Alexander v. South Carolina State Conference of the NAACP

602 U.S. 1

Opinions

Opinion

CASES ADJUDGED IN THE

SUPREME COURT OF THE UNITED STATES AT

OCTOBER TERM, 2023

ALEXANDER, PRESIDENT OF THE SOUTH CARO- LINA SENATE, et al. v. SOUTH CAROLINA STATE CONFERENCE OF THE NAACP et al.

appeal from the united states district court for the Page Proof Pending Publication district of south carolina No. 22–807. Argued October 11, 2023—Decided May 23, 2024 The Constitution entrusts state legislatures with the primary responsibility for drawing congressional districts, and legislative redistricting is an inescapably political enterprise. Claims that a map is unconstitutional because it was drawn to achieve a partisan end are not justiciable in federal court. By contrast, if a legislature gives race a predominant role in redistricting decisions, the resulting map is subjected to strict scrutiny and may be held unconstitutional. These doctrinal lines collide when race and partisan preference are highly correlated. This Court has endorsed two related propositions when navigating this tension. First, a party challenging a map's constitutionality must disentangle race and politics to show that race was the legislature's “predominant” motivating factor. Miller v. Johnson, 515 U. S. 900, 916. Second, the Court starts with a presumption that the legislature acted in good faith. To disentangle race from other permissible considerations, plaintiffs may employ some combination of direct and circumstantial evidence. Cooper v. Harris, 581 U. S. 285, 291. Where race and politics are highly correlated, a map that has been gerrymandered to achieve a partisan end can look very similar to a racially gerrymandered map. Thus, in Easley v. Cromartie, 532 U. S. 234, the Court held that the plaintiffs

1

Syllabus

failed to meet the high bar for a racial-gerrymandering claim when they failed to produce an alternative map showing that a rational legislature sincerely driven by its professed partisan goals would have drawn a different map with greater racial balance. Id., at 258. Without an alternative map, the Court also found it diffcult for plaintiffs to defeat the starting presumption that the legislature acted in good faith.

Following the 2020 census, South Carolina was tasked with redrawing its congressional district maps because of population shifts in two of its seven districts—Districts 1 and 6. The State Senate subcommittee responsible for drawing the new map issued a statement explaining that the process would be guided by traditional districting principles along with the goal of creating a stronger Republican tilt in District 1. To draw the new maps, the Senate turned to Will Roberts, a nonpartisan staffer with experience in drawing reapportionment plans. Roberts's plan (the Enacted Plan) achieved the legislature's political goal by increasing District 1's projected Republican vote share by 1.36% to 54.39%. The plan also raised the black voting-age population (BVAP) from 16.56% to 16.72%. The legislature adopted the plan, and the Governor signed it into law.

The National Association for the Advancement of Colored People and

Page Proof Pending Publication District 1 voter Taiwan Scott (the Challengers) challenged the plan, alleging that it resulted in racial gerrymanders in certain districts and in the dilution of the electoral power of the State's black voters. The three-judge District Court held that the State drew District 1 with a 17% BVAP target in mind in violation of the Equal Protection Clause and that this putative use of race to draw District 1 unlawfully diluted the black vote.

Held:

1. The District Court's fnding that race predominated in the design of District 1 in the Enacted Plan was clearly erroneous. Pp. 17–38.

(a) Because the State's principal legal argument—that the District Court did not properly disentangle race from politics—is an attack on the factual basis of the District Court's fndings, this case can be disposed on clear-error grounds. The District Court clearly erred because the Challengers did not satisfy the demanding burden of showing that the “legislature subordinated traditional race-neutral districting principles . . . to racial considerations.” Miller, 515 U. S., at 916. The Challengers provided no direct evidence of a racial gerrymander, and their circumstantial evidence is very weak. Instead the Challengers relied on deeply fawed expert reports. And the Challengers did not offer a single alternative map to show that the legislature's partisan goal could be achieved while raising the BVAP in District 1. Pp. 17–18.

3

(b) The District Court's factual fndings in this case are reviewed for clear error. Because the racial predominance test has a very substantial legal component that must take account of the Court's prior relevant decisions, special care must be exercised in reviewing the relevant fndings of fact. Pp. 18–19.

(c) The District Court's heavy reliance on four pieces of evidence was seriously misguided in light of the appropriate legal standard and repeated instructions that a court in a case such as this must rule out the possibility that politics drove the districting process. None of the facts on which the District Court relied to infer a racial motive is suffcient to support an inference that can overcome the presumption of legislative good faith. First, the District Court concluded that the legislature deliberately sought to maintain a particular BVAP because the maps that produced the sought-after partisan goal all had roughly the same BVAP. But the mere fact that District 1's BVAP remained around 17%, despite all the changes made during the redistricting process , proves very little. The tight correlation between the legislature's partisan aim and District 1's BVAP is substantiated by the District Court's own fndings. The Challengers could not point to a single map in the record that would satisfy the legislature's political aim with a BVAP above 17%. The District Court disregarded the presumption of Page Proof Pending Publication legislative good faith by drawing an inference that the State acted in bad faith based on the racial consequences of a political gerrymander in a jurisdiction in which race and partisan preference are very closely related. Second, the District Court inferred a racial motive from the fact that the Enacted Plan moved more voters out of District 1 than were needed to comply with the one person, one vote rule, and that the Enacted Plan split a few counties. But the high priority that the legislature gave to its partisan aim can explain these decisions. Third, the District Court clearly erred when it concluded that the legislature's real aim was racial based on the movement of certain predominantly black Charleston precincts from District 1 to District 6. Again, the legislature's partisan goal can easily explain this decision. Fourth, the District Court placed excessive weight on the fact that several legislative staffers admitted to viewing racial data at some point during the redistricting process. The District Court cited no evidence that could not also support the inference that politics drove the mapmaking process and provided no explanation why a mapmaker who wanted to produce a version of District 1 that would be safely Republican would use data about voters' race rather than their political preferences. Pp. 19–24.

