Selecting Reasonably Good Redistricting Plans in a Criteria-Agnostic Manner Using the Efficient Frontier

Keywords: redistricting, data envelopment analysis, public participation
Session Type: Virtual Paper Abstract
Day: Saturday
Session Start / End Time: 2/26/2022 02:00 PM (Eastern Time (US & Canada)) - 2/26/2022 03:20 PM (Eastern Time (US & Canada))
Room: Virtual 30

Abstract

As part of a widespread frustration with partisan gerrymandering, many states are considering redistricting reforms that include citizen participation in proposing and evaluating redistricting plans. Evaluating plans will overwhelm redistricting commissions if citizen participation becomes popular or if political actors flood the commission submissions with (potentially bad faith) submissions. Low quality plans will need to be discarded quickly. But what is a low quality plan? Many redistricting criteria exist, but their merit is relative (only contiguity and population equality have bright line rules), and most states do not prioritize among competing redistricting criteria, such as political subdivision splits, compactness, communities of interest, core preservation, and others. Furthermore, improving on one criterion may necessarily compromise another one, such as the tension between compactness and subdivision splits. Analytical methods are needed to reduce the amount of time devoted to evaluating low-quality plans in a criteria-agnostic manner, that is, in a manner that does not impose a prioritization among criteria not found in statute or case law. Using a modified version of data envelopment analysis (DEA), we present a nonpartisan approach that can substantially reduce the number of maps that need to be considered by a redistricting body. An efficient frontier is constructed from an actual set of proposed plans, and a single score represents how close a plan is to the frontier. We report on the application of DEA to a set of 11,000 randomly generated plans for Pennsylvania, and are seeking opportunities for a real-world implementation prior to the conference.