Allen v. Milligan, 599 U.S. ___ (2023)

NOTICE: This opinion is subject to formal revision before publication in the United States Reports. Readers are requested to notify the Reporter of Decisions, Supreme Court of the United States, Washington, D. C. 20543, pio@supremecourt.gov, of any typographical or other formal errors.

SUPREME COURT OF THE UNITED STATES

_________________

Nos. 21–1086 and 21–1087

_________________

WES ALLEN, ALABAMA SECRETARY OF STATE, et al., APPELLANTS

21–1086v.

EVAN MILLIGAN, et al.

on appeal from the united states district court for the northern district of alabama

WES ALLEN, ALABAMA SECRETARY OF STATE, et al., PETITIONERS

21–1087v.

MARCUS CASTER, et al.

on writ of certiorari before judgment to the united states court of appeals for the eleventh circuit

[June 8, 2023]

Chief Justice Roberts delivered the opinion of the Court, except as to Part III–B–1.[1]*

In January 2022, a three-judge District Court sitting in Alabama preliminarily enjoined the State from using the districting plan it had recently adopted for the 2022 congressional elections, finding that the plan likely violated Section 2 of the Voting Rights Act, 52 U. S. C. §10301. This Court stayed the District Court’s order pending further review. 595 U. S. ___ (2022). After conducting that review, we now affirm.

I

A

Shortly after the Civil War, Congress passed and the States ratified the Fifteenth Amendment, providing that “[t]he right of citizens of the United States to vote shall not be denied or abridged . . . on account of race, color, or previous condition of servitude.” U. S. Const., Amdt. 15, §1. In the century that followed, however, the Amendment proved little more than a parchment promise. Jim Crow laws like literacy tests, poll taxes, and “good-morals” requirements abounded, South Carolina v. Katzenbach, 383 U.S. 301, 312–313 (1966), “render[ing] the right to vote illusory for blacks,” Northwest Austin Municipal Util. Dist. No. One v. Holder, 557 U.S. 193, 220–221 (2009) (Thomas, J., concurring in judgment in part and dissenting in part). Congress stood up to little of it; “[t]he first century of congressional enforcement of the [Fifteenth] Amendment . . . can only be regarded as a failure.” Id., at 197 (majority opinion).

That changed in 1965. Spurred by the Civil Rights movement, Congress enacted and President Johnson signed into law the Voting Rights Act. 79Stat. 437, as amended, 52 U. S. C. §10301 et seq. The Act “create[d] stringent new remedies for voting discrimination,” attempting to forever “banish the blight of racial discrimination in voting.” Katzenbach, 383 U. S., at 308. By 1981, in only sixteen years’ time, many considered the VRA “the most successful civil rights statute in the history of the Nation.” S. Rep. No. 97–417, p. 111 (1982) (Senate Report).

These cases concern Section 2 of that Act. In its original form, “§2 closely tracked the language of the [Fifteenth] Amendment” and, as a result, had little independent force. Brnovich v. Democratic National Committee, 594 U. S. ___, ___ (2021) (slip op., at 3).[2] Our leading case on §2 at the time was City of Mobile v. Bolden, which involved a claim by black voters that the City’s at-large election system effectively excluded them from participating in the election of city commissioners. 446 U.S. 55 (1980). The commission had three seats, black voters comprised one-third of the City’s population, but no black-preferred candidate had ever won election.

The Court ruled against the plaintiffs. The Fifteenth Amendment—and thus §2—prohibits States from acting with a “racially discriminatory motivation” or an “invidious purpose” to discriminate. Id., at 61–65 (plurality opinion). But it does not prohibit laws that are discriminatory only in effect. Ibid. The Mobile plaintiffs could “register and vote without hindrance”—“their freedom to vote ha[d] not been denied or abridged by anyone.” Id., at 65. The fact that they happened to lose frequently was beside the point. Nothing the City had done “purposeful[ly] exclu[ded]” them “from participati[ng] in the election process.” Id., at 64.

Almost immediately after it was decided, Mobile “produced an avalanche of criticism, both in the media and within the civil rights community.” T. Boyd & S. Markman, The 1982 Amendments to the Voting Rights Act: A Legislative History, 40 Wash. & Lee L. Rev. 1347, 1355 (1983) (Boyd & Markman). The New York Times wrote that the decision represented “the biggest step backwards in civil rights to come from the Nixon Court.” N. Y. Times, Apr. 23, 1980, p. A22. And the Washington Post described Mobile as a “major defeat for blacks and other minorities fighting electoral schemes that exclude them from office.” Washington Post, Apr. 23, 1980, p. A5. By focusing on discriminatory intent and ignoring disparate effect, critics argued, the Court had abrogated “the standard used by the courts to determine whether [racial] discrimination existed . . . : Whether such discrimination existed.” It’s Results That Count, Philadelphia Inquirer, Mar. 3, 1982, p. 8–A.

But Mobile had its defenders, too. In their view, abandoning the intent test in favor of an effects test would inevitably require a focus on proportionality—wherever a minority group won fewer seats in the legislature than its share of the population, the charge could be made that the State law had a discriminatory effect. That, after all, was the type of claim brought in Mobile. But mandating racial proportionality in elections was regarded by many as intolerable. Doing so, wrote Senator Orrin Hatch in the Washington Star, would be “strongly resented by the American public.” Washington Star, Sept. 30, 1980, p. A–9. The Wall Street Journal offered similar criticism. An effects test would generate “more, not less, racial and ethnic polarization.” Wall Street Journal, Jan. 19, 1982, p. 28.

This sharp debate arrived at Congress’s doorstep in 1981. The question whether to broaden §2 or keep it as is, said Hatch—by then Chairman of the Senate Subcommittee before which §2 would be debated—“involve[d] one of the most substantial constitutional issues ever to come before this body.” 2 Hearings before the Subcommittee on the Constitution of the Senate Committee on the Judiciary, 97th Cong., 2d Sess., pt. 1, p. 1 (1982).

Proceedings in Congress mirrored the disagreement that had developed around the country. In April 1981, Congressman Peter W. Rodino, Jr.—longtime chairman of the House Judiciary Committee—introduced a bill to amend the VRA, proposing that the words “to deny or abridge” in §2 be replaced with the phrase “in a manner which results in a denial or abridgement.” H. R. 3112, 97th Cong., 1st Sess., 2 (as introduced) (emphasis added). This was the effects test that Mobile’s detractors sought.

But those wary of proportionality were not far behind. Senator Hatch argued that the effects test “was intelligible only to the extent that it approximated a standard of proportional representation by race.” Boyd & Markman 1392. The Attorney General had the same concern. The effects test “would be triggered whenever election results did not mirror the population mix of a particular community,” he wrote, producing “essentially a quota system for electoral politics.” N. Y. Times, Mar. 27, 1982, p. 23.

The impasse was not resolved until late April 1982, when Senator Bob Dole proposed a compromise. Boyd & Markman 1414. Section 2 would include the effects test that many desired but also a robust disclaimer against proportionality. Seeking to navigate any tension between the two, the Dole Amendment borrowed language from a Fourteenth Amendment case of ours, White v. Regester, 412 U.S. 755 (1973), which many in Congress believed would allow courts to consider effects but avoid proportionality. The standard for liability in voting cases, White explained, was whether “the political processes leading to nomination and election were not equally open to participation by the group in question—[in] that its members had less opportunity than did other residents in the district to participate in the political processes and to elect legislators of their choice.” Id., at 766.

The Dole compromise won bipartisan support and, on June 18, the Senate passed the 1982 amendments by an overwhelming margin, 85–8. Eleven days later, President Reagan signed the Act into law. The amended §2 reads as follows:

“(a) No voting qualification or prerequisite to voting or standard, practice, or procedure shall be imposed or applied by any State or political subdivision in a manner which results in a denial or abridgement of the right of any citizen of the United States to vote on account of race or color . . . as provided in subsection (b).

“(b) A violation of subsection (a) is established if, based on the totality of circumstances, it is shown that the political processes leading to nomination or election in the State or political subdivision are not equally open to participation by members of a class of citizens . . . in that its members have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice. The extent to which members of a protected class have been elected to office in the State or political subdivision is one circumstance which may be considered: Provided, That nothing in this section establishes a right to have members of a protected class elected in numbers equal to their proportion in the population.” 52 U. S. C. §10301.

B

For the first 115 years following Reconstruction, the State of Alabama elected no black Representatives to Congress. See Singleton v. Merrill, 582 F. Supp. 3d 924, 947 (ND Ala. 2022) ( per curiam). In 1992, several plaintiffs sued the State, alleging that it had been impermissibly diluting the votes of black Alabamians in violation of §2. See Wesch v. Hunt, 785 F. Supp. 1491, 1493 (SD Ala.). The lawsuit produced a majority-black district in Alabama for the first time in decades. Id., at 1499. And that fall, Birmingham lawyer Earl Hillard became the first black Representative from Alabama since 1877. 582 F. Supp. 3d, at 947.

Alabama’s congressional map has “remained remarkably similar” after Wesch. Brief for Appellants in No. 21–1086 etc., p. 9 (Brief for Alabama). The map contains seven congressional districts, each with a single representative. See Supp. App. 205–211; 582 F. Supp. 3d, at 951. District 1 encompasses the Gulf Coast region in the southwest; District 2—known as the Wiregrass region—occupies the southeast; District 3 covers the eastern-central part of the State; Districts 4 and 5 stretch width-wise across the north, with the latter layered atop the former; District 6 is right in the State’s middle; and District 7 spans the central west. Id., at 951.

In 2020, the decennial census revealed that Alabama’s population had grown by 5.1%. See 1 App. 86. A group of plaintiffs led by Alabama legislator Bobby Singleton sued the State, arguing that the existing congressional map was malapportioned and racially gerrymandered in violation of the Equal Protection Clause. 582 F. Supp. 3d, at 938–939. While litigation was proceeding, the Alabama Legislature’s Committee on Reapportionment began creating a new districting map. Ibid. Although the prior decade’s population growth did not change the number of seats that Alabama would receive in the House, the growth had been unevenly distributed across the State, and the existing map was thus out of date.

To solve the problem, the State turned to experienced mapmaker Randy Hinaman, who had created several districting maps that Alabama used over the past 30 years. Id., at 947–948. The starting point for Hinaman was the then-existing 2011 congressional map, itself a product of the 2001 map that Hinaman had also created. Civ. No. 21–1530 (ND Ala.), ECF Doc. 70–2, pp. 40, 93–94; see also 582 F. Supp. 3d, at 950. Hinaman worked to adjust the 2011 map in accordance with the redistricting guidelines set by the legislature’s Reapportionment Committee. Id., at 948–950; 1 App. 275. Those guidelines prioritized population equality, contiguity, compactness, and avoiding dilution of minority voting strength. 582 F. Supp. 3d, at 1035–1036. They also encouraged, as a secondary matter, avoiding incumbent pairings, respecting communities of interest, minimizing the number of counties in each district, and preserving cores of existing districts. Id., at 1036–1037.

The resulting map Hinaman drew largely resembled the 2011 map, again producing only one district in which black voters constituted a majority of the voting age population. Supp. App. 205–211. The Alabama Legislature enacted Hinaman’s map under the name HB1. 582 F. Supp. 3d, at 935, 950–951. Governor Ivey signed HB1 into law on November 4, 2021. Id., at 950.

C

Three groups of plaintiffs brought suit seeking to stop Alabama’s Secretary of State from conducting congressional elections under HB1. The first group was led by Dr. Marcus Caster, a resident of Washington County, who challenged HB1 as invalid under §2. Id., at 934–935, 980. The second group, led by Montgomery County resident Evan Milligan, brought claims under §2 and the Equal Protection Clause of the Fourteenth Amendment. Id., at 939–940, 966. Finally, the Singleton plaintiffs, who had previously sued to enjoin Alabama’s 2011 congressional map, amended their complaint to challenge HB1 as an impermissible racial gerrymander under the Equal Protection Clause. Id., at 938–939.

A three-judge District Court was convened, comprised of Circuit Judge Marcus and District Judges Manasco and Moorer. The Singleton and Milligan actions were consolidated before the three-judge Court for purposes of preliminary injunction proceedings, while Caster proceeded before Judge Manasco on a parallel track. 582 F. Supp. 3d, at 934–935. A preliminary injunction hearing began on January 4, 2022, and concluded on January 12. Id., at 943. In that time, the three-judge District Court received live testimony from 17 witnesses, reviewed more than 1000 pages of briefing and upwards of 350 exhibits, and considered arguments from the 43 different lawyers who had appeared in the litigation. Id., at 935–936. After reviewing that extensive record, the Court concluded in a 227-page opinion that the question whether HB1 likely violated §2 was not “a close one.” It did. Id., at 1026. The Court thus preliminarily enjoined Alabama from using HB1 in forthcoming elections. Id., at 936.[3]

Four days later, on January 28, Alabama moved in this Court for a stay of the District Court’s injunction. This Court granted a stay and scheduled the cases for argument, noting probable jurisdiction in Milligan and granting certiorari before judgment in Caster. 595 U. S. ___ (2022).

II

The District Court found that plaintiffs demonstrated a reasonable likelihood of success on their claim that HB1 violates §2. We affirm that determination.

A

For the past forty years, we have evaluated claims brought under §2 using the three-part framework developed in our decision Thornburg v. Gingles, 478 U.S. 30 (1986). Gingles concerned a challenge to North Carolina’s multimember districting scheme, which allegedly diluted the vote of its black citizens. Id., at 34–36. The case presented the first opportunity since the 1982 amendments to address how the new §2 would operate.

Gingles began by describing what §2 guards against. “The essence of a §2 claim,” the Court explained, “is that a certain electoral law, practice, or structure interacts with social and historical conditions to cause an inequality in the opportunities enjoyed by black and white voters.” Id., at 47. That occurs where an “electoral structure operates to minimize or cancel out” minority voters’ “ability to elect their preferred candidates.” Id., at 48. Such a risk is greatest “where minority and majority voters consistently prefer different candidates” and where minority voters are submerged in a majority voting population that “regularly defeat[s]” their choices. Ibid.

To succeed in proving a §2 violation under Gingles, plaintiffs must satisfy three “preconditions.” Id., at 50. First, the “minority group must be sufficiently large and [geographically] compact to constitute a majority in a reasonably configured district.” Wisconsin Legislature v. Wisconsin Elections Comm’n, 595 U. S. ___, ___ (2022) (per curiam) (slip op., at 3) (citing Gingles, 478 U. S., at 46–51). A district will be reasonably configured, our cases explain, if it comports with traditional districting criteria, such as being contiguous and reasonably compact. See Alabama Legislative Black Caucus v. Alabama, 575 U.S. 254, 272 (2015). “Second, the minority group must be able to show that it is politically cohesive.” Gingles, 478 U. S., at 51. And third, “the minority must be able to demonstrate that the white majority votes sufficiently as a bloc to enable it . . . to defeat the minority’s preferred candidate.” Ibid. Finally, a plaintiff who demonstrates the three preconditions must also show, under the “totality of circumstances,” that the political process is not “equally open” to minority voters. Id., at 45–46; see also id., at 36–38 (identifying several factors relevant to the totality of circumstances inquiry, including “the extent of any history of official discrimination in the state . . . that touched the right of the members of the minority group to register, to vote, or otherwise to participate in the democratic process”).

Each Gingles precondition serves a different purpose. The first, focused on geographical compactness and numerosity, is “needed to establish that the minority has the potential to elect a representative of its own choice in some single-member district.” Growe v. Emison, 507 U.S. 25, 40 (1993). The second, concerning the political cohesiveness of the minority group, shows that a representative of its choice would in fact be elected. See ibid. The third precondition, focused on racially polarized voting, “establish[es] that the challenged districting thwarts a distinctive minority vote” at least plausibly on account of race. Ibid. And finally, the totality of circumstances inquiry recognizes that application of the Gingles factors is “peculiarly dependent upon the facts of each case.” 478 U. S., at 79. Before courts can find a violation of §2, therefore, they must conduct “an intensely local appraisal” of the electoral mechanism at issue, as well as a “searching practical evaluation of the ‘past and present reality.’ ” Ibid.

Gingles has governed our Voting Rights Act jurisprudence since it was decided 37 years ago. Congress has never disturbed our understanding of §2 as Gingles construed it. And we have applied Gingles in one §2 case after another, to different kinds of electoral systems and to different jurisdictions in States all over the country. See Voinovich v. Quilter, 507 U.S. 146 (1993) (Ohio); Growe, 507 U. S., at 25 (Minnesota); Johnson v. De Grandy, 512 U.S. 997 (1994) (Florida); Holder v. Hall, 512 U.S. 874 (1994) (Georgia); Abrams v. Johnson, 521 U.S. 74 (1997) (Georgia); League of United Latin American Citizens v. Perry, 548 U.S. 399, 423 (2006) (LULAC) (Texas); Bartlett v. Strickland, 556 U.S. 1 (2009) (plurality opinion) (North Carolina); Cooper v. Harris, 581 U.S. 285 (2017) (North Carolina); Abbott v. Perez, 585 U. S. ___ (2018) (Texas); Wisconsin Legislature, 595 U. S. ___ (Wisconsin).

B

As noted, the District Court concluded that plaintiffs’ §2 claim was likely to succeed under Gingles. 582 F. Supp. 3d, at 1026. Based on our review of the record, we agree.

With respect to the first Gingles precondition, the District Court correctly found that black voters could constitute a majority in a second district that was “reasonably configured.” 1 App. to Emergency Application for Stay in No. 21–1086 etc., p. 253 (MSA). The plaintiffs adduced eleven illustrative maps—that is, example districting maps that Alabama could enact—each of which contained two majority-black districts that comported with traditional districting criteria. With respect to compactness, for example, the District Court explained that the maps submitted by one of plaintiffs’ experts, Dr. Moon Duchin, “perform[ed] generally better on average than” did HB1. 582 F. Supp. 3d, at 1009. A map offered by another of plaintiffs’ experts, Bill Cooper, produced districts roughly as compact as the existing plan. Ibid. And none of plaintiffs’ maps contained any “tentacles, appendages, bizarre shapes, or any other obvious irregularities that would make it difficult to find” them sufficiently compact. Id., at 1011. Plaintiffs’ maps also satisfied other traditional districting criteria. They contained equal populations, were contiguous, and respected existing political subdivisions, such as counties, cities, and towns. Id., at 1011, 1016. Indeed, some of plaintiffs’ proposed maps split the same number of county lines as (or even fewer county lines than) the State’s map. Id., at 1011–1012. We agree with the District Court, therefore, that plaintiffs’ illustrative maps “strongly suggest[ed] that Black voters in Alabama” could constitute a majority in a second, reasonably configured, district. Id., at 1010.