(d) The four expert reports relied upon by the Challengers are fawed because they ignored traditional districting criteria such as geographical constraints and the legislature's partisan interests. Allen v.

4

Milligan, 599 U. S. 1, 34. The report of Dr. Kosuke Imai made no effort to disentangle race from politics. It also failed to consider “core district retention,” a term referring to “the proportion of districts that remain when a State transitions from one districting plan to another.” Id., at 21. The report of Dr. Jordan Ragusa did attempt to disentangle race from politics, but its analysis has two serious defects. First, each of his three models failed to control for contiguity or compactness. Second, he used an inferior method of measuring a precinct's partisan leanings by counting absolute votes rather than a party's relative share of the vote. The report of Dr. Baodong Liu purported to show that race rather than politics explains District 1's design, but Dr. Liu's methodology was plainly fawed. Like Dr. Ragusa, Dr. Liu failed to account for contiguity and compactness. And while this defect alone is suffcient to preclude reliance, Dr. Liu also used inferior data to measure a district's partisan tilt—i. e., data from the 2018 off-cycle gubernatorial primaries. Finally, the report of Dr. Moon Duchin, like that of Dr. Imai, did not account for partisanship or core retention and was based on an assessment of the map as a whole rather than District 1 in particular. Thus, her report has no probative force with respect to the Challengers' racial- gerrymandering claim regarding District 1's boundaries. Pp. 24–33.

(e) The District Court also critically erred by failing to draw an Page Proof Pending Publication adverse inference against the Challengers for not providing an adequate alternative map. By showing that a rational legislature, driven only by its professed mapmaking criteria, could have produced a different map with “greater racial balance,” Cromartie, 532 U. S., at 258, an alternative map can perform the critical task of distinguishing between racial and political motivations when race and partisanship are closely entwined . Moreover, an alternative map is easy to produce. The District Court mistakenly held that an alternative map is relevant only for the purpose of showing that a remedy is plausible. A plaintiff's failure to submit an alternative map should be interpreted by courts as an implicit concession that the plaintiff cannot draw a map that undermines the legislature's defense. Pp. 34–35.

2. Because the same fndings of fact and reasoning that guided the court's racial-gerrymandering analysis also guided the analysis of the Challengers' independent vote-dilution claim, that conclusion also cannot stand. The District Court also erred in confating the two claims. A plaintiff pressing a vote-dilution claim cannot prevail simply by showing that race played a predominant role in the districting process, but rather must show that the State “enacted a particular voting scheme as a purposeful device to minimize or cancel out the voting potential of racial or ethnic minorities.” Miller, 515 U. S., at 911. In other words, the plaintiff must show that the State's districting plan “has the purpose

5

and effect” of diluting the minority vote. Shaw v. Reno, 509 U. S. 630, 649. In light of these two errors in the District Court's analysis, a remand is appropriate. Pp. 38–39. Reversed in part and remanded in part.

Alito, J., delivered the opinion of the Court, in which Roberts, C. J., and Gorsuch, Kavanaugh, and Barrett, JJ., joined, and in which Thomas, J., joined as to all but Part III–C. Thomas, J., fled an opinion concurring in part, post, p. 39. Kagan, J., fled a dissenting opinion, in which Sotomayor and Jackson, JJ., joined, post, p. 66.

John M. Gore argued the cause for appellants. With him on the briefs were Joseph P. Falvey, Robert E. Tyson, Jr., William W. Wilkins, Andrew A. Mathias, Mark C. Moore, Hamilton B. Barber, M. Elizabeth Crum, and Michael R. Burchstead.

Leah C. Aden argued the cause for appellees. With her on the brief were Adriel I. Cepeda Derieux, Davin Rosborough, Sophia Lin Lakin, David D. Cole, Cecillia D. Wang, Janai S. Page Proof Pending Publication Nelson, Samuel Spital, Raymond Audain, John S. Cusick, Antonio L. Ingram II, John A. Freedman, Elisabeth S. Theodore , and Stephen K. Wirth.

