Introduction
The law of redistricting is based on the potential for tradeoffs among line-drawing criteria. Better performance along one dimension—compactness, partisan fairness, minority representation, and so on—is often thought to compel worse performance across one or more other parameters. Consider partisan gerrymandering, which the Supreme Court deemed nonjusticiable (under the federal Constitution) in the 2019 case of Rucho v. Common Cause.
One of the Court’s rationales was that, according to some litigants, disregard for traditional (or form‑related) criteria should be the standard for unlawful gerrymandering.
“If compliance with traditional districting criteria is the fairness touchstone,” however, “how should mapdrawers prioritize competing criteria?”
Conflicts among traditional criteria supposedly make it impossible for courts to determine whether district plans adequately satisfy these requirements.
Analogously, courts that are open to partisan gerrymandering claims (like federal courts before Rucho and many state courts to this day) frequently ask “whether a plan’s partisan effect is justifiable”—that is, “whether it can be explained by the legitimate state prerogatives and neutral factors that are implicated in the districting process.”
This justification stage is, at its core, an evaluation of tradeoffs. The issue is whether a challenged map’s partisan bias results from efforts to achieve permissible nonpartisan objectives. If so, a tradeoff exists between partisan fairness and valid nonpartisan goals, and the map is lawful. If not, partisan balance is possible without compromising other aims, and the map is unconstitutional.
The prospect of tradeoffs is also central to racial vote dilution doctrine under the Voting Rights Act (VRA). To succeed in this sort of suit, a plaintiff must show not only that an additional majority-minority district could be drawn, but that this hypothetical district could be “reasonably configured” in that “it comports with traditional districting criteria.”
This requirement presumes that, in at least some cases, greater minority representation is attainable only through the creation of unreasonably configured districts that violate traditional criteria. When noncompliance with traditional criteria is the price of more minority representation, this element can’t be established and the plaintiff loses.
A similar potential tradeoff is the crux of the constitutional claim of racial gerrymandering. To demonstrate that a district is a racial gerrymander—one unjustifiably crafted for a racially predominant purpose—“a plaintiff must prove that the State ‘subordinated’ race-neutral districting criteria . . . to ‘racial considerations.’”
Such subordination occurs when race-neutral criteria are sacrificed on race-conscious grounds like targeting a particular demographic composition for a district. When race-neutral and race-conscious ends are both realized, racial predominance is absent and no liability ensues.
Given the importance of tradeoffs in redistricting law, one might expect the academic literature to focus intently on them. One might think there would be studies galore on the existence and extent of tradeoffs among traditional criteria, among partisan fairness and nonpartisan goals, and among minority representation and nonracial aims. But that hunch would be wrong.
In fact, no legal scholarship has previously identified redistricting tradeoffs as a discrete topic of interest. A handful of political science articles have done so, but these pieces, while helpful, have been quite limited in their scope. They have mostly addressed just one state (not many of them)
and just one pair of criteria (not a wider array of partisan, racial, and other factors).
These few existing works have also analyzed small or unrepresentative sets of maps—not ones generated at scale to reflect accurately the relevant map universe.
This Article is therefore the first to assess redistricting tradeoffs systematically. I cover congressional and state legislative maps for seven priority states, as well as congressional maps alone for all forty-four states with two or more congressional districts. This is a much larger swath of American redistricting than any prior study in this genre has surveyed. I examine most measurable line-drawing criteria: traditional requirements like compactness and adherence to county boundaries (each calculated in multiple ways), electoral variables like partisan fairness and competitiveness (also quantified through multiple metrics), and minority representation (estimated by ecological inference on a national scale, not crude demographic shortcuts). And to conduct this investigation, I use a total of more than fourteen billion district maps generated at random by cutting-edge computer algorithms.
The new map ensembles on which this project relies are themselves noteworthy. To my knowledge, they’re the largest and highest-quality sets of maps yet created through the emerging method of computational redistricting. Unlike most other map ensembles, each one is demonstrably representative of the applicable map universe for the state and electoral level. And all these new sets of maps are publicly available—in numerous versions, to boot, corresponding to different degrees of population equality, compactness, and county-splitting—so other researchers may study and learn from them.
The Article’s principal finding is that tradeoffs among redistricting criteria are generally weak to nonexistent. More specifically, within the vast majority of randomly generated maps that are in the heartland of each bivariate distribution, substantial improvement along one dimension can almost always be achieved without any decline in terms of the other parameter. Typically, the correlation between each pair of criteria is also close to zero, and no meaningful link appears when alternative ensembles or measures are used or other variables are included as regression controls. These patterns hold among traditional criteria like compactness and adherence to county boundaries, which can be simultaneously increased in most cases. Likewise, partisan fairness is usually unrelated to traditional criteria and to minority representation—and it’s often positively associated with competitiveness, meaning that these goods tend to be complements, not substitutes. Minority representation, too, mostly has little to no connection to the compliance of minority-opportunity districts with traditional criteria.
One might worry that these results are idiosyncratic to the new map ensembles on which they’re based. To allay this concern, I rerun the analyses using the sets of congressional maps published in 2022 by the Algorithm-Assisted Redistricting Methodology (ALARM) Project.