The State nevertheless argues that plaintiffs’ maps were not reasonably configured because they failed to keep together a traditional community of interest within Alabama. See, e.g., id., at 1012. A “community of interest,” according to Alabama’s districting guidelines, is an “area with recognized similarities of interests, including but not limited to ethnic, racial, economic, tribal, social, geographic, or historical identities.” Ibid. Alabama argues that the Gulf Coast region in the southwest of the State is such a community of interest, and that plaintiffs’ maps erred by separating it into two different districts. Ibid.

We do not find the State’s argument persuasive. Only two witnesses testified that the Gulf Coast was a community of interest. Id., at 1015. The testimony provided by one of those witnesses was “partial, selectively informed, and poorly supported.” Ibid. The other witness, meanwhile, justified keeping the Gulf Coast together “simply” to preserve “political advantage[ ]”: “You start splitting counties,” he testified, “and that county loses its influence. That’s why I don’t want Mobile County to be split.” Id., at 990, 1015. The District Court understandably found this testimony insufficient to sustain Alabama’s “overdrawn argument that there can be no legitimate reason to split” the Gulf Coast region. Id., at 1015.

Even if the Gulf Coast did constitute a community of interest, moreover, the District Court found that plaintiffs’ maps would still be reasonably configured because they joined together a different community of interest called the Black Belt. Id., at 1012–1014. Named for its fertile soil, the Black Belt contains a high proportion of black voters, who “share a rural geography, concentrated poverty, unequal access to government services, . . . lack of adequate healthcare,” and a lineal connection to “the many enslaved people brought there to work in the antebellum period.” Id., at 1012–1013; see also 1 App. 299–304. The District Court concluded—correctly, under our precedent—that it did not have to conduct a “beauty contest[ ]” between plaintiffs’ maps and the State’s. There would be a split community of interest in both. 582 F. Supp. 3d, at 1012 (quoting Bush v. Vera, 517 U.S. 952, 977–978 (1996) (plurality opinion)).

The State also makes a related argument based on “core retention”—a term that refers to the proportion of districts that remain when a State transitions from one districting plan to another. See, e.g., Brief for Alabama 25, 61. Here, by largely mirroring Alabama’s 2011 districting plan, HB1 performs well on the core retention metric. Plaintiffs’ illustrative plans, by contrast, naturally fare worse because they change where the 2011 district lines were drawn. See e.g., Supp. App. 164–173. But this Court has never held that a State’s adherence to a previously used districting plan can defeat a §2 claim. If that were the rule, a State could immunize from challenge a new racially discriminatory redistricting plan simply by claiming that it resembled an old racially discriminatory plan. That is not the law: §2 does not permit a State to provide some voters “less opportunity . . . to participate in the political process” just because the State has done it before. 52 U. S. C. §10301(b).

As to the second and third Gingles preconditions, the District Court determined that there was “no serious dispute that Black voters are politically cohesive, nor that the challenged districts’ white majority votes sufficiently as a bloc to usually defeat Black voters’ preferred candidate.” 582 F. Supp. 3d, at 1016 (internal quotation marks omitted). The Court noted that, “on average, Black voters supported their candidates of choice with 92.3% of the vote” while “white voters supported Black-preferred candidates with 15.4% of the vote.” Id., at 1017 (internal quotation marks omitted). Plaintiffs’ experts described the evidence of racially polarized voting in Alabama as “intens[e],” “very strong,” and “very clear.” Ibid. Even Alabama’s expert conceded “that the candidates preferred by white voters in the areas that he looked at regularly defeat the candidates preferred by Black voters.” Id., at 1018.

Finally, the District Court concluded that plaintiffs had carried their burden at the totality of circumstances stage. The Court observed that elections in Alabama were racially polarized; that “Black Alabamians enjoy virtually zero success in statewide elections”; that political campaigns in Alabama had been “characterized by overt or subtle racial appeals”; and that “Alabama’s extensive history of repugnant racial and voting-related discrimination is undeniable and well documented.” Id., at 1018–1024.

We see no reason to disturb the District Court’s careful factual findings, which are subject to clear error review and have gone unchallenged by Alabama in any event. See Cooper, 581 U. S., at 309. Nor is there a basis to upset the District Court’s legal conclusions. The Court faithfully applied our precedents and correctly determined that, under existing law, HB1 violated §2.

III

The heart of these cases is not about the law as it exists. It is about Alabama’s attempt to remake our §2 jurisprudence anew.

The centerpiece of the State’s effort is what it calls the “race-neutral benchmark.” The theory behind it is this: Using modern computer technology, mapmakers can now generate millions of possible districting maps for a given State. The maps can be designed to comply with traditional districting criteria but to not consider race. The mapmaker can determine how many majority-minority districts exist in each map, and can then calculate the median or average number of majority-minority districts in the entire multimillion-map set. That number is called the race-neutral benchmark.

The State contends that this benchmark should serve as the point of comparison in §2 cases. The benchmark, the State says, was derived from maps that were “race-blind”—maps that cannot have “deni[ed] or abridge[d]” anyone’s right to vote “on account of race” because they never took race into “account” in the first place. 52 U. S. C. §10301(a). Courts in §2 cases should therefore compare the number of majority-minority districts in the State’s plan to the benchmark. If those numbers are similar—if the State’s map “resembles” the benchmark in this way—then, Alabama argues, the State’s map also cannot have “deni[ed] or abridge[d]” anyone’s right to vote “on account of race.” Ibid.

Alabama contends that its approach should be adopted for two reasons. First, the State argues that a race-neutral benchmark best matches the text of the Voting Rights Act. Section 2 requires that the political processes be “equally open.” §10301(b). What that means, the State asserts, is that the State’s map cannot impose “obstacles or burdens that block or seriously hinder voting on account of race.” Brief for Alabama 43. These obstacles do not exist, in the State’s view, where its map resembles a map that never took race into “account.” Ibid. Second, Alabama argues that the Gingles framework ends up requiring racial proportionality in districting. According to the State, Gingles demands that where “another majority-black district could be drawn, it must be drawn.” Brief for Alabama 71 (emphasis deleted). And that sort of proportionality, Alabama continues, is inconsistent with the compromise that Congress struck, with the text of §2, and with the Constitution’s prohibition on racial discrimination in voting.

To apply the race-neutral benchmark in practice, Alabama would require §2 plaintiffs to make at least three showings. First, the illustrative plan that plaintiffs adduce for the first Gingles precondition cannot have been “based” on race. Brief for Alabama 56. Second, plaintiffs must show at the totality of circumstances stage that the State’s enacted plan diverges from the average plan that would be drawn without taking race into account. And finally, plaintiffs must ultimately prove that any deviation between the State’s plan and a race-neutral plan is explainable “only” by race—not, for example, by “the State’s naturally occurring geography and demography.” Id., at 46.

As we explain below, we find Alabama’s new approach to §2 compelling neither in theory nor in practice. We accordingly decline to recast our §2 case law as Alabama requests.

A

1

Section 2 prohibits States from imposing any “standard, practice, or procedure . . . in a manner which results in a denial or abridgement of the right of any citizen . . . to vote on account of race or color.” 52 U. S. C. §10301(a). What that means, §2 goes on to explain, is that the political processes in the State must be “equally open,” such that minority voters do not “have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice.” §10301(b).

We have understood the language of §2 against the background of the hard-fought compromise that Congress struck. To that end, we have reiterated that §2 turns on the presence of discriminatory effects, not discriminatory intent. See, e.g., Chisom v. Roemer, 501 U.S. 380, 403–404 (1991). And we have explained that “[i]t is patently clear that Congress has used the words ‘on account of race or color’ in the Act to mean ‘with respect to’ race or color, and not to connote any required purpose of racial discrimination.” Gingles, 478 U. S., at 71, n. 34 (plurality opinion) (some alterations omitted). Individuals thus lack an equal opportunity to participate in the political process when a State’s electoral structure operates in a manner that “minimize[s] or cancel[s] out the[ir] voting strength.” Id., at 47. That occurs where an individual is disabled from “enter[ing] into the political process in a reliable and meaningful manner” “in the light of past and present reality, political and otherwise.” White, 412 U. S., at 767, 770. A district is not equally open, in other words, when minority voters face—unlike their majority peers—bloc voting along racial lines, arising against the backdrop of substantial racial discrimination within the State, that renders a minority vote unequal to a vote by a nonminority voter.

The State’s reading of §2, by contrast, runs headlong into our precedent. Alabama asserts that a State’s map does not “abridge[ ]” a person’s right to vote “on account of race” if the map resembles a sufficient number of race-neutral alternatives. See Brief for Alabama 54–56. But our cases have consistently focused, for purposes of litigation, on the specific illustrative maps that a plaintiff adduces. Deviation from that map shows it is possible that the State’s map has a disparate effect on account of race. The remainder of the Gingles test helps determine whether that possibility is reality by looking to polarized voting preferences and the frequency of racially discriminatory actions taken by the State, past and present.

A State’s liability under §2, moreover, must be determined “based on the totality of circumstances.” 52 U. S. C. §10301(b). Yet Alabama suggests there is only one “circumstance[ ]” that matters—how the State’s map stacks up relative to the benchmark. That single-minded view of §2 cannot be squared with the VRA’s demand that courts employ a more refined approach. And we decline to adopt an interpretation of §2 that would “revise and reformulate the Gingles threshold inquiry that has been the baseline of our §2 jurisprudence” for nearly forty years. Bartlett, 556 U. S., at 16 (plurality opinion); see also Wisconsin Legislature, 595 U. S., at ___ (slip op., at 7) (faulting lower court for “improperly reduc[ing] Gingles’ totality-of-circumstances analysis to a single factor”); De Grandy, 512 U. S., at 1018 (“An inflexible rule would run counter to the textual command of §2, that the presence or absence of a violation be assessed ‘based on the totality of circumstances.’ ”).[4]

2

Alabama also argues that the race-neutral benchmark is required because our existing §2 jurisprudence inevitably demands racial proportionality in districting, contrary to the last sentence of §2(b). But properly applied, the Gingles framework itself imposes meaningful constraints on proportionality, as our decisions have frequently demonstrated.

In Shaw v. Reno, for example, we considered the permissibility of a second majority-minority district in North Carolina, which at the time had 12 seats in the U. S. House of Representatives and a 20% black voting age population. 509 U.S. 630, 633–634 (1993). The second majority-minority district North Carolina drew was “160 miles long and, for much of its length, no wider than the [interstate] corridor.” Id., at 635. The district wound “in snakelike fashion through tobacco country, financial centers, and manufacturing areas until it gobble[d] in enough enclaves of black neighborhoods.” Id., at 635–636. Indeed, the district was drawn so imaginatively that one state legislator remarked: “[I]f you drove down the interstate with both car doors open, you’d kill most of the people in the district.” Id., at 636.

Though North Carolina believed the additional district was required by §2, we rejected that conclusion, finding instead that those challenging the map stated a claim of impermissible racial gerrymandering under the Equal Protection Clause. Id., at 655, 658. In so holding, we relied on the fact that the proposed district was not reasonably compact. Id., at 647. North Carolina had “concentrated a dispersed minority population in a single district by disregarding traditional districting principles such as compactness, contiguity, and respect for political subdivisions.” Ibid. (emphasis added). And “[a] reapportionment plan that includes in one district individuals who belong to the same race, but who are otherwise separated by geographical and political boundaries,” we said, raised serious constitutional concerns. Ibid. (emphasis added).

The same theme emerged in our 1995 decision Miller v. Johnson, where we upheld a district court’s finding that one of Georgia’s ten congressional districts was the product of an impermissible racial gerrymander. 515 U.S. 900, 906, 910–911. At the time, Georgia’s black voting age population was 27%, but there was only one majority-minority district. Id., at 906. To comply with the VRA, Georgia thought it necessary to create two more majority-minority districts—achieving proportionality. Id., at 920–921. But like North Carolina in Shaw, Georgia could not create the districts without flouting traditional criteria. One district “centered around four discrete, widely spaced urban centers that ha[d] absolutely nothing to do with each other, and stretch[ed] the district hundreds of miles across rural counties and narrow swamp corridors.” 515 U. S., at 908. “Geographically,” we said of the map, “it is a monstrosity.” Id., at 909.

In Bush v. Vera, a plurality of the Court again explained how traditional districting criteria limited any tendency of the VRA to compel proportionality. The case concerned Texas’s creation of three additional majority-minority districts. 517 U. S., at 957. Though the districts brought the State closer to proportional representation, we nevertheless held that they constituted racial gerrymanders in violation of the Fourteenth Amendment. That was because the districts had “no integrity in terms of traditional, neutral redistricting criteria.” Id., at 960. One of the majority-black districts consisted “of narrow and bizarrely shaped tentacles.” Id., at 965. The proposed majority-Hispanic district resembled “a sacred Mayan bird” with “[s]pindly legs reach[ing] south” and a “plumed head ris[ing] northward.” Id., at 974.

The point of all this is a simple one. Forcing proportional representation is unlawful and inconsistent with this Court’s approach to implementing §2. The numbers bear the point out well. At the congressional level, the fraction of districts in which black-preferred candidates are likely to win “is currently below the Black share of the eligible voter population in every state but three.” Brief for Professors Jowei Chen et al. as Amici Curiae 3 (Chen Brief ). Only one State in the country, meanwhile, “has attained a proportional share” of districts in which Hispanic-preferred candidates are likely to prevail. Id., at 3–4. That is because as residential segregation decreases—as it has “sharply” done since the 1970s—satisfying traditional districting criteria such as the compactness requirement “becomes more difficult.” T. Crum, Reconstructing Racially Polarized Voting, 70 Duke L. J. 261, 279, and n. 105 (2020).

Indeed, as amici supporting the appellees emphasize, §2 litigation in recent years has rarely been successful for just that reason. See Chen Brief 3–4. Since 2010, plaintiffs nationwide have apparently succeeded in fewer than ten §2 suits. Id., at 7. And “the only state legislative or congressional districts that were redrawn because of successful Section 2 challenges were a handful of state house districts near Milwaukee and Houston.” Id., at 7–8. By contrast, “[n]umerous lower courts” have upheld districting maps “where, due to minority populations’ geographic diffusion, plaintiffs couldn’t design an additional majority-minority district” or satisfy the compactness requirement. Id., at 15–16 (collecting cases). The same has been true of recent litigation in this Court. See Abbott, 585 U. S., at ___–___ (slip op., at 33–34) (finding a Texas district did not violate §2 because “the geography and demographics of south and west Texas do not permit the creation of any more than the seven Latino . . . districts that exist under the current plan”).[5]

Reapportionment, we have repeatedly observed, “is primarily the duty and responsibility of the State[s],” not the federal courts. Id., at ___ (slip op., at 21). Properly applied, the Gingles factors help ensure that remains the case. As respondents themselves emphasize, §2 “never require[s] adoption of districts that violate traditional redistricting principles.” Brief for Respondents in No. 21–1087, p. 3. Its exacting requirements, instead, limit judicial intervention to “those instances of intensive racial politics” where the “excessive role [of race] in the electoral process . . . den[ies] minority voters equal opportunity to participate.” Senate Report 33–34.

B

Although we are content to reject Alabama’s invitation to change existing law on the ground that the State misunderstands §2 and our decisions implementing it, we also address how the race-neutral benchmark would operate in practice. Alabama’s approach fares poorly on that score, which further counsels against our adopting it.

1

The first change to existing law that Alabama would require is prohibiting the illustrative maps that plaintiffs submit to satisfy the first Gingles precondition from being “based” on race. Brief for Alabama 56. Although Alabama is not entirely clear whether, under its view, plaintiffs’ illustrative plans must not take race into account at all or whether they must just not “prioritize” race, ibid., we see no reason to impose such a new rule.

When it comes to considering race in the context of districting, we have made clear that there is a difference “between being aware of racial considerations and being motivated by them.” Miller, 515 U. S., at 916; see also North Carolina v. Covington, 585 U. S. ___, ___ (2018) ( per curiam) (slip op., at 8). The former is permissible; the latter is usually not. That is because “[r]edistricting legislatures will . . . almost always be aware of racial demographics,” Miller, 515 U. S., at 916, but such “race consciousness does not lead inevitably to impermissible race discrimination,” Shaw, 509 U. S., at 646. Section 2 itself “demands consideration of race.” Abbott, 581 U. S., at ___ (slip op., at 4). The question whether additional majority-minority districts can be drawn, after all, involves a “quintessentially race-conscious calculus.” De Grandy, 512 U. S., at 1020.

At the same time, however, race may not be “the predominant factor in drawing district lines unless [there is] a compelling reason.” Cooper, 581 U. S., at 291. Race predominates in the drawing of district lines, our cases explain, when “race-neutral considerations [come] into play only after the race-based decision had been made.” Bethune-Hill v. Virginia State Bd. of Elections, 580 U.S. 178, 189 (2017) (internal quotation marks omitted). That may occur where “race for its own sake is the overriding reason for choosing one map over others.” Id., at 190.

While the line between racial predominance and racial consciousness can be difficult to discern, see Miller, 515 U. S., at 916, it was not breached here. The Caster plaintiffs relied on illustrative maps produced by expert Bill Cooper. See 2 App. 591–592. Cooper testified that while it was necessary for him to consider race, he also took several other factors into account, such as compactness, contiguity, and population equality. Ibid. Cooper testified that he gave all these factors “equal weighting.” Id., at 594. And when asked squarely whether race predominated in his development of the illustrative plans, Cooper responded: “No. It was a consideration. This is a Section 2 lawsuit, after all. But it did not predominate or dominate.” Id., at 595.

