Do voters use voting advice from online tools to update their political preferences? The case of Denmark

Voting Advice Applications (VAAs) are online tools that allow a large number of voters to get an estimate of how much their policy preferences match those of the parties and candidates they are eligible to vote for. They have millions of users in many countries, but there are still gaps in our knowledge on how these devices impact their users. Some previous studies have used experimental designs that enable them to causally identify the effects of VAAs on vote volatility and fragmentation, but not on the probability of switching to a recommended party. Others have used observational data where causal identification is difficult to achieve. In this paper, we address both gaps using an experimental design that randomizes the order of the VAA advice using data from a real VAA from a major Danish news site. This allows us to compare the political preferences of real VAA users who received similar advice, but where only some groups had been exposed to the recommendation at the time they expressed their vote intention.