Ever since the Warren Court constitutionalized the right to vote, judges have recognized that an electoral system in which every adult citizen votes without hindrance may nonetheless be fundamentally unfair, owing to the mechanisms for aggregating votes into outcomes. Votes cast by citizens with distinct political interests may be “diluted”—“minimiz[ed] or cancel[ed] out”—by, for example, the design of legislative districts or the choice between at-large and districted elections. Congress drew on this insight in 1982 when it amended § 2 of the Voting Rights Act (VRA) to provide a statutory remedy for racial vote dilution. 

At the heart of vote dilution law is the concept of racially polarized voting. Voting is polarized when (1) the political prefer- ences of majority-race and minority-race voters diverge substantially and the racial majority votes with enough cohesion to usually defeat the minority’s candidates of choice. We call these the “preference polarization” and “voting power” requirements. Since the Supreme Court’s 1986 decision in Thornburg v Gingles, plaintiffs have had to satisfy both conditions and propose a remedial district at the outset of their case. Only if this threshold Gingles showing has been made does the court apply the liability standard prescribed by statute: whether the “totality of circumstances” indicates that plaintiff-race voters “have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice.”

The Supreme Court understands the Gingles test to serve a dual purpose. The test keeps vote dilution law manageable by limiting the number of cases in which courts must make politically delicate totality-of-the-circumstances judgment calls about racial fairness in the distribution of political opportunity. An important premise of the manageability story is that the polarization test is objective and rule-like—and thus that it is likely to be applied consistently by judges whose “personal political views” may diverge. The second purpose is normative diagnosis. 

The polarization test helps courts to make quick, rough judgments about whether serious harms within the meaning of § 2 are likely to be present. It “ensures that clearly meritorious claims will survive summary judgment . . . while appropriately closing the courthouse to marginal cases.”

Implementation of the Gingles test is subject to a constitutionally derived side constraint (or parallel objective): judges are not to indulge “prohibited,” racially essentialist assumptions, such as assuming that voters of the same race “think alike, share the same political interests, and will prefer the same candidates at the polls.”

Responding to the Supreme Court, we argue that the lower courts cannot—under current conditions—implement the racial polarization test so as to satisfy the Court’s stated objectives concurrently. The test cannot be at once diagnostic of liability, constraining of judicial discretion, and free of strong racial assumptions.

There are two reasons for this. One is the normatively unsettled state of vote dilution law. Several competing theories of racial vote dilution each find some support in Supreme Court precedent, and these theories have radically different implications for the racial-polarization test. That there is no generally accepted theory of racial vote dilution is common knowledge among legal academics, but, with limited exceptions, academics have not considered how normative disagreements shape judicial application of the polarization test. If anything, law professors have tacitly assumed that the putatively objective polarization test covers for the lack of a theory, allowing judges who may have very different normative understandings of racial vote dilution (or no understanding at all) to make reasonably consistent decisions. This is mistaken. We show that the polarization test leaves district judges with broad discretion and that many of the lower courts’ disputes about how to exercise this discretion correspond to often-unspoken normative disagreements about the meaning of racial vote dilution.

The second problem is the long-established convention, encouraged though not compelled by the Supreme Court, of grounding racial-polarization findings on “voting preferences expressed in actual elections.” The working assumptions have been (1) that the average level of bloc voting among coethnic vot- ers in some class of typical elections reliably indicates the extent of within-group preference homogeneity and between-group preference divergence, and (2) that the level of racial bloc voting in any given election can be estimated from precinct-level vote totals and demographic data without making strong assumptions about political homogeneity within racial groups. Neither premise is tenable.

We show theoretically and with evidence from survey experiments that the presence (or absence) of racial polarization in vote shares is an unreliable indicator of preference polarization. The root of the problem is strategic behavior—by candidates, parties, donors, and voters. Because of strategic behavior, the relationship between polarization in vote shares and polarization in underlying political preferences is highly contingent. This threatens to render the Gingles test quite arbitrary, unless judges either make very strong assumptions (some racial) or else abandon the notion of an objective, quantitative polarization test in favor of a subjective inquiry that requires close attention to the very thing that the creators of the Gingles test wanted the courts to ignore: “the political stories behind the election returns.” Thirty years of racial-polarization law in the lower courts bear witness to this problem, even as commentators continue to describe the Gingles framework as objective and constraining.

Moreover, so long as polarization findings continue to be based on “voting preferences expressed in actual elections,” those preferences must be estimated, and the estimation of candidates’ vote shares by racial group from ballots cast in actual elections depends on strong assumptions about political homogeneity within racial groups across geographic areas. These assumptions are close kin to those the Supreme Court disavowed in the recent case of League of United Latin American Citizens v Perry (“LULAC”). 

What follows for vote dilution law? The most important implications concern the Supreme Court’s ongoing campaign to bolster its manageability and prevent § 2 from “infus[ing] race into virtually every redistricting.” The current strategy of the Court—or at least of a decisive plurality of the justices—is to circumscribe the geographic reach of §2 through limiting, bright-line constructions of the Gingles conditions. In Bartlett v Strickland, the controlling opinion held that vote dilution claims may be brought only by plaintiffs whose racial group would compose a literal, numeric majority of the voting-age population in a compact remedial district.

Strickland also encourages the establishment of bright-line rules for the preference-polarization component of the Gingles test. The plurality ventured in dicta that claims should probably fail as a matter of law if white voters typically “cross over” and support minority-preferred candidates at levels exceeding some numeric threshold of legal significance, perhaps 20 percent. At oral argument, five justices voiced support for this idea, suggesting cutoffs ranging from 15 percent to 40 percent. Presumably a vote dilution claim would also fail if minority voters defected from minority-preferred candidates at similar rates. “Legally significant” preference polarization would exist only if, say, at least 60 percent (or 70 percent, or 80 percent) of plaintiff-race voters typically voted for the minority-preferred candidate and at least 60 percent (or 70 percent, or 80 percent) of other voters typically supported the opposing candidate. Though the lower courts have so far declined invitations to create such vote-share cutoffs, leading academic commentators regard their establishment as a natural next step in the evolution of vote dilution law.

Our analysis casts serious doubt on the vote-share-cutoff idea, and further suggests that if appellate courts do adopt cutoffs, this will induce fact finders to delve ever more deeply into “the political stories behind the election returns,” or else to rely even more heavily on strong assumptions, some of which are “racial” in nature. As such, the establishment of vote-share cutoffs would, in important respects, hinder rather than advance the Supreme Court’s manageability and constitutional objectives for the racial-polarization test. 

But what is the alternative? One option is for courts to invite preference-polarization showings based on survey data rather than on votes. Surveys can be designed to yield comparable information across jurisdictions, and survey data can be analyzed without imposing strong assumptions about homogeneity within racial groups.

Alternatively, the Supreme Court could simply drop the preference-polarization requirement. Claims could instead be screened with a test focused on minority political incorporation, or perhaps with a test that probed for indirect evidence of intentionally discriminatory districting. A court adopting the former approach might ask whether there is active two-party competition for control of the legislative body and active recruitment of minority-race candidates, or the court might focus on whether minority-race candidates have in fact been elected. A court more concerned with discriminatory intent by conventional state actors might ask, as Judge Frank Easterbrook recently proposed, whether there are fewer majority-minority districts than would likely have been drawn by an automated redistricting algorithm.