Polarization of What? A Model of Elections with Endogenous Valence

We connect three characteristics of political candidates: their preferences, their platforms, and their valence. To do so, we define and study three types of elite polarization: preference polarization, valence polarization, and platform polarization. In our model, policy is represented as a position in a unidimensional space; candidates are policy motivated; valence is represented as a dimension orthogonal to policy; and candidates can increase their valence by paying a cost. We find that candidates will in general not converge to one another in the policy dimension or in the valence dimension. A candidate’s preferences are positively, but not perfectly, correlated with her platform. Strikingly, under some circumstances, candidates’ platforms will diverge when their preferences converge. And there is an unfortunate trade-off regarding candidates’ attributes: they will display high valence or low polarization, but not both. One of our empirical predictions is a positive correlation between campaign spending and platform polarization.