In 2012, the French “Boussole présidentielle” was the first ever Voting Advice Application for a two round presidential election. Thanks to multiple media partners, it drew a large audience: in the four weeks before the first round, more than 550,000 useful questionnaires (after filters) gathered information about respondents’ background, positions on 30 issues, evaluation of candidates and propensities to vote for each of them. Immediately after the VAA result displayed the candidates’ positions and their own location within the political space, roughly 20% filled out an additional questionnaire about political attitudes and behaviour, including their likely vote. More than 3% volunteered their email-address to be later contacted for further research purposes. Four days before the second round, 22,000 of them were sent an online “follow up” survey recording if and how they voted in the first round and exploring, inter alia, their reactions to the VAA. Several potential effects of the tool were explored: information seeking, an increase of inter-personal communication about the election and campaign activities and the impact on their vote. For the latter, users responded to a subjective scale ranging from no effect to an actual vote change, with intermediate options (considering an alternative candidate, pondering about one’s choice). This measure can be compared to the objective trajectory of intended/reported vote. An experiment was also included to assess how well users remembered their position according to the VAA, since this memorization might be a precondition of any tool impact. Nearly 6,300 users answered this follow-up questionnaire and this paper analyses their profile and electoral behaviour. We test 4 hypotheses: VAA effects are not homogeneous but conditional 1) on age and level of education 2) on prior politicization and partisanship 3) on the congruence between intended vote and VAA results 4) on being (un)decided before using the tool.
