More Choice, Higher Turnout? The Impact of Consideration Set Size and Homogeneity on Political Participation

Abstract

Central to the emerging scholarship on how political supply influences electoral behavior is the claim that more choice leads to higher turnout. However, empirical tests of this proposition have been limited to the aggregate level. This article examines the relationship between the properties of electoral choice sets, as perceived by the voters, and electoral participation. Following recent advances in choice research, the article distinguishes between an awareness set, consisting of all choice options known to the voter, and a consideration set which includes only those alternatives that are seriously considered by the voter. We hypothesize that the cardinality, ideological homogeneity and distinctiveness of individual consideration sets are positively associated with electoral participation. The expectation is tested with individual-level data from three waves of the European Elections Study. Our results suggest that the relationship between the structure of political supply and participation is complex: while the number of choice alternatives in the consideration set is positively associated with turnout, the ideological diversity of choice options suppresses electoral participation.

Disclosure Statement

No potential conflict of interest was reported by the authors.

Notes

1. Extending the pooled dataset beyond the 1999 wave is not feasible due to the absence of key indicators used in our operationalization. Notice that we have dropped observations from enforced compulsory voting countries Belgium and Luxembourg (all years) and Greece (1999) (http://www.idea.int/vt/compulsory_voting.cfm).

2. The EES survey question is formulated as follows. “Some people are quite certain that they will always vote for the same party. Others reconsider in each case to which party they will give their vote. I shall mention a number of parties. Would you indicate for each party how probable it is that you will ever vote for that party?” The respondent is provided with the list of parties in the respective polity with a scale ranging from 0 to 10, where 0 means “Will certainly never vote for this party” and 10 means “Will certainly vote for this party at some time”.

3. We replicated the analysis that follows also by grouping parties with a PTV between 4–7 and 0–3. The effects were not statistically significant, indicating that the PTV cut-off point suggested by van der Eijk and Oppenhuis (1991) is an empirically valid threshold separating parties that seem to matter most for the vote choice.

4. Q: And about where would you place the following parties on this (left-right) scale? How about the (Party a)? Which number from 0 to 10 [1–10 in 1999 and 2004], where 0 [1 in 1999 and 2004] means “left” and 10 means “right” best describes (Party a)? Q: And about where would you place the following parties on this scale (EU)? How about the (Party a)? Which number from 0 to 10 [1–10 in 1999 and 2004], where 0 [1 in 1999 and 2004] means “already gone too far” and 10 means “should be pushed further” best describes (Party a)?

5. Missing party positions were imputed using the iterative Markov chain Monte Carlo method assuming an underlining multivariate normal model (using mi impute mvn command in STATA 12). We imputed the socio-demographic controls and the party positions on both dimensions and did so for each country in each year separately. Both the left-right and the EU-integration dimension tended to have more missing values on smaller parties and the EU-dimension had more missing values in general indicating clear non-random missingness. We did not impute, however, the PTVs, as these formed the core of the operationalization of the choice sets and lacking any choices due to not wanting to assign PTVs suggest the individual is not even aware of the party which differs from being unable to place a given party on an ideological dimension. We produced five datasets with imputations to arrive at the recommended 90% efficiency level (Cole 2008, 224).

6. Computed based on data from the European Election Datatabase http://tinyurl.com/n9ggeuq. Independent candidates were treated as one party and only parties receiving more than 1% of the vote were included in the calculations.

7. Multicollinearity diagnostics were within the acceptable limits, the largest variance inflation factor was 3.35. The measures of fit which we can report are severely limited due do the analysis being performed on multiply imputed datasets.

8. Given that the number of parties in the consideration set could be partly determined by the overall number of parties in the given country, we also ran the model with an alternative specification of the cardinality variable (the share of parties included in the consideration set out of all parties). The effect was similarly positive and significant, though somewhat stronger than the number of parties expressed as a count. We still use count as the preferred measure as it corresponds more closely to our theoretical approach.

9. We also ran the model on old and new democracies separately and found that all of the effects hold for old democracies, while the number of parties in the consideration set and the ideological span did not play a significant role in new democracies. The distinctiveness of the consideration had a comparable and persistent effect in both groups.