EU legislative analysis has been enriched by insightful controversies over the interpretation of the policy process. This debate has concentrated on the interpretation of the process by focusing on the identification of the agenda setter and the relevance of voting weights, but little attention has been paid to the accurate specification of the second component of spatial analysis, the preferences of the actors involved. Although a misspecification can seriously distort the predictions of spatial theory, empirical applications often tend to reduce the number of dimensions, exclude actors' saliencies and assume continuous policy issues. Using computer simulation we show that spatial models are more robust to a misinterpretation of the policy process than to a misspecification of actors' preferences, and that their institutional elements are less decisive for the models' outcome predictions. Our empirical analysis confirms these results and provides detailed insights into the impact of the institutional and the preference component of spatial theory. We conclude that scholars should pay more attention to the accurate specification of the preference component of the models to improve our understanding of legislative decision making in the EU.
