A Unified Model of Spatial Voting∗

Experimental research shows that while most voters have some form of spatial preferences, individuals differ in the type of spatial preferences they have: many voters prefer candidates closer to themselves in a policy space (proximity voting), others prefer candidates that are simply on the same side of an issue as themselves (directional voting), and still others prefer those who will move policy closest to them (discounted proximity voting). No existing theory explains this variation. I propose a theory based on the idea that that people categorize candidates and have preferences defined over categories. As a voter gains political experience, she makes finer distinctions between candidates, and the set of categories grows. In this way, voters move from either-or conceptions of politics that approximate directional preferences toward more detailed conceptions consistent with proximity preferences, with some cases approximating discounted proximity voting as well. I show that the categorization model accurately predicts the observed frequencies of different voting types as well as observed comparative statics. I also show that the comparative statics results explain observed differences in the distribution of voting types across different policy areas. ∗I thank John Bullock, Jonathan Bendor, Matt Levendusky, and Ken Shotts for helpful comments.