Voting Advice Applications Effects on Voting Behaviour: Experimental evidence from Luxembourg

Luxembourg has an open-list electoral system which allows voters to express a list vote, or to cast preferential votes for candidates of either a single party or several different parties. With 12 parties competing in four constituencies (where respectively 7, 9, 21 and 23 MPs are elected), this institutional setting is ideal for testing the effect of a VAA providing voters with information on both parties and individual candidates' policy positions (and resulting advices/recommendations for parties and candidates). In order to address the well-known self-selection problems of most of the existing literature on VAA effects, we followed the recent approach of embedding a randomized invitation to use a VAA within an online survey. In this case, the VAA was a real, publicly available and widely used tool (smartwielen), and our sample of respondents was nationally representative of Luxembourg voters. In addition, we were able to track the answers provided, and the advice received by members of the treatment group. Finally, the post-electoral wave of the panel survey allowed us to measure VAA effects on actual voting behaviour, rather than short-term effects of VAA use on voting intentions. With this setup we contribute to an emerging literature using panel surveys, embedded experiments and the tracking of users' behaviour and the personal advice they received on the VAA, to study whether VAA use actually affects voter choices when electoral systems allow for complex voting behaviour.