The District Court agreed. It found “Cooper’s testimony highly credible” and commended Cooper for “work[ing] hard to give ‘equal weight[ ]’ to all traditional redistricting criteria.” 582 F. Supp. 3d, at 1005–1006; see also id., at 978–979. The court also explained that Alabama’s evidence of racial predominance in Cooper’s maps was exceedingly thin. Alabama’s expert, Thomas Bryan, “testified that he never reviewed the exhibits to Mr. Cooper’s report” and “that he never reviewed” one of the illustrative plans that Cooper submitted. Id., at 1006. Bryan further testified that he could offer no “conclusions or opinions as to the apparent basis of any individual line drawing decisions in Cooper’s illustrative plans.” 2 App. 740. By his own admission, Bryan’s analysis of any race predominance in Cooper’s maps “was pretty light.” Id., at 739. The District Court did not err in finding that race did not predominate in Cooper’s maps in light of the evidence before it.[6]

The dissent contends that race nevertheless predominated in both Cooper’s and Duchin’s maps because they were designed to hit “ ‘express racial target[s]’ ”—namely, two “50%-plus majority-black districts.” Post, at 15 (opinion of Thomas, J.) (quoting Bethune-Hill, 580 U. S., at 192). This argument fails in multiple ways. First, the dissent’s reliance on Bethune-Hill is mistaken. In that case, this Court was unwilling to conclude that a State’s maps were produced in a racially predominant manner. Instead, we remanded for the lower court to conduct the predominance analysis itself, explaining that “the use of an express racial target” was just one factor among others that the court would have to consider as part of “[a] holistic analysis.” Id., at 192. Justice Thomas dissented in relevant part, contending that because “the legislature sought to achieve a [black voting-age population] of at least 55%,” race necessarily predominated in its decisionmaking. Id., at 198 (opinion concurring in part and dissenting in part). But the Court did not join in that view, and Justice Thomas again dissents along the same lines today.

The second flaw in the dissent’s proposed approach is its inescapable consequence: Gingles must be overruled. According to the dissent, racial predominance plagues every single illustrative map ever adduced at the first step of Gingles. For all those maps were created with an express target in mind—they were created to show, as our cases require, that an additional majority-minority district could be drawn. That is the whole point of the enterprise. The upshot of the approach the dissent urges is not to change how Gingles is applied, but to reject its framework outright.

The contention that mapmakers must be entirely “blind” to race has no footing in our §2 case law. The line that we have long drawn is between consciousness and predominance. Plaintiffs adduced at least one illustrative map that comported with our precedents. They were required to do no more to satisfy the first step of Gingles.

2

The next condition Alabama would graft onto §2 is a requirement that plaintiffs demonstrate, at the totality of circumstances stage, that the State’s enacted plan contains fewer majority-minority districts than the race-neutral benchmark. Brief for Alabama 43. If it does not, then §2 should drop out of the picture. Id., at 44.

Alabama argues that is what should have happened here. It notes that one of plaintiffs’ experts, Dr. Duchin, used an algorithm to create “2 million districting plans for Alabama . . . without taking race into account in any way in the generation process.” 2 App. 710. Of these two million “race-blind” plans, none contained two majority-black districts while many plans did not contain any. Ibid. Alabama also points to a “race-neutral” computer simulation conducted by another one of plaintiffs’ experts, Dr. Kosuke Imai, which produced 30,000 potential maps. Brief for Alabama 55. As with Dr. Duchin’s maps, none of the maps that Dr. Imai created contained two majority-black districts. See 2 App. 571–572. Alabama thus contends that because HB1 sufficiently “resembles” the “race-neutral” maps created by Dr. Duchin and Dr. Imai—all of the maps lack two majority-black districts—HB1 does not violate §2. Brief for Alabama 54.

Alabama’s reliance on the maps created by Dr. Duchin and Dr. Imai is misplaced. For one, neither Duchin’s nor Imai’s maps accurately represented the districting process in Alabama. Dr. Duchin’s maps were based on old census data—from 2010 instead of 2020—and ignored certain traditional districting criteria, such as keeping together communities of interest, political subdivisions, or municipalities.[7] And Dr. Imai’s 30,000 maps failed to incorporate Alabama’s own districting guidelines, including keeping together communities of interest and preserving municipal boundaries. See Supp. App. 58–59.[8]

But even if the maps created by Dr. Duchin and Dr. Imai were adequate comparators, we could not adopt the map-comparison test that Alabama proposes. The test is flawed in its fundamentals. Districting involves myriad considerations—compactness, contiguity, political subdivisions, natural geographic boundaries, county lines, pairing of incumbents, communities of interest, and population equality. See Miller, 515 U. S., at 916. Yet “[q]uantifying, measuring, prioritizing, and reconciling these criteria” requires map drawers to “make difficult, contestable choices.” Brief for Computational Redistricting Experts as Amici Curiae 8 (Redistricting Brief ). And “[i]t is easy to imagine how different criteria could move the median map toward different . . . distributions,” meaning that “the same map could be [lawful] or not depending solely on what the mapmakers said they set out to do.” Rucho v. Common Cause, 588 U. S. ___, ___–___ (2019) (slip op., at 27–28). For example, “the scientific literature contains dozens of competing metrics” on the issue of compactness. Redistricting Brief 8. Which one of these metrics should be used? What happens when the maps they produce yield different benchmark results? How are courts to decide?

Alabama does not say; it offers no rule or standard for determining which of these choices are better than others. Nothing in §2 provides an answer either. In 1982, the computerized mapmaking software that Alabama contends plaintiffs must use to demonstrate an (unspecified) level of deviation did not even exist. See, e.g., J. Chen & N. Stephanopoulos, The Race-Blind Future of Voting Rights, 130 Yale L. J. 862, 881–882 (2021) (Chen & Stephanopoulos). And neither the text of §2 nor the fraught debate that produced it suggests that “equal access” to the fundamental right of voting turns on computer simulations that are technically complicated, expensive to produce, and available to “[o]nly a small cadre of university researchers [that] have the resources and expertise to run” them. Brief for United States as Amicus Curiae 28 (citing Chen & Stephanopoulos 882–884).[9]

One final point bears mentioning. Throughout these cases, Alabama has repeatedly emphasized that HB1 cannot have violated §2 because none of plaintiffs’ two million odd maps contained more than one majority-minority district. See, e.g., Brief for Alabama 1, 23, 30, 31, 54–56, 70, 79. The point is that two million is a very big number and that sheer volume matters. But as elsewhere, Alabama misconceives the math project that it expects courts to oversee. A brief submitted by three computational redistricting experts explains that the number of possible districting maps in Alabama is at least in the “trillion trillions.” Redistricting Brief 6, n. 7. Another publication reports that the number of potential maps may be orders of magnitude higher: “the universe of all possible connected, population-balanced districting plans that satisfy the state’s requirements,” it explains, “is likely in the range of googols.” Duchin & Spencer 768. Two million maps, in other words, is not many maps at all. And Alabama’s insistent reliance on that number, however powerful it may sound in the abstract, is thus close to irrelevant in practice. What would the next million maps show? The next billion? The first trillion of the trillion trillions? Answerless questions all. See, e.g., Redistricting Brief 2 (“[I]t is computationally intractable, and thus effectively impossible, to generate a complete enumeration of all potential districting plans. [Even] algorithms that attempt to create a manageable sample of that astronomically large universe do not consistently identify an average or median map.”); Duchin & Spencer 768 (“[A] comprehensive survey of [all districting plans within a State] is impossible.”).

Section 2 cannot require courts to judge a contest of computers when there is no reliable way to determine who wins, or even where the finish line is.

3

Alabama’s final contention with respect to the race-neutral benchmark is that it requires plaintiffs to demonstrate that any deviations between the State’s enacted plan and race-neutral alternatives “can be explained only by racial discrimination.” Brief for Alabama 44 (emphasis added).

We again find little merit in Alabama’s proposal. As we have already explained, our precedents and the legislative compromise struck in the 1982 amendments clearly rejected treating discriminatory intent as a requirement for liability under §2. See, e.g., Chisom, 501 U. S., at 403–404; Shaw, 509 U. S., at 641; Reno v. Bossier Parish School Bd., 520 U.S. 471, 481–482 (1997). Yet Alabama’s proposal is even more demanding than the intent test Congress jettisoned. Demonstrating discriminatory intent, we have long held, “does not require a plaintiff to prove that the challenged action rested solely on racially discriminatory purpose[ ].” Arlington Heights v. Metropolitan Housing Development Corp., 429 U.S. 252, 265 (1977) (emphasis added); see also Reno, 520 U. S., at 488. Alabama’s proposed approach stands in sharp contrast to all this, injecting into the effects test of §2 an evidentiary standard that even our purposeful discrimination cases eschew.

C

Alabama finally asserts that the Court should outright stop applying §2 in cases like these because the text of §2 does not apply to single-member redistricting and because §2 is unconstitutional as the District Court applied it here. We disagree on both counts.

Alabama first argues that §2 does not apply to single-member redistricting. Echoing Justice Thomas’s concurrence in Holder v. Hall, Alabama reads §2’s reference to “standard, practice, or procedure” to mean only the “methods for conducting a part of the voting process that might . . . be used to interfere with a citizen’s ability to cast his vote.” 512 U. S., at 917–918 (opinion concurring in judgment). Examples of covered activities would include “registration requirements, . . . the locations of polling places, the times polls are open, the use of paper ballots as opposed to voting machines, and other similar aspects of the voting process.” Id., at 922. But not “a single-member districting system or the selection of one set of districting lines over another.” Id., at 923.

This understanding of §2 cannot be reconciled with our precedent. As recounted above, we have applied §2 to States’ districting maps in an unbroken line of decisions stretching four decades. See supra, at 11; see also Brnovich, 594 U. S., at ___, n. 5 (slip op., at 7, n. 5) (collecting cases). In doing so, we have unanimously held that §2 and Gingles “[c]ertainly . . . apply” to claims challenging single-member districts. Growe, 507 U. S., at 40. And we have even invalidated portions of a State’s single-district map under §2. See LULAC, 548 U. S., at 427–429.[10] Alabama’s approach would require “abandoning” this precedent, “overruling the interpretation of §2” as set out in nearly a dozen of our cases. Holder, 512 U. S., at 944 (opinion of Thomas, J.).

We decline to take that step. Congress is undoubtedly aware of our construing §2 to apply to districting challenges. It can change that if it likes. But until and unless it does, statutory stare decisis counsels our staying the course. See, e.g., Kimble v. Marvel Entertainment, LLC, 576 U.S. 446, 456 (2015).[11]

The statutory text in any event supports the conclusion that §2 applies to single-member districts. Alabama’s own proffered definition of a “procedure is the manner or method of proceeding in a process or course of action.” Brief for Alabama 51 (internal quotation marks omitted). But the manner of proceeding in the act of voting entails determining in which districts voters will vote. The fact that the term “procedure” is preceded by the phrase “qualification or prerequisite to voting,” 52 U. S. C. §10301(a), does not change its meaning. It is hard to imagine many more fundamental “prerequisites” to voting than determining where to cast your ballot or who you are eligible to vote for. Perhaps for that reason, even Alabama does not bear the courage of its conviction on this point. It refuses to argue that §2 is inapplicable to multimember districting, though its textual arguments apply with equal force in that context.

The dissent, by contrast, goes where even Alabama does not dare, arguing that §2 is wholly inapplicable to districting because it “focuses on ballot access and counting” only. Post, at 2 (opinion of Thomas, J.). But the statutory text upon which the dissent relies supports the exact opposite conclusion. The relevant section provides that “[t]he terms ‘vote’ or ‘voting’ shall include all action necessary to make a vote effective.” Ibid. (quoting 52 U. S. C. §10310(c)(1); emphasis added). Those actions “includ[e], but [are] not limited to, . . . action[s] required by law prerequisite to voting, casting a ballot, and having such ballot counted properly and included in the appropriate totals of votes cast.” §10310(c)(1). It would be anomalous to read the broad language of the statute—“all action necessary,” “including but not limited to”—to have the crabbed reach that Justice Thomas posits. And we have already discussed why determining where to cast a ballot constitutes a “prerequisite” to voting, as the statute requires.

The dissent also contends that “applying §2 to districting rests on systematic neglect of . . . the ballot-access focus of the 1960s’ voting-rights struggles.” Post, at 3 (opinion of Thomas, J.). But history did not stop in 1960. As we have explained, Congress adopted the amended §2 in response to the 1980 decision City of Mobile, a case about districting. And—as the dissent itself acknowledges—“Congress drew §2(b)’s current operative language” from the 1973 decision White v. Regester, post, at 4, n. 3 (opinion of Thomas, J.), a case that was also about districting (in fact, a case that invalidated two multimember districts in Texas and ordered them redrawn into single-member districts, 412 U. S., at 765). This was not lost on anyone when §2 was amended. Indeed, it was the precise reason that the contentious debates over proportionality raged—debates that would have made little sense if §2 covered only poll taxes and the like, as the dissent contends.

We also reject Alabama’s argument that §2 as applied to redistricting is unconstitutional under the Fifteenth Amendment. According to Alabama, that Amendment permits Congress to legislate against only purposeful discrimination by States. See Brief for Alabama 73. But we held over 40 years ago “that, even if §1 of the [Fifteenth] Amendment prohibits only purposeful discrimination, the prior decisions of this Court foreclose any argument that Congress may not, pursuant to §2 [of the Fifteenth Amendment] outlaw voting practices that are discriminatory in effect.” City of Rome v. United States, 446 U.S. 156, 173 (1980). The VRA’s “ban on electoral changes that are discriminatory in effect,” we emphasized, “is an appropriate method of promoting the purposes of the Fifteenth Amendment.” Id., at 177. As City of Rome recognized, we had reached the very same conclusion in South Carolina v. Katzenbach, a decision issued right after the VRA was first enacted. 383 U. S., at 308–309, 329–337; see also Brnovich, 594 U. S., at ___ (slip op., at 3).

Alabama further argues that, even if the Fifteenth Amendment authorizes the effects test of §2, that Amendment does not authorize race-based redistricting as a remedy for §2 violations. But for the last four decades, this Court and the lower federal courts have repeatedly applied the effects test of §2 as interpreted in Gingles and, under certain circumstances, have authorized race-based redistricting as a remedy for state districting maps that violate §2. See, e.g., supra, at 11; cf. Mississippi Republican Executive Committee v. Brooks, 469 U.S. 1002 (1984). In light of that precedent, including City of Rome, we are not persuaded by Alabama’s arguments that §2 as interpreted in Gingles exceeds the remedial authority of Congress.

The concern that §2 may impermissibly elevate race in the allocation of political power within the States is, of course, not new. See, e.g., Shaw, 509 U. S., at 657 (“Racial gerrymandering, even for remedial purposes, may balkanize us into competing racial factions; it threatens to carry us further from the goal of a political system in which race no longer matters.”). Our opinion today does not diminish or disregard these concerns. It simply holds that a faithful application of our precedents and a fair reading of the record before us do not bear them out here.

*  *  *

The judgments of the District Court for the Northern District of Alabama in the Caster case, and of the three-judge District Court in the Milligan case, are affirmed.

It is so ordered.

SUPREME COURT OF THE UNITED STATES

_________________

Nos. 21–1086 and 21–1087

_________________

WES ALLEN, ALABAMA SECRETARY OF STATE, et al., APPELLANTS

21–1086v.

EVAN MILLIGAN, et al.

on appeal from the united states district court for the northern district of alabama

WES ALLEN, ALABAMA SECRETARY OF STATE, et al., PETITIONERS

21–1087v.

MARCUS CASTER, et al.

on writ of certiorari before judgment to the united states court of appeals for the eleventh circuit

[June 8, 2023]

Justice Kavanaugh, concurring in all but Part III–B–1.

I agree with the Court that Alabama’s redistricting plan violates §2 of the Voting Rights Act as interpreted in Thornburg v. Gingles, 478 U.S. 30 (1986). I write separately to emphasize four points.

First, the upshot of Alabama’s argument is that the Court should overrule Gingles. But the stare decisis standard for this Court to overrule a statutory precedent, as distinct from a constitutional precedent, is comparatively strict. Unlike with constitutional precedents, Congress and the President may enact new legislation to alter statutory precedents such as Gingles. In the past 37 years, however, Congress and the President have not disturbed Gingles, even as they have made other changes to the Voting Rights Act. Although statutory stare decisis is not absolute, “the Court has ordinarily left the updating or correction of erroneous statutory precedents to the legislative process.” Ramos v. Louisiana, 590 U. S. ___, ___ (2020) (Kavanaugh, J., concurring in part) (slip op., at 4); see also, e.g., Kimble v. Marvel Entertainment, LLC, 576 U.S. 446, 456 (2015); Patterson v. McLean Credit Union, 491 U.S. 164, 172–173 (1989); Flood v. Kuhn, 407 U.S. 258, 283–284 (1972); Burnet v. Coronado Oil & Gas Co., 285 U.S. 393, 406 (1932) (Brandeis, J., dissenting).[1]

Second, Alabama contends that Gingles inevitably requires a proportional number of majority-minority districts, which in turn contravenes the proportionality disclaimer in §2(b) of the Voting Rights Act. 52 U. S. C. §10301(b). But Alabama’s premise is wrong. As the Court’s precedents make clear, Gingles does not mandate a proportional number of majority-minority districts. Gingles requires the creation of a majority-minority district only when, among other things, (i) a State’s redistricting map cracks or packs a large and “geographically compact” minority population and (ii) a plaintiff ’s proposed alternative map and proposed majority-minority district are “reasonably configured”—namely, by respecting compactness principles and other traditional districting criteria such as county, city, and town lines. See, e.g., Cooper v. Harris, 581 U.S. 285, 301–302 (2017); Voinovich v. Quilter, 507 U.S. 146, 153–154 (1993); ante, at 10–12, 18–22.

If Gingles demanded a proportional number of majority-minority districts, States would be forced to group together geographically dispersed minority voters into unusually shaped districts, without concern for traditional districting criteria such as county, city, and town lines. But Gingles and this Court’s later decisions have flatly rejected that approach. See, e.g., Abbott v. Perez, 585 U. S. ___, ___–___ (2018) (slip op., at 33–34); Bush v. Vera, 517 U.S. 952, 979 (1996) (plurality opinion); Gingles, 478 U. S., at 50; see also Miller v. Johnson, 515 U.S. 900, 917–920 (1995); Shaw v. Reno, 509 U.S. 630, 644–649 (1993); ante, at 18–22.[2]

Third, Alabama argues that courts should rely on race-blind computer simulations of redistricting maps to assess whether a State’s plan abridges the right to vote on account of race. It is true that computer simulations might help detect the presence or absence of intentional discrimination. For example, if all of the computer simulations generated only one majority-minority district, it might be difficult to say that a State had intentionally discriminated on the basis of race by failing to draw a second majority-minority district.

