Estimating policy positions of local parties in elections with multi-vote ballots

ABSTRACT

This paper investigates the potential for estimating policy positions from electoral results in elections with multiple votes. When voters can spread their votes across multiple party lists in open list elections, they are more likely to select candidates from parties with similar policy positions. The electoral results can therefore be exploited to infer parties’ preferences based on the structure of vote combinations. The proposed data provide a valuable tool for analysing party behaviour in circumstances where ordinary methods for estimating policy positions fail, most importantly in electoral contexts with local competitors. Applying an ideal point model for count data, party preferences are estimated for a German municipality.

KEYWORDS:

• Electoral behaviour
• party positions
• measurement
• panachage

Acknowledgements

I am grateful to Mariyana Angelova, Thomas Gschwend, Arndt Leininger, Sven-Oliver Proksch, Susumu Shikano and the participants of the Democratic Governance Colloquium at the Department of Government, University of Vienna, for valuable feedback on the ideas presented in this paper. I am also grateful to Stefan Lenz for providing access to the data analyzed in the empirical part of the contribution.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1. The visual representation differs from the actual data source for reasons of simplicity. Notably, the actual matrices need not be – and are not – perfectly symmetric. The actual data matrix for the empirical analysis is provided in Table A1 in the Appendix.

2. The party positions are estimated with a mean value of 0 and a standard deviation of 1 (Slapin and Proksch 2008, 710). A cardinal interpretation of the estimated space is possible such that distances between the party positions can be compared.

3. The data matrix underlying the estimation is provided in Table A1 in the Appendix. Estimations were performed using version 0.2 of the Austin package in R.

4. There is still the caveat that up to this point, there is no evidence on the degree to which the estimates are driven by voter misperception (see Validation of the estimates).

5. Nevertheless, one might have expected the AfD to be assigned a more extreme policy position within the conservative party camp. The comparatively moderate position of the AfD in the estimated space might indicate that a second dimension would be necessary to accurately portray the policy profile of this particular party. As the AfD receives votes from the entire political spectrum, it might score highly on an anti-system dimension. Alternatively, one might consider whether the concurrent elections to the European Parliament have shifted the estimated dimension to reflect a European integration dimension. This scenario seems unlikely since the AfD should not be placed midrange on a European integration dimension.

6. As the underlying data matrix is not symmetric, the estimated policy positions need not be perfectly aligned from a technical point of view. However, there is no plausible theoretical argument for preferring the position estimates of the receiver over those of the sender or vice versa as the voters’ incentives to link two ideologically close parties should work in both directions. Therefore, notable differences between the position estimates of senders and receivers would discredit the proposed data source – or the appropriateness of the applied ideal point model at the very least.

7. Estimations were performed using version 1.46 of the pscl package in R. The vote advice application provides estimates of the state-level party positions. It contains no information on purely local competitors.

8. It was not possible to acquire a party manifesto from AfD, generation.hd or HPE.

9. The manifestos were preprocessed prior to the analysis in accordance with common practices (Munzert et al. 2014, ch. 10). The content was transformed to lower case; numbers, punctuation and common words were removed along with terms that did not appear in at least one-third of the texts in order to make the estimates less dependent on infrequent words.

Additional information

Notes on contributors

Dominic Nyhuis

Dominic Nyhuis is a researcher at the Department of Politics at the Goethe University, Frankfurt, Germany. His research focuses on party politics, candidates, comparative municipal politics and small-area policy preferences.