Rationalizable Voting

We derive necessary and sufficient conditions in order for a finite number of binary voting choices to be consistent with the hypothesis that voters have preferences that admit concave utility representations. When the location of the voting alternatives is known, we apply these conditions in order to derive simple, nontrivial testable restrictions on the location of voters’ ideal points, and in order to predict individual voting behavior. If, on the other hand, the location of voting alternatives is unrestricted then voting decisions impose no testable restrictions on the joint location of voter ideal points, even if the space of alternatives is one dimensional. Furthermore, two dimensions are always sufficient to represent or fold the voting records of any number of voters while endowing all these voters with strictly concave preferences and arbitrary ideal points. The analysis readily generalizes to choice situations over any finite sets of alternatives.