Measures of Partisan Bias for Legislating Fair Elections

TL;DR

This paper reviews and introduces measures of partisan bias in single-member districts, applying them to real and modeled elections to evaluate fairness and inform election law reform.

Contribution

It introduces a new geometric measure of partisan bias and compares it with existing measures across various election data and models.

Findings

Different bias measures generally agree for most cases.

Partisan bias measures are useful for election law reform discussions.

The new geometric bias measure offers a novel perspective.

Abstract

Several measures of partisan bias are reviewed for single member districts with two dominant parties. These include variants of the simple bias that considers only deviation of seats from 50% at statewide 50% vote. Also included are equalization of losing votes and equalization of wasted votes, both of which apply directly when the statewide vote is not 50% and which require, not just partisan symmetry, but specific forms of the seats-votes curve. A new measure of bias is introduced, based on the geometric area between the seats-vote curve and the symmetrically inverted seats-votes curve. These measures are applied to recent Pennsylvania congressional elections and to abstract models of the seats-votes curves. The numerical values obtained from the various measures of bias are compared and contrasted. Each bias measure has merits for different seats-votes curves and for different elections, but all essentially agree for most cases when applied to measure only partisan bias, not conflated with competitiveness. This supports the inclusion of partisan fairness as a fundamental element for election law reform, and some options are discussed.