Caroline A. Flynn argued the cause for the United States as amicus curiae supporting neither party. With her on the brief were Solicitor General Prelogar, Assistant Attorney General Clarke, Deputy Solicitor General Fletcher, and Nicolas Y. Riley.*

*Briefs of amici curiae urging reversal were fled for the State of Alabama et al. by Steve Marshall, Attorney General of Alabama, Edmund G. LaCour, Jr., Solicitor General, and Bethany C. Lee, Assistant Solicitor General, by Angela Colmenero, Provisional Attorney General of Texas, and by the Attorneys General for their respective States as follows: Treg Taylor of Alaska, Tim Griffn of Arkansas, Ashley Moody of Florida, Chris Carr of Georgia, Theodore E. Rokita of Indiana, Brenna Bird of Iowa, Jeff Landry of Louisiana, Lynn Fitch of Mississippi, Austin Knudsen of Montana, Michael T. Hilgers of Nebraska, Alan Wilson of South Carolina, Jonathan Skrmetti of Tennessee, Sean D. Reyes of Utah, and Patrick Morrisey of West Virginia; for the Fair Lines America Foundation by Richard B. Raile, Efrem Marshall Braden, and Katherine L. Mc-

Opinion of the Court

Justice Alito delivered the opinion of the Court.

I

The Constitution entrusts state legislatures with the primary responsibility for drawing congressional districts, and redistricting is an inescapably political enterprise. Legislators are almost always aware of the political ramifcations of the maps they adopt, and claims that a map is unconstitutional because it was drawn to achieve a partisan end are not justiciable in federal court. Thus, as far as the Federal Constitution is concerned, a legislature may pursue partisan ends when it engages in redistricting. By contrast, if a legislature gives race a predominant role in redistricting decisions , the resulting map is subjected to strict scrutiny and may be held unconstitutional.

These doctrinal lines collide when race and partisan preference are highly correlated. We have navigated this ten- Page Proof Pending Publication sion by endorsing two related propositions. First, a party challenging a map's constitutionality must disentangle race and politics if it wishes to prove that the legislature was motivated by race as opposed to partisanship. Second, in assessing a legislature's work, we start with a presumption that the legislature acted in good faith.

Knight; for Judicial Watch, Inc., et al. by T. Russell Nobile, Robert D. Popper, and H. Christopher Coates; for the National Republican Redistricting Trust by Phillip M. Gordon; for Nancy Mace et al. by Jason B. Torchinsky; and for Gov. Henry McMaster by Thomas A. Limehouse, Jr., and William Grayson Lambert.

Briefs of amici curiae urging affrmance were fled for the Constitutional Accountability Center by Elizabeth B. Wydra, Brianne J. Gorod, and David H. Gans; for Historians by Michelle K. Moriarty and Godfre O. Blackman; for the Lawyers' Committee for Civil Rights Under Law et al. by Julie Veroff, Kathleen Hartnett, Damon T. Hewitt, Jon Greenbaum , Ezra D. Rosenberg, and Adam S. Gershenson; for the League of Women Voters of South Carolina et al. by Brian C. Duffy; for Political Science Professors by Angela M. Liu; for Cong. James E. Clyburn by John Graubert and Christopher Kimmel; and for Nicholas O. Stephanopoulos et al. by Ruth Greenwood.

7

In this case, which features a challenge to South Carolina's redistricting efforts in the wake of the 2020 census, the three-judge District Court paid only lip service to these propositions. That misguided approach infected the District Court's fndings of fact, which were clearly erroneous under the appropriate legal standard. We therefore reverse the trial court in part and remand for further proceedings.

II

A

Redistricting constitutes a traditional domain of state legislative authority. See Moore v. Harper, 600 U. S. 1 (2023); see also U. S. Const., Art. I, § 4, cl. 1. The Fourteenth Amendment introduces one constraint by prohibiting a State from engaging in a racial gerrymander unless it can satisfy strict scrutiny. But given “the complex interplay of forces that enter a legislature's redistricting calculus,” we have re- Page Proof Pending Publication peatedly emphasized that federal courts must “exercise extraordinary caution in adjudicating claims that a State has drawn district lines on the basis of race.” Miller v. Johnson , 515 U. S. 900, 915–916 (1995). Such caution is necessary because “[f]ederal-court review of districting legislation represents a serious intrusion on the most vital of local functions .” Id., at 915. To untangle race from other permissible considerations, we require the plaintiff to show that race was the “predominant factor motivating the legislature's decision to place a signifcant number of voters within or without a particular district.” Id., at 916.

To make that showing, a plaintiff must prove that the State “subordinated” race-neutral districting criteria such as compactness , contiguity, and core preservation to “racial considerations .” Ibid. Racial considerations predominate when “[r]ace was the criterion that, in the State's view, could not be compromised” in the drawing of district lines.1 Shaw v.

1 A plaintiff can also establish racial predominance by showing that the legislature used “race as a proxy” for “political interest[s].” Miller, 515 U. S., at 914; see also Cooper v. Harris, 581 U. S. 285, 291, n. 1 (2017)

8

Hunt, 517 U. S. 899, 907 (1996). We have recognized that, “[a]s a practical matter,” challengers will often need to show that the State's chosen map conficts with traditional redistricting criteria. Bethune-Hill v. Virginia State Bd. of Elections, 580 U. S. 178, 190 (2017). That is because it may otherwise “be diffcult for challengers to fnd other evidence suffcient to show that race was the overriding factor causing neutral considerations to be cast aside.” Ibid.

