Estimating policy positions of political actors across countries and time

Numerous empirical studies in comparative and international politics use estimates of policy positions of parties and political elites to analyse how preferences shape outcomes of political decision-making processes. In the past decade, data availability and methodological advances have fostered the shift from a static, cross-sectional view on positions and ideologies to a dynamic, longitudinal perspective on changing preferences and political change. A large part of the ongoing debate on what kind of data can be used for longitudinal analysis revolves around the pros and cons of using fully computerized content analysis of political texts as compared to the yardstick dataset in comparative government, the 'Comparative Manifesto Project' (CMP) data of hand-coded party manifestos. While a large part of this discussion is methodological, we know little how the different methods compare empirically. This is what we do in this paper. We compare estimates of parties' policy positions from CMP with positions derived from Wordscores for 13 Western European countries in the time period between 1980 and 2010. Our analysis shows that by and large, the CMP and the Wordscores approach produce similar estimates of parties' positions on a general left-right dimension. Yet, the degree of congruence differs considerably over countries and also varies with the type of party and party manifesto the estimate is based on. We examine outliers from the overall pattern and discuss possible reasons for them.