https://orcid.org/0000-0002-5069-7964
![Sample our Politics & International Relations journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5947/TOC-Subject-Sample-2021-Politics-International-Relations.jpg"})
ABSTRACT
Electoral competition is a cornerstone of representative democracies. However, measuring its extent and intensity constitutes a challenging task for the discipline. Based on multilevel conceptualizations, we discuss three different measures of political competition (electoral volatility, vote switching, and voters’ availability) and their relation to each other. We argue that electoral volatility and vote switching as indicators of electoral competitiveness tend to misestimate the degree of competition in multiparty systems. As an alternative, we propose focusing on the individual’s propensity to vote for different parties, i.e. electoral availability. Using data provided by the European Election Studies, we compare availability to electoral volatility and vote switching in the framework of necessary and sufficient conditions. Our regression results show that operationalizing electoral competitiveness based on voter availability – which is increasingly retrievable from cross-national voter surveys – helps to avoid type-II errors, i.e. identifying competitive elections as less or non-competitive.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 It is worth mentioning, however, that we do not apply a strict understanding of necessary and sufficient conditions. A narrower definition of necessary/sufficient would imply that there are no deviant cases, i.e. no cases with low availability and moderate or high levels of vote switching. Against the background of possible measurement error in the data, we will outline our (more stochastic-oriented) usage of the terms below.
2 The corresponding survey question reads as follows: “How probable is it that you will ever vote for the following parties?” Answers range from 0 (not at all probable) to 10 (very probable). For our purposes, we recoded all PTVs to range from zero to one. Note that in earlier election studies, the numbers ranged from 1 to 10. The rescaling to 0–1 still allows for comparison.
3 Naturally, the party the person has the highest propensity to vote for is not included in the n parties mentioned in the equation.
4 We also calculated the median to take care of potentially influential outliers (cf. ). We thank the anonymous reviewer for raising this point. Additionally, we estimated the models restricting the individual-level sample to voters only (cf. ). Neither specification changed the results significantly.
5 We therefore do not consider non-voters in the following analyses. Note that in elections with highly fluctuating voter turnout, volatility can increase despite moderate levels of vote switching. The depth of documentation of non-voters in the surveys used, however, does not allow for further investigation of this possibility. We thank one anonymous reviewer for making us aware of this point.
6 The dataset contains information on Belgium, Denmark, France, Germany, Greece, Ireland, Italy, Luxembourg, Portugal, Spain, the Netherlands, and the United Kingdom.
7 PTVs do relate to actual voting decisions (cf. van der Eijk et al. Citation2006). In 2014, 76 percent of the voters reported vote choice for the party with the highest PTV in the last national election and 88 percent planned to do so in the next national election. Using European Election data implies considering the second-order nature of these elections (e.g., Giebler and Wagner Citation2015). However, potential distortions due to the second-order nature should equally apply to all three measures. As we are interested first and foremost in the relationship between volatility, vote switching, and availability, the second-order nature of the elections does not affect the substance of our results.
8 Although including survey year dummies to the regression model risks overfitting the model, we did so in a robustness test. Additionally, we added variables controlling for the timing of the European Election Study within the national election cycles However, controlling for these time variables does not change the substantial interpretation of our empirical results (see table A.1 in the Appendix).
9 We also analyzed the relationship between volatility and availability. As expected, availability constitutes a necessary but insufficient condition of volatility. See and as well as in the Appendix. As an additional robustness test, we estimated the relationship without online panel election studies (Italy, Belgium, Ireland, and the Netherlands in 2004). Here, response patterns (e.g., straight lining) might bias the measurement of citizens’ availability. However, the results (not presented here) remain the same once we drop these cases.
10 Note that, at the same time, the collection of citizens’ PTV scores is less costly than collecting panel data which would give us a less biased picture of voters’ switching behavior.
11 The following countries are included in the analysis: Austria, Belgium, Brazil, Bulgaria, Taiwan, Croatia, Czech Republic, Denmark, Estonia, Finland, Germany, Greece, Hungary, Iceland, Israel, Italy, Latvia, Montenegro, the Netherlands, New Zealand, Norway, Portugal, Slovakia, Slovenia, South Africa, Spain, Sweden, Switzerland, Uruguay.
12 As in the main text, outliers were excluded following Cook’s Distance criterion. We also exclude elections with unreasonably high switching scores (above 60%) as we expect errors in the party coding provided by CSES. This includes Bulgaria (2001), Israel (2006 and 2013), Latvia (2010), Poland (2001 and 2005), Romania (2004), and Serbia (2012).