Do Voting Advice Applications Shape Voting Behavior? Evidence from a Pre-Registered Experiment in Belgium During the Election Campaign

Vote Advice Applications (VAAs) are digital tools designed to assist voters in aligning their preferences with parties’ policy positions. While previous research has documented the role of VAAs in fostering informed voting and party position or issue knowledge (Munzert et al., 2021; Schultze, 2014), their impact on actual voting behavior has been less often explored during a real electoral setting: Do voters follow the advice they receive from a VAA? Next to studies focusing on turnout (Gemenis & Rosema, 2014) and extending work on effects on voting behavior (Alvarez et al., 2014; Kamoen et al., 2018; Munzert et al., 2018). In this study, we examine in a systematical way whether exposure to VAAs influences the propensity to vote for political parties, particularly among young and first-time voters. In Belgium, elections were held on June 9, 2024. Youngsters who had reached the age of 16 were given the right to vote for the European Parliament. Using data from a randomized pre-registered experiment embedded in a 5-Wave panel survey tracking youngsters across the electoral campaign (n = 925) conducted in Flanders, we analyze the electoral effects of VAA exposure. The VAAs we study, namely “De Stemtest” (generic VAA) and “De Jongerenstemtest” (tailored to youngsters) presents participants an end screen with a match (i.e., a score or percentage indicating how well a party fits with voters’ policy positions, based on an average of over 35 statements on various policy issues). Participants were either exposed during the campaign to a VAA or not, followed by an assessment of their vote intentions for various parties immediately after taking the VAA (in Wave 2) while the effect is tracked across the intensive campaign phase (in Wave 3) and in the post-electoral Wave (Wave 4), where we assess the actual vote. Preliminary findings verify that there is a significant effect of VAA exposure on voting behavior in line with the advice, although some disparities are found across parties and voting behavior measurements. Overall, we find the strongest effects immediately after exposure on the propensity to vote for a party and the actual reported vote in the post-electoral wave, even when controlling for pre-treatment vote intentions. Still, there is evidence that the effect seems to wear off after time; moreover, we find differences across parties, with some evidence that there are less strong effects on the vote for the radical right. Interestingly, we also find evidence for negative partisanship where participants do not vote for parties that are ranked lowest. It also seems that the voting decisions are also made in relation to the poll standing of parties. These effects hold while controlling for prior predispositions (i.e., left-right orientation, pre-treatment vote intentions) and we find the largest effects on first-time voters. Finally, the severity of a ‘representational deficit’ plays a role, especially participants with a strong overlap with parties (i.e., when this deficit is low) are strongly affected.