This showing can be made through some combination of direct and circumstantial evidence. See Cooper v. Harris, 581 U. S. 285, 291 (2017). Direct evidence often comes in the form of a relevant state actor's express acknowledgment that race played a role in the drawing of district lines. Such concessions are not uncommon because States often admit to considering race for the purpose of satisfying our precedent interpreting the Voting Rights Act of 1965. See, e. g., Alabama Legislative Black Caucus v. Alabama, 575 U. S. 254, Page Proof Pending Publication 259–260 (2015). Direct evidence can also be smoked out over the course of litigation. In Cooper, for instance, we offered the hypothetical example of a plaintiff fnding “scores of leaked e-mails from state offcials instructing their mapmaker to pack as many black voters as possible into a district.” 581 U. S., at 318. In such instances, if the State cannot satisfy strict scrutiny, direct evidence of this sort amounts to a confession of error.

Proving racial predominance with circumstantial evidence alone is much more diffcult. Although we have never invalidated an electoral map in a case in which the plaintiff failed to adduce any direct evidence, we have, at least in theory, kept the door open for those rare instances in which a district 's shape is “so bizarre on its face that it discloses a racial design” absent any alternative explanation. Miller, 515

(noting that strict scrutiny is warranted when “a legislature elevated race to the predominant criterion in order to advance other goals, including political ones”).

9

U. S., at 914; see also Shaw v. Reno, 509 U. S. 630, 643–645 (1993) (Shaw I).

A circumstantial-evidence-only case is especially diffcult when the State raises a partisan-gerrymandering defense. That is because partisan and racial gerrymanders “are capable of yielding similar oddities in a district's boundaries” when there is a high correlation between race and partisan preference. Cooper, 581 U. S., at 308. And that is the situation in this case, as the 2020 Presidential election illustrated . Exit polls found that at least 90% of black voters voted for the Democratic candidate in South Carolina and throughout the Nation.2 When partisanship and race correlate , it naturally follows that a map that has been gerrymandered to achieve a partisan end can look very similar to a racially gerrymandered map. For that reason, “[o]ur prior decisions have made clear that a jurisdiction may engage in constitutional political gerrymandering, even if it so happens that the most loyal Democrats happen to be black Democrats Page Proof Pending Publication and even if the State were conscious of that fact.” Hunt v. Cromartie, 526 U. S. 541, 551 (1999) (Cromartie I); see also Rucho v. Common Cause, 588 U. S. 684, 721 (2019) (concluding that federal judges lack the license to evaluate partisan- gerrymandering claims). We have noted that a State's partisan-gerrymandering defense therefore raises “special challenges” for plaintiffs. Cooper, 581 U. S., at 308. To prevail , a plaintiff must “disentangle race from politics” by proving “that the former drove a district's lines.” Ibid. (emphasis added). That means, among other things, ruling out the competing explanation that political considerations domi-

2 See, e. g., Pew Research Center, Behind Biden's 2020 Victory (June 30, 2021), https://www.pewresearch.org/politics/2021/06/30/ behind-bidens- 2020-victory/; NBC News, South Carolina Presidential Election Results 2020 (Nov. 3, 2020), https://www.nbcnews.com/politics/2020-elections/ south-carolina-president-results/; N. Y. Times, South Carolina Exit Polls: How Different Groups Voted (Nov. 3, 2020), https://www.nytimes.com/ interactive/2020/11/03/us/elections/exit-polls-south-carolina.html.

10

nated the legislature's redistricting efforts. If either politics or race could explain a district's contours, the plaintiff has not cleared its bar.

Our decision in Easley v. Cromartie, 532 U. S. 234 (2001) (Cromartie II), illustrates the diffculties that plaintiffs must overcome in this context. There, the plaintiffs' case hinged on circumstantial evidence of a racial gerrymander such as expert testimony and discrepancies between the relevant district lines and traditional districting criteria. Id., at 240–241; see also Cooper, 581 U. S., at 321 (describing the direct evidence in Cromartie II as “extremely weak”). After the State asserted a partisan-gerrymandering defense, we faulted the plaintiffs for failing to show “that the legislature could have achieved its legitimate political objectives in alternative ways that are comparably consistent with traditional districting principles.” Cromartie II, 532 U. S., at 258. In other words, the plaintiffs failed to meet the high bar for a racial- Page Proof Pending Publication gerrymandering claim by failing to produce, among other things, an alternative map showing that a rational legislature sincerely driven by its professed partisan goals would have drawn a different map with greater racial balance. Since our decision in Cromartie II, any plaintiff with a strong case has had every incentive to produce such an alternative map.

Without an alternative map, it is diffcult for plaintiffs to defeat our starting presumption that the legislature acted in good faith. This presumption of legislative good faith directs district courts to draw the inference that cuts in the legislature's favor when confronted with evidence that could plausibly support multiple conclusions. See, e. g., Abbott v. Perez, 585 U. S. 579, 610–612 (2018). This approach ensures that “race for its own sake, and not other districting principles , was the legislature's dominant and controlling rationale in drawing its district lines.” Miller, 515 U. S., at 913; see also Cromartie I, 526 U. S., at 546 (noting that strict scrutiny is warranted when a map is “unexplainable on grounds other than race” (internal quotation marks omitted)).