But as this Court has long recognized—and as all Members of this Court today agree—the text of §2 establishes an effects test, not an intent test. See ante, at 17; post, at 7 (Thomas, J., dissenting); post, at 16 (Alito, J., dissenting). And the effects test, as applied by Gingles to redistricting, requires in certain circumstances that courts account for the race of voters so as to prevent the cracking or packing—whether intentional or not—of large and geographically compact minority populations. See Abbott, 585 U. S., at ___ (slip op., at 4); Johnson v. De Grandy, 512 U.S. 997, 1006–1007, 1020 (1994); Voinovich, 507 U. S., at 153–154; see generally Brnovich v. Democratic National Committee, 594 U. S. ___, ___ (2021) (slip op., at 22) (“§2 does not demand proof of discriminatory purpose”); Reno v. Bossier Parish School Bd., 520 U.S. 471, 482 (1997) (Congress “clearly expressed its desire that §2 not have an intent component”); Holder v. Hall, 512 U.S. 874, 923–924 (1994) (Thomas, J., concurring in judgment) (§2 adopts a “ ‘results’ test, rather than an ‘intent’ test”); Chisom v. Roemer, 501 U.S. 380, 394, 404 (1991) (“proof of intent is no longer required to prove a §2 violation” as “Congress made clear that a violation of §2 could be established by proof of discriminatory results alone”); Gingles, 478 U. S., at 71, n. 34 (plurality opinion) (§2 does not require “ ‘purpose of racial discrimination’ ”).

Fourth, Alabama asserts that §2, as construed by Gingles to require race-based redistricting in certain circumstances, exceeds Congress’s remedial or preventive authority under the Fourteenth and Fifteenth Amendments. As the Court explains, the constitutional argument presented by Alabama is not persuasive in light of the Court’s precedents. See ante, at 33–34; see also City of Rome v. United States, 446 U.S. 156, 177–178 (1980). Justice Thomas notes, however, that even if Congress in 1982 could constitutionally authorize race-based redistricting under §2 for some period of time, the authority to conduct race-based redistricting cannot extend indefinitely into the future. See post, at 44–45 (dissenting opinion). But Alabama did not raise that temporal argument in this Court, and I therefore would not consider it at this time.

For those reasons, I vote to affirm, and I concur in all but Part III–B–1 of the Court’s opinion.

SUPREME COURT OF THE UNITED STATES

_________________

Nos. 21–1086 and 21–1087

_________________

WES ALLEN, ALABAMA SECRETARY OF STATE, et al., APPELLANTS

21–1086v.

EVAN MILLIGAN, et al.

on appeal from the united states district court for the northern district of alabama

WES ALLEN, ALABAMA SECRETARY OF STATE, et al., PETITIONERS

21–1087v.

MARCUS CASTER, et al.

on writ of certiorari before judgment to the united states court of appeals for the eleventh circuit

[June 8, 2023]

Justice Alito, with whom Justice Gorsuch joins, dissenting.

Based on a flawed understanding of the framework adopted in Thornburg v. Gingles, 478 U.S. 30 (1986), the Court now holds that the congressional districting map adopted by the Alabama Legislature violates §2 of the Voting Rights Act. Like the Court, I am happy to apply Gingles in these cases. But I would interpret that precedent in a way that heeds what §2 actually says, and I would take constitutional requirements into account. When the Gingles framework is viewed in this way, it is apparent that the decisions below must be vacated.

I

A

Gingles marked the Court’s first encounter with the amended version of §2 that Congress enacted in 1982, and the Court’s opinion set out an elaborate framework that has since been used to analyze a variety of §2 claims. Under that framework, a plaintiff must satisfy three “preconditions.” Id., at 50. As summarized in more recent opinions, they are as follows:

“First, [the] ‘minority group’ [whose interest the plaintiff represents] must be ‘sufficiently large and geographically compact to constitute a majority’ in some reasonably configured legislative district. Second, the minority group must be ‘politically cohesive.’ And third, a district’s white majority must ‘vote[ ] sufficiently as a bloc’ to usually ‘defeat the minority’s preferred candidate.’ ” Cooper v. Harris, 581 U.S. 285, 301–302 (2017) (citations omitted).

See also Wisconsin Legislature v. Wisconsin Elections Comm’n, 595 U. S. ___, ___ (2022) (per curiam) (slip op., at 3); Merrill v. Milligan, 595 U. S. ___, ___ (2022) (Kagan, J., dissenting from grant of applications for stays) (slip op., at 3–4).

If a §2 plaintiff can satisfy all these preconditions, the court must then decide whether, based on the totality of the circumstances, the plaintiff ’s right to vote was diluted. See Gingles, 478 U. S., at 46–48, 79. And to aid in that inquiry, Gingles approved consideration of a long list of factors set out in the Senate Judiciary Committee’s Majority Report on the 1982 VRA amendments. Id., at 44–45 (citing S. Rep. No. 97–417, pp. 28–30 (1982)).

B

My fundamental disagreement with the Court concerns the first Gingles precondition. In cases like these, where the claim is that §2 requires the creation of an additional majority-minority district, the first precondition means that the plaintiff must produce an additional illustrative majority-minority district that is “reasonably configured.” Cooper, 581 U. S., at 301; Wisconsin Legislature, 595 U. S., at ___ (slip op., at 3); see also Gingles, 478 U. S., at 50.

The Court’s basic error is that it misunderstands what it means for a district to be “reasonably configured.” Our cases make it clear that “reasonably configured” is not a synonym for “compact.” We have explained that the first precondition also takes into account other traditional districting criteria like attempting to avoid the splitting of political subdivisions and “communities of interest.” League of United Latin American Citizens v. Perry, 548 U.S. 399, 433–434 (2006) (LULAC).

To its credit, the Court recognizes that compactness is not enough and that a district is not reasonably configured if it flouts other “traditional districting criteria.” Ante, at 10. At various points in its opinion it names quite a few: minimizing the splitting of counties and other political subdivisions, keeping “communities of interest” together where possible, and avoiding the creation of new districts that require two incumbents to run against each other. Ante, at 12, 26–27. In addition, the Court acknowledges that a district is not “reasonably configured” if it does not comport with the Equal Protection Clause’s one-person, one-vote requirement. Ante, at 27. But the Court fails to explain why compliance with “traditional districting criteria” matters under §2 or why the only relevant equal protection principle is the one-person, one-vote requirement. If the Court had attempted to answer these questions, the defect in its understanding of the first Gingles precondition would be unmistakable.

To explain this, I begin with what is probably the most frequently mentioned traditional districting criterion and ask why it should matter under §2 whether a proposed majority-minority district is “compact.” Neither the Voting Rights Act (VRA) nor the Constitution imposes a compactness requirement. The Court notes that we have struck down bizarrely shaped districts, ante, at 19–20, but we did not do that for esthetic reasons. Compactness in and of itself is not a legal requirement—or even necessarily an esthetic one. (Some may find fancifully shaped districts more pleasing to the eye than boring squares.)

The same is true of departures from other traditional districting criteria. Again, nothing in the Constitution or the VRA demands compliance with these criteria. If a whimsical state legislature cavalierly disregards county and municipal lines and communities of interest, draws weirdly shaped districts, departs radically from a prior map solely for the purpose of change, and forces many incumbents to run against each other, neither the Constitution nor the VRA would make any of that illegal per se. Bizarrely shaped districts and other marked departures from traditional districting criteria matter because mapmakers usually heed these criteria, and when it is evident that they have not done so, there is reason to suspect that something untoward—specifically, unconstitutional racial gerrymandering—is afoot. See, e.g., Shaw v. Reno, 509 U.S. 630, 643–644 (1993); Bush v. Vera, 517 U.S. 952, 979 (1996) (plurality opinion); cf. LULAC, 548 U. S., at 433–435.

Conspicuous violations of traditional districting criteria constitute strong circumstantial evidence of unconstitutionality. And when it is shown that the configuration of a district is attributable predominantly to race, that is more than circumstantial evidence that the district is unlawful. That is direct evidence of illegality because, as we have often held, race may not “predominate” in the drawing of district lines. See, e.g., Cooper, 581 U. S., at 292; Bethune-Hill v. Virginia State Bd. of Elections, 580 U.S. 178, 191–192 (2017); Shaw v. Hunt, 517 U.S. 899, 906–907 (1996) (Shaw II); Miller v. Johnson, 515 U.S. 900, 920 (1995).[1]

Because non-predominance is a longstanding and vital feature of districting law, it must be honored in a Gingles plaintiff ’s illustrative district. If race predominated in the creation of such a district, the plaintiff has failed to satisfy both our precedent, which requires “reasonably configured” districts, and the terms of §2, which demand equal openness. Two Terms ago, we engaged in a close analysis of the text of §2 and explained that its “key requirement” is that the political processes leading to nomination or election must be “ ‘equally open to participation’ by members of a protected class.” Brnovich v. Democratic National Committee, 594 U. S. ___, ___ (2021) (slip op., at 6, 15) (quoting 52 U. S. C. §10301(b); emphasis deleted). “[E]qual openness,” we stressed, must be our “touchstone” in interpreting and applying that provision. 594 U. S., at ___ (slip op., at 15).

When the race of one group is the predominant factor in the creation of a district, that district goes beyond making the electoral process equally open to the members of the group in question. It gives the members of that group an advantage that §2 does not require and that the Constitution may forbid. And because the creation of majority- minority districts is something of a zero-sum endeavor, giving an advantage to one minority group may disadvantage others.

C

What all this means is that a §2 plaintiff who claims that a districting map violates §2 because it fails to include an additional majority-minority district must show at the outset that such a district can be created without making race the predominant factor in its creation. The plaintiff bears both the burden of production and the burden of persuasion on this issue, see Voinovich v. Quilter, 507 U.S. 146, 155–156 (1993); White v. Regester, 412 U.S. 755, 766 (1973), but a plaintiff can satisfy the former burden simply by adducing evidence—in any acceptable form—that race did not predominate.

A plaintiff need not offer computer-related evidence. Once upon a time, legislative maps were drawn without using a computer, and nothing prevents a §2 plaintiff from taking this old-school approach in creating an illustrative district. See, e.g., M. Altman, K. McDonald, & M. McDonald, From Crayons to Computers: The Evolution of Computer Use in Redistricting, 23 Soc. Sci. Computer Rev. 334, 335–336 (2005). In that event, the plaintiff can simply call upon the mapmaker to testify about the process he or she used and the role, if any, that race played in that process. The defendant may seek to refute that testimony in any way that the rules of civil procedure and evidence allow.

If, as will often be the case today, a §2 plaintiff ’s mapmaker uses a computer program, the expert can testify about the weight, if any, that the program gives to race. The plaintiff will presumably argue that any role assigned to race was not predominant, and the defendant can contest this by cross-examining the plaintiff ’s expert, seeking the actual program in discovery, and calling its own expert to testify about the program’s treatment of race. After this, the trial court will be in a position to determine whether the program gave race a “predominant” role.

This is an entirely workable scheme. It does not obligate either party to offer computer evidence, and it minimizes the likelihood of a clash between what §2 requires and what the Constitution forbids. We have long assumed that §2 is consistent with the Constitution. See, e.g., Cooper, 581 U. S., at 301 (assuming States have a compelling interest in complying with §2); Shaw II, 517 U. S., at 915 (same); Vera, 517 U. S., at 977 (plurality opinion) (same). But that cannot mean that every conceivable interpretation of §2 is constitutional, and I do not understand the majority’s analysis of Alabama’s constitutional claim to suggest otherwise. Ante, at 33–34; ante, at 4 (Kavanaugh, J., concurring in part).

Our cases make it perfectly clear that using race as a “predominant factor” in drawing legislative districts is unconstitutional unless the stringent requirements of strict scrutiny can be satisfied,[2] and therefore if §2 can be found to require the adoption of an additional majority-minority district that was created under a process that assigned race a “predominant” role, §2 and the Constitution would be headed for a collision.

II

When the meaning of a “reasonably configured” district is properly understood, it is apparent that the decisions below must be vacated and that the cases must be remanded for the application of the proper test. In its analysis of whether the plaintiffs satisfied the first Gingles precondition, the District Court gave much attention to some traditional districting criteria—specifically, compactness and avoiding the splitting of political subdivisions and communities of interest—but it failed to consider whether the plaintiffs had shown that their illustrative districts were created without giving race a “predominant role.” Singleton v. Merrill, 582 F. Supp. 3d 924, 1008–1016 (ND Ala. 2022). For this reason, the District Court’s §2 analysis was deficient.

It is true that the District Court addressed the question of race-predominance when it discussed and rejected the State’s argument that the plaintiffs’ maps violated the Equal Protection Clause, but the court’s understanding of predominance was deeply flawed. The court began this part of its opinion with this revealing statement:

“Dr. Duchin and Mr. Cooper [plaintiffs’ experts] testified that they prioritized race only for the purpose of determining and to the extent necessary to determine whether it was possible for the Milligan plaintiffs and the Caster plaintiffs to state a Section Two claim. As soon as they determined the answer to that question, they assigned greater weight to other traditional redistricting criteria.” Id., at 1029–1030 (emphasis added).

This statement overlooks the obvious point that by “prioritiz[ing] race” at the outset, Dr. Duchin and Mr. Cooper gave race a predominant role.

The next step in the District Court’s analysis was even more troubling. The court wrote, “Dr. Duchin’s testimony that she considered two majority-Black districts as ‘non- negotiable’ does not” show that race played a predominant role in her districting process. Id., at 1030. But if achieving a certain objective is “non-negotiable,” then achieving that objective will necessarily play a predominant role. Suppose that a couple are relocating to the Washington, D. C., metropolitan area, and suppose that one says to the other, “I’m flexible about where we live, but it has to be in Maryland. That’s non-negotiable.” Could anyone say that finding a home in Maryland was not a “predominant” factor in the couple’s search? Or suppose that a person looking for a flight tells a travel agent, “It has to be non-stop. That’s non-negotiable.” Could it be said that the number of stops between the city of origin and the destination was not a “predominant” factor in the search for a good flight? The obvious answer to both these questions is no, and the same is true about the role of race in the creation of a new district. If it is “non-negotiable” that the district be majority black, then race is given a predominant role.

The District Court wrapped up this portion of its opinion with a passage that highlighted its misunderstanding of the first Gingles precondition. The court thought that a §2 plaintiff cannot proffer a reasonably configured majority-minority district without first attempting to see if it is possible to create such a district—that is, by first making the identification of such a district “non-negotiable.” Ibid. But that is simply not so. A plaintiff ’s expert can first create maps using only criteria that do not give race a predominant role and then determine how many contain the desired number of majority-minority districts.

One final observation about the District Court’s opinion is in order. The opinion gives substantial weight to the disparity between the percentage of majority-black House districts in the legislature’s plan (14%) and the percentage of black voting-age Alabamians (27%), while the percentage in the plaintiffs’ plan (29%) came closer to that 27% mark. See, e.g., id., at 946, 1016, 1018, 1025–1026; see also id., at 958–959, 969, 976, 982, 991–992, 996–997. Section 2 of the VRA, however, states expressly that no group has a right to representation “in numbers equal to their proportion in the population.” 52 U. S. C. §10301(b). This provision was a critical component of the compromise that led to the adoption of the 1982 amendments, as the Court unanimously agreed two Terms ago. See Brnovich, 594 U. S., at ___, and n. 14 (slip op., at 22, and n. 14); id., at ___, n. 6 (Kagan, J., dissenting) (slip op., at 19, n. 6). The District Court’s reasoning contravened this statutory proviso. See ante, at 11–12, 28–30 (Thomas, J., dissenting).

III

The Court spends much of its opinion attacking what it takes to be the argument that Alabama has advanced in this litigation. I will not debate whether the Court’s characterization of that argument is entirely correct, but as applied to the analysis I have just set out, the Court’s criticisms miss the mark.

A

The major theme of this part of the Court’s opinion is that Alabama’s argument, in effect, is that “Gingles must be overruled.” Ante, at 25. But as I wrote at the beginning of this opinion, I would decide these cases under the Gingles framework. We should recognize, however, that the Gingles framework is not the same thing as a statutory provision, and it is a mistake to regard it as such. National Pork Producers Council v. Ross, 598 U. S. ___, ___ (2023) (slip op., at 9) (“[T]he language of an opinion is not always to be parsed as though we were dealing with language of a statute” (quoting Reiter v. Sonotone Corp., 442 U.S. 330, 341 (1979))). In applying that framework today, we should keep in mind subsequent developments in our case law.

One important development has been a sharpening of the methodology used in interpreting statutes. Gingles was decided at a time when the Court’s statutory interpretation decisions sometimes paid less attention to the actual text of the statute than to its legislative history, and Gingles falls into that category. The Court quoted §2 but then moved briskly to the Senate Report. See 478 U. S., at 36–37, 43, and n. 7. Today, our statutory interpretation decisions focus squarely on the statutory text. National Assn. of Mfrs. v. Department of Defense, 583 U.S. 109, 127 (2018); Puerto Rico v. Franklin Cal. Tax-Free Trust, 579 U.S. 115, 125 (2016); cf. Brnovich, 594 U. S., at ___ (slip op., at 14). And as we held in Brnovich, “[t]he key requirement” set out in the text of §2 is that a State’s electoral process must be “ ‘equally open’ ” to members of all racial groups. Id., at ___ (slip op., at 15). The Gingles framework should be interpreted in a way that gives effect to this standard.

Another development that we should not ignore concerns our case law on racial predominance. Post-Gingles decisions like Miller, 515 U. S., at 920, Shaw II, 517 U. S., at 906–907, and Vera, 517 U. S., at 979 (plurality opinion), made it clear that it is unconstitutional to use race as a “predominant” factor in legislative districting. “[W]hen statutory language is susceptible of multiple interpretations, a court may shun an interpretation that raises serious constitutional doubts and instead may adopt an alternative that avoids those problems.” Jennings v. Rodriguez, 583 U. S. ___, ___ (2018) (slip op., at 2). This same principle logically applies with even greater force when we interpret language in one of our prior opinions. It therefore goes without question that we should apply the Gingles framework in a way that does not set up a confrontation between §2 and the Constitution, and understanding the first Gingles precondition in the way I have outlined achieves that result.[3]

B

The Court’s subsidiary criticisms of Alabama’s arguments are likewise inapplicable to my analysis. The Court suggests that the “centerpiece” of Alabama’s argument regarding the role race can permissibly play in a plaintiff ’s illustrative map seeks the imposition of “a new rule.” Ante, at 15, 22. But I would require only what our cases already demand: that all legislative districts be produced without giving race a “predominant” role.[4]

The Court maintains that Alabama’s benchmark scheme would be unworkable because of the huge number of different race-neutral maps that could be drawn. As the Court notes, there are apparently numerous “competing metrics on the issue of compactness” alone, and each race-neutral computer program may assign different values to each traditional districting criterion. Ante, at 27 (internal quotation marks omitted).

My analysis does not create such problems. If a §2 plaintiff chooses to use a computer program to create an illustrative district, the court need ask only whether that program assigned race a predominant role.

