Participation in hard times: how constrained government depresses turnout among the highly educated

Abstract

Existing studies on electoral turnout in times of economic crisis have predominantly focused on disadvantaged voters. However, during the recent economic crisis, turnout among highly educated citizens has strongly declined as well. Existing resource-based theories of political participation cannot account for this. This article suggests that the anticipation of government inefficacy is an important driver of abstention among highly educated. Where governments are severely constrained, these citizens anticipate that the hands of future governments will be tied. Hence they are more likely to abstain out of frustration or rational calculations. The study uses the recent economic crisis as test case, as it entails particularly acute constraints on several European governments. The cross-sectional and longitudinal evidence – based on ESS survey data and different measures of government constraint in 28 European countries – provides ample support for the argument.

Keywords:

• European politics
• elections
• turnout
• participation
• economic crisis
• Eurocrisis

Acknowledgements

Previous versions of this manuscript were presented at the EUDO conference, 2013; at the EPSA, 2014; at the Comparative Political Economy Workshop in Konstanz, 2014; at the BIGSSS workshop in Bremen, 2015; and at the publication seminar in Zurich 2015. We would like to thank the participants of these seminars for their valuable comments, and in particular we would like to thank Stefanie Walter, Stefaan Walgrave, Hanspeter Kriesi, Marius Busemeyer, and Julian Garritzmann for helpful feedback. We are also grateful to the anonymous reviewers for their insightful comments.

Notes

1. Data relates to parliamentary elections, retrieved from the IDEA database: http://www.idea.int.

2. The following countries are defined as strongly affected by the crisis, because they showed bond yields of 6% or higher, deficits of 10% or higher, or the presence of conditionality agreements in one or more years from 2006 until 2012 (also see section on data and methods): The Southern European countries Cyprus, Spain, Greece and Portugal, the Baltic states (Estonia, Latvia and Lithuania), Iceland, Ireland, Bulgaria, Poland and Hungary. For Romania, Iceland and Latvia, however, only one ESS round is available so we could not calculate their participation rates over time.

3. The empirical evidence for the economic vote overall is somewhat inconclusive and effects are substantively small (Fraile and Lewis-Beck 2014; Kayser and Peress 2012; Kayser and Wlezien 2010).

4. International trade flows, in contrast, do not induce systematic constraints on governments, since they are less volatile and thus less consequential for the domestic economy.

5. Mair describes a long-term, structural process of political disaffection, not a cyclical, situational reaction that we study in this article. However, these two empirical views (long-term gradual vs. cyclical) are entirely compatible. Not by accident, the crisis has hit those countries hardest which have been affected by institutional, political and economic rigidities for a very long time. The political elites in these countries have been severely constrained in their actions by both institutional legacies and oftentimes economically dysfunctional power relations since the 1980s (Beramendi et al. 2015). Hence, in our attempt to understand electoral participation under constrained government, the crisis in itself is not a rival, alternative factor to such structural rigidities. Rather, the crisis works as a magnifying lens that brings existing constraints into the light, exacerbates them and adds even further (external) limitations to what governments can do.

6. In the section below on ‘Data and methods’, ‘How constrained government affects the relationship between education and participation’, we develop and discuss the respective indicators to control for programmatic convergence in our models. The correlation between our joint factor of economic constraints and Dalton’s (2008) polarisation index measuring party system convergence is –0.15. The correlations between the separate constraint measures – deficit, conditionality and bond-yields – are –0.27, –0.08 and –0.15, respectively.

7. The correlation between our joint factor of economic constraints and the restricted polarisation index measuring mainstream party convergence is –0.29. As for the separate constraint measures, the correlation is –0.35 for deficit, 0.10 for conditionality and –0.29 for bond yields.

8. Note that oftentimes not all ESS waves are available for a country.

9. More precisely, revenue (including grants) minus expense, minus net acquisition of nonfinancial assets.

10. Greece, of course, is subject to constant conditionality agreements from 2010 on. However, it is not included in our sample in 2012.

11. Pearson’s R of –0.38 (deficit and bond yields), –0.27 (deficit and conditionality) and 0.66 (bond yields and conditionality).

12. In the following tables, the coefficients, standard errors and levels of significance of the following indicators are reported at the higher levels of countries and years: deficit, conditionality, government bond yields, the factor of the three just listed indicators, growth in GDP, party system or mainstream party polarisation, effective number of parties, party system disproportionality and recent democratisation.

13. We show the predicted probabilities with the control variables fixed at zero (dichotomous variables) or at their mean (continuous variables).

14. This moderation is quite linear over the different educational levels, as regression analyses with an ordinal specification of education indicate (see Table A4 in the online appendix).

15. The age cohort argument brought forward by Franklin (2004) and further specified by Franklin and Hobolt (2011) can be traced in our analysis (see Tables A8 and A9 in the online appendix). The impact of government constraints on the relationship between education and turnout is significant for both younger and older age cohorts. However, it is stronger negative among citizens 35 years and under (odds ratio of 0.85) than among citizens above 35 (odds ratio of 0.95). Hence, younger citizens who have not yet fully developed their habit of voting are more affected by the macroeconomic context of elections.

16. We only show results of one measure for government constraints, the factor resulting from the other three indicators, noting that the application of alternative indicators, yields very similar effects (see Table A5 in the online appendix). In addition, for the sake of interpretability, we abstain from presenting the dynamic analysis as a three-way interaction between education, government constraints and the four years. Such a specification, however, leads to robust findings regarding the continuous change in the moderation effect of government constraints (see Table A6 in the online appendix).

17. Here we rely on the variable chpldmc in the ESS 2012, which asks people to what extent (0–10) they think that: ‘in [country] government changes policies in response to what most people think’.

18. Due to the different scale of the dependent variables, linear multilevel models instead of hierarchical ordered logit models have been estimated. See Table A10 in the online appendix for the full results. Table A10 also includes a linear probability model for our main dependent variable, participation, to set the effects into perspective and demonstrate that the magnitude is comparable.