Understanding people’s choice when they have two votes

ABSTRACT

This paper introduces a model of vote choice in mixed-member proportional representation systems where electors cast two votes. Despite the growing popularity of mixed systems around the world, a recent stream of literature suggests that the candidate vote contaminates the list vote, inducing the type of behavior observed under majority rule. We propose a new approach to account for these so-called “contamination” effects, a phenomenon that we define as a causal influence making choices more similar across the vote decisions. Since causality entails a time ordering, we argue that contamination arises only when voters choose sequentially. By making use of new survey questions asking respondents about the timing of vote decisions, we can estimate the magnitude of these contamination effects directly. The model is tested using Bayesian multinomial probit models with survey data from the 2013 federal election in Germany. A key contribution of this paper is to show that contamination effects are present only among voters with lower levels of education, and work primarily from the list vote to the candidate vote. We also test a number of predictions about the determinants of the two vote choices in mixed systems.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes on contributors

Ludovic Rheault is an assistant professor in political science at the University of Toronto, where he works in the fields of data analysis, computer-assisted textual analysis and political behavior. Recently he contributed to the Lipad project (www.lipad.ca), a digitized collection of Canadian parliamentary debates covering more than a century of history.

Andre Blais is a professor in the Department of Political Science at the University of Montreal where he holds a Research Chair in Electoral Studies. Recently he co-authored Multi-Level Electoral Politics: Beyond the Second-Order Election Model (Oxford University Press 2017) and co-edited The Many Faces of Strategic Voting: Tactical Voting in Electoral Systems Around the World (University of Michigan Press 2018).

John H. Aldrich is Pfizer-Pratt University Professor of political science at Duke University. He recently co-authored Why Parties Matter? Political Competition and Democracy in the American South (University of Chicago Press, 2018) and Change and Continuity in the 2016 Elections (CQ Press 2018), and co-edited The Many Faces of Strategic Voting: Tactical Voting in Electoral Systems Around the World (University of Michigan Press 2018).

Thomas Gschwend is a professor of political science at the University of Mannheim, Germany. He is co-author of Multi-Level Electoral Politics: Beyond the Second-Order Election Model (Oxford University Press 2017) and has published widely on the topics of elections and public opinion.

Notes

1 Data retrieved from https://www.idea.int/data-tools/data/electoral-system-design on 19 March2018. Note that the definition used to classify mixed systems may vary, and some include additional countries into this category (see Carter and Farrell 2010; Massicotte and Blais 1999; Shugart and Wattenberg 2001).

2 While “spillover” has fewer negative connotations in common parlance, “contamination” has become the dominant term used in the literature.

3 A reason for the larger proportion choosing the candidate vote first may be the fact that this vote appears first on the German ballot. However, for the theoretical reasons outlined above, we do not expect the candidate vote to induce stronger contamination effects, all else being equal.

4 Notice that the contamination effect parameters are not specific to party. In multi-party elections, such models will likely be estimated with multinomial models in which party-specific variables have constrained coefficients across the alternatives. As a result, there would be only one parameter for each pair of elections {s, t}.

5 The evidence in favor of the hypothesis is considered very strong if 2 log(B₁₀) is greater than 10, strong if between 6 and 10, positive if between 2 and 6, and barely worth mentioning if between 0 and 2 (Kass and Raftery 1995, 777). The null is supported if the value is negative.

6 Note that the Bayes factor cannot be computed in one case for which 100% of the draws are greater in size than the median in the other vote model.

7 The effect of coalition preferences on the vote is an important question that would deserve a thorough analysis, one that we could not include in this paper (on the topic, see Meffert and Gschwend 2010; Plescia and Aichholzer 2017).

Additional information

Funding

This work was supported by the Social Sciences and Humanities Research Council of Canada [grant number 412-2009-1004].