dummy variable, coding those who voted no as 1 and all others as 0. To assess general salience we used the VOX-question: How important was the XY initiative for our country? Here, the answer categories were from 0 (no importance) to 10 (very great importance). Similarly the measure for individual salience was the question: How important for you personally was the XY initiative? To indicate the left-right self placement of the voters we referred to the VOX-question: Where do you position yourself on the left-right scale. In this case 0 referred to a selfplacement on the extreme left and 10 on the extreme right. Finally, the ideological congruence between citizens and their preferred party was taken into account. For this we measured the absolute distance between the left-right self placement of the citizens (same as above) and the mean party placement, which we derived from the 2003 Swiss Election Study. The created scale ranged from 0 (congruent) to 8 (not congruent). 13

Table 2: Descriptive statistics of variables included in the model (N=19 051) M SD Code Partisan-party conguence.74.44 0 = not congruent; 1 = congruent Social policy.19.39 0 = others, 1= social policy issues; Ecology.15.36 0 = others, 1= ecological issues; Optional referenda.32.47 0 = others, 1= optional referenda, Popular initiative.48.50 0 =others, 1= initiative Compulsory referenda.14.34 0 = others, 1= compulsory referenda; Counter project.05.23 0= others, 1 = counter project Left Coalition (SPS, GPS, CVP).07.25 0 = others, 1= Left Coalition Right Coalition (CVP, FDP, SVP).49.50 0 = others, 1= Right Coalition SVP against all others.16.37 0 = others, 1= SVP vs. all others All parties agree.13.34 0 = others, 1= all agree Other coalitions.11.31 0 = others,1= other coalitions Party homogeneity.44.73 0 = no homogeneity to 3 = a lot of homogeneity FDP.09.28 0 = others, 1 = FDP SVP.11.31 0 = others, 1 = SVP SPS.15.36 0 = others, 1 = SPS GPS.02.15 0 = others, 1 = GPS CVP.06.24 0 = others, 1 = CVP Political competence.03.52-2 (not very competent) to 2 (very competent) Voting no.28.49 0 other, 1= voting no General salience 6.93 2.52 0 = no importance to 10 = high importance Individual salience 5.57 3.12 0 = no importance to 10 = high importance l-r self-placement 4.92 1.79 0 to 10 Ideological congruence 1.23 1.12 0 to 8 (congruent to not congruent) Age 48.9 17.6 age in years (18 to 97) Gender 1.51.50 1= male; 2 = female Rural/non-rural 1.33.47 1= urban; 2 = rural French language area.24.43 0= other, 1= French German language area.70.46 0= other, 1= German Italian language area.60.24 0= other, 1= Italian Control variables. In the analysis we also considered that socio-demographic factors might have an influence on the voting decision. Therefore, we integrated the variables age (in years), gender (51.1% of those polled were women), as well as the association to an urban (67.2%) or rural (32.8%) residence and to a specific language area in Switzerland in the model. We differentiated between the French (24.1 %), German (69.8%) and Italian (6.1%) language area and created three dummy variables. In the model the French language area was used as a reference category. 14

4.3 Analysis A logistic regression model to estimate congruence between a citizen and his or her preferred party s recommendation was set up (see Table 3). Coefficients can be interpreted by taking the antilog (e B ) to determine how strong the odds of congruence increase or decrease when the independent variable increases by one unit (e.g. from 0 to 1 for the dummy variables). 5 Results Table 3 presents the results of the logistic regression on partisan-party congruence in popular votes between 2000 and 2007 in Switzerland. The aim of this analysis is to identify circumstances that influence the outcome of the citizens votes either in line with or against the political elite. On the contextual level we first viewed the influence of the topic of the vote. Here the model shows that ecological issues are more likely to result in congruent voting behaviour than social policy (RQ 1a). In accordance to our expectations, this is even more the case for voters of the Green party. On the central political topic of this party - ecology - their partisans tend to vote more in line with their party as compared to other parties partisans (RQ 1b). Similar to the salience argument, this result indicates that partisans and parties may tend to agree more on topics that are most important for both, those that characterize or represent the party s profile. As to the type of vote, the model indicates that citizens are most likely to vote in congruence with their party in initiatives and less on optional referenda or counter projects. The odds for voting in line with the party are reduced the most by compulsory referenda (RQ 2). This stands in contrast to the results of Trechsel and Sciarini, who found that party s recommendations are most influential in the case of compulsory referendums (see 1998: 120). Since their analysis covered votes in an earlier decade, it is possible that these mechanisms have changed since then. 15

