This paper studies the relation between party institutionalization and intra-party preference homogeneity in democracies. In weakly institutionalized parties, it cannot be taken for granted that party actors have similar policy views because they lack the capability or motivation to coordinate agreement and to recruit personnel in line with this agreement. This should matter most when other safeguards against preference heterogeneity are missing. Empirically, we explore the association between institutionalization and intra-party preference homogeneity at the level of candidates to the national legislature based on survey data. In a single-country study, we first look at the case of Germany in 2013 and 2017, contrasting the young and weakly institutionalized Alternative for Germany (AfD) with the older, established parties. In a second step, we study the link between party institutionalization and preference homogeneity in a cross-country analysis of 19 established democracies. We find that parties with high value infusion—parties whose candidates are committed to the party—are generally more homogenous in their policy preferences. Moreover, value infusion is more consequential when the issues in question are not constitutive for the party and when candidates are selected in a decentralized way. Similarly, routinization of internal party behavior—the second dimension of institutionalization that we account for—seems to contribute to preference homogeneity only when parties are less policy oriented and have decentralized candidate selection procedures.
Die Zeitschrift für Vergleichende Politikwissenschaft. Comparative Governance and Politics (ZfVP) ist die erste deutschsprachige Zeitschrift für zentrale Themen und innovative Forschungsergebnisse aus dem Bereich der Vergleichenden Politikwissenschaft.
In some circumstances, however, speaking with a single voice may be undesirable. For example, parties may send blurred messages to different audiences and take ambiguous positions in order to attract votes from different groups (e.g., Rovny 2012; Bräuninger and Giger 2018). Parties where actors hold divergent policy preferences can presumably send such mixed signals with more credibility and better target specific audiences (Tromborg 2019). Furthermore, intra-party heterogeneity in policy preferences may sometimes be an asset in coalition negotiations, allowing parties to achieve better policy compromises (Baumann et al. 2017).
While our main focus here lies on the potential effects of value infusion on preference homogeneity, is seems likely that there is in fact a reciprocal relationship between the two concepts. The willingness to follow the party line might be influenced by the degree to which party actors agree about policies. If agreement is generally high, it seems much easier—and thus more likely—to follow the party line in the rare occasion that there is disagreement. Since our research design does not allow to control for the direction of influence, our results concerning the effects of value infusion in preference homogeneity might be inflated. We come back to this issue in the conclusion.
We exclude the CSU because of the low number of candidate observations (24 in 2013, 12 in 2017). We also excluded the Pirates Party, which was included only in the 2013 GLES candidate survey.
These are Australia (2007), Austria (2008), Belgium (2007, 2010), Czech Republic (2006), Denmark (2011), Finland (2011), Germany (2005, 2009), Greece (2007, 2009, 2012), Hungary (2010), Iceland (2009), Ireland (2007), Italy (2013), Netherlands (2006), Norway (2009), Portugal (2009, 2011), Romania (2012), Sweden (2010), Switzerland (2007, 2011), and United Kingdom (2010).
Table A1 in the online appendix reports question wording and response categories of all items used here. To summarize: Positions on the economic dimension were measured with statements on whether government should provide social security, redistribute wealth and income, and intervene in the economy. The socio-cultural dimension aggregates views on whether immigrants should adjust to the customs of the country, on stiffer sentences for criminal offenders, on the use of torture to prevent terrorism, on same sex marriages, and on abortion. Preferences about European integration were derived from assessments of EU membership, the preferred level of European unification, and satisfaction with the way democracy works in the EU.
We calculated a party heterogeneity score only for parties from which at least ten candidates reported a preference—the measures would likely be unreliable if calculated on the basis of fewer candidates.
Another limitation is that the DALP data are from a one-time expert survey conducted in 2008 and 2009. While this timing corresponds reasonably well with the timing of the CCS data we use (see footnote 4), we have to invoke the—we believe: reasonable—assumption that routinization and other variables drawn from DALP (see below) are mostly stable over the short- to medium-run. Note that our results are robust to including only one election per country and thus to merging the DALP information only once (see the online appendix).
To test Hypothesis 3 we also need corresponding measures of preference homogeneity regarding these constitutive issues. We measured homogeneity on environment protection with an item asking whether stronger measures should be taken to protect the environment. Based on the responses to this item, van der Eijk’s (2001) measure of agreement for ordered rating scales was computed. For homogeneity on social justice, we first computed an additive index composed of the two available social justice items (social security and income redistribution) and then calculated the within-party standard deviation. Both measures were reversed and rescaled to range from zero to one.
In the online appendix, we present result from models that deal with the clustering of observations differently: Fixed effect models which include only one (the last) election per country and multilevel models with random intercepts at the election and country level. These alternative estimations lead to similar findings.
Note that the pattern for environment protection is more in line with our expectation in that the line for ecological parties is basically flat: Value infusion is not relevant when issues are constitutive. The convergence is exclusively driven by value infusion increasing homogeneity for the other parties. In contrast, the convergence for social justice at high values of value infusion is partly a result of a contra-intuitive downward sloping line for the socialist/social democratic parties.
To avoid over-specifying the regression, we relied on separate models to test H4 and H5 (while also continuing to test for the effects of routinization and value infusion in different models). The results for value infusion and party routinization, respectively, remain substantively the same, however, when the interactions with policy orientation and centralization of candidate selection are included in one model. See Figures A1 and A2 in the online appendix.