Existing spatial models of multiparty spatial competition have difficulty accounting for two related features of party behavior: that parties typically propose divergent policy programs, and that these programs are similar to, but more extreme than, the policies of their supporters. We expand and elaborate Iversen's "representational policy leadership" model, which combines insights from the proximity and directional theories of voting, in an effort to explain this phenomenon. By developing a spatial model of party competition that incorporates voters' nonpolicy motivations (i.e., sociodemographic characteristics, retrospective economic evaluations, partisanship) and a mixed directional-proximity policy utility metric, we significantly improve our ability to explain parties' policy strategies, compared to alternative models that ignore either nonpolicy motivations or the directional component of voters' policy utilities. We illustrate and test our arguments with conditional legit analysis of voter choice for 1989 Norwegian election data. We show that a model that includes nonpolicy motivations in addition to proximity and directional motivations accurately accounts for the actual locations of parties but that the proximity and directional components alone are insufficient. Voters' nonpolicy motivations are crucial for explaining parties' policy strategies and also for understanding the connection between parties' policy positions and the positions of their supporters.