The Court argues that Alabama’s focus on race-neutral maps cannot be squared with a totality-of-the- circumstances test because “Alabama suggests there is only one ‘circumstance[ ]’ that matters—how the State’s map stacks up relative to the benchmark” maps. Ante, at 18. My analysis, however, simply follows the Gingles framework, under which a court must first determine whether a §2 plaintiff has satisfied three “preconditions” before moving on to consider the remainder of relevant circumstances. See Growe v. Emison, 507 U.S. 25, 40–41 (1993) (unless plaintiffs establish all three preconditions, there “neither has been a wrong nor can be a remedy”).

IV

As noted, I would vacate and remand for the District Court to apply the correct understanding of Gingles in the first instance. Such a remand would require the District Court to determine whether the plaintiffs have shown that their illustrative maps did not give race a predominant role, and I will therefore comment briefly on my understanding of the relevant evidence in the record as it now stands.

A

In my view, there is strong evidence that race played a predominant role in the production of the plaintiffs’ illustrative maps and that it is most unlikely that a map with more than one majority-black district could be created without giving race such a role. An expert hired by the Milligan plaintiffs, Dr. Kosuke Imai, used a computer algorithm to create 30,000 potential maps, none of which contained two majority-black districts. See 2 App. 571–572; Supp. App. 59, 72. In fact, in 20,000 of those simulations, Dr. Imai intentionally created one majority-minority district, and yet even with one majority-minority district guaranteed as a baseline, none of those 20,000 attempts produced a second one. See 2 App. 571–572; Supp. App. 72.

Similarly, Dr. Moon Duchin, another expert hired by the Milligan plaintiffs, opined that “it is hard to draw two majority-black districts by accident.” 2 App. 714. Dr. Duchin also referred to a study where she generated two million maps of potential district configurations in Alabama, none of which contained a second majority-minority district. Id., at 710. And the first team of trained mapmakers that plaintiff Milligan consulted was literally unable to draw a two-majority-black-district map, even when they tried. Id., at 511–512. Milligan concluded at the time that the feat was impossible. Id., at 512.

The majority quibbles about the strength of this evidence, protesting that Dr. Imai’s studies failed to include as controls certain redistricting criteria and that Dr. Duchin’s two-million-map study was based on 2010 census data, see ante, at 26–27, and nn. 6–7, but this is unconvincing for several reasons. It is plaintiffs’ burden to produce evidence and satisfy the Gingles preconditions, so if their experts’ maps were deficient, that is no strike against Alabama. And the racial demographics of the State changed little between 2010 and 2020, Supp. App. 82, which is presumably why Dr. Duchin herself raised the older study in answering questions about her work in this litigation, see 2 App. 710. If it was impossible to draw two such districts in 2010, it surely at least requires a great deal of intentional effort now.

The Court suggests that little can be inferred from Dr. Duchin’s two-million-map study because two million maps are not that many in comparison to the “trillion trillion” maps that are possible. See ante, at 28–29, and n. 9. In making this argument, the Court relies entirely on an amicus brief submitted by three computational redistricting experts in support of the appellees. See Brief for Computational Redistricting Experts 2, 6, n. 7. These experts’ argument concerns a complicated statistical issue, and I think it is unwise for the Court to make their argument part of our case law based solely on this brief. By the time this amicus brief was submitted, the appellants had already filed their main brief, and it was too late for any experts with contrary views to submit an amicus brief in support of appellants. Computer simulations are widely used today to make predictions about many important matters, and I would not place stringent limits on their use in VRA litigation without being quite sure of our ground. If the cases were remanded, the parties could take up this issue if they wished and call experts to support their positions on the extent to which the two million maps in the study are or can be probative of the full universe of maps.

In sum, based on my understanding of the current record, I am doubtful that the plaintiffs could get by the first Gingles precondition, but I would let the District Court sort this matter out on remand.

B

Despite the strong evidence that two majority-minority districts cannot be drawn without singular emphasis on race, a plurality nonetheless concludes that race did not predominate in the drawing of the plaintiffs’ illustrative maps. See ante, at 22–25. Their conclusion, however, rests on a faulty view of what non-predominance means.

The plurality’s position seems to be that race does not predominate in the creation of a districting map so long as the map does not violate other traditional districting criteria such as compactness, contiguity, equally populated districts, minimizing county splits, etc. Ibid. But this conclusion is irreconcilable with our cases. In Miller, for instance, we acknowledged that the particular district at issue was not “shape[d] . . . bizarre[ly] on its face,” but we nonetheless held that race predominated because of the legislature’s “overriding desire to assign black populations” in a way that would create an additional “majority-black district.” 515 U. S., at 917.

Later cases drove home the point that conformity with traditional districting principles does not necessarily mean that a district was created without giving race a predominant role. In Cooper, we held that once it was shown that race was “ ‘the overriding reason’ ” for the selection of a particular map, “a further showing of ‘inconsistency between the enacted plan and traditional redistricting criteria’ is unnecessary to a finding of racial predominance.” 581 U. S., at 301, n. 3 (quoting Bethune-Hill, 580 U. S., at 190). We noted that the contrary argument was “foreclosed almost as soon as it was raised in this Court.” Cooper, 581 U. S., at 301, n. 3; see also Vera, 517 U. S., at 966 (plurality opinion) (race may still predominate even if “traditional districting principle[s] do correlate to some extent with the district’s layout”). “Traditional redistricting principles . . . are numerous and malleable. . . . By deploying those factors in various combinations and permutations, a [mapmaker] could construct a plethora of potential maps that look consistent with traditional, race-neutral principles.” Bethune-Hill, 580 U. S., at 190. Here, a plurality allows plaintiffs to do precisely what we warned against in Bethune-Hill.

The plurality’s analysis of predominance contravenes our precedents in another way. We have been sensitive to the gravity of “ ‘trapp[ing]’ ” States “ ‘between the competing hazards of liability’ ” imposed by the Constitution and the VRA. Id., at 196 (quoting Vera, 517 U. S., at 977). The VRA’s demand that States not unintentionally “dilute” the votes of particular groups must be reconciled with the Constitution’s demand that States generally avoid intentional augmentation of the political power of any one racial group (and thus the diminution of the power of other groups). The plurality’s predominance analysis shreds that prudential concern. If a private plaintiff can demonstrate §2 liability based on the production of a map that the State has every reason to believe it could not constitutionally draw, we have left “state legislatures too little breathing room” and virtually guaranteed that they will be on the losing end of a federal court’s judgment. Bethune-Hill, 580 U. S., at 196.

*  *  *

The Court’s treatment of Gingles is inconsistent with the text of §2, our precedents on racial predominance, and the fundamental principle that States are almost always prohibited from basing decisions on race. Today’s decision unnecessarily sets the VRA on a perilous and unfortunate path. I respectfully dissent.

SUPREME COURT OF THE UNITED STATES

_________________

Nos. 21–1086 and 21–1087

_________________

WES ALLEN, ALABAMA SECRETARY OF STATE, et al., APPELLANTS

21–1086v.

EVAN MILLIGAN, et al.

on appeal from the united states district court for the northern district of alabama

WES ALLEN, ALABAMA SECRETARY OF STATE, et al., PETITIONERS

21–1087v.

MARCUS CASTER, et al.

on writ of certiorari before judgment to the united states court of appeals for the eleventh circuit

[June 8, 2023]

Justice Thomas, with whom Justice Gorsuch joins, with whom Justice Barrett joins as to Parts II and III, and with whom Justice Alito joins as to Parts II–A and II–B, dissenting.

These cases “are yet another installment in the ‘disastrous misadventure’ of this Court’s voting rights jurisprudence.” Alabama Legislative Black Caucus v. Alabama, 575 U.S. 254, 294 (2015) (Thomas, J., dissenting) (quoting Holder v. Hall, 512 U.S. 874, 893 (1994) (Thomas, J., concurring in judgment)). What distinguishes them is the uncommon clarity with which they lay bare the gulf between our “color-blind” Constitution, Plessy v. Ferguson, 163 U.S. 537, 559 (1896) (Harlan, J., dissenting), and “the consciously segregated districting system currently being constructed in the name of the Voting Rights Act.” Holder, 512 U. S., at 907 (opinion of Thomas, J.). The question presented is whether §2 of the Act, as amended, requires the State of Alabama to intentionally redraw its longstanding congressional districts so that black voters can control a number of seats roughly proportional to the black share of the State’s population. Section 2 demands no such thing, and, if it did, the Constitution would not permit it.

I

At the outset, I would resolve these cases in a way that would not require the Federal Judiciary to decide the correct racial apportionment of Alabama’s congressional seats. Under the statutory text, a §2 challenge must target a “voting qualification or prerequisite to voting or standard, practice, or procedure.” 52 U. S. C. §10301(a). I have long been convinced that those words reach only “enactments that regulate citizens’ access to the ballot or the processes for counting a ballot”; they “do not include a State’s . . . choice of one districting scheme over another.” Holder, 512 U. S., at 945 (opinion of Thomas, J.). “Thus, §2 cannot provide a basis for invalidating any district.” Abbott v. Perez, 585 U. S. ___, ___ (2018) (Thomas, J., concurring) (slip op., at 1).

While I will not repeat all the arguments that led me to this conclusion nearly three decades ago, see Holder, 512 U. S., at 914–930 (opinion concurring in judgment), the Court’s belated appeal to the statutory text is not persuasive. See ante, at 31–32. Whatever words like “practice” and “procedure” are capable of meaning in a vacuum, the prohibitions of §2 apply to practices and procedures that affect “voting” and “the right . . . to vote.” §10301(a). “Vote” and “voting” are defined terms under the Act, and the Act’s definition plainly focuses on ballot access and counting:

“The terms ‘vote’ or ‘voting’ shall include all action necessary to make a vote effective in any primary, special, or general election, including, but not limited to, registration, listing pursuant to this chapter, or other action required by law prerequisite to voting, casting a ballot, and having such ballot counted properly and included in the appropriate totals of votes cast with respect to candidates for public or party office and propositions for which votes are received in an election.” §10310(c)(1).

In enacting the original Voting Rights Act in 1965, Congress copied this definition almost verbatim from Title VI of the Civil Rights Act of 1960—a law designed to protect access to the ballot in jurisdictions with patterns or practices of denying such access based on race, and which cannot be construed to authorize so-called vote-dilution claims. See 74Stat. 91–92 (codified in relevant part at 52 U. S. C. §10101(e)). Title I of the Civil Rights Act of 1964, which cross-referenced the 1960 Act’s definition of “vote,” likewise protects ballot access alone and cannot be read to address vote dilution. See 78Stat. 241 (codified in relevant part at 52 U. S. C. §10101(a)). Tellingly, the 1964 Act also used the words “standard, practice, or procedure” to refer specifically to voting qualifications for individuals and the actions of state and local officials in administering such requirements.[1] Our entire enterprise of applying §2 to districting rests on systematic neglect of these statutory antecedents and, more broadly, of the ballot-access focus of the 1960s’ voting-rights struggles. See, e.g., Brnovich v. Democratic National Committee, 594 U. S. ___, ___ (2021) (slip op., at 2) (describing the “notorious methods” by which, prior to the

Voting Rights Act, States and localities deprived black Americans of the ballot: “poll taxes, literacy tests, property qualifications, white primaries, and grandfather clauses” (alterations and internal quotation marks omitted)).[2]

Moreover, the majority drastically overstates the stare decisis support for applying §2 to single-member districting plans like the one at issue here.[3] As the majority implicitly acknowledges, this Court has only applied §2 to invalidate one single-member district in one case. See League of United Latin American Citizens v. Perry, 548 U.S. 399, 447 (2006) (LULAC) (opinion of Kennedy, J.). And no party in that case argued that the plaintiffs’ vote-dilution claim was not cognizable. As for Growe v. Emison, 507 U.S. 25 (1993), it held only that the threshold preconditions for challenging multimember and at-large plans must limit challenges to single-member districts with at least the same force, as “[i]t would be peculiar [if] a vote-dilution challenge to the (more dangerous) multimember district require[d] a higher threshold showing than a vote-fragmentation challenge to a single-member district.” Id., at 40. Growe did not consider (or, thus, reject) an argument that §2 does not apply to single-member districts.

In any event, stare decisis should be no barrier to reconsidering a line of cases that “was based on a flawed method of statutory construction from its inception,” has proved incapable of principled application after nearly four decades of experience, and puts federal courts in the business of “methodically carving the country into racially designated electoral districts.” Holder, 512 U. S., at 945 (opinion of Thomas, J.). This Court has “never applied stare decisis mechanically to prohibit overruling our earlier decisions determining the meaning of statutes,” and it should not do so here. Monell v. New York City Dept. of Social Servs., 436 U.S. 658, 695 (1978). Stare decisis did not save “separate but equal,” despite its repeated reaffirmation in this Court and the pervasive reliance States had placed upon it for decades. See, e.g., Brief for Appellees in Brown v. Board of Education, O. T. 1953, No. 1, pp. 18–30. It should not rescue modern-day forms of de jure racial balkanization—which, as these cases show, is exactly where our §2 vote-dilution jurisprudence has led.[4]

II

Even if §2 applies here, however, Alabama should prevail. The District Court found that Alabama’s congressional districting map “dilutes” black residents’ votes because, while it is possible to draw two majority-black districts, Alabama’s map only has one.[5] But the critical question in all vote-dilution cases is: “Diluted relative to what benchmark?” Gonzalez v. Aurora, 535 F.3d 594, 598 (CA7 2008) (Easterbrook, C. J.). Neither the District Court nor the majority has any defensible answer. The text of §2 and the logic of vote-dilution claims require a meaningfully race-neutral benchmark, and no race-neutral benchmark can justify the District Court’s finding of vote dilution in these cases. The only benchmark that can justify it—and the one that the District Court demonstrably applied—is the decidedly nonneutral benchmark of proportional allocation of political power based on race.

A

As we have long recognized, “the very concept of vote dilution implies—and, indeed, necessitates—the existence of an ‘undiluted’ practice against which the fact of dilution may be measured.” Reno v. Bossier Parish School Bd., 520 U.S. 471, 480 (1997). In a challenge to a districting plan, a court must be able to compare a State’s enacted plan with “a hypothetical, undiluted plan,” ibid., ascertained by an “objective and workable standard.” Holder, 512 U. S., at 881 (plurality opinion); see also id., at 887 (opinion of O’Connor, J.) (noting the “general agreement” on this point).

To be sure, it is no easy task to identify an objective, “undiluted” benchmark against which to judge a districting plan. As we recently held in the analogous context of partisan gerrymandering, “federal courts are not equipped to apportion political power as a matter of fairness.” Rucho v. Common Cause, 588 U. S. ___, ___ (2019) (slip op., at 17). Yet §2 vote-dilution cases require nothing less. If §2 prohibited only intentional racial discrimination, there would be no difficulty in finding a clear and workable rule of decision. But the “results test” that Congress wrote into §2 to supersede Mobile v. Bolden, 446 U.S. 55 (1980), eschews intent as the criterion of liability. See Bossier Parish School Bd., 520 U. S., at 482. Accordingly, a §2 vote-dilution claim does not simply “as[k] . . . for the elimination of a racial classification.” Rucho, 588 U. S., at ___ (slip op., at 21). It asks, instead, “for a fair share of political power and influence, with all the justiciability conundrums that entails.” Ibid. Nevertheless, if §2 applies to single-member districts, we must accept that some “objective and workable standard for choosing a reasonable benchmark” exists; otherwise, single-member districts “cannot be challenged as dilutive under §2.” Holder, 512 U. S., at 881 (plurality opinion).

Given the diverse circumstances of different jurisdictions, it would be fanciful to expect a one-size-fits-all definition of the appropriate benchmark. Cf. Thornburg v. Gingles, 478 U.S. 30, 79 (1986) (explaining that the vote-dilution inquiry “is peculiarly dependent upon the facts of each case and requires an intensely local appraisal” (citation and internal quotation marks omitted)). One overriding principle, however, should be obvious. A proper districting benchmark must be race neutral: It must not assume, a priori, that an acceptable plan should include any particular number or proportion of minority-controlled districts.

I begin with §2’s text. As relevant here, §2(a) prohibits a State from “impos[ing] or appl[ying]” any electoral rule “in a manner which results in a denial or abridgement of the right . . . to vote on account of race or color.” §10301(a). Section 2(b) then provides that §2(a) is violated

“if, based on the totality of circumstances, . . . the political processes leading to nomination or election in the State . . . are not equally open to participation by members of [a protected class] in that its members have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice. The extent to which members of a protected class have been elected to office in the State . . Provided, That nothing in this section establishes a right to have members of a protected class elected in numbers equal to their proportion in the population.” §10301(b).

As we held two Terms ago in Brnovich, the “equal openness” requirement is “the core” and “touchstone” of §2(b),

with “equal opportunity” serving an ancillary function.[6] 594 U. S., at ___ (slip op., at 15). Relying significantly on §2(b)’s disclaimer of a right to proportional representation, we also held that §2 does not enact a “freewheeling disparate-impact regime.” Id., at ___, and n. 14 (slip op., at 22, and n. 14). Brnovich further stressed the value of “benchmarks with which . . . challenged [electoral] rule[s] can be compared,” id., at ___ (slip op., at 17), and that “a meaningful comparison is essential” in judging the significance of any challenged scheme’s racially disparate impact. Id., at ___ (slip op., at 18). To the extent §2 applies to districting plans, then, it requires that they be “equally open to participation” by voters of all races, but it is not a pure disparate-impact statute and does not guarantee proportional representation.

In its main argument here, Alabama simply carries these principles to their logical conclusion: Any vote-dilution benchmark must be race neutral. See Brief for Appellants 32–46. Whatever “equal openness” means in the context of single-member districting, no “meaningful comparison” is possible using a benchmark that builds in a presumption in favor of minority-controlled districts. Indeed, any benchmark other than a race-neutral one would render the vote-dilution inquiry fundamentally circular, allowing courts to conclude that a districting plan “dilutes” a minority’s voting strength “on account of race” merely because it does not measure up to an ideal already defined in racial terms. Such a question-begging standard would not answer our precedents’ demand for an “objective,” “reasonable benchmark.” Holder, 512 U. S., at 881 (plurality opinion) (emphasis added). Nor could any nonneutral benchmark be reconciled with Brnovich’s rejection of a disparate-impact regime or the text’s disclaimer of a right to proportional representation. 594 U. S., at ___, and n. 14 (slip op., at 22, and n. 14).

