top of page
  • Max Stearns

Gill v. Whitford, Partisan Gerrymandering, and the "Efficiency Gap"

A recent theme of this blog has centered on the comparing how the Constitution's Framers envisioned our political structures might operate, on one side, and how the Constitution actually does operate, on the other. An important case, pending in the Supreme Court, is illustrative. Gill v. Whitford presents the fourth Equal Protection challenge in just over three decades to the prevalent state practice of partisan, or political, gerrymandering. (The case also raises a less probable First Amendment claim.) The tortured doctrinal history of partisan gerrymandering was played out in Davis v. Bandemer (1986), Veith v. Jubelirer (2004), and League of United Latin American Citizens v. Perry (2006).

To somewhat abbreviate too long a story, in the aftermath of a claimed partisan gerrymander substantially benefitting the Republican party in Indiana, a plurality in Davis rejected the lower court standard used to find a Fourteenth Amendment Equal Protection violation. Justice White insisted upon proof that the political process consistently degrade the representation of the subordinate party (the Democrats) in the state as a whole. In a swing state such as Indiana, the standard was nearly impossible to meet. In each of the two subsequent cases, Veith and Perry, Anthony Kennedy was the controlling justice. In Veith, Kennedy rejected White's consistent-degradation standard along with several others that his liberal colleagues offered in their separate dissenting opinions. Even so, both in Veith and in Perry, Kennedy remained undeterred, declining to find the equal protection challenge beyond the judicial competence of the courts, or in technical legal jargon, a "non-justiciable political question," even as he also declined to articulate the governing standard. Instead, Kennedy expressed the hope that a future litigant might someday bring to the Court a standard capable of meaningful measurement and application, perhaps one arising from new or developing technology.

Typically, the Court chooses its own constitutional standards, rather than leaving it to future litigants to identify them on the Court's behalf. Justice Scalia, writing separately in Veith, went so far as to describe Justice Kennedy's approach as "not legally available." But since Kennedy, the Court's median jurist in both Veith and Perry, embraced that position, it wasn't merely available; it was controlling. (This latter point implicates a separate pending Supreme Court case, Hughes v. United States, and the narrowest grounds rule; I filed an Amicus brief in that case, joined by nine other law professors).

A popular measure to assess partisan gerrymanders is "partisan bias." This measure takes a given districting scheme and inquires what the outcome, in terms of seat percentages, would have been had each major party cast an equal number of votes. The divergence from the 50/50 hypothetical baseline to the actual seat percentages reflects the partisan bias distortion. The distortion arises through a series of practices commonly known as "cracking," "packing," and "stacking." Controlling state legislative majorities are able to favor their own by providing as many near simple-majority seats to members of their party as possible, and by densely packing the other party's seats into super-majority districts. The consequence is to "waste" more votes of the subordinate party, beyond the simple majorities needed to elect preferred representatives, and to waste fewer votes of the controlling party, spread across the minority party's densely packed districts. These practices effectively disallow the subordinate party the benefit of spreading its votes across a larger number of districts to tip the balance of more districts in its favor with simple majorities. The partisan bias measures is based on a stylized 50/50 measure, and with a different voting ratio, the measure risks under- or over-stating the effective distortion.

Along come Nicholas O. Stephanopoulos, a law professor at the University of Chicago Law School, and Eric M. McGhee, a political scientist at the Public Policy Institute of California. The theoretical contribution of Stephanopoulos and McGhee has played an important role throughout the Wisconsin districting challenge that gave rise to the Whitford case. Stephanopoulis and McGhee describe their methodology in considerable detail in an article titled Partisan Gerrymandering and the Efficiency Gap, published in the University of Chicago Law Review, and available here. The authors' proposed "Efficiency Gap ("EG")" measures "the difference between the parties' respective wasted votes, divided by the total number of votes cast in the election." Id. at 851. (The actual formula: "Efficiency Gap=Seat Margin - (2 x Vote Margin).") Id. at 853. The EG approximates the difference between the percentage of wasted votes for the advantaged versus disadvantaged party as reflected in the difference in gained seats, while acknowledging what the authors refer to as the "winner's bonus." This bonus reflects winner-take-all districted elections; a party needs just over half the votes to gain the next marginal seat.

