A data-driven approach to scheduling the US presidential primary elections

In this article we formulate and solve a model for choosing the grouping and ordering of state primary elections for a United States presidential election cycle. Our work fills a gap in the existing literature by defining a reproducible procedure for deriving an entire schedule of U.S. primary elections in a way that meets quantifiable goals of efficiency, equity, and effectiveness. Our model considers each state's attributes and gives higher priority to states whose voting and demographic profiles more closely reflect those of the county as a whole. It also clusters states into primary dates in a way that spreads the electorate relatively evenly over the primary season. Finally, the clustering process takes proximity into account, so that political campaigns can focus more effectively on a particular region without incurring excessive travel time. Since these criteria may conflict, the preferred ordering depends on the tradeoffs made among the goals. We highlight two solutions in which states with voting and demographic attributes close to those of the national profile are scheduled early on, while the geographic grouping and spread of voters over the schedule are significantly better than those of the status quo.