Matching voters with political parties and candidates: an empirical test of four algorithms

Voting Advice Applications (VAAs) have enjoyed a growing popularity across Europe in recent years and, increasingly, in other parts of the globe too. In response a growing literature is emerging around the promises and potential perils associated with VAAs. This paper contributes to these debates by addressing a core methodological aspect of VAA design: how voters' policy preferences are aggregated to produce measures of concordance with parties/candidates. To this end, the paper analyses the performance of four VAA models that are based on competing algorithms. The findings, which draw from the analysis of VAA–generated data from four experiments, raise a number of concerns for the future design of VAA systems.