The EG has quirks, largely attributable to geographical districting. Take, for example, a five-district state, with three Republican and two Democratic districts, and with a 60/40 ratio of Republican to Democratic voters. EG measures a 10 percent distortion (60-50)-(2x(60-50))=-10 even though the result is precisely proportional. To remove the distortion, the Republicans would have to hold 70% of the seats, even though that would undermine proportionality, giving 70% Republican seats for 60% votes (as would be possible by doubling the districts to ten).

If the goal is to capture representational distortion based on per party votes cast, there are more intuitive approaches. We could, for example, directly compare the percentage of seats per party to the percentage of votes per party. Or, we could ask: assuming proportional party representation, as in parliamentary list voting, and given the per party votes cast, what would the outcome have been? The EG seems is a kind of analytical hybrid that seeks to capture something approximating these intuitions about voter distortion, while eschewing a direct measure of proportionality. Most likely this is owing to specific language in governing Supreme Court case law. In City of Mobile v. Bolden (1980), which involved an equal protection challenge to an at large municipal voting scheme in Mobile, Alabama, operating to the detriment of African American voters, the Supreme Court stated: "The Equal Protection Clause of the Fourteenth Amendment does not require proportional representation as an imperative of political organization." Perhaps to avoid claiming otherwise, Stephanopoulis and McGhee have advanced a less intuitive distortion proxy. Even so, Paul Smith, counsel for Appellee, had to defend against a claim of proportional representation at oral argument:

"CHIEF JUSTICE ROBERTS: And if you need a convenient label for that approach, you can call it proportional representation, which has never been accepted as a political principle in the history of this country.

"MR. SMITH: Your Honor, we are not arguing for proportional representation. We are arguing for partisan symmetry, a map which within rough bounds at least treats the two parties relatively equal in terms of their ability to translate votes into seats. That's -

"CHIEF JUSTICE ROBERTS: That sounds exactly like proportional representation to me."

See here (oral argument transcript from October 3, 2017).

If the Court accepts EG, it will further need to identify the percentage distortion that constitutes a constitutional violation. Assuming the Court does so, it would have a generally consistent, if imperfect, measure by which to compare the fairness of electoral mapping that would presumably motivate state legislatures to avoid violations by districting within the permissible error range.

Although I have some quibbles with the EG methodology, it's a plausible measure within the highly imperfect world of districted voting. Beyond the measurement issues, there are other concerns, which roughly fall into three groupings: (1) legitimacy, (2) constitutional (and statutory) premises, and (3) endogeneity.

The first concern is legitimacy. The conservatives on the Court have long wished to ditch this entire area of law, finding it a non-justiciable political question. These jurists are apt to believe that Justice Kennedy's quest for an administrable standard has led litigants on something of a fool's errand. As expressed in Baker v. Carr (1962), in addition to separation of powers, the political question doctrine rests on the concern for a "lack of judicially discoverable and manageable standards." Both features matter. After Baker held that a decades-old state apportionment scheme that yielded woefully unequally populated districts created the basis for an equal protection violation, Reynolds v. Sims (1964) announced one-person, one-vote rule as the governing test. That standard was a pure doctrinal invention, but it was also eminently administrable. Both features might also characterize EG, a made-up, yet administrable, standard.

The second concern goes to constitutional and statutory premises. In a thoughtful National Review article, Kyle Sammin expressed the concern that a problem with EG is that within winner-take-all districted elections, voters do not cast their ballots based on the party's state's aggregate House delegation; instead, they vote based on the candidates running in their districts. Sammin is right, and this gets at the heart of the problem. We are stuck with districting based on a combination of federal statutes tracing to 1842, and early historical assumptions dating to the Framing period that don't translate well into the twenty-first century. Within parliamentary systems, voters vote on delegations as a whole. Per-party votes are aggregated and the percentage each party receives dictates how far down the party list each delegation goes. In that system, there are no wasted votes, and there is no distortion; party votes align, as much as feasible, to the percentage of seats obtained. The earliest practice governing House of Representative elections were also at large, but winner take all for the state as a whole, thereby allowing the dominant party to fence out the minority party, thus inviting the 1842 statutory reform. See here (historical review by UMBC Professor Thomas F. Schaller on UVA Professor Larry Sabado's "Sabado's Crystal Ball" website).

