From ideology to bargaining space: estimating the overlaps of party positions

This article raises some of the problems with common practices in measuring party positions, namely dimension reduction and ideal-point estimation. While serving as a guidance for political communication between parties, voters, and experts, ideal party positions do not convey much information about the dynamics of inter-party and intra-party negotiations, which are of utmost importance to understanding party competition. On the other hand, when reducing multidimensional party positions to a few latent dimensions, much of the between-issue variance is lost in the process, making predicting policy outcomes less accurate. As a replacement, I introduce bargaining space, a new approach to scaling positions that emphasizes the fluidity, multidimensionality, and interdependence of party ideologies. Applying BERT, the state-of-the-art language model in natural language processing, I construct bargaining space and bargaining ranges from parliamentary speeches delivered in the German Bundestag from 1991 – 2017. Using overlaps of bargaining ranges on various policy issues to predict the number of instances each pair of parties vote together, I show that this alternative approach has a lot of potentials, especially for future studies wishing to study the intricate dynamics of party competition or use party positions for explanatory purposes.