Abstract
A large body of literature has investigated the factors that lead to abstention on the one hand or vote volatility on the other hand. We argue in this paper that the most fruitful approach is to simultaneously consider the exit, voice, and loyalty options. The analyses are based on data from the Comparative Study of Electoral Systems and cover a broad set of advanced democracies. We demonstrate that while party switchers and abstainers have a lot in common, switching parties can be considered a more positive choice. Most importantly, contrary to what previous research suggested and in contrast to abstainers, party switchers cannot be characterized as frustrated with politics. Furthermore, the supply side does to a certain extent affect whether voters choose to remain loyal, switch parties, or abstain from voting.
Acknowledgements
A previous version of this manuscript was presented at the Annual Conference of the Midwest Political Science Association (MPSA), Chicago, 2013. Ruth Dassonneville acknowledges the Research Foundation Flanders (FWO) for its generous support.
Supplementary material
Supplemental data for this article can be accessed at http://dx.doi.org/10.1080/17457289.2015.1016528.
Notes
1. Given that Belgium can be considered a country with two separate party systems, the Dutch- and French-speaking samples for Belgium were considered different election-samples.
2. A full overview of reasons for excluding some elections is provided in Online Appendix 2.
3. Only election samples for which no single party's vote share was over 7.5 points off the official results and for which this bias strongly diverged from the bias in the survey year (over 5 points difference) were retained. Additional tests indicated that being stricter in which election samples to include for the analyses does not alter our conclusions. See Online Appendix 4 for detailed information on biases and Online Appendix 2 on which election samples were retained.
4. An extensive description of the coding of party switching for all countries included can be found in Online Appendix 3.
5. Note that the addition of more control variables reduces the estimation sample at both the individual and the contextual level, see Online Appendix 5.