Chief Justice Roberts is right to assert that under the United States Constitution, voting is not based on proportional representation; it is based on districts, which are based on geography. During much of our early history, geography was a paramount consideration. Although the Constitution's Framers did not mandate districted elections, they understood geography to affect our lives in ways myriad ways: influencing our professions, our religious practices, our family bonds, our sense of community, and fundamental aspects of how we lived and who we are are. In modern times, this is increasingly fictitious. True, urban voters might tend to be more liberal, and coastal elites tend to hold divergent views from those dominant within in the great red center. Even so, liberals and conservatives are routinely interspersed as minorities within venues dominated by the other side. Geographical districting involves pulling together sufficient numbers of like-minded people, either to create a packed district, or to forge a simple majority district, with a process that too often resembles the childhood game of Twister. (The game is all the more complex when we add the layer of race, including the effect of Shelby County v. Holder (2015) on state obligations under the Voting Rights Act). With actual party-list voting, much of this would be beside the point; there would be no need for districting, and thus no need for contiguity. Twister would translate to the less entertaining exercise of cutting out and piling the dots, with no more need for physical contortions to connect them. The EG measure is trying to get at this intuition without abandoning the pretense of geographical districting.

EG's third problem is endogeneity. The measure assumes our present, troubled, state of politics, namely a dysfunctional party duopoly. The test only performs well in a two-party world. But imagine a state with 45 percent Democrats, 40 percent Republicans, and 15 percent other, perhaps Green Party, Libertarian, or something new. And further imagine that the latter group is joined by internet chat groups, FB postings, and the like, but not geography; they are literally all over the map! No credible district can conceivably draw them in together. They are EG losers each and every time, whereas under list voting, whatever their voting percentage would translate to percentage representation.

To truly avoid distortions, we would need something far more radical than EG. A more radical proportionality measure might break the duopoly, inviting credible third parties. Chief Justice Roberts is, of course, right that equal protection doesn't demand proportionality. And it is likely that subsequent statutes to the original 1842 Act, the most recent one enacted in 1967, would prohibit returning to winner-take-all at large House elections. The really interesting question might be whether equal protection, and the federal statutes, might allow a kind of hybrid. Imagine, for example, that some states chose to experiment with a variation on list voting. Imagine that each party's candidates run on a list, but those chosen from each list are then assigned to previously dawn geographical districts, with proportionality disallowing the dominant party to sweep all seats. (This avoids the problem motivating the 1842 Act, although it is arguably in tension with the wording of the 1967 Act.) Within this scheme, third party candidates could have a shot. And if there were enough successful third party candidates, it is even possible that neither major party could claim immediate control of the House. The President would still be elected directly (albeit with the Electoral College filter), and so two parties would continue to dominate our national political scene. Even so, the President might find himself forced, as the price of a securing a favorable House coalition, conceding to one or more third parties on some major issue. (A Republican President might, for example, be forced to relax his stance on gun control or tax policy; a Democratic President might be forced to relax her stance on any number of favored regulatory commitments. Either party's President might have to make concessions on who might be appointed to the Supreme Court). This could really shake things up! Although EG might seem radical, if the goal is to make our system more representative, and less broken, is it just possible that the measure is not quite radical enough?

A final note: Justice Kennedy will not remain on the Court forever. The liberals on the Court-Justices Ginsburg, Breyer, Kagan, Sotomayor-certainly recognize that if Justice Kennedy were to sign onto EG, finding this his long-awaited test, it might be their final opportunity to create a precedent that allows challenges to partisan gerrymanders to proceed (that is, before Trump replaces Kennedy or one of the liberal Justices, with the effect of forging a majority capable of finding the whole area of law non-justiciable). The EG measure is imperfect. And it's a bit radical. But just maybe it's small step in the right direction.

(Special thanks to Alan Miller for helpful technical clarifications and illustrations. Alan is coauthor, with Christopher P. Chambers, Alan D. Miller, and Joel Sobel of an article, found here, which thoughtfully responds to the Stephanopoulos and McGhee law review article proposing the EG measure. Thanks also to Mark Graber for some help on the early history of electoral districting.)

I welcome your comments.

81 views0 comments
bottom of page