11

Three additional reasons justify this presumption. First, this presumption refects the Federal Judiciary's due respect for the judgment of state legislators, who are similarly bound by an oath to follow the Constitution. Second, when a federal court fnds that race drove a legislature's districting decisions , it is declaring that the legislature engaged in “offensive and demeaning” conduct, Miller, 515 U. S., at 912, that “bears an uncomfortable resemblance to political apartheid,” Shaw I, 509 U. S., at 647. We should not be quick to hurl such accusations at the political branches. Third, we must be wary of plaintiffs who seek to transform federal courts into “weapons of political warfare” that will deliver victories that eluded them “in the political arena.” Cooper, 581 U. S., at 335 (Alito, J., concurring in judgment in part and dissenting in part). The presumption of good faith furthers each of these constitutional interests. It also explains why we have held that the plaintiff's evidentiary burden in these cases is especially stringent. See Cromartie II, 532 U. S., at 241. Page Proof Pending Publication If a plaintiff can demonstrate that race drove the mapping of district lines, then the burden shifts to the State to prove that the map can overcome the daunting requirements of strict scrutiny. Under this standard, we begin by asking whether the State's decision to sort voters on the basis of race furthers a compelling governmental interest. Cooper, 581 U. S., at 292. We then determine whether the State's use of race is “narrowly tailored”—i. e., “necessary”—to achieve that interest. This standard is extraordinarily onerous because the Fourteenth Amendment was designed to eradicate race-based state action. Students for Fair Admissions , Inc. v. President and Fellows of Harvard College, 600 U. S. 181, 206 (2023).

B

South Carolina has seven congressional districts, and this case concerns two of them, Districts 1 and 6. District 1 covers the State's southeast region, while District 6 covers its southwest and central regions. South Carolina's prior map,

12

which was enacted in 2011, split several counties between Districts 1 and 6, including Beaufort, Berkeley, Charleston, Colleton, and Dorchester Counties. See Figure 1, infra, at 16. The Department of Justice precleared the 2011 map, and a three-judge District Court upheld it against racial- gerrymandering and intentional vote-dilution claims after fnding that the legislature “demonstrat[ed] that [it] adhered to traditional race-neutral principles.” Backus v. South Carolina, 857 F. Supp. 2d 553, 560 (SC), summarily aff'd, 568 U. S. 801 (2012). The relevant part of that map is shown in Figure 1, infra, at 16.

Over the next decade, the 2011 map consistently yielded a 6-to-1 Republican-Democratic delegation—with one exception . In 2018, the Democratic candidate, with 50.7% of the votes, narrowly won District 1, which had previously elected Republican candidates.3 But in 2020, when the Republican Presidential candidate handily won the State, the Republican congressional candidate retook District 1 by a slender mar- Page Proof Pending Publication gin, winning 50.6% of the votes.4 South Carolina had to redraw its map after the 2020 census because two of the State's seven districts saw major population shifts. District 1 was overpopulated by 87,689 residents while District 6 was underpopulated by 84,741 residents . South Carolina therefore had to add voters to District 6 while subtracting voters from District 1 in order to comply with the principle of one person, one vote. The remaining districts also had to be modifed in order to bring the whole map into compliance with that requirement.

In September 2021, the Senate subcommittee tasked with drawing the new map issued guidance explaining that tradi-

3 N. Y. Times, South Carolina Election Results: First House District (Jan. 28, 2019), https://www.nytimes.com/elections/results/south-carolina- house-district-1.

4 N. Y. Times, South Carolina Election Results: First Congressional District (Nov. 3, 2020), https://www.nytimes.com/interactive/2020/11/03/us/ elections/results-south-carolina-house-district-1.html.

13

tional districting principles, such as respect for contiguity and incumbent protection, would guide the mapmaking process along with the strict equal-population requirement. At the same time, the Republican-controlled legislature also made it clear that it would aim to create a stronger Republican tilt in District 1. Senate Majority Leader Shane Massey , for instance, testifed at trial that partisanship was “one of the most important factors” in the process and that the Republican Party was “not going to pass a plan that sacrifced [District 1].” J. S. A. 265a. As he put it, the legislature 's adoption of any map that improved the Democrats' chance of reclaiming District 1 would constitute “political malpractice.” Id., at 276a. Contemporaneous evidence confrms that leaders in the legislature sought to “create a stronger Republican tilt” in District 1 while “honoring” other race-neutral, traditional districting criteria. 649 F. Supp. 3d 177, 187 (SC 2023); J. S. A. 333a–334a.

To draw its maps, the Senate turned to Will Roberts, a Page Proof Pending Publication nonpartisan staffer with 20 years of experience in state government . Roberts had “worked with the three-judge panel in Backus” and had routinely prepared “reapportionment plans for counties, cities[,] and school boards across the state.” 649 F. Supp. 3d, at 188. During the trial of this case, one of the judges praised Roberts's expertise and honesty on the record.5 Under the Senate's open-door policy, Roberts drew maps upon request for Republican and Democratic Senators alike. In making these maps, Roberts relied on political data from the 2020 Presidential election along with traditional districting criteria and input from various lawmakers, including Representative Jim Clyburn, whose

5 During the proceedings, one of the judges described Roberts as “a very precise guy” and a “good man.” J. S. A. 74a, 421a. That judge also remarked that he “always liked asking [Roberts] questions,” that “the legislature's blessed to have Mr. Roberts,” and that if Roberts says a report is not accurate, “that's good enough for [him].” Id., at 74a–75a, 254a, 263a.

