Political information and retrospective voting

Abstract

This study aims to provide a comprehensive view of the role of political information in retrospective voting by simultaneously investigating the effects of levels of information on the individual level and the availability of information on the contextual level. It is argued that the sophistication-gap in retrospective voting is confined to those contexts in which it is hard for voters to assign responsibility for the country’s state of affairs to the correct political party/-ies. Contrary to this expectation, analyses using the CSES data show that there is a larger difference in retrospective voting between voters with different levels of political information in high-clarity contexts, while voters do not seem to hold incumbent parties to account at the polls in low-clarity contexts – irrespective of their level of political knowledge. However, additional analyses show variation in this result depending on the data and measures used.

Keywords:

• Retrospective voting
• political information
• clarity of responsibility
• economic voting
• political sophistication

Acknowledgements

This article has greatly benefitted from constructive feedback on several occasions. I thank all participants of the Montréal-Leuven Winter School on Elections (Montréal, 2018), and a research seminar at KU Leuven, for helpful comments and feedback. Furthermore, I thank Ruth Dassonneville, Shane Singh, Marc Hooghe, Martin Okolikj, Irene Esteban and Zoltan Fazekas in particular for their constructive comments. I also thank the editors and anonymous reviewers of West European Politics for their very useful suggestions.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1 In this study, I follow Jessee (2017: 711), noting that: ‘Following Luskin (2002), I use “knowledge,” “sophistication,” and “information” interchangeably’.

2 Note that, while in this study the focus is on the extent to which voters with different levels of sophistication hold the government accountable, other research has shown that voters with different levels of sophistication are likely to use other kinds of performance evaluations in deciding which party to vote for (i.e. egotropic versus sociotropic assessments). See, most notably, Gomez and Wilson (2001, 2006). However, see Godbout and Bélanger (2007) and Gomez and Wilson (2007).

3 Focusing on retrospective voting on party level, Stiers (2019b), however, does not find evidence of a sophistication-gap in retrospective voting.

4 It needs to be noted that de Vries and Giger (2014) have attention for the different systems they investigate, as they report their findings for high- and low-clarity contexts separately in the online appendix to their article. However, they do not provide a theoretical reason as to why the sophistication-gap in retrospective voting would be different between these two sub-samples, nor a formal test of the significance of the difference between the results in the two groups respectively. This study provides these contributions.

5 While I use the data of the second and third module in additional analyses reported in the discussion, I do not use the data of the first module of the CSES. In the first module, the knowledge questions varied over the different countries, making them less comparable, especially in combination with the data of the fourth module, in which the same questions were asked in all countries.

6 More specifically, the following elections are included in the analyses: Australia (2013), Austria (2013), Bulgaria (2014), Brazil (2014), Czech Republic (2013), Finland (2015), Germany (2013), Great-Britain (2015), Greece (2012), Hong Kong (2012), Iceland (2013), Israel (2013), Japan (2013), Kenya (2013), Mexico (2012, 2015), Montenegro (2012), Norway (2013), New Zealand (2011, 2014), Philippines (2016), Poland (2011), Portugal (2015), Romania (2015), Serbia (2012), Slovakia (2016), Slovenia (2011), Sweden (2014), Switzerland (2011), Turkey (2015), and U.S.A. (2012).

7 As previous studies have shown that, in coalition governments, it is mostly the party of the Prime Minister that receives credit or blame for the government’s performance, I also conducted the analyses with only the main party coded as incumbent. The results, reported in online appendix B, support the conclusions presented here. (The results of this robustness test also hold using an adapted measure of clarity of responsibility, in which the indicator of coalition government or not is excluded.)

8 As the models predict individuals’ voting behaviour, I use economic perceptions rather than aggregate-level economic indicators. See Stevenson and Duch (2013) for a discussion on the use of economic perceptions in the study of retrospective voting.

9 More pragmatically, it also needs to be noted that only sociotropic measures were available in the data sets used here, and hence it is not possible to replicate the analyses using egotropic perceptions.

10 I prefer this index over the more recent index of Dassonneville and Lewis-Beck (2017), because the latter use aggregate-level economic data in their analyses, while Hobolt et al. (2013) use individual-level perceptions, which is more in line with the models presented here.

11 As a test, I also conducted the analyses using the institutional measure of clarity of responsibility of Hobolt et al. (2013) as well as a measure combining both dimensions. (However, coding difficulties led to the exclusion of committee structures as an indicator.) The results of these analyses are summarised in online appendix C (institutional measure) and online appendix D (combined measure). The classic economic voting thesis and the moderation by voter knowledge are supported. In line with the results of Hobolt et al. (2013), institutional clarity does not significantly moderate economic voting, and the results show that voters vote retrospectively in the lowest-clarity contexts as well. The combined measure does seem to moderate economic voting, and using this measure, there is no evidence for economic voting in the lowest clarity contexts. The difference in moderation of voter knowledge between different contexts is not significant with any of the alternative measures. However, using the institutional measure, the lowest knowledgeable voters do not seem to vote retrospectively in low institutional clarity contexts, while the most knowledgeable voters do. Using the combined measure, there is no evidence for economic voting in the lowest clarity context – irrespective of the voter’s level of knowledge.

12 See online appendix E for an overview of all elections under investigation and their corresponding level of clarity of responsibility.

13 Note that Hobolt et al. (2013: 184) do not include control variables in the analyses. Therefore, as a robustness test, the models were also estimated without the control variables. The results are summarised in online appendix F and are substantially the same as the results presented here.

14 However, it needs to be noted that including party identification in the model risks biasing the coefficients, as it can affect several associations between the variables of interest in this specific model. Party identification is one of the main determinants of the vote (Campbell et al. 1960), and some argue that it is a tautological measure to the vote. In the CSES data at hand, 82.72% of the observations can be correctly classified using this indicator only. Furthermore, it is also possible that party identification is in itself a result of retrospective performance evaluations (Fiorina 1981). Moreover, there are indications that the impact of partisanship as a heuristic differs between voters with different levels of political sophistication (Brader and Tucker 2012; Marthaler 2008). As the association between performance evaluations, voter knowledge, and the vote, are of primary interest in this study, a model without party identification is preferred as main model, while the results including partisanship are included in online appendix G.

15 An important limitation of these analyses is that it remains unclear whether the different results can be fully attributed to the different measure of performance evaluations, or whether it can be partially explained by the different countries and elections under investigation. Unfortunately, none of the CSES modules so far include both economic and general measures, making it impossible to test whether the results would be equivalent on an identical set of countries and elections.

16 Note that these analyses are not presented as the main results because they rely on a vote intention measure as dependent variable, rather than actual reported voting behaviour as the results presented in the main text. See online appendix I for more information.

Additional information

Funding

Fonds Wetenschappelijk Onderzoek.

Notes on contributors

Dieter Stiers

Dieter Stiers is a PhD candidate in political science at KU Leuven (Belgium). He focuses mainly on electoral behaviour, and on accountability mechanisms and the congruence of the vote more specifically. His research has been published in, among others, the European Journal of Political Research, Electoral Studies, Political Behavior, and Party Politics. [[email protected]]