The Effect of Voting Advice Applications (VAAs) on Political Preferences - Evidence from a Randomized Field Experiment

Abstract

Can online Voting Advice Applications (VAAs) that match voters to candidates and parties based on issue congruence influence voters' political preferences? To address this question, we conducted a field experiment with University students during the 2011 Swiss federal elections and examine whether voters who used the Swiss VAA smartvote prior to the elections were more likely to adapt their initial party preferences.

smartvote is an online web tool that enables voters to match their issue preferences to those of parties and candidates running for office, providing voters with custom made voting recommendations. Moreover, the tool allows voters to visualize their political position in the political landscape and compare their position to the political offer. The VAA gives unprecedented access to detailed political information and lets voters systematically compare electoral alternatives. Given these unique opportunities offered by the VAA, we analyze whether the revelation of "objective" preferences by the voting recommendation initiates changes in "subjective" preferences of voters.

Decades of research on political behavior indicate that voters' knowledge of and interest in politics is fairly limited and that political preferences are therefore highly amenable to (additional) information. Hence, we expect that voters who have their policy matches revealed by smartvote to be more likely to reconsider previously held convictions and consequently show changes in their political preference structure. We argue that VAAs can reinforce and rearrange pre-existing preferences. Employing the common political preference measure of an individual's propensity to vote (ptv) for specific parties, we ascertain whether the smartvote voting recommendation caused voters to adapt their ptv scores for the running parties in the election.

To empirically test these assumptions, we conducted a randomized field experiment during the 2011 Swiss federal elections among 2'000 University students in Switzerland. The treatment group received * Contact: pianzola@nccr-democracy.uzh.ch an email invitation to use smartvote before elections, providing survey participants with a personalized login for the website. The control group did not receive such an invitation. We measured the baseline characteristics and political preferences of both groups before and after the elections by means of an online survey.

Given the randomized treatment assignment and the panel structure of the experiment, the data allows us to estimate the causal effect of smartvote use on political preferences. Due to non-compliance in the sample, we identify local average treatment effects for compliers. We find significant changes in the awarded scores for the top party preference among smartvote users. The assigned ptv score for the top party preferences increases significantly among those voters who used smartvote. Among those who changed their top party ptv score, most remained with the same party choice. Thus, in case of the most preferred party, smartvote use seems to reinforce pre-existing preferences. We do not only find a strengthening of the top party preference but also a multiplication thereof. Voters who consulted the tool report higher likelihoods for considering alternative choice options at elections. smartvote users, compared to non-users, are significantly more likely to change their initial preferences from a single most favored party to multiple highly preferred parties. In other words, being exposed to detailed information about vote alternatives seems to incline voters to consider these alternative options more closely and include the closest ones as part of their future choice set. These systematic shifts among smartvote users present empirical evidence for a causal effect of VAA use on political preferences.

Files

Additional details