The Differentiated Effects of Direct Mobilisation on Turnout: Evidence from the 2013 Austrian Parliamentary Election

Abstract

In recent years several European countries have experienced a significant decrease in turnout, even as the level of campaign professionalisation has been increasing. Since scholars have recognised mobilisation as a key aspect in determining turnout, this article aims to disentangle the effects of mobilisation on electoral participation. It will do this by focussing on the 2013 Parliamentary Election in Austria, which saw the lowest turnout rate ever recorded in that country. The data used are taken from the Austrian National Election Study Rolling Cross-Section Panel 2013. This data source allows for the dynamic analysis of campaign effects through LOWESS estimations of the daily means of the measures of interest. Moreover, its panel structure permits us to take under control the individual likelihood to turn out before the elections whilst studying the effects of various forms of party contact on self-reported turnout. The findings show that only personal forms of party contact significantly increase electoral participation. Furthermore, these forms are more effective in increasing the turnout among low-propensity voters. Overall, one of the main contributions of the work is to provide differentiated estimations of the effects of various forms of party contact on turnout, computed on the same sample and concerning the same electoral context.

ACKNOWLEDGEMENTS

The author wishes to thank the two anonymous reviewers and the reviewers of his dissertation Sylvia Kritzinger, Ruud Luijkx, and Cristian Vaccari for having provided insightful comments to a previous version of this paper.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the author.

Notes

1 Nonetheless, these forms of contact are not analysed in GOTV research.

2 The effect was estimated in terms of Complier Average Causal Effect (CACE). It represents the average treatment effect estimated among the compliers, namely the individuals receiving the treatment when assigned to the treatment group and not receiving the treatment when assigned to the control group (Gerber and Green 2012, 137).

3 Since researchers do not know whether the individuals have actually been treated, namely whether they have received the relevant mail, this effect is calculated in terms of intent to treat (ITT) and not in terms of CACE.

4 According to Zaller’s (1992) framework, intensity itself could be interpreted as a function of media attention to a political news/issue (for instance, in terms of broadcast time or front-page space; see also Brady, Johnston, and Sides 2006).

5 In 2017 Austrian parliamentary elections, turnout rose to 80.0%.

6 It is the first rolling cross-section survey carried out in the Austrian context.

7 In Austria, political advertisements are only allowed on private TV channels.

8 In his study on Germany, Schmitt-Beck (2016) argues that these are more connected to the personal modes of contact, thus are included in the count index of personal party contact.

9 This is computed by considering the average of two items, dealing respectively with how many politicians are honest with voters and how many politicians are in politics to achieve as much personal gains as possible, which is reversed to preserve the same semantic polarity. Both questions include 5 answer categories, from 0 (almost no one) to 4 (almost all).

10 Because of sampling issues, the first four days are not considered in these analyses.

11 For an additional robustness check, data were pre-processed through Coarsened Exact Matching (CEM), a non-parametric matching technique aimed at reducing the imbalance in covariates between the treatment and control group (Iacus, King, and Porro 2012). Since being contacted by a party could depend on several individual characteristics (Karp, Banducci, and Bowler 2008), the CEM technique in our framework minimises the effect of selection bias when analysing the relationship between the various forms of party contact and turnout. Findings (see Table A4 in the Appendix) provide further evidence in support of Hypothesis 1, since only talking to a party member and being visited by a politician significantly (at the 90% level) enhance the likelihood that citizens will vote.

12 Surprisingly, somewhat religious people are significantly more likely to vote than non-religious ones, but it is not the same for very religious people compared to non-religious.

13 In Figure 4, Hypothesis 2 is tested only for the two forms of party contact which proved to be significant at the 90% level in the previous analyses, namely talking to a party member and being visited by a politician. Average marginal interaction effects between propensity to turn out and the other four types of party contact are reported in the Appendix (Figure A1).

Additional information

Notes on contributors

Riccardo Ladini

Riccardo Ladini is postdoctoral fellow at the Department of Social and Political Sciences, University of Milan. His main research interests include voting behaviour and participation, public opinion research, experimental and survey methods in social research. He is member of the Italian team of the European Values Study.