There is yet another compelling reason to insist on a race-neutral benchmark. “The Constitution abhors classifications based on race.” Grutter v. Bollinger, 539 U.S. 306, 353 (2003) (Thomas, J., concurring in part and dissenting in part). Redistricting is no exception. “Just as the State may not, absent extraordinary justification, segregate citizens on the basis of race in its public parks, buses, golf courses, beaches, and schools,” the State also “may not separate its citizens into different voting districts on the basis of race.” Miller v. Johnson, 515 U.S. 900, 911 (1995) (citations omitted). “[D]istricting maps that sort voters on the basis of race ‘ “are by their very nature odious.” ’ ” Wisconsin Legislature v. Wisconsin Elections Comm’n, 595 U. S. ___, ___ (2022) (per curiam) (slip op., at 2) (quoting Shaw v. Reno, 509 U.S. 630, 643 (1993) (Shaw I)). Accordingly, our precedents apply strict scrutiny whenever race was “the predominant factor motivating [the placement of] a significant number of voters within or without a particular district,” Miller, 515 U. S., at 916, or, put another way, whenever “[r]ace was the criterion that . . . could not be compromised” in a district’s formation. Shaw v. Hunt, 517 U.S. 899, 907 (1996) (Shaw II).

Because “[r]acial gerrymandering, even for remedial purposes, may balkanize us into competing racial factions” and undermine “the goal of a political system in which race no longer matters,” Shaw I, 509 U. S., at 657, our cases have long recognized the need to interpret §2 to avoid “unnecessarily infus[ing] race into virtually every redistricting” plan. LULAC, 548 U. S., at 446 (opinion of Kennedy, J.); accord, Bartlett v. Strickland, 556 U.S. 1, 21 (2009) (plurality opinion). Plainly, however, that “infusion” is the inevitable result of any race-based benchmark. Any interpretation of §2 that permits courts to condemn enacted districting plans as dilutive relative to a nonneutral benchmark “would result in a substantial increase in the number of mandatory districts drawn with race as ‘the predominant factor motivating the legislature’s decision,’ ” thus “ ‘raising serious constitutional questions.’ ” Id., at 21–22 (first quoting Miller, 515 U. S., at 916, then quoting LULAC, 548 U. S., at 446). To avoid setting §2 on a collision course with the Constitution, courts must apply a race-neutral benchmark in assessing any claim that a districting plan unlawfully dilutes a racial minority’s voting strength.

B

The plaintiffs in these cases seek a “proportional allocation of political power according to race.” Holder, 512 U. S., at 936 (opinion of Thomas, J.). According to the 2020 census, black Alabamians account for 27.16% of the State’s total population and 25.9% of its voting-age population, both figures slightly less than two-sevenths. Of Alabama’s seven existing congressional districts, one, District 7, is majority-black.[7] These cases were brought to compel “the creation of two majority-minority congressional districts”—roughly proportional control. 1 App. 135 (emphasis added); see also id., at 314 (“Plaintiffs seek an order . . . ordering a congressional redistricting plan that includes two majority-Black congressional districts”).

Remarkably, the majority fails to acknowledge that two minority-controlled districts would mean proportionality, or even that black Alabamians are about two-sevenths of the State. Yet that context is critical to the issues before us, not least because it explains the extent of the racial sorting the plaintiffs’ goal would require. “[A]s a matter of mathematics,” single-member districting “tends to deal out representation far short of proportionality to virtually all minorities, from environmentalists in Alaska to Republicans in Massachusetts.” M. Duchin & D. Spencer, Models, Race, and the Law, 130 Yale L. J. Forum 744, 752 (2021) (Duchin & Spencer). As such, creating two majority-black districts would require Alabama to aggressively “sort voters on the basis of race.” Wisconsin Legislature, 595 U. S., at ___ (slip op., at 2).

The plaintiffs’ 11 illustrative maps make that clear. All 11 maps refashion existing District 2 into a majority-black district while preserving the current black majority in District 7. They all follow the same approach: Starting with majority-black areas of populous Montgomery County, they

expand District 2 east and west to encompass predominantly majority-black areas throughout the rural “Black Belt.” In the process, the plans are careful to leave enough of the Black Belt for District 7 to maintain its black majority. Then—and critically—the plans have District 2 extend a southwestern tendril into Mobile County to capture a dense, high-population majority-black cluster in urban Mobile.[8] See Supp. App. 184, 186, 188, 190, 193, 195, 197, 199, 201, 203; see also id., at 149.

Those black Mobilians currently reside in the urban heart of District 1. For 50 years, District 1 has occupied the southwestern pocket of Alabama, consisting of the State’s two populous Gulf Coast counties (Mobile and Baldwin) as well as some less populous areas to the immediate north and east. See id., at 205–211. It is indisputable that the Gulf Coast region is the sort of community of interest that the Alabama Legislature might reasonably think a congressional district should be built around. It contains Alabama’s only coastline, its fourth largest city, and the Port of Mobile. Its physical geography runs north along the Alabama and Mobile Rivers, whose paths District 1 follows. Its economy is tied to the Gulf—to shipping, shipbuilding, tourism, and commercial fishing. See Brief for Coastal Alabama Partnership as Amicus Curiae 13–15.

But, for the plaintiffs to secure their majority-black District 2, this longstanding, compact, and eminently sensible district must be radically transformed. In the Gulf Coast region, the newly drawn District 1 would retain only the majority-white areas that District 2 did not absorb on its path to Mobile’s large majority-black population. To make up the lost population, District 1 would have to extend eastward through largely majority-white rural counties along the length of Alabama’s border with the Florida panhandle. The plaintiffs do not assert that white residents on the Gulf Coast have anything special in common with white residents in those communities, and the District Court made no such finding. The plaintiffs’ maps would thus reduce District 1 to the leftover white communities of the southern fringe of the State, its shape and constituents defined almost entirely by the need to make District 2 majority-black while also retaining a majority-black District 7.

The plaintiffs’ mapmaking experts left little doubt that their plans prioritized race over neutral districting criteria. Dr. Moon Duchin, who devised four of the plans, testified that achieving “two majority-black districts” was a “nonnegotiable principl[e]” in her eyes, a status shared only by our precedents’ “population balance” requirement. 2 App. 634; see also id., at 665, 678. Only “after” those two “nonnegotiable[s]” were satisfied did Dr. Duchin then give lower priority to “contiguity” and “compactness.” Id., at 634. The architect of the other seven maps, William Cooper, considered “minority voting strengt[h]” a “traditional redistricting principl[e]” in its own right, id., at 591, and treated “the minority population in and of itself ” as the paramount community of interest in his plans, id., at 601.

Statistical evidence also underscored the illustrative maps’ extreme racial sorting. Another of the plaintiffs’ experts, Dr. Kosuke Imai, computer generated 10,000 districting plans using a race-blind algorithm programmed to observe several objective districting criteria. Supp. App. 58–59. None of those plans contained even one majority-black district. Id., at 61. Dr. Imai generated another 20,000 plans using the same algorithm, but with the additional constraint that they must contain at least one majority- black district; none of those plans contained a second majority-black district, or even a second district with a black voting-age population above 40%. Id., at 54, 67, 71–72. In a similar vein, Dr. Duchin testified about an academic study in which she had randomly “generated 2 million districting plans for Alabama” using a race-neutral algorithm that gave priority to compactness and contiguity. 2 App. 710; see Duchin & Spencer 765. She “found some [plans] with one majority-black district, but never found a second . . . majority-black district in 2 million attempts.” 2 App. 710. “[T]hat it is hard to draw two majority-black districts by accident,” Dr. Duchin explained, “show[ed] the importance of doing so on purpose.” Id., at 714.[9]

The plurality of Justices who join Part III–B–I of The Chief Justice’s opinion appear to agree that the plaintiffs could not prove the first precondition of their statewide vote-dilution claim—that black Alabamians could constitute a majority in two “reasonably configured” districts, Wisconsin Legislature, 595 U. S., at ___ (slip op., at 3)—by drawing an illustrative map in which race was predominant. See ante, at 25. That should be the end of these cases, as the illustrative maps here are palpable racial gerrymanders. The plaintiffs’ experts clearly applied “express racial target[s]” by setting out to create 50%-plus majority-black districts in both Districts 2 and 7. Bethune-Hill v. Virginia State Bd. of Elections, 580 U.S. 178, 192 (2017). And it is impossible to conceive of the State adopting the illustrative maps without pursuing the same racially motivated goals. Again, the maps’ key design features are: (1) making District 2 majority-black by connecting black residents in one metropolitan area (Montgomery) with parts of the rural Black Belt and black residents in another metropolitan area (Mobile); (2) leaving enough of the Black Belt’s majority-black rural areas for District 7 to maintain its majority-black status; and (3) reducing District 1 to the white remainder of the southern third of the State.

If the State did this, we would call it a racial gerrymander, and rightly so. We would have no difficulty recognizing race as “the predominant factor motivating [the placement of] significant number[s] of voters within or without” Districts 1, 2, and 7. Miller, 515 U. S., at 916. The “stark splits in the racial composition of populations moved into and out of ” Districts 1 and 2 would make that obvious. Bethune-Hill, 580 U. S., at 192. So would the manifest absence of any nonracial justification for the new District 1. And so would the State’s clear intent to ensure that both Districts 2 and 7 hit their preordained racial targets. See ibid. (noting that “pursu[it of] a common redistricting policy toward multiple districts” may show predominance). That the plan delivered proportional control for a particular minority—a statistical anomaly that over 2 million race-blind simulations did not yield and 20,000 race-conscious simulations did not even approximate—would be still further confirmation.

The State could not justify such a plan simply by arguing that it was less bizarre to the naked eye than other, more elaborate racial gerrymanders we have encountered. See ante, at 19–20 (discussing cases). As we held in Miller, visual “bizarreness” is not “a necessary element of the constitutional wrong,” only “persuasive circumstantial evidence.” 515 U. S., at 912–913.[10]

Nor could such a plan be explained by supposed respect for the Black Belt. For present purposes, I accept the District Court’s finding that the Black Belt is a significant community of interest. But the entire black population of the Black Belt—some 300,000 black residents, see Supp. App. 33—is too small to provide a majority in a single congressional district, let alone two.[11] The black residents needed to populate majority-black versions of Districts 2 and 7 are overwhelmingly concentrated in the urban counties of Jefferson (i.e., the Birmingham metropolitan area, with about 290,000 black residents), Mobile (about 152,000 black residents), and Montgomery (about 134,000 black residents). Id., at 83. Of the three, only Montgomery County is in the Black Belt. The plaintiffs’ maps, therefore, cannot and do not achieve their goal of two majority-black districts by “join[ing] together” the Black Belt, as the majority seems wrongly to believe. Ante, at 13. Rather, their majority-black districts are anchored by three separate high-density clusters of black residents in three separate metropolitan areas, two of them outside the Black Belt. The Black Belt’s largely rural remainder is then divided between the two districts to the extent needed to fill out their population numbers with black majorities in both. Respect for the Black Belt as a community of interest cannot explain this approach. The only explanation is the plaintiffs’ express racial target: two majority-black districts and statewide proportionality.

The District Court nonetheless found that race did not predominate in the plaintiffs’ illustrative maps because Dr. Duchin and Mr. Cooper “prioritized race only as necessary . . . to draw two reasonably compact majority-Black congressional districts,” as opposed to “maximiz[ing] the number of majority-Black districts, or the BVAP [black voting-age population] in any particular majority-Black district.” Singleton v. Merrill, 582 F. Supp. 3d 924, 1029–1030 (ND Ala. 2022) (per curiam). This reasoning shows a profound misunderstanding of our racial-gerrymandering precedents. As explained above, what triggers strict scrutiny is the intentional use of a racial classification in placing “a significant number of voters within or without a particular district.” Miller, 515 U. S., at 916. Thus, any plan whose predominant purpose is to achieve a nonnegotiable, predetermined racial target in a nonnegotiable, predetermined number of districts is a racial gerrymander subject to strict scrutiny. The precise fraction used as the racial target, and the number of districts it is applied to, are irrelevant.

In affirming the District Court’s nonpredominance finding, the plurality glosses over these plain legal errors,[12] and it entirely ignores Dr. Duchin’s plans—presumably because her own explanation of her method sounds too much like textbook racial predominance. Compare 2 App. 634 (“[A]fter . . . what I took to be nonnegotiable principles of population balance and seeking two majority-black districts, after that, I took contiguity as a requirement and compactness as paramount” (emphasis added)) and id., at 635 (“I took . . . county integrity to take precedence over the level of [black voting-age population] once that level was past 50 percent” (emphasis added)), with Bethune-Hill, 580 U. S., at 189 (explaining that race predominates when it “ ‘was the criterion that . . . could not be compromised,’ and race-neutral considerations ‘came into play only after the race-based decision had been made’ ” (quoting Shaw II, 517 U. S., at 907)), and Miller, 515 U. S., at 916 (explaining that race predominates when “the [mapmaker] subordinated traditional race-neutral districting principles . . . to racial considerations”). The plurality thus affirms the District Court’s finding only in part and with regard to Mr. Cooper’s plans alone.

In doing so, the plurality acts as if the only relevant evidence were Mr. Cooper’s testimony about his own mental state and the State’s expert’s analysis of Mr. Cooper’s maps. See ante, at 23–24. Such a blinkered view of the issue is unjustifiable. All 11 illustrative maps follow the same approach to creating two majority-black districts. The essential design features of Mr. Cooper’s maps are indistinguishable from Dr. Duchin’s, and it is those very design features that would require race to predominate. None of the plaintiffs’ maps could possibly be drawn by a mapmaker who was merely “aware of,” rather than motivated by, “racial demographics.” Miller, 515 U. S., at 916. They could only ever be drawn by a mapmaker whose predominant motive was hitting the “express racial target” of two majority-black districts. Bethune-Hill, 580 U. S., at 192.[13]

The plurality endeavors in vain to blunt the force of this obvious fact. See ante, at 24–25. Contrary to the plurality’s apparent understanding, nothing in Bethune-Hill suggests that “an express racial target” is not highly probative evidence of racial predominance. 580 U. S., at 192 (placing “express racial target[s]” alongside “stark splits in the racial composition of [redistricted] populations” as “relevant districtwide evidence”). That the Bethune-Hill majority “decline[d]” to act as a “ ‘court of . . . first view,’ ” instead leaving the ultimate issue of predominance for remand, cannot be transmuted into such an implausible holding or, in truth, any holding at all. Id., at 193.

The plurality is also mistaken that my predominance analysis would doom every illustrative map a §2 plaintiff “ever adduced.” Ante, at 25 (emphasis deleted). Rather, it would mean only that—because §2 requires a race-neutral benchmark—plaintiffs cannot satisfy their threshold burden of showing a reasonably configured alternative plan with a proposal that could only be viewed as a racial gerrymander if enacted by the State. This rule would not bar a showing, in an appropriate case, that a State could create an additional majority-minority district through a reasonable redistricting process in which race did not predominate. It would, on the other hand, screen out efforts to use §2 to push racially proportional districting to the limits of what a State’s geography and demography make possible—the approach taken by the illustrative maps here.

C

The foregoing analysis should be enough to resolve these cases: If the plaintiffs have not shown that Alabama could create two majority-black districts without resorting to a racial gerrymander, they cannot have shown that Alabama’s one-majority-black-district map “dilutes” black Alabamians’ voting strength relative to any meaningfully race-neutral benchmark. The inverse, however, is not true: Even if it were possible to regard the illustrative maps as not requiring racial predominance, it would not necessarily follow that a two-majority-black-district map was an appropriate benchmark. All that might follow is that the illustrative maps were reasonably configured—in other words, that they were consistent with some reasonable application of traditional districting criteria in which race did not predominate. See LULAC, 548 U. S., at 433. But, in virtually all jurisdictions, there are countless possible districting schemes that could be considered reasonable in that sense. The mere fact that a plaintiff ’s illustrative map is one of them cannot justify making it the benchmark against which other plans should be judged. Cf. Rucho, 588 U. S., at ___–___ (slip op., at 19–20) (explaining the lack of judicially manageable standards for evaluating the relative fairness of different applications of traditional districting criteria).

That conceptual gap—between “reasonable” and “benchmark”—is highly relevant here. Suppose, for argument’s sake, that Alabama reasonably could decide to create two majority-black districts by (1) connecting Montgomery’s black residents with Mobile’s black residents, (2) dividing up the rural parts of the Black Belt between that district and another district with its population core in the majority-black parts of the Birmingham area, and (3) accepting the extreme disruption to District 1 and the Gulf Coast that this approach would require. The plaintiffs prefer that approach because it allows the creation of two majority-black districts, which they think Alabama should have. But even if that approach were reasonable, there is hardly any compelling race-neutral reason to elevate such a plan to a benchmark against which all other plans must be measured. Nothing in Alabama’s geography or demography makes it clearly the best way, or even a particularly attractive way, to draw three of seven equally populous districts. The State has obvious legitimate, race-neutral reasons to prefer its own map—most notably, its interest in “preserving the cores of prior districts” and the Gulf Coast community of interest in District 1. Karcher v. Daggett, 462 U.S. 725, 740 (1983). And even discounting those interests would not yield a race-neutral case for treating the plaintiffs’ approach as a suitable benchmark: Absent core retention, there is no apparent race-neutral reason to insist that District 7 remain a majority-black district uniting Birmingham’s majority-black neighborhoods with majority-black rural areas in the Black Belt.

Finally, it is surely probative that over 2 million race-neutral simulations did not yield a single plan with two majority-black districts, and even 20,000 simulations with a one-majority-black-district floor did not yield a second district with a black voting-age population over 40%. If any plan with two majority-black districts would be an “out-out-out-outlier” within the likely universe of race-neutral districting plans, Rucho, 588 U. S., at ___ (Kagan, J., dissenting) (slip op., at 19), it is hard to see how the mere possibility of drawing two majority-black districts could show that a one-district map diluted black Alabamians’ votes relative to any appropriate benchmark.[14]

D

Given all this, by what benchmark did the District Court find that Alabama’s enacted plan was dilutive? The answer is as simple as it is unlawful: The District Court applied a benchmark of proportional control based on race. To be sure, that benchmark was camouflaged by the elaborate vote-dilution framework we have inherited from Gingles. But nothing else in that framework or in the District Court’s reasoning supplies an alternative benchmark capable of explaining the District Court’s bottom line: that Alabama’s one-majority-black-district map dilutes black voters’ fair share of political power.

