We investigate the performance of a voting procedure in which voters cast multiple votes, with some votes positive and some negative. The procedure is a variant of the “D21” method proposed by Janecek (2016). The context of the investigation is the 2018 presidential election in the Czech Republic. Evidence is provided through a survey of secondary school students who were asked to cast ballots in a conventional single vote procedure and through an alternative multi-vote procedure. Using results from the survey, we estimate a spatial voting model to uncover ideological positions of candidates and respondents. Results suggest that the multi-vote procedure may limit the success of extreme candidates in elections. We also investigate how survey respondents alter their voting behavior in response to an informational intervention.