All (electoral) politics is local? Candidate's regional roots and vote choice

ABSTRACT

Many authors argue that candidates are more popular among voters from their own region. Two potential explanations have been suggested: voters’ identification with their home region, and the representation of regional interests. The information on candidates’ residence can be transmitted to voters in different ways, the most easily accessible way being information printed on the ballot paper. However, most studies on “friends and neighbour voting” use aggregate data. Studies that rely on individual level data usually put respondents in hypothetical situations and confront them with synthetic candidates, reducing their realism. To bridge this gap and to test the effect of providing information on the candidates’ residence, we use data from a survey experiment to analyze voters’ responses to ballot paper information on the regional background of real candidates in the 2014 European election in Germany. We find that voters in an open list PR election are more likely to support regional candidates if ballot paper information on the candidates’ geographic background helps them to do so. The appeal of personal ties is a stronger explanation for vote preference than the one based on regional interests.

Acknowledgements

The experiment was conducted as part of the Making Electoral Democracy Work project (http://electoraldemocracy.com), founded by the Social Sciences and Humanities Research Council of Canada. The authors thank the three anonymous reviewers, Sarah C. Dingler, Andreas Dür, Jessica Fortin-Rittberger, Corinna Kröber, Philip Manow, Valentin Schröder, and Thomas Zittel for helpful comments and suggestions. Karine Van Der Straeten acknowledges funding from the French National Research Agency (ANR) under the Investments for the Future programme (Investissements d'Avenir, grant ANR-17-EURE-0010).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

Replication Data and Syntax are available at Harvard Dataverse, https://doi.org/10.7910/DVN/HLVXRM

 Google Scholar

.

Notes

1 Indeed, one could argue that voters living in regions with higher levels of local identification care more than others about what happens to their region as compared to the rest of Germany. The reason why they support local candidates might than be to foster the prospects of their region. This explanation would combine instrumental motives and a particular attachment to one’s region. Our definition therefore emphasises the self-centred character of instrumental geographic voting.

2 In addition, EP election is referred to as “second order elections” (Reif and Schmitt 1980). The second order character of EP elections might reinforce the effect of candidate characteristics if voters are less interested, collect less information on what is at stake in the election, and rely more on ballot paper cues.

3 Note that this realism might come at the cost of validity. Given that we rely on real candidates (instead of fictitious), we cannot randomize their geographical background, and hence we cannot ensure that this geographic background is independent from other unobserved characteristics like their quality as politicians.

4 Randomisation of list order and random composition of the CDU/CSU list might have frustrated some participants and influenced their behaviour.

5 Occupation is left out in order to be able to concentrate the following analyses on geographic cues. See A1 for a sample of the original ballot paper and A2 for a screenshot of the online ballot used in the experiment.

6 The questionnaire is documented in A13.

7 See A5 in the Online Appendix for full models. Note that interaction models have to interpreted with precaution (Hainmueller, Mummolo, and Xu 2019). This caveat specifically applies to three-way interactions that we use here. However, for the base model and its two-way-interaction (Model 2 in Table A14 in the Online Appendix), we can exclude lack of common support, one of the two problems Hainmueller, Mummolo, and Xu (2019) refer to. As Tables 2 and 3 above show, we base our estimates on a sufficient number of observations in all categories of our dependent variable and sufficient variation in the treatment.

8 In order to better represent the impact of political knowledge graphically, we dichotomise the variable: 0 = political knowledge low (0 to 3 correct answers to five survey questions on EP elections), 1 = political knowledge high (4 or 5 correct answers). Compared to other European participants in the EVP survey, Germans score relatively low on the five-point knowledge scale (see A6).

9 Less knowledgeable voters might use positive and negative votes to a lesser extent than the more knowledgeable ones. In that case, the observed effects would then be due to a higher likelihood to cast preference votes (both positive and negative) among more knowledgeable voters. This, however, is not the case in the EVP sample. For a detailed discussion see A12.

Additional information

Funding

This work was supported by Agence Nationale de la Recherche: [Grant Number ANR-17-EURE-0010].