The main purpose of representative democracy is to enact the will of the people indirectly through political parties. Voters choose their representatives based on policy pledges made during the election campaign. However, once elected, voters have very limited influence over politicians’ actions. The question of whether democratic elections ensure good representation therefore lies at the very heart of representative theories. Mandate theory states that for successful representation the policy outcome generated during a legislative period should match the promises made by political parties before the election. Previous research designs on pledge-fulfillment were often limited to single-party governments such as the USA or Great Britain, studied only one policy dimension or based their analysis on a single election. Additionally, these studies partly relied on crude measurements of core independent variables such as policy distance or saliency. The present study attempts at addressing those shortcomings and brings in a new case. It examines the program-to-policy linkage in Austria, where a multiparty-system and proportional representation make coalition governments the rule rather than the exception. To ensure comparability with the existing literature, we follow Terry Royed (1996) who defines pledges as objectively testable commitments by political parties. The analysis is based on a dataset generated in three steps: First, we conduct a quantitative text analysis to identify concrete pledges in party manifestos of governing parties. Second, we compare these promises to the coalition bargaining outcome as written down in the coalition agreement. In the last step we check which pledges are fulfilled at the end of the legislative period. We specify multivariate models to test at the pledge-level a number of hypotheses pertaining to consensual pledges, the status quo, the policy distance between parties, opposition support, the saliency of single pledges, and their enactment in the coalition agreement. We believe that especially our improvements concerning the measurement of the independent variables will highlight empirical relationships that former models might have missed.