14

recommendations would have preserved the strong Democratic tilt in his district (District 6) and included a version of District 1 with a black voting-age population (BVAP) of 15.48%. J. S. A. 127a.

The eventual map (Enacted Plan), see Figure 2, infra, at 17, differed from the 2011 map in three important respects that refected the legislature's priorities. First, the Enacted Plan unifed Beaufort and Berkeley Counties within District 1. This move enhanced the Republican advantage in District 1 because the moved-in portions of those counties leaned Republican. Second, to further increase the Republican lead in District 1, Roberts also put more of Dorchester County in District 1. These changes exacerbated the population imbalance between District 1 and District 6. Third, to cure this problem, Roberts moved a series of precincts in Charleston from District 1 to District 6. In keeping with the legislature's partisan objectives, the pre- Page Proof Pending Publication cincts moved out of District 1 had a 58.8% Democratic vote share.

By design, the legislature divided Charleston between Districts 1 and 6. This split was seen as in Charleston's best interests because it meant that the county would have two Representatives in the House—one Democrat, Representative Clyburn, who has represented District 6 since 1993 and has held important House leadership positions, and one Republican representing District 1. Republican Senator Chip Campsen, who spearheaded the mapmaking process , testified that Charleston benefits from bipartisan congressional representation on “bread-and-butter things” like port maintenance and “infuence with the incumbent administration.” J. S. A. 338a. As he explained, “I am tickled to death that Jim Clyburn represents Charleston County,” id., at 371a, because “Clyburn has more infuence with the Biden Administration perhaps than anyone in the nation,” id., at 338a. To achieve all these objectives, Rob-

15

erts moved roughly 193,000 residents between the districts with a net migration of 87,690 people into District 6. Id., at 439a, 443a.

The Enacted Plan achieved the legislature's political goal by increasing District 1's projected Republican vote share by 1.36% to 54.39%. The version of District 1 in the Enacted Plan also had a slightly higher BVAP, rising from 16.56% to 16.72%. The legislature voted to adopt the Enacted Plan, and the Governor signed it into law in January 2022. While the Enacted Plan was still in the making, the plaintiff-appellees in this case—the National Association for the Advancement of Colored People (NAACP) and Taiwan Scott, a voter in District 1 (collectively, the Challengers)— sued to contest the 2011 map on the ground that, in light of the 2020 census, it violated the one person, one vote requirement . After South Carolina passed the Enacted Plan, the Challengers amended their complaint to attack that map in- Page Proof Pending Publication stead. The Challengers alleged that Districts 1, 2, and 5 were racially gerrymandered and that these districts diluted the electoral power of the State's black voters. A three- judge District Court rejected these claims with respect to Districts 2 and 5. But the court held that South Carolina drew District 1 with a 17% BVAP “target” in mind and that this violated the Equal Protection Clause. For similar reasons , the court also found that the State's putative use of race to draw District 1 unlawfully diluted the black vote. The court permanently enjoined South Carolina from conducting elections in District 1 until it approved a new map. The State appealed to this Court, and we noted probable jurisdiction. 598 U. S. ––– (2023).

16

Page Proof Figure 1. 2011 Pending Map—Districts 1 and 6 (Exh. 1 toPublication State's Motion for Summary Judgment in South Carolina State Conference of the NAACP v. McMaster, No. 3:21–cv–3302 (D SC, Aug. 19, 2022), ECF Doc. 323–1, p. 2).

17

Page Proof Pending Publication Figure 2. Enacted Plan—Districts 1 and 6 (South Carolina House of Representatives, S. 865 Passed—As Signed by the Governor, https:// redistricting.schouse.gov/docs/plans/cpg/conpassed%20map.pdf).

III

The State contends that the District Court committed both legal error and clear factual error in concluding that race played a predominant role in the legislature's design of District 1. The State's principal legal argument is that the District Court did not properly disentangle race from politics. Because this argument, at bottom, attacks the factual basis of the District Court's fndings, we dispose of this case on clear-error grounds.

Under our case law, the Challengers bore the burden of showing that the “legislature subordinated traditional race- neutral districting principles . . . to racial considerations.” Miller, 515 U. S., at 916. In this case, the District Court clearly erred because the Challengers did not meet this “de-

18

manding” standard. Id., at 928 (O'Connor, J., concurring). They provided no direct evidence of a racial gerrymander, and their circumstantial evidence is very weak. Instead, the Challengers relied on deeply fawed expert reports. And while these experts produced tens of thousands of maps with differently confgured districts, they did not offer a single map that achieved the legislature's partisan goal while including a higher BVAP in District 1. Faced with this record, we must reverse the District Court on the racial- gerrymandering claim.

We divide our analysis into four parts. First, we set out the appropriate legal standard for reviewing a district court's factual findings in racial-gerrymandering cases. Second, we explain why the District Court's factual fndings are clearly fawed with respect to the Challengers' circumstantial evidence. Third, we examine the four expert reports that the Challengers presented below. And fnally, we explain that the District Court erred by not drawing an ad- Page Proof Pending Publication verse inference from the Challengers' failure to submit an alternative map that would have allowed the State to achieve its districting goals while maintaining a higher BVAP in District 1.

