The lower equality standard reflected in current Constitutional interpretations is well understood among both election law experts and political scientists. Justice Breyer has noted the tension between single-seat districts and proportional representation. The most prominent metric of partisan advantage in political science, Gelman and King’s partisan asymmetry statistic, is a measure of vote dilution that accounts for disproportionalities inherent in our single-seat, winner-take-all elections. Asymmetry measures the difference in seat shares that each party’s voters receive in a plan for the same statewide vote share, say 50%. A second component of the Gelman and King model, responsiveness, captures that inherent disproportionality that emerges when voters shift support from one party to another, typically resulting in a “winner’s bonus.” A system that is less responsive to shifts in support is evidence of a more durable gerrymander.
And it is within this asymmetry framework that we find a test that parallels the three-prong approach established for deciding racial gerrymandering cases in Thornburg v Gingles. It is described in the Amicus Brief submitted in the North Carolina case (in favor of neither party) by professors Bernard Grofman and Keith Gaddie. Their proposed test requires that plaintiffs first demonstrate that the opposition party has been deprived of partisan advantage in at least one district in the enacted plan in the same manner as Gingles: targeted voters must be a large and compact enough group to create a majority district without diluting their advantage in other districts. Second, the opposition voters must exhibit polarized partisan voting. Voters who regularly shift support between parties would not provide an advantage to either.
Third, plaintiffs must demonstrate vote dilution at the district level, through either a district where the opposition party regularly loses, or in a majority district where voters could be allocated more efficiently across districts. Finally, because changing parties is an option whereas changing race is not, the responsiveness statistic can be used to demonstrate the durability of a partisan gerrymander.
Justice Kagan’s expressed concern that the Maryland plan “flips the composition of the district from 47 percent Republicans and 36 percent Democrats to, instead, 45 percent Democrats and 34 percent Republicans, effectively ensuring that Republicans will never win this seat again…” and is excessive can be answered by including durability in the test.
Using responsiveness as a means of estimating durability is especially appealing as it is distinct from proportional outcome expectations. In fact, this final element of the test could require that very disproportional plans be upheld. Consider an extreme hypothetical where the difference in party affiliation in Maryland’s eight districts is only one person in each district. That is, each district is nearly perfectly split 50/50. In such a case, one person changing their vote changes control of the district, and if each pivotal voter in the district votes Democratic, Democrats win all eight seats, which is the least proportional plan. However, responsiveness is also maximized if it only takes eight people to flip every district, demonstrating that the plan is not a durable gerrymander.