Finding Pareto Dominant Redistricting Plans Using DEA: A Comparison With Ranking by Human Judges

Abstract

Changing technology now allows for increased citizen participation in the redistricting process. Bodies responsible for redistricting, including legislatures, commissions, and courts, will need new analytical methods to evaluate large numbers of maps. We propose the use of data envelopment analysis (DEA) as a decision-support tool that will help evaluators identify high quality maps proposed by citizens and citizen groups. DEA does not rely upon an a priori ranking of redistricting criteria, but scores maps based on the values that emerge from the pool of submitted maps. We apply DEA to 403 maps submitted to Draw the Lines (DTL), a series of redistricting contests held in Pennsylvania prior to the 2021 redistricting round. We compare DEA scores with contest outcomes, in particular comparing high scoring maps and semifinalists chosen by DTL judges. We explore the incomplete overlap between the maps that performed well in the contest and those that scored highly on DEA. We argue that DEA can aid in reducing the number of maps to be considered by human judges, for any process in which a large number of maps, whether created by humans or computers, must be evaluated for the selection of a final redistricting plan. These include public contests, public comment to legislatures and redistricting commissions, and, increasingly, cases in which the courts have had to select redistricting plans after the failure of the political process.