Social and Behavioral Sciences Commons^™

To analyze the performance of the U.S. presidential primary system, we develop a model with horizontally and vertically differentiated candidates. Voters are uncertain about candidates' valences, and use results in earlier elections to update their beliefs. The temporal organization of primaries affects both voter learning and vote splitting, i.e., several candidates in the same policy position competing for the same voter pool. Sequential voting minimizes vote-splitting in late districts, but voters may coordinate on the wrong candidate. We structurally estimate the model using the 2008 Democratic presidential primaries. Using the parameter estimates, we conduct policy experiments such as replacing the …