A

We review the District Court's factual fndings for clear error. That means we may not set those fndings aside unless , after examining the entire record, we are “left with the defnite and frm conviction that a mistake has been committed .” Cooper, 581 U. S., at 309 (internal quotation marks omitted). This is a demanding test, but it is not a rubber stamp.

Moreover, in a case like this, there is a special danger that a misunderstanding of what the law requires may infect what is labeled a fnding of fact. “[I]f [a] trial court bases its fndings upon a mistaken impression of applicable legal principles , the reviewing court is not bound by the clearly erroneous standard.” Inwood Laboratories, Inc. v. Ives

19

Laboratories, Inc., 456 U. S. 844, 855, n. 15 (1982); see also Abbott, 585 U. S., at 607. Here, the standard of proof that the three-judge court was required to apply, i. e., the racial- predominance test, has a very substantial legal component that must take account of our prior relevant decisions.6 And the application of this test calls for particular care when the defense contends that the driving force in its critical districting decisions (namely, partisanship) was a factor that is closely correlated with race. Thus, in a case like this, we must exercise special care in reviewing the relevant fndings of fact.

B

The District Court found that South Carolina drew District 1 with a racial “target,” namely, the maintenance of a 17% BVAP, and it concluded that this deliberate use of race rendered District 1's lines unlawful. See Bethune-Hill, 580 U. S., at 183–185. But the Challengers did not offer any di- Page Proof Pending Publication rect evidence to support that conclusion, and indeed, the direct evidence that is in the record is to the contrary. Roberts, the non-partisan career employee who drew the Enacted Plan, testifed that he used only political data, and his colleagues likewise steadfastly denied using race in drawing the Enacted Plan. None of the facts on which the District Court relied to infer a racial motive is suffcient to sup-

6 The dissent is correct to note that it is not enough for a plaintiff to show that race was a mere factor in the State's redistricting calculus. Rather, the plaintiff must show that race played a “ `predominant' ” role in shaping a district's lines. Post, at 82, n. 4 (opinion of Kagan, J.) (quoting Miller, 515 U. S., at 916). But the dissent then retreats from this standard because the State denied relying at all on racial data. Post, at 82, n. 4. That is a puzzling argument. Parties can stipulate to issues of fact, but they cannot by stipulation amend the law. See, e. g., United States Nat. Bank of Ore. v. Independent Ins. Agents of America, Inc., 508 U. S. 439, 447 (1993). And it would be uniquely perverse to deprive the State of a more generous constitutional standard simply because it made the laudable effort to disregard race altogether in the redistricting process.

20

port an inference that can overcome the presumption of legislative good faith.

First, the District Court inferred a racial motive from the fact that District 1's BVAP stayed around 17% “[d]espite all of th[e] changes” that South Carolina made during the redistricting process. 649 F. Supp. 3d, at 191. But where race and partisan preferences are very closely tied, as they are here, the mere fact that District 1's BVAP stayed more or less constant proves very little. If 100% of black voters voted for Democratic candidates, it is obvious that any map with the partisan breakdown that the legislature sought in District 1—something in the range of 54% Republican to 46% Democratic—would inevitably involve the removal of a disproportionate number of black voters. And since roughly 90% of black voters cast their ballots for Democratic candidates , the same phenomenon is very likely.

The District Court's own fndings substantiate the tight Page Proof Pending Publication correlation between the legislature's partisan aim and District 1's BVAP. During the redistricting process, the State considered a variety of maps, including those submitted by the Challengers. Maps with a Democratic-leaning District 1 had BVAP percentages that generally ranged between 21% to 24%. See App. 83; J. S. A. Supp. 142a. The District Court itself concluded that a 17% BVAP “produced a Republican tilt,” a 20% BVAP “produced a `toss up district,' ” and a 21% to 24% BVAP “produced a Democratic tilt.” 649 F. Supp. 3d, at 188. And the Challengers cannot point to even one map in the record that would have satisfed the legislature's political aim and had a BVAP above 17%. Thus, there is strong evidence that the district's BVAP of 17% was simply a side effect of the legislature's partisan goal. And certainly nothing rules out that possibility. In light of the presumption of legislative good faith, that possibility is dispositive.

The District Court's reasoning, however, is fatly inconsistent with that presumption. And what the court did—

21

inferring bad faith based on the racial effects of a political gerrymander in a jurisdiction in which race and partisan preference are very closely correlated—would, if accepted, provide a convenient way for future litigants and lower courts to sidestep our holding in Rucho that partisan- gerrymandering claims are not justiciable in federal court. Under the District Court's reasoning, a litigant could repackage a partisan-gerrymandering claim as a racial- gerrymandering claim by exploiting the tight link between race and political preference. Instead of claiming that a State impermissibly set a target Republican-Democratic breakdown, a plaintiff could simply reverse-engineer the partisan data into racial data and argue that the State impermissibly set a particular BVAP target. Our decisions cannot be evaded with such ease. For that reason, the District Court clearly erred in fnding that the legislature deliberately sought to maintain a particular BVAP just because the Page Proof Pending Publication maps that produced the sought-after partisan goal all had roughly the same BVAP.

