In this article, we test rival theories of party competition. Previous research predicting party policy positions is limited to a small (and arguably biased) sample of countries. Using new data from the 2008 Austrian National Election Study, we test the proximity and directional model of voting, Grofman’s discounting model and Kedar’s compensational model of voting. We find that Grofman’s discounting model performs best predicting party policy positions. Yet, not all party policy positions are predicted equally well. The most obvious misrepresentation is that the rightist FPÖ is placed close to the centre of the policy space. This centripetal bias, which has also been found in earlier work, may result from simplifying assumptions of the algorithm used to derive Nash equilibria of party policy positions that need to be challenged. Our findings emphasize the necessity to test models of party competition using new data and to motivate further research explaining party position-taking.