An optimization based heuristic for political districting

Redistricting, the redrawing of congressional district boundaries within the states, may occur every 10 years on the basis of the population census. Many redistricting plans are designed with partisan politics in mind, resulting in disputes and forcing judges to intervene. We address this problem from a nonpolitical viewpoint and present an optimization based heuristic incorporating universally agreed upon characteristics. We model the problem as a constrained graph partitioning problem and develop a specialized branch-and-price based solution methodology. We demonstrate the feasibility of our methodology by showing how to satisfy the one-person, one-vote principle with compact and contiguous districts for the state of South Carolina.