In this paper, we present a method based on the CMP data to identify parties' ideological signals. Although our approach uses the same data as others, our method is different in that we distinguish the methods for the directional and proximity models which have long been relevant for the theoretical literature. We apply our method for all German Bundestag parties between 1949 and 2005. One problem with policy preferences is that various – and sometimes only slightly deviating – approaches lead to rather different results. Thus, we test our method in terms of its robustness using the Jackknife method. We also show confidence intervals around policy positions.