The ALARM map ensembles are produced by a completely different computer algorithm. Yet they give rise to identical substantive conclusions. Again, within the heartland of maps in each bivariate distribution, concurrent progress along both dimensions is generally possible. And again, the correlation between each pair of criteria is, in the main, statistically and practically insignificant.
Why do courts—and many scholars
—wrongly suppose that redistricting tradeoffs are pervasive? A likely answer is that their mental models of mapmaking are too limited. They may fixate on a particular plan, which could be subject to material tradeoffs if it’s amended, and overlook the near-infinite number of other district configurations, some of which almost certainly dominate the benchmark plan. To make this point empirically, I start with a recent court-drawn plan that epitomizes how humans balance redistricting criteria. When a computer algorithm initially adjusts this plan, tradeoffs among criteria indeed appear. Crucially, however, these tradeoffs lessen as the algorithm continues to run, moving further away from the original plan. When the algorithm runs for long enough, the tradeoffs disappear entirely as many maps are churned out that are superior to the judicial handiwork.
Courts and scholars may also intuit that, at some point, parallel improvement along two (or more) parameters becomes impossible. Eventually, a Pareto frontier must be reached where progress on one axis requires regression on another.
In the Article’s last empirical contribution, I identify this frontier for each pair of redistricting criteria for each state and electoral level. Two facts about these frontiers are most striking. First, their slopes tend to be gentle, not steep. In other words, only a minor setback toward one objective is typically necessary for a major gain toward another goal. Second, enacted plans—plans crafted and ratified by humans—are rarely at these frontiers. In most cases, enacted plans are actually quite distant from the areas where compromises among criteria are unavoidable. These zones’ existence, then, doesn’t vindicate claims of ubiquitous tradeoffs because, as far as most mapmaking is concerned, the frontiers might as well be absent.
These findings have sweeping implications for redistricting law. Begin with the Rucho Court’s argument that disregard for traditional criteria can’t be the standard for unlawful partisan gerrymandering because these requirements usually conflict.
In fact, in most distributions of maps for most states and electoral levels, these requirements are rarely at odds. So, under this approach, courts wouldn’t need to decide how to “prioritize competing criteria”
because these criteria are seldom in competition. Instead, courts could insist on strong performance with respect to all traditional requirements. Plans would then be invalid if they fell short along any of these dimensions, because this subpar score normally could not be explained by an aim to satisfy another criterion.
In states where partisan gerrymandering suits can still be brought, these results also augur poorly for defendants at the justification stage of the inquiry. At this stage, a defendant maintains that a plan’s bias is attributable to the pursuit of some legitimate nonpartisan end.
In the map ensembles, however, partisan fairness is infrequently in tension with nonpartisan criteria like compactness, adherence to county boundaries, competitiveness, and minority representation. Most often, partisan fairness is uncorrelated with these criteria and can be improved without losing any ground in these respects. The upshot is that, in the mine run of litigation, a defendant should be hard-pressed to justify a biased plan. Data like that presented here should generally help to refute any attempted justification.
Turning from party to race, the Article bears good news for racial vote dilution plaintiffs. One of their obligations is to show that minority representation could be bolstered without unreasonably worsening the compliance of minority-opportunity districts with traditional criteria.
The map ensembles illustrate that this is possible in many circumstances. Furthermore, opportunity districts don’t typically become less mindful of traditional criteria as their volume rises. Accordingly, the element that has been the stumbling block for many past VRA litigants
could now be easier to establish with modern redistricting technology. Indeed, these plaintiffs might consult maps produced for this project for ideas about how to prove this pillar of their cases.
In the racial gerrymandering context, finally (and uniquely), the Article’s findings benefit defendants. Here, a district is subject to strict scrutiny if it subordinates race-neutral criteria to racial considerations.
The map ensembles indicate that, as a rule, such subordination is unnecessary. By and large, racial considerations like reaching certain levels of minority representation can be achieved while satisfying race-neutral criteria to the same extent. Consequently, just as VRA plaintiffs may wish to peruse this project’s maps for offensive purposes, potential racial gerrymandering defendants could filter the maps for ones that accomplish their race-conscious objectives without sacrificing other goals that are race-neutral. If a jurisdiction adopts this kind of map, a court is unlikely to find that any district was designed for a racially predominant reason.
The Article is organized as follows. Part I explains how redistricting law (and much redistricting scholarship) assumes the existence of widespread tradeoffs among line-drawing criteria. Part II summarizes the small extant literature on redistricting tradeoffs and describes the Article’s empirical strategy. Part III presents the Article’s results. These span: (1) a congressional map ensemble for an illustrative state; (2) congressional and state legislative map ensembles for seven priority states; (3) the ALARM congressional map ensembles for all states with two or more congressional districts; (4) local tradeoffs for a recent court-drawn plan; and (5) Pareto frontiers for all map ensembles. Part IV discusses the implications of these analyses. They unsettle several redistricting subfields and reveal an available future in which plans abide by traditional criteria while yielding fair representation for voters of all partisan and racial affiliations.