Discrete geometry for electoral geography

Abstract

"Compactness," or the use of shape as a proxy for fairness, has been a long-running theme in the scrutiny of electoral districts; badly-shaped districts are often flagged as examples of the abuse of power known as gerrymandering. The most popular compactness metrics in the redistricting literature belong to a class of scores that we call contour-based, making heavy use of area and perimeter. This entire class of district scores has some common drawbacks, outlined here. We make the case for discrete shape scores and offer two promising ideas: a cut score and a spanning tree score. We use recent United States redistricting history as a source of examples. No shape metric can work alone as a seal of fairness, but we argue that discrete metrics are better aligned both with the grounding of the redistricting problem in geography and with the computational tools that have recently gained significant traction in the courtroom.