The Electoral Trade-Off: How Issues and Ideology Affect Party Preference Formation in Europe

Abstract

Political science has shown increasing interest in cross-country differences in the extent of structural voting, ideological voting and issue voting. Several studies have identified how voting behaviour is structured in post-communist democracies of East-Central Europe and established democracies of Western Europe. This article looks beyond a simple East–West distinction by developing a more sophisticated general model to explain cross-country variations in the effects of issues and left–right on party support. We demonstrate that the more issues are related to left–right, the stronger is the effect of left–right on party preferences. This effect occurs at the expense of the effects of issues on party preferences, which become weaker. These general findings help explain why the effect of left–right on party preferences is weaker in post-communist democracies than in more established democracies. Our proposition is empirically substantiated in a two-stage analysis using the European Election Study 2009.

Acknowledgements

The author would like to thank the Marie Curie Initial Training Network (ITN) (FP7-PEOPLE-ITN-2008, Grant number: 238607) for funding this research.

Notes

Belgium is effectively a two-system country as it is not possible for voters in the Flemish region to cast a ballot for Wallonian parties and vice versa. For this reason, Flanders and Wallonia are treated as two separate systems.

Opinion leadership has been cited as another component of political sophistication (De Vries et al., 2011). It relates to respondents' active participation in following the campaign before the elections to the national or the European Parliament. As we are concerned with voters' general interest in and knowledge about politics and not their specific behaviour to particular elections, we exclude the aspect of opinion leadership from the analysis.

Statements related to the EU are: “Switzerland is a member of the EU”, “The European Union has 25 member states”, “Every country in the EU elects the same number of representatives to the European Parliament”, “Every six months, a different Member State becomes president of the Council of the European Union”. Statements on national politics are country-specific and relate to the name of one of the ministers in the country's government, the required age for being eligible to stand as a candidate in the national elections and the number of the members of the country's parliament.

Mokken scaling technique has an advantage over other data reduction methods such as factor analysis as it overcomes the assumption of the same frequency distribution of item responses. Factor analysis assumes that answers to each item are equally distributed, i.e. the level of difficulty of each item is the same. Mokken scaling corrects for this deficiency, which makes it more appropriate for assessing latent dimensions of political knowledge questions that we use in this analysis.

Finland is the only country where two questions form a common latent scale. In most of the countries, three, four or five questions form a common scale. In Italy and Slovenia a scale is formed out of six items, while in Cyprus, Malta ad Portugal it is formed out of seven items.

The correct answers to political knowledge questions have been coded with 1, 0 otherwise. The political interest question has been collapsed into a binary one, where 1 stands for “very” and “somewhat” interested in politics and 0 otherwise.

If the respondent did not answer the question on the position of a particular party, we replaced the missing value with the national sample mean of the perceived party position. In this way, we lost only respondents who failed to place themselves on the left–right scale.

As literature on attitude formation and party choice postulates the presence of clear socio-economic, socio-cultural and integration-demarcation dimensions in Western Europe (Kitschelt & McGann, 1995; Kriesi et al., 2006, 2008), we analysed by means of non-parametric Mokken scaling whether each set of issues forms a common scale in each country under study. In most of the political systems, especially those in East-Central Europe, no dimensions were found. In some countries attitudes dimensions were found, but the issue items forming them varied significantly across countries. Therefore, we treat each issue separately in the analysis.

However, excluding the left–right distance as a control variable from our analysis does not lead to a substantive change in the interpretation of our results.

In these analyses we do not control for the effects of issues since the relationship between issues and left–right is reciprocal. In countries where the effect of left–right strongly reflects issue preferences, we would then underestimate the total contribution of left–right to party preferences.

As we see in Figure 4, Spain is an outlier. When we remove it from the analysis, we obtain a significant relationship. In this case, the value of the coefficient is 0.64 with p <0.05 and R2 of 0.23.