At an ever increasing rate voting advice applications (VAAs) are deployed during election campaigns. Often framed as citizen empowerment tools, VAAs are closely linked to hopes of boosting turnout. In this paper, we assess the electoral mobilization effect of the Swiss VAA smartvote, one of the most established of its kind. For severe problems associated with an experimental approach, we settle for an observational design using data from the 2007 Swiss Electoral Study (Selects). To counter self-selection into treatment, we employ entropy balancing, a recently introduced pre-processing technique that directly incorporates covariate balance into a unit weights function. We find that even in the most conservative of our models, smartvote users are statistically significantly more likely to vote compared to non-users. The mobilization effect withholds nearly a dozen sensitivity analyses, including placebo tests to test for selection on observables.