Under Gingles, the majority explains, there are three “preconditions” to a vote-dilution claim: (1) the relevant “minority group must be sufficiently large and geographically compact to constitute a majority in a reasonably configured district”; (2) the minority group must be “politically cohesive”; and (3) the majority group must “vot[e] sufficiently as a bloc to enable it to defeat the minority’s preferred candidate[s].” Ante, at 10 (alterations and internal quotation marks omitted). If these preconditions are satisfied, Gingles instructs courts to “consider the totality of the circumstances and to determine, based upon a searching practical evaluation of the past and present reality, whether the political process is equally open to minority voters.” 478 U. S., at 79 (citation and internal quotation marks omitted).

The majority gives the impression that, in applying this framework, the District Court merely followed a set of well-settled, determinate legal principles. But it is widely acknowledged that “Gingles and its progeny have engendered considerable disagreement and uncertainty regarding the nature and contours of a vote dilution claim,” with commentators “noting the lack of any ‘authoritative resolution of the basic questions one would need to answer to make sense of [§2’s] results test.’ ” Merrill v. Milligan, 595 U. S. ___, ___–___ (2022) (Roberts, C. J., dissenting from grant of applications for stays) (slip op., at 1–2) (quoting C. Elmendorf, Making Sense of Section 2: Of Biased Votes, Unconstitutional Elections, and Common Law Statutes, 160 U. Pa. L. Rev. 377, 389 (2012)). If there is any “area of law notorious for its many unsolved puzzles,” this is it. J. Chen & N. Stephanopoulos, The Race-Blind Future of Voting Rights, 130 Yale L. J. 862, 871 (2021); see also Duchin & Spencer 758 (“Vote dilution on the basis of group membership is a crucial instance of the lack of a prescribed ideal”).

The source of this confusion is fundamental: Quite simply, we have never succeeded in translating the Gingles framework into an objective and workable method of identifying the undiluted benchmark. The second and third preconditions are all but irrelevant to the task. They essentially collapse into one question: Is voting racially polarized such that minority-preferred candidates consistently lose to majority-preferred ones? See Gingles, 478 U. S., at 51. Even if the answer is yes, that tells a court nothing about “how hard it ‘should’ be for minority voters to elect their preferred candidates under an acceptable system.” Id., at 88 (O’Connor, J., concurring in judgment). Perhaps an acceptable system is one in which the minority simply cannot elect its preferred candidates; it is, after all, a minority. Rejecting that outcome as “dilutive” requires a value judgment relative to a benchmark that polarization alone cannot provide.

The first Gingles precondition is only marginally more useful. True, the benchmark in a redistricting challenge must be “a hypothetical, undiluted plan,” Bossier Parish School Bd., 520 U. S., at 480, and the first precondition at least requires plaintiffs to identify some hypothetical alternative plan. Yet that alternative plan need only be “reasonably configured,” and—as explained above—to say that a plan is reasonable is a far cry from establishing an objective standard of fairness.

That leaves only the Gingles framework’s final stage: the totality-of-circumstances determination whether a State’s “political process is equally open to minority voters.” 478 U. S., at 79. But this formulation is mere verbiage unless one knows what an “equally open” system should look like—in other words, what the benchmark is. And, our cases offer no substantive guidance on how to identify the undiluted benchmark at the totality stage. The best they have to offer is a grab bag of amorphous “factors”—widely known as the Senate factors, after the Senate Judiciary Committee Report accompanying the 1982 amendments to §2—that Gingles said “typically may be relevant to a §2 claim.” See id., at 44–45. Those factors, however, amount to no more than “a list of possible considerations that might be consulted by a court attempting to develop a gestalt view of the political and racial climate in a jurisdiction.” Holder, 512 U. S., at 938 (opinion of Thomas, J.). Such a gestalt view is far removed from the necessary benchmark of a hypothetical, undiluted districting plan.

To see this, one need only consider the District Court’s use of the Senate factors here. See 582 F. Supp. 3d, at 1018–1024. The court began its totality-stage analysis by reiterating what nobody disputes: that voting in Alabama is racially polarized, with black voters overwhelmingly preferring Democrats and white voters largely preferring Republicans. To rebut the State’s argument that this pattern is attributable to politics, not race per se, the court noted that Donald Trump (who is white) prevailed over Ben Carson (who is black) in the 2016 Republican Presidential primary. Next, the court observed that black candidates rarely win statewide elections in Alabama and that black state legislators overwhelmingly come from majority-minority districts. The court then reviewed Alabama’s history of racial discrimination, noted other voting-rights cases in which the State was found liable, and cataloged socioeconomic disparities between black and white Alabamians in everything from car ownership to health insurance coverage. The court attributed these disparities “at least in part” to the State’s history of discrimination and found that they hinder black residents from participating in politics today, notwithstanding the fact that black and white Alabamians register and turn out to vote at similar rates. Id., at 1021–1022. Last, the court interpreted a handful of comments by three white politicians as “racial campaign appeals.” Id., at 1023–1024.

In reviewing this march through the Senate factors, it is impossible to discern any overarching standard or central question, only what might be called an impressionistic moral audit of Alabama’s racial past and present. Nor is it possible to determine any logical nexus between this audit and the remedy ordered: a congressional districting plan in which black Alabamians can control more than one seat. Given the District Court’s finding that two reasonably configured majority-black districts could be drawn, would Alabama’s one-district map have been acceptable if Ben Carson had won the 2016 primary, or if a greater number of black Alabamians owned cars?

The idea that such factors could explain the District Court’s judgment line is absurd. The plaintiffs’ claims pose one simple question: What is the “right” number of Alabama’s congressional seats that black voters who support Democrats “should” control? Neither the Senate factors nor the Gingles framework as a whole offers any principled answer.

In reality, the limits of the Gingles preconditions and the aimlessness of the totality-of-circumstances inquiry left the District Court only one obvious and readily administrable option: a benchmark of “allocation of seats in direct proportion to the minority group’s percentage in the population.” Holder, 512 U. S., at 937 (opinion of Thomas, J.). True, as disussed above, that benchmark is impossible to square with what the majority calls §2(b)’s “robust disclaimer against proportionality,” ante, at 5, and it runs headlong into grave constitutional problems. See Parents Involved in Community Schools v. Seattle School Dist. No. 1, 551 U.S. 701, 730 (2007) (plurality opinion). Nonetheless, the intuitive pull of proportionality is undeniable. “Once one accepts the proposition that the effectiveness of votes is measured in terms of the control of seats, the core of any vote dilution claim” “is inherently based on ratios between the numbers of the minority in the population and the numbers of seats controlled,” and there is no more logical ratio than direct proportionality. Holder, 512 U. S., at 902 (opinion of Thomas, J.). Combine that intuitive appeal with the “lack of any better alternative” identified in our case law to date, id., at 937, and we should not be surprised to learn that proportionality generally explains the results of §2 cases after the Gingles preconditions are satisfied. See E. Katz, M. Aisenbrey, A. Baldwin, E. Cheuse, & A. Weisbrodt, Documenting Discrimination in Voting: Judicial Findings Under Section 2 of the Voting Rights Act Since 1982, 39 U. Mich. J. L. Reform 643, 730–732 (2006) (surveying lower court cases and finding a near-perfect correlation between proportionality findings and liability results).

Thus, in the absence of an alternative benchmark, the vote-dilution inquiry has a strong and demonstrated tendency to collapse into a rough two-part test: (1) Does the challenged districting plan give the relevant minority group control of a proportional share of seats? (2) If not, has the plaintiff shown that some reasonably configured districting plan could better approximate proportional control? In this approach, proportionality is the ultimate benchmark, and the first Gingles precondition becomes a proxy for whether that benchmark is reasonably attainable in practice.

Beneath all the trappings of the Gingles framework, that two-part test describes how the District Court applied §2 here. The gravitational force of proportionality is obvious throughout its opinion. At the front end, the District Court even built proportionality into its understanding of Gingles’ first precondition, finding the plaintiffs’ illustrative maps to be reasonably configured in part because they “provide[d] a number of majority-Black districts . . . roughly proportional to the Black percentage of the population.” 582 F. Supp. 3d, at 1016. At the back end, the District Court concluded its “totality” analysis by revisiting proportionality and finding that it “weigh[ed] decidedly in favor of the plaintiffs.” Id., at 1025. While the District Court disclaimed giving overriding significance to proportionality, the fact remains that nothing else in its reasoning provides a logical nexus to its finding of a districting wrong and a need for a districting remedy. Finally, as if to leave no doubt about its implicit benchmark, the court admonished the State that “any remedial plan will need to include two districts in which Black voters either comprise a voting-age majority or something quite close.” Id., at 1033. In sum, the District Court’s thinly disguised benchmark was proportionality: Black Alabamians are about two-sevenths of the State’s population, so they should control two of the State’s seven congressional seats.

That was error—perhaps an understandable error given the limitations of the Gingles framework, but error nonetheless. As explained earlier, any principled application of §2 to cases such as these requires a meaningfully race-neutral benchmark. The benchmark cannot be an a priori thumb on the scale for racially proportional control.

E

The majority opinion does not acknowledge the District Court’s express proportionality-based reasoning. That omission is of a piece with its earlier noted failures to acknowledge the well-known indeterminacy of the Gingles framework, that black Alabamians are about two-sevenths of the State’s population, and that the plaintiffs here are thus seeking statewide proportionality. Through this pattern of omissions, the majority obscures the burning question in these cases. The District Court’s vote-dilution finding can be justified only by a racially loaded benchmark—specifically, a benchmark of proportional control based on race. Is that the benchmark the statute demands? The majority fails to confront this question head on, and it studiously avoids mentioning anything that would require it to do so.

The same nonresponsiveness infects the majority’s analysis, which is largely devoted to rebutting an argument nobody makes. Contrary to the majority’s telling, Alabama does not equate the “race-neutral benchmark” with “the median or average number of majority-minority districts” in a large computer-generated set of race-blind districting plans. Ante, at 15. The State’s argument for a race-neutral benchmark is rooted in the text of §2, the logic of vote- dilution claims, and the constitutional problems with any nonneutral benchmark. See Brief for Appellants 32–46. It then relies on the computer evidence in these cases, among other facts, to argue that the plaintiffs have not shown dilution relative to any race-neutral benchmark. See id., at 54–56. But the idea that “race-neutral benchmark” means the composite average of many computer-generated plans is the majority’s alone.

After thus straw-manning Alabama’s arguments at the outset, the majority muddles its own response. In a perfunctory footnote, it disclaims any holding that “algorithmic map making” evidence “is categorically irrelevant” in §2 cases. Ante, at 28, n. 8. That conclusion, however, is the obvious implication of the majority’s reasoning and rhetoric. See ante, at 27 (decrying a “map-comparison test” as “flawed in its fundamentals” even if it involves concededly “adequate comparators”); see also ante, at 17–18 (stating that the “focu[s]” of §2 analysis is “on the specific illustrative maps that a plaintiff adduces,” leaving unstated the implication that other algorithmically generated maps are irrelevant). The majority in effect, if not in word, thus forecloses any meaningful use of computer evidence to help locate the undiluted benchmark.

There are two critical problems with this fiat. The first, which the majority seems to recognize yet fails to resolve, is that excluding such computer evidence from view cannot be reconciled with §2’s command to consider “the totality of circumstances.”[15] Second—and more fundamentally—the reasons that the majority gives for downplaying the relevance of computer evidence would more logically support a holding that there is no judicially manageable way of applying §2’s results test to single-member districts. The majority waxes about the “myriad considerations” that go into districting, the “difficult, contestable choices” those considerations require, and how “[n]othing in §2 provides an answer” to the question of how well any given algorithm approximates the correct benchmark. Ante, at 27–28 (internal quotation marks omitted). In the end, it concludes, “Section 2 cannot require courts to judge a contest of computers” in which “there is no reliable way to determine who wins, or even where the finish line is.” Ante, at 29.

The majority fails to recognize that whether vote-dilution claims require an undiluted benchmark is not up for debate. If §2 applies to single-member districting plans, courts cannot dispense with an undiluted benchmark for comparison, ascertained by an objective and workable method. Bossier Parish School Bd., 520 U. S., at 480; Holder, 512 U. S., at 881 (plurality opinion). Of course, I would be the last person to deny that defining the undiluted benchmark is difficult. See id., at 892 (opinion of Thomas, J.) (arguing that it “immerse[s] the federal courts in a hopeless project of weighing questions of political theory”). But the “myriad considerations” and “[a]nswerless questions” the majority frets about, ante, at 27, 29, are inherent in the very enterprise of applying §2 to single-member districts. Everything the majority says about the difficulty of defining the undiluted benchmark with computer evidence applies with equal or greater force to the task of defining it without such evidence. At their core, the majority’s workability concerns are an isolated demand for rigor against the backdrop of a legal regime that has long been “ ‘inherently standardless,’ ” and must remain so until the Court either discovers a principled and objective method of identifying the undiluted benchmark, Holder, 512 U. S., at 885 (plurality opinion), or abandons this enterprise altogether, see id., at 945 (opinion of Thomas, J.).

Ultimately, the majority has very little to say about the appropriate benchmark. What little it does say suggests that the majority sees no real alternative to the District Court’s proportional-control benchmark, though it appears unwilling to say so outright. For example, in a nod to the statutory text and its “equal openness” requirement, the majority asserts that “[a] district is not equally open . . . when minority voters face—unlike their majority peers—bloc voting along racial lines, arising against the backdrop of substantial racial discrimination within the State, that renders a minority vote unequal to a vote by a nonminority voter.” Ante, at 17. But again, we have held that dilution cannot be shown without an objective, undiluted benchmark, and this verbiage offers no guidance for how to determine it.[16] Later, the majority asserts that “the Gingles framework itself imposes meaningful constraints on proportionality.” Ante, at 18–19. But the only constraint on proportionality the majority articulates is that it is often difficult to achieve—which, quite obviously, is no principled limitation at all. Ante, at 20–22.

Thus, the end result of the majority’s reasoning is no different from the District Court’s: The ultimate benchmark is a racially proportional allocation of seats, and the main question on which liability turns is whether a closer approximation to proportionality is possible under any reasonable application of traditional districting criteria.[17] This approach, moreover, is consistent with how the majority describes the role of plaintiffs’ illustrative maps, as well as an unjustified practical asymmetry to which its rejection of computer evidence gives rise. Courts are to “focu[s] . . . on the specific illustrative maps that a plaintiff adduces,” ante, at 17–18, by which the majority means that courts should not “focu[s]” on statistical evidence showing those maps to be outliers. Thus, plaintiffs may use an algorithm to generate any number of maps that meet specified districting criteria and a preferred racial target; then, they need only produce one of those maps to “sho[w] it is possible that the State’s map” is dilutive. Ante, at 18 (emphasis in original). But the State may not use algorithmic evidence to suggest that the plaintiffs’ map is an unsuitable benchmark for comparison—not even, apparently, if it can prove that the illustrative map is an outlier among “billion[s]” or “trillion[s]” of concededly “adequate comparators.” Ante, at 27, 29; see also ante, at 29 (rejecting sampling algorithms). This arbitrary restriction amounts to a thumb on the scale for §2 plaintiffs—an unearned presumption that any “reasonable” map they put forward constitutes a benchmark against which the State’s map can be deemed dilutive. And, once the comparison is framed in that way, the only workable rule of decision is proportionality. See Holder, 512 U. S., at 941–943 (opinion of Thomas, J.).

By affirming the District Court, the majority thus approves its benchmark of proportional control limited only by feasibility, and it entrenches the most perverse tendencies of our vote-dilution jurisprudence. It guarantees that courts will continue to approach vote-dilution claims just as the District Court here did: with no principled way of determining how many seats a minority “should” control and with a strong temptation to bless every incremental step toward a racially proportional allocation that plaintiffs can pass off as consistent with any reasonable map.

III

As noted earlier, the Court has long recognized the need to avoid interpretations of §2 that “ ‘would unnecessarily infuse race into virtually every redistricting, raising serious constitutional questions.’ ” Bartlett, 556 U. S., at 21 (plurality opinion) (quoting LULAC, 548 U. S., at 446 (opinion of Kennedy, J.)). Today, however, by approving the plaintiffs’ racially gerrymandered maps as reasonably configured, refusing to ground §2 vote-dilution claims in a race-neutral benchmark, and affirming a vote-dilution finding that can only be justified by a benchmark of proportional control, the majority holds, in substance, that race belongs in virtually every redistricting. It thus drives headlong into the very constitutional problems that the Court has long sought to avoid. The result of this collision is unmistakable: If the District Court’s application of §2 was correct as a statutory matter, §2 is unconstitutional as applied here.

Because the Constitution “restricts consideration of race and the [Voting Rights Act] demands consideration of race,” Abbott, 585 U. S., at ___ (slip op., at 4), strict scrutiny is implicated wherever, as here, §2 is applied to require a State to adopt or reject any districting plan on the basis of race. See Bartlett, 556 U. S., at 21–22 (plurality opinion). At this point, it is necessary to confront directly one of the more confused notions inhabiting our redistricting jurisprudence. In several cases, we have “assumed” that compliance with §2 of the Voting Rights Act could be a compelling state interest, before proceeding to reject race-predominant plans or districts as insufficiently tailored to that asserted interest. See, e.g., Wisconsin Legislature, 595 U. S., at ___ (slip op., at 3); Cooper v. Harris, 581 U.S. 285, 292 (2017); Shaw II, 517 U. S., at 915; Miller, 515 U. S., at 921. But we have never applied this assumption to uphold a districting plan that would otherwise violate the Constitution, and the slightest reflection on first principles should make clear why it would be problematic to do so.[18] The Constitution is supreme over statutes, not vice versa. Marbury v. Madison, 1 Cranch 137, 178 (1803). Therefore, if complying with a federal statute would require a State to engage in unconstitutional racial discrimination, the proper conclusion is not that the statute excuses the State’s discrimination, but that the statute is invalid.

If Congress has any power at all to require States to sort voters into congressional districts based on race, that power must flow from its authority to “enforce” the Fourteenth and Fifteenth Amendments “by appropriate legislation.” Amdt. 14, §5; Amdt. 15, §2. Since Congress in 1982 replaced intent with effects as the criterion of liability, however, “a violation of §2 is no longer a fortiori a violation of ” either Amendment. Bossier Parish School Bd., 520 U. S., at 482. Thus, §2 can be justified only under Congress’ power to “enact reasonably prophylactic legislation to deter constitutional harm.” Allen v. Cooper, 589 U. S. ___, ___ (2020) (slip op., at 11) (alteration and internal quotation marks omitted); see City of Boerne v. Flores, 521 U.S. 507, 517–529 (1997). Because Congress’ prophylactic- enforcement authority is “remedial, rather than substantive,” “[t]here must be a congruence and proportionality between the injury to be prevented or remedied and the means adopted to that end.”[19] Id., at 520. Congress’ chosen means, moreover, must “ ‘consist with the letter and spirit of the constitution.’ ” Shelby County v. Holder, 570 U.S. 529, 555 (2013) (quoting McCulloch v. Maryland, 4 Wheat. 316, 421 (1819)); accord, Miller, 515 U. S., at 927.