Table 3: Logistic regression on partisan-party congruence in popular votes B SE B e B Social policy -.073.049.930 Ecology.201***.058 1.223 Compulsory referenda a -.200***.060.819 Optional referenda a -.055.042.946 Counter project a -.081.087.922 Left coalition b -.096.070.909 SVP against all others b.456***.052 1.578 All parties agree.662***.064 1.939 Other coalitions.102.064 1.107 Party homogeneity.593***.024 1.809 FDP c.190**.062 1.209 SVP c.369***.063 1.446 SPS c -.251***.060.778 GPS c -.103.095.902 Political competence.287***.034 1.332 Voting no -.320***.037.726 General salience.030***.009 1.030 Individual salience.013.008 1.013 l-r self-placement -.051***.011.950 Ideological congruence -.095***.016.910 Age -.001.001.999 Gender -.059.035.943 Rural/non-rural -.035.037.966 German language area d.066.046 1.068 Italian language area d -.492***.090.611 Constant 1.464***.143 4.321 x 2 1202.6 df 25 a. Reference category is the initiative, b. Reference category is the right coalition, c. Reference category is the CVP, d. Reference category is the French language area *p<.05 **p<.01 ***p<.001; 81.0% of the cases were correctly predicted with this model. 16

Compared to the situation of a right coalition, voters are less likely to vote in line with their party s recommendation when a left coalition exists, whereas when the SVP stands against all other parties the likelihood for congruence increases. This shows the high degree of polarization of voters that correspond to this type of conflict structure among parties, with voters rather following their preferred party s recommendation. Unsurprisingly, when there is no conflict among the different parties, the odds for citizens to vote in line with their party s recommendation massively increases (RQ 3a). Similarly, the degree of consensus within a party also results in more congruent voting: we find that the more agreement within a party, the higher the odds for voting in line with the party (RQ 3b). These results confirm earlier research that also found that homogeneous positions of the political elite enhance the likelihood that voters will agree with the elite s stand. We find this to be true not only for the government and parliament positions but also for the degree of agreement within each party. The effect that such unity has is however less strong for the more competent voters, who do not seem to be influenced as much by external demonstrations of agreement (RQ 3c) (see Table 4 in the annex for the regression model including interaction terms). On the individual level we first focused on differences between partisans of the five parties. We find that compared to the CVP the voters of the FDP and SVP are more likely to vote congruently. In contrast, partisans of the SPS are less likely to be congruent with their party compared to the CVP voters and voters of the right parties in general. No significant difference exists between voters of the GPS and the CVP in terms of voting in line with their party s recommendation (RQ 4a). As mentioned above, the analysis also indicated that partisans differ not only in terms of their overall likelihood of voting congruently, but also in their voting behaviour on specific topics. (GPS partisans have increased odds of voting congruently on ecological issues compared to CVP voters) (RQ 4b). With regard to the influence of political competence of the individual voters, the model shows that the more competent a citizen is the more likely is congruent voting behaviour (RQ 5a). Thus, similar to results of Kobi (1993), partisans who are less competent have a higher likelihood of voting against their party. As referred to above, the more competent voters are also less affected by conflict within the elites, which otherwise reduces the odds for congruence. Under these circumstances they still have a higher likelihood of congruent voting than their less competent neighbours. The model s results also confirm the existence of the so called Neinsager effect which has been documented in other pieces of research (RQ 5b). This term refers to the fact that the population in general and especially less competent citizens tend to oppose change more than the elites or in other words tend to follow the status quo heuristics when making a vote choice (Kriesi 2005: 146; Passy 1993: 223). In general we also find that 17

citizens vote no more often than what their preferred party recommends. This results in a decreased probability to vote in line with the preferred party when voting no as compared to a situation in which a citizen votes yes. With respect to salience, the model indicates that the more important or salient an individual believes the vote to be for the general public the more likely it is that he or she will vote congruently. There is however no significant effect of personal importance on the likelihood for congruent voting (RQ 6a). This indicates that the voting decision is not influenced as much by the personal importance but rather by the believed general importance of a vote (RQ 6b). While salience is often measured through media or campaign analysis the indicator chosen here reflects the individual s own evaluation of importance. We believe this is a more valid proxy for salience, as what matters for the vote choice itself is not so much what is promoted in the public but how each individual evaluates the vote s importance. As to the influence of voters position within the political spectrum on their tendency to agree with their party or not, we find that the further to the right of the political spectrum the less likely do partisans vote congruently. This goes against observations of Donovan and Karp (2006), who found that left wing citizens tend more to voting in line with their party s recommendation (RQ 7a). One of the possible reasons for this result in the Swiss case is that left-wing parties find themselves more often in minority situations than right wing parties and thus are more likely to recommend voting against the Federal Council who s voice is generally loudly heard. As to the ideological closeness of the individual to the preferred party (measured on the left-right scale), this also significantly increase the odds for congruent voting behaviour on a specific issue (RQ 7b). Thus, it is seems that it is not only the distribution on the left-right continuum but also the absolute distance to the preferred party on this scale which affects the likelihood for congruence. The socio-demographic controls we included in the model were mostly insignificant. Effects were merely found between the French and the Italian language area. Though the stable rift through political Switzerland has been described between French and German language area, we indeed find that citizens form the Italian language area are less likely to vote congruently than the those from the Romandie (RQ 8). As congruence does not refer to political preferences but is more a measure for political representation, we tentively surmise that voters from the Tessin are not as close to their preferred national party as voters from the other parts of Switzerland. 6 Conclusion and discussion In this study, we examined voting behaviour in popular votes, focussing on circumstances that foster congruence between a citizen s vote and his or her preferred party s recommendation. We viewed this relationship to gain insight into the interrelation 18

