Role of Self Selection in Estimating the Effects of Voting Advice Applications: Empirical Evidence on the Basis of Swiss Smartvote Data

Voting advice applications (VAA) proliferate across Europe and beyond. These non-partisan internet applications in which the political offer is matched with voters’ preferences assist voters in their decisions to which party to vote for. With the growing number of VAA users in several European polities, recent years have also seen emerging scholarly interest in detecting the impact of VAAs on voting behavior. However, dominant research in this field offers contradictory evidence for it suffers from poor data quality and self-selection biases. To remedy these problems I employ panel data from Switzerland and demonstrate how using a Heckman selection model allows to correct for selection biases that are inherent to these data. In so doing I first replicate the results found by a number of studies relying on Swiss Smartvote data and confirm that the voting advice indeed has a sizable effect on individual level vote choice. These findings are then compared with the results where the potential selection mechanism is controlled for. Preliminary findings suggest that the naive estimator is likely to overreport the effect size, since it is driven by the subgroup of the entire universe of Smartvote users for whom the effect may be higher than on average. Indeed, the results from the Heckman model confirm that after controlling for the non-random event of individual participation in the panel survey, VAAs effect on vote choice diminishes considerably.