Here, as with everything else in our vote-dilution jurisprudence, the task of sound analysis is encumbered by the lack of clear principles defining §2 liability in districting. It is awkward to examine the “congruence” and “proportionality” of a statutory rule whose very meaning exists in a perpetual state of uncertainty. The majority makes clear, however, that the primary factual predicate of a vote-dilution claim is “bloc voting along racial lines” that results in majority-preferred candidates defeating minority-preferred ones. Ante, at 17; accord, Gingles, 478 U. S., at 48 (“The theoretical basis for [vote-dilution claims] is that where minority and majority voters consistently prefer different candidates, the majority, by virtue of its numerical superiority, will regularly defeat the choices of minority voters”). And, as I have shown, the remedial logic with which the District Court’s construction of §2 addresses that “wrong” rests on a proportional-control benchmark limited only by feasibility. Thus, the relevant statutory rule may be approximately stated as follows: If voting is racially polarized in a jurisdiction, and if there exists any more or less reasonably configured districting plan that would enable the minority group to constitute a majority in a number of districts roughly proportional to its share of the population, then the jurisdiction must ensure that its districting plan includes that number of majority-minority districts “or something quite close.”[20] 582 F. Supp. 3d, at 1033. Thus construed and applied, §2 is not congruent and proportional to any provisions of the Reconstruction Amendments.

To determine the congruence and proportionality of a measure, we must begin by “identify[ing] with some precision the scope of the constitutional right at issue.” Board of Trustees of Univ. of Ala. v. Garrett, 531 U.S. 356, 365 (2001). The Reconstruction Amendments “forbi[d], so far as civil and political rights are concerned, discrimination . . . against any citizen because of his race,” ensuring that “[a]ll citizens are equal before the law.” Gibson v. Mississippi, 162 U.S. 565, 591 (1896) (Harlan, J.). They dictate “that the Government must treat citizens as individuals, not as simply components of a racial, religious, sexual or national class.” Miller, 515 U. S., at 911 (internal quotation marks omitted). These principles are why the Constitution presumptively forbids race-predominant districting, “even for remedial purposes.” Shaw I, 509 U. S., at 657.

These same principles foreclose a construction of the Amendments that would entitle members of racial minorities, qua racial minorities, to have their preferred candidates win elections. Nor do the Amendments limit the rights of members of a racial majority to support their preferred candidates—regardless of whether minorities prefer different candidates and of whether “the majority, by virtue of its numerical superiority,” regularly prevails. Gingles, 478 U. S., at 48. Nor, finally, do the Amendments establish a norm of proportional control of elected offices on the basis of race. See Parents Involved, 551 U. S., at 730–731 (plurality opinion); Shaw I, 509 U. S., at 657. And these notions are not merely foreign to the Amendments. Rather, they are radically inconsistent with the Amendments’ command that government treat citizens as individuals and their “goal of a political system in which race no longer matters.” Ibid.

Those notions are, however, the values at the heart of §2 as construed by the District Court and the majority. As applied here, the statute effectively considers it a legal wrong by the State if white Alabamians vote for candidates from one political party at high enough rates, provided that black Alabamians vote for candidates from the other party at a still higher rate. And the statute remedies that wrong by requiring the State to engage in race-based redistricting in the direction of proportional control.

I am not certain that Congress’ enforcement power could ever justify a statute so at odds “ ‘with the letter and spirit of the constitution.’ ” Shelby County, 570 U. S., at 555. If it could, it must be because Congress “identified a history and pattern” of actual constitutional violations that, for some reason, required extraordinary prophylactic remedies. Garrett, 531 U. S., at 368. But the legislative record of the 1982 amendments is devoid of any showing that might justify §2’s blunt approximation of a “racial register for allocating representation on the basis of race.” Holder, 512 U. S., at 908 (opinion of Thomas, J.). To be sure, the Senate Judiciary Committee Report that accompanied the 1982 amendment to the Voting Rights Act “listed many examples of what the Committee took to be unconstitutional vote dilution.” Brnovich, 594 U. S., at ___ (slip op., at 6) (emphasis added). But the Report also showed the Committee’s fundamental lack of “concern with whether” those examples reflected the “intentional” discrimination required “to raise a constitutional issue.” Allen, 589 U. S., at ___ (slip op., at 15). The Committee’s “principal reason” for rejecting discriminatory purpose was simply that it preferred an alternative legal standard; it thought Mobile’s intent test was “the wrong question,” and that courts should instead ask whether a State’s election laws offered minorities “a fair opportunity to participate” in the political process. S. Rep. No. 97–417, p. 36.

As applied here, the amended §2 thus falls on the wrong side of “the line between measures that remedy or prevent unconstitutional actions and measures that make a substantive change in the governing law.” City of Boerne, 521 U. S., at 519. It replaces the constitutional right against intentionally discriminatory districting with an amorphous race-based right to a “fair” distribution of political power, a “right” that cannot be implemented without requiring the very evils the Constitution forbids.

If that alone were not fatal, §2’s “reach and scope” further belie any congruence and proportionality between its districting-related commands, on the one hand, and actionable constitutional wrongs, on the other. Id., at 532. Its “[s]weeping coverage ensures its intrusion at every level of government” and in every electoral system. Ibid. It “has no termination date or termination mechanism.” Ibid. Thus, the amended §2 is not spatially or temporally “limited to those cases in which constitutional violations [are] most likely.” Id., at 533. Nor does the statute limit its reach to “attac[k] a particular type” of electoral mechanism “with a long history as a ‘notorious means to deny and abridge voting rights on racial grounds.’ ” Ibid. (quoting South Carolina v. Katzenbach, 383 U.S. 301, 355 (1966) (Black, J., concurring and dissenting)). In view of this “indiscriminate scope,” “it simply cannot be said that ‘many of [the districting plans] affected by the congressional enactment have a significant likelihood of being unconstitutional.’ ” Florida Prepaid Postsecondary Ed. Expense Bd. v. College Savings Bank, 527 U.S. 627, 647 (1999) (quoting City of Boerne, 521 U. S., at 532).

Of course, under the logically unbounded totality-of-circumstances inquiry, a court applying §2 can always embroider its vote-dilution determination with findings about past or present unconstitutional discrimination. But this possibility does nothing to heal either the fundamental contradictions between §2 and the Constitution or its extreme overbreadth relative to actual constitutional wrongs. “A generalized assertion of past discrimination” cannot justify race-based redistricting, “because it provides no guidance for a legislative body to determine the precise scope of the injury it seeks to remedy.” Shaw II, 517 U. S., at 909 (internal quotation marks omitted). To justify a statute tending toward the proportional allocation of political power by race throughout the Nation, it cannot be enough that a court can recite some indefinite quantum of discrimination in the relevant jurisdiction. If it were, courts “could uphold [race-based] remedies that are ageless in their reach into the past, and timeless in their ability to affect the future.” Wygant v. Jackson Bd. of Ed., 476 U.S. 267, 276 (1986) (plurality opinion). That logic “would effectively assure that race will always be relevant in [redistricting], and that the ultimate goal of eliminating entirely from governmental decisionmaking such irrelevant factors as a human being’s race will never be achieved.” Parents Involved, 551 U. S., at 730 (plurality opinion) (alteration and internal quotation marks omitted).

For an example of these baleful results, we need look no further than the congressional districts at issue here. In 1992, Alabama and a group of §2 plaintiffs, whom a federal court chose to regard as the representatives “of all African-American citizens of the State of Alabama,” stipulated that the State’s black population was “ ‘sufficiently compact and contiguous to comprise a single member significant majority (65% or more) African American Congressional district,’ ” and that, “ ‘[c]onsequently,’ ” such a “ ‘district should be created.’ ” Wesch v. Hunt, 785 F. Supp. 1491, 1493, 1498 (SD Ala.). Accepting that stipulation, the court reworked District 7 into an irregularly shaped supermajority-black district—one that scooped up populous clusters of black voters in the disparate urban centers of Birmingham and Montgomery to connect them across a swath of largely majority-black rural areas—without even “decid[ing] whether the creation of a majority African-American district [was] mandated by either §2 or the Constitution.” Id., at 1499; see n. 7, supra. It did not occur to the court that the Constitution might forbid such an extreme racial gerrymander, as it quite obviously did. But, once District 7 had come into being as a racial gerrymander thought necessary to satisfy §2, it became an all-but-immovable fixture of Alabama’s districting scheme.

Now, 30 years later, the plaintiffs here demand that Alabama carve up not two but three of its main urban centers on the basis of race, and that it configure those urban centers’ black neighborhoods with the outlying majority-black rural areas so that black voters can control not one but two of the State’s seven districts. The Federal Judiciary now upholds their demand—overriding the State’s undoubted interest in preserving the core of its existing districts, its plainly reasonable desire to maintain the Gulf Coast region as a cohesive political unit, and its persuasive arguments that a race-neutral districting process would not produce anything like the districts the plaintiffs seek. Our reasons for doing so boil down to these: that the plaintiffs’ proposed districts are more or less within the vast universe of reasonable districting outcomes; that Alabama’s white voters do not support the black minority’s preferred candidates; that Alabama’s racial climate, taken as a rarefied whole, crosses some indefinable line justifying our interference; and, last but certainly not least, that black Alabamians are about two-sevenths of the State’s overall population.

By applying §2 in this way to claims of this kind, we encourage a conception of politics as a struggle for power between “competing racial factions.” Shaw I, 509 U. S., at 657. We indulge the pernicious tendency of assigning Americans to “creditor” and “debtor race[s],” even to the point of redistributing political power on that basis. Adarand Constructors, Inc. v. Peña, 515 U.S. 200, 239 (1995) (Scalia, J., concurring in part and concurring in judgment). We ensure that the race-based redistricting we impose on Alabama now will bear divisive consequences long into the future, just as the initial creation of District 7 segregated Jefferson County for decades and minted the template for crafting black “political homelands” in Alabama. Holder, 512 U. S., at 905 (opinion of Thomas, J.). We place States in the impossible position of having to weigh just how much racial sorting is necessary to avoid the “competing hazards” of violating §2 and violating the Constitution. Abbott, 585 U. S., at ___ (slip op., at 4) (internal quotation marks omitted). We have even put ourselves in the ridiculous position of “assuming” that compliance with a statute can excuse disobedience to the Constitution. Worst of all, by making it clear that there are political dividends to be gained in the discovery of new ways to sort voters along racial lines, we prolong immeasurably the day when the “sordid business” of “divvying us up by race” is no more. LULAC, 548 U. S., at 511 (Roberts, C. J., concurring in part, concurring in judgment in part, and dissenting in part). To the extent §2 requires any of this, it is unconstitutional.

The majority deflects this conclusion by appealing to two of our older Voting Rights Act cases, City of Rome v. United States, 446 U.S. 156 (1980), and South Carolina v. Katzenbach, 383 U.S. 301, that did not address §2 at all and, indeed, predate Congress’ adoption of the results test. Ante, at 33–34. That maneuver is untenable. Katzenbach upheld §5’s preclearance requirements, §4(b)’s original coverage formula, and other related provisions aimed at “a small number of States and political subdivisions” where “systematic resistance to the Fifteenth Amendment” had long been flagrant. 383 U. S., at 328; see also id., at 315–317 (describing the limited issues presented). Fourteen years later, City of Rome upheld the 1975 Act extending §5’s preclearance provisions for another seven years. See 446 U. S., at 172–173. The majority’s reliance on these cases to validate a statutory rule not there at issue could make sense only if we assessed the congruence and proportionality of the Voting Rights Act’s rules wholesale, without considering their individual features, or if Katzenbach and City of Rome meant that Congress has plenary power to enact whatever rules it chooses to characterize as combating “discriminatory . . . effect[s].” Ante, at 33 (internal quotation marks omitted). Neither proposition makes any conceptual sense or is consistent with our cases. See, e.g., Shelby County, 570 U. S., at 550–557 (holding the 2006 preclearance coverage formula unconstitutional); Northwest Austin Municipal Util. Dist. No. One v. Holder, 557 U.S. 193, 203 (2009) (emphasizing the distinctness of §§2 and 5); City of Boerne, 521 U. S., at 533 (discussing City of Rome as a paradigm case of congruence-and-proportionality review of remedial legislation); Miller, 515 U. S., at 927 (stressing that construing §5 to require “that States engage in presumptively unconstitutional race-based districting” would raise “troubling and difficult constitutional questions,” notwithstanding City of Rome).

In fact, the majority’s cases confirm the very limits on Congress’ enforcement powers that are fatal to the District Court’s construction of §2. City of Rome, for example, immediately after one of the sentences quoted by the majority, explained the remedial rationale for its approval of the 1975 preclearance extension: “Congress could rationally have concluded that, because electoral changes by jurisdictions with a demonstrable history of intentional racial discrimination in voting create the risk of purposeful discrimination, it was proper to prohibit changes that have a discriminatory impact.” 446 U. S., at 177 (emphasis added; footnote omitted). The next section of City of Rome then separately examined and upheld the reasonableness of the extension’s 7-year time period. See id., at 181–182. City of Rome thus stands for precisely the propositions for which City of Boerne cited it: Congress may adopt “[p]reventive measures . . . when there is reason to believe that many of the laws affected by the congressional enactment have a significant likelihood of being unconstitutional,” 521 U. S., at 532, particularly when it employs “termination dates, geographic restrictions, or egregious predicates” that “tend to ensure Congress’ means are proportionate to ends legitimate,” id., at 533; see also id., at 532–533 (analyzing Katzenbach in similar terms); Shelby County, 570 U. S., at 535, 545–546 (same). Again, however, the amended §2 lacks any such salutary limiting principles; it is unbounded in time, place, and subject matter, and its districting-related commands have no nexus to any likely constitutional wrongs.

In short, as construed by the District Court, §2 does not remedy or deter unconstitutional discrimination in districting in any way, shape, or form. On the contrary, it requires it, hijacking the districting process to pursue a goal that has no legitimate claim under our constitutional system: the proportional allocation of political power on the basis of race. Such a statute “cannot be considered remedial, preventive legislation,” and the race-based redistricting it would command cannot be upheld under the Constitution. City of Boerne, 521 U. S., at 532.[21]

IV

These cases are not close. The plaintiffs did not prove that Alabama’s districting plan “impose[s] or applie[s]” any “voting qualification or prerequisite to voting or standard, practice, or procedure” that effects “a denial or abridgement of the[ir] right . . . to vote on account of race or color.” §10301(a). Nor did they prove that Alabama’s congressional districts “are not equally open to participation” by black Alabamians. §10301(b). The plaintiffs did not even prove that it is possible to achieve two majority-black districts without resorting to a racial gerrymander. The most that they can be said to have shown is that sophisticated mapmakers can proportionally allocate Alabama’s congressional districts based on race in a way that exceeds the Federal Judiciary’s ability to recognize as a racial gerrymander with the naked eye. The District Court held that this showing, plus racially polarized voting and its gestalt view of Alabama’s racial climate, was enough to require the State to redraw its districting plan on the basis of race. If that is the benchmark for vote dilution under §2, then §2 is nothing more than a racial entitlement to roughly proportional control of elective offices—limited only by feasibility—wherever different racial groups consistently prefer different candidates.

If that is what §2 means, the Court should hold that it is unconstitutional. If that is not what it means, but §2 applies to districting, then the Court should hold that vote-dilution challenges require a race-neutral benchmark that bears no resemblance to unconstitutional racial registers. On the other hand, if the Court believes that finding a race-neutral benchmark is as impossible as much of its rhetoric suggests, it should hold that §2 cannot be applied to single-member districting plans for want of an “objective and workable standard for choosing a reasonable benchmark.” Holder, 512 U. S., at 881 (plurality opinion). Better yet, it could adopt the correct interpretation of §2 and hold that a single-member districting plan is not a “voting qualification,” a “prerequsite to voting,” or a “standard, practice, or procedure,” as the Act uses those terms. One way or another, the District Court should be reversed.

The majority goes to great lengths to decline all of these options and, in doing so, to fossilize all of the worst aspects of our long-deplorable vote-dilution jurisprudence. The majority recites Gingles’ shopworn phrases as if their meaning were self-evident, and as if it were not common knowledge that they have spawned intractable difficulties of definition and application. It goes out of its way to reaffirm §2’s applicability to single-member districting plans both as a purported original matter and on highly exaggerated stare decisis grounds. It virtually ignores Alabama’s primary argument—that, whatever the benchmark is, it must be race neutral—choosing, instead, to quixotically joust with an imaginary adversary. In the process, it uses special pleading to close the door on the hope cherished by some thoughtful observers, see Gonzalez, 535 F. 3d, at 599–600, that computational redistricting methods might offer a principled, race-neutral way out of the thicket Gingles carried us into. Finally, it dismisses grave constitutional questions with an insupportably broad holding based on demonstrably inapposite cases.[22]

I find it difficult to understand these maneuvers except as proceeding from a perception that what the District Court did here is essentially no different from what many courts have done for decades under this Court’s superintendence, joined with a sentiment that it would be unthinkable to disturb that approach to the Voting Rights Act in any way. I share the perception, but I cannot understand the sentiment. It is true that, “under our direction, federal courts [have been] engaged in methodically carving the country into racially designated electoral districts” for decades now. Holder, 512 U. S., at 945 (opinion of Thomas, J.). But that fact should inspire us to repentance, not resignation. I am even more convinced of the opinion that I formed 29 years ago:

“In my view, our current practice should not continue. Not for another Term, not until the next case, not for another day. The disastrous implications of the policies we have adopted under the Act are too grave; the dissembling in our approach to the Act too damaging to the credibility of the Federal Judiciary. The ‘inherent tension’—indeed, I would call it an irreconcilable conflict—between the standards we have adopted for evaluating vote dilution claims and the text of the Voting Rights Act would itself be sufficient in my view to warrant overruling the interpretation of §2 set out in Gingles. When that obvious conflict is combined with the destructive effects our expansive reading of the Act has had in involving the Federal Judiciary in the project of dividing the Nation into racially segregated electoral districts, I can see no reasonable alternative to abandoning our current unfortunate understanding of the Act.” Id., at 944.

I respectfully dissent.