of direct democratic procedures and representative institutions, specifically between citizens popular voting behaviour and parties recommendations in Switzerland s semidirect democracy. The above presented model shows that contextual as well as individual factors affect partisan-party congruence in popular votes. Factors that have before been shown to influence voting decisions in general also increase or decrease the likelihood of voting in line with the preferred party on a specific topic. Overall, the model indicates that congruence depends on the issue voted on, the type of vote (initiative or not) but even more on the coalition that has formed or how parties are divided on the issue. Additionally, we have found support that internal party heterogeneity negatively influences congruence, and that partisans of the five parties differ in their relationship to their party. Also, political competence appears to increase the likelihood for congruent voting as does believed general salience, whereas the rating of individual importance does not. In contrast, ideological closeness or ideological congruence between partisans and party raises the likelihood for congruence in a specific situation. Overall, it is interesting to view in parallel the findings in relation to the effect of salience on congruent voting as well as the fact that Green voters tend to vote more in line with their party on ecological issues. Both these observations relate to the complex relationship between the choice of representatives and the voting decision in the popular vote. When voting in elections, citizens are likely to choose the party that most closely mirrors their preferences on a wide range of topics. However, due notably to the fact that citizens have in general less structured policy preferences than their representatives, it is very unlikely that they will find a party with which they would agree on all issues. Therefore, citizens tend to choose the party that is closest to them on those issues that are most salient to them (see Jones and Baumgartner 2004). Or as in the case of the Greens to choose a party that puts a specific emphasis on those issues. In turn, it is then very probable that the party, with which the citizen identifies, represents this citizen s opinions well on precisely the issues that are salient to him/her. Therefore, our results regarding saliency and issue specific voting can be read in the broader perspective of issue specific representation. Additionally, our results give evidence that direct democratic procedures in Switzerland are worthwhile, as the elected elites and especially parties do not automatically represent citizens wishes well (though ideologically congruent) with regard to each political issue. However, in the period we studies they did this in 88% percent of the cases, which acknowledges that Swiss parties are rather successful at representing their partisans preferences. In the remaining cases, popular votes resulted in different outcomes than if the parliament alone had decided, thereby correcting the elites pursued path and thus giving parties input on how to reframe their policy activities to accommodate their partisans preferences better. 19

We also argue that it is sensible to introduce a conceptual differentiation between ideological congruence (which might be stable over time) and situational congruence, which is affected by various contextual and individual factors. In Switzerland s semidirect democratic system both types of congruence exist, parallel to the representative (ideological) and direct-democratic (situational) procedures. Thus, the proposed gradual differentiation of congruence is in line with Budge s argument that the common distinction between popular policy decision and representative deliberation is rather a gradual one between individual policy voting and package policy voting (2005: 20). It was our aim to use the detailed information available on popular votes to study situational congruence, also collecting evidence that might be applied to the overall concept of ideological congruence. If this generalization of the contextual and individual affects we found is possible will need to be tested in a next step by comparing congruence in a direct democratic with congruence in a representative system. Coming back to our recent example of the Minaret vote, the question remains if it was predictable that many partisans of the SPS, CVP, FDP and GPS would not vote in congruence with their preferred party or the parliament s recommendation. Our results would rather point the opposite direction. Indeed the case of an initiative being voted on under the circumstances of a SVP against all other parties conflict configuration among political elites seem to be precisely a case in which citizens tend to vote most in congruence with their preferred party. However, if we turn to a more qualitative analysis of all the votes in the years 2000 to 2007 that brought a different outcome as compared to parliamentary decision, the Minarets vote does not appear to be a complete outlier. The 9 cases (12%) that lead to a popular decision contrary to that of the parliament can be split into two broad categories. All four optional referenda that were eventually rejected by citizens were fought against by a left coalition composed of the Social-Democratic Party and the Greens and were on that were broadly related to the liberalisation of the market and the welfare state. (This coalition of parties also led to the success of an initiative against genetically modified organisms in food products.). The four remaining votes were held under the circumstance of a SVP against all other parties configuration. Three of them (two compulsory referenda and one initiative) were related to citizenship rules or tougher sentences for criminals, in other words votes that lead to an increased authoritarianism of the state. To summarise, it seems that citizens had at the same time a more authoritarian stand than the parliament on cultural issues and also a more pro-state intervention view than their representatives on matters related to the economy. The case of the Minarets clearly falls into the first category and does not have to be seen as an exception but rather a case that integrates itself in a series of others that follow a systematic pattern. As always there are a number of limitations of this paper that at the same time stimulate 20