Campaign Effects in Direct-Democratic Votes in Switzerland

Abstract

The purpose of this article is to analyse the conditions under which referendum campaigns have an impact on voting choices. Based on a model of opinion formation that integrates both campaign effects and partisan effects, we argue that campaign effects vary according to the context of the popular vote (size and type of conflict among the party elite and intensity and direction of the referendum campaign). We test our hypotheses with two-step estimations for hierarchical models on data covering 25 popular votes on foreign, European and immigration policy in Switzerland. Our results show strong campaign effects and they suggest that their strength and nature are indeed highly conditional on the context of the vote: the type of party coalition pre-structures the patterns of individual voting choices, campaign effects are higher when the campaign is highly intense and they are more symmetric when it is balanced.

Acknowledgments

This article is part of a research project supported by the Swiss National Science Foundation (grant no. 100012-112171).

Notes

In addition, campaign effects are also likely to vary across voters, e.g. depending on the timing of the vote decision (Chaffee & Rimal, 1996; Fournier et al., 2004; Lachat & Sciarini, 2002). However, we do not include voters' heterogeneity in the present article, focusing instead on the role of the context.

Hobolt's (2007) study of 14 EU referendums is the only exception we are aware of. There are, of course, other comparative studies (e.g. LeDuc, 2002b), but these studies limit themselves to aggregate level analyses. In the Swiss literature, by contrast, campaign effects are included in various studies, albeit not as the main focus (e.g. Bützer & Marquis, 2002; Kriesi, 2005; Marquis & Sciarini, 1999; Sciarini et al., 2007).

Since the rejection of the agreement on the European Economic Area in a popular vote in 1992, Switzerland has concluded more than 15 bilateral agreements with the EC/EU in various fields. The Swiss people had to (and did) ratify five of them in a referendum. As a result, over the last decade Switzerland has reached a level of integration that is characterized as “customized quasi-membership” (Kriesi & Trechsel, 2008: 189).

Ironically, then, while it is not a member of the EU, Switzerland has had more popular votes on European integration than any EU member state.

Note that the expected absence of variations across political competence for voters who are in a hard learning situation may also be interpreted as a partisan effect: the one-sidedness of the campaign, which conforms to their partisan orientations, can only reinforce the role of these orientations; in that specific case, as in the case of an overall low-intensity campaign, opinion formation is mostly driven by political predispositions.

Of these votes, four were compulsory referendums, 15 were optional referendums and six were popular initiatives. We do not account for the institutional logic or the specific policy area because we have no compelling theoretical reason to do so.

It is also a more valid measure than left–right ideology, which does not clearly correspond to the openness–closedness dimension and would hardly enable us to distinguish moderate right from conservative right voters.

While the number of non-partisans varies from vote to vote, the relative proportion of left, moderate right and conservative right voters remains more stable. Variations in their respective share mainly reflect the increasing electoral force of the Swiss People's Party from the late 1990s onwards.

While Zaller (1992) relies on a more general measure of awareness, he does so mostly for reasons of data-availability and as a second-best as compared to a more focused conception.

Note that Swiss parties are known for their cantonal sectionalism. Hence, it is not uncommon to see cantonal sections deviating from the recommendations of their national party. In cases where more than half of all cantonal sections adopted a deviating vote recommendation, we recoded the type of party coalition accordingly. In our sample, three votes were concerned. First, 13 cantonal sections of the Swiss People's Party opposed the Bilateral Agreements with the EU against the recommendation of the national party; hence the vote was classified as a centre-left coalition instead of a grand coalition. Second, 13 cantonal sections of the Swiss People's Party deviated from the national party and rejected the popular initiative for the regulation of immigration, making this vote a case of a grand coalition instead of a centre-left coalition. Third, concerning the popular initiative “Yes to Europe”, 18 cantonal sections of the Christian Democratic Party favoured a “no” vote despite the positive recommendation of the national party; accordingly, the vote was classified as a centre-right coalition instead of a grand coalition.

The newspapers are Le Matin, Le Journal de Genève/Le Temps, La Tribune de Genève, Neue Zürcher Zeitung, Tages-Anzeiger, Blick. We thank Hanspeter Kriesi for sharing these data with us.

While our model is parsimonious and includes a limited number of variables at the individual level (competence and predispositions), these variables are embedded in a fairly complex set of interactions between them and with context-level variables. This, together with the need to control for additional individual variables and the high number of popular votes included in our analysis, prevents us from relying on more simple techniques such as bivariate or tri-variate analyses.

Such large and independent samples of individuals per macro-level unit offer good estimations of the effects, variances and covariances of first-level parameters. Therefore, it is not necessary to borrow statistical strength across individual-level samples (in our case, across votes).

Given that our dependent variable is dichotomous (approval or rejection of ballot proposals related to questions of international openness), we relied on binary logistic regressions. We refrain from reporting these 25 tables of results because they are only of secondary relevance. They are available from the authors upon request. In a preliminary analysis, we also estimated our model on a pooled dataset covering all 25 popular votes. This analysis confirms our general assumption that political competence and political predispositions influence support to international openness both separately and in interaction with each other.

Grand coalitions are excluded from the models estimating the impact of the intensity and direction of the campaign. In popular votes with a grand coalition, all four governing parties support international openness. Due to this consensus among the major parties, voting campaigns are always of low intensity and the dominant message is never in favour of closedness.

This interpretation is compatible with the strong impact on the voters' decision of the variable education, which we included as a control in our model. To the extent that one's level of education can be considered as an indicator of overall (as opposed to issue-specific) competence (e.g. Kriesi, 2005), our result confirms that the voting choice on integrative policies is still strongly dependent on cognition. Sciarini (2000) found a similar education effect in a study based on four Swiss foreign policy votes of the late 1980s and early 1990s.