A theoretical model is developed for predicting the relative effectiveness of different electoral systems for reducing government corruption. We consider voting games in which parties with known corruption levels and known positions on a major policy question are competing of legislative seats. We find that approval voting and proportional representation are fully effective, in the sense that all equilibria exclude corrupt parties from legislative seats. Plurality voting is partly effective, in the sense that there always exist some equilibria that exclude corrupt parties. Borda voting is ineffective because, for some political situations, no equilibria can guarantee the exclusion of corrupt parties.