Chief Justice Roberts’s decision in Rucho v. Common Cause expresses a paradox: (a) gerrymandering a map for political advantage goes against the Constitution, yet (b) the five-vote majority couldn’t find a way to ascertain when it has occurred. This stance has the consequence of disregarding the careful work of five lower courts in Wisconsin, North Carolina, Maryland, Ohio, and Michigan.
Resolving this conflict is not simply of academic significance. State courts may yet act to curb partisan gerrymandering under their laws and constitutions. Surely some of the same questions will arise.
Today I want to answer Roberts’s concern about “proportional representation,” which was also expressed by Justice Kavanaugh in oral arguments. I think this particular problem arises from poor communication between scientists and lawyers.
Justice Roberts’s disproportionate error
Roberts’s decision cited the problem of “proportional representation” as an impediment to establishing a standard of fairness. He asserts that claims of partisan gerrymandering rely on an assumption about what would be appropriate levels of representation. Specifically, he says that the number of seats is not guaranteed for a given number of votes cast statewide – a quota of sorts.
Smart lawyers may object: nobody is arguing for proportional representation. But I don’t think Roberts and Kavanaugh meant literal proportionality, i.e. 40% of the vote should translate to 40% of the seats. Roberts objects to any diagnostic tool that uses the number of seats won by either party. This recalls Justice O’Connor’s warning that a standard might “evolve towards some loose form of proportionality.”
I suggest that we can make a distinction between (a) those ways of determining fairness that count wins and losses, and (b) those that don’t. This may fly in the face of intuition. After all, partisans gerrymander to win as many seats as possible. But suspend that for a moment.
To be sure, some tests use wins and losses for their calculation. The efficiency gap comes to mind, as well as partisan bias. In the Harvard Law Review Blog and Election Law Journal I argue that these tests measure inequity of outcome, which answers the question “did the parties get a fair share of seats?” In other words, a loose form of proportionality.
But another concept that can be tested: inequity of opportunity, i.e. did both parties have similar opportunity to win seats? This concept echoes existing concepts in discrimination law and fits with previous doctrine on racial gerrymandering – a must if we are ever to have a coherent doctrine of fair districting.
Defining a racial gerrymander is a mature subject. Was a group packed? Then it would have lopsidedly large wins. Was a group cracked between districts? Then its opponents would have carefully arranged, narrow wins. A court can inspect individual districts for evidence of packing and cracking by race. It can also do so for party.
If a court is willing to look at an aggregated statewide pattern of legislative-district-level election returns, here are examples of tests that probe for inequity of opportunity:
Lopsided wins. in a closely-divided state like North Carolina, are one side’s wins more lopsided than the other’s? The t-test, perhaps the most widely used statistical test in all the sciences, does this.
Uniform wins. In a party-dominated state like Maryland, were the wins engineered to be very uniform, thus protecting the majority party? The chi-square test does this.
Map-simulation methods. Wesley Pegden, Moon Duchin, Jonathan Mattingly, and others have done large-scale map simulation to detect if a plan produces an outlier in terms of the number of seats. Those methods could just as easily detect an outlier for some other quantity, like the lopsidedness of one side’s wins. Map-simulation methods have the advantage of using a state’s specific geography and thus adapting to peculiar circumstances such as the clustering of one side’s voters.
An opportunity-and-outcome framework can be used in state courts. In a forthcoming article, Rick Ober, Ben Williams, and I review how all fifty state constitutions contain rights and protections which can be used to bring a partisan gerrymandering claim. These include analogues of the First and Fourteenth Amendments; guarantees of pure, free, and fair elections; and redistricting-specific guarantees such as geographic compactness. Because each of these protections involves either individualized or associational harms, the Roberts and Kagan opinions in Gill v. Whitford offer state courts guidance for how to apply their own constitutional provisions to a partisan gerrymandering claim. And even Roberts’s opinion in Rucho invites the states to act.
Of course, some mental adjustments are necessary. Legal scholarship on election law has focused on federal courts. I think it’s time to focus that intellectual firepower on state courts – and to let go of what scientists mean when they say “proportional.”