VAAs aid people in finding out which parties reflect best their political preferences. The overlap between single user and the party is calculated by a simple algorithm. Commonly these are based on either the Manhattan or Euclidean distance. In this paper we take the matching algorithms of three popular VAAs in Europe - Wahl-O-Mat, Smartvote, and EU-Profiler - and examine and compare the impact of a specific matching algorithm on eventual matches produced. Additionally we test a number alternative matching algorithms. Preliminary findings suggest that linear distance metrics perform better than non-linear ones, although large opinion heterogeneity between voters and parties may produce marked differences between the algorithms used. We also find universally relevant conditions that should be avoided when calculating the matches. Our findings have implications toward the way in which voting aids are interpreted and they alert VAA developers to attend problematic domains.