Second, the District Court inferred a racial motive from certain changes that the State made in redrawing District 1, namely, the Enacted Plan moved more voters out of District 1 (approximately 140,000) than were needed to comply with the one person, one vote rule (about 88,000), and the Enacted Plan split Charleston and a few other counties even though the avoidance of such splits is a traditional redistricting objective. But here, again, the State's avowed partisan objective easily explains these facts. The State claims it sought to ensure that District 1 had a reliable Republican majority, and simply removing 88,000 voters without regard to their party preferences would not have satisfed that objective. Similarly, the high priority that the legislature gave to its partisan goal provides an entirely reasonable explanation for the subordination of other objectives such as the avoidance of county splits. See Cooper, 581 U. S., at 308 (“[P]olitical and racial [gerrymanders]

22

are capable of yielding similar oddities in a district's boundaries”).

Third, the District Court found it telling that many predominantly black Charleston precincts were moved out of District 1 and into District 6. But because of the tight correlation between race and partisan preferences, this fact does little to show that race, not politics, drove the legislature's choice. The Charleston County precincts that were removed are 58.8% Democratic. Thus, the legislature 's stated partisan goal can easily explain this decision, and the District Court therefore erred in crediting the less charitable conclusion that the legislature's real aim was racial.

Fourth, the District Court placed too much weight on the fact that several legislative staffers, including Roberts, viewed racial data at some point during the redistricting process. This acknowledgment means little on its own be- Page Proof Pending Publication cause we expect that “[r]edistricting legislatures will . . . almost always be aware of racial demographics.” Miller, 515 U. S., at 916. Here, Roberts testifed without contradiction that he considered the relevant racial data only after he had drawn the Enacted Plan and that he generated that data solely for a lawful purpose, namely, to check that the maps he produced complied with our Voting Rights Act precedent. J. S. A. 92a, 205a, 379a.

The District Court discredited this testimony, but it cited no evidence that could not also support the inference that politics drove the mapmaking process. And the court provided no explanation why a mapmaker who wanted to produce a version of District 1 that would be safely Republican would use data about voters' race rather than their political preferences. Why would Roberts have used racial data—with the associated legal risks—as a proxy for partisan data when he had access to refned, sub-precinct-level political data that accounted for voter turnout and electoral

23

preferences? The District Court provided no answer to this obvious question.7 The Challengers look to plug this gap by arguing that Roberts must have used racial data because the political data he claimed to have used was blatantly unsatisfactory. For support, they cite the testimony of Dale Oldham, a political consultant who did not participate in drawing the Enacted Plan. Oldham testifed that he believed the standard data South Carolina used for measuring partisanship is unreliable because it does not accurately refect the partisan preferences of absentee voters. Oldham opined that a new type of composite data that frst became available in 2020 does a better job in that regard. J. S. A. Supp. 417a–418a, 420a.

This criticism is entitled to little weight. One consultant's opinion about the quality of South Carolina's political data obviously does not settle the question whether the State's political data was inferior. And in any event, the relevant Page Proof Pending Publication question is not whether the State used the best available data but whether it is reasonable to infer that the mapmakers ' political data was so obviously fawed that they must have surreptitiously used racial data. Oldham's testimony falls far short of establishing that the State cannot plausibly have believed that its own political data was suffcient. Nothing in our case law requires the State to adopt novel methodologies in analyzing election data. Indeed, the State plausibly argues that its data was more than good enough for its purposes because it showed partisan preferences at

7 The dissent argues that racial data is superior because black Democrats are more loyal to the party than white Democrats. Post, at 86–87. But whether or not this is true (and the dissent relies solely on the say-so of one witness), studies show that non-white voters turn out at a much lower rate than white voters. See Brennan Center for Justice, K. Morris & C. Grange, Large Racial Turnout Gap Persisted in 2020 Election (Aug. 6, 2021), https://www.brennancenter.org/our-work/analysis-opinion/ large-racial-turnout-gap-persisted-2020-election.

24

the sub-precinct level and also accounted for variations in voter turnout. Reply Brief 9, 11; J. S. A. 93a.

In sum, the District Court's heavy reliance on these four pieces of evidence was seriously misguided in light of the appropriate legal standard and our repeated instructions that a court in a case such as this must rule out the possibility that politics drove the districting process.

C

Once these weak inferences are set aside, all that the Challengers have left are four expert reports. But these reports are fawed because they “ignored certain traditional districting criteria” such as geographical constraints and the legislature 's partisan interests. Allen v. Milligan, 599 U. S. 1, 34 (2023). Because these reports do not replicate the “myriad considerations” that a legislature must balance as part of its redistricting efforts, they cannot sustain a fnding that race Page Proof Pending Publication played a predominant role in the drawing of District 1's lines. Id., at 35. We will discuss each of the Challengers' four experts in turn.

Dr. Kosuke Imai. The report of the Challengers' frst expert , Dr. Kosuke Imai, provides no support for the decision below because Dr. Imai made no effort to disentangle race from politics. Dr. Imai developed a computer algorithm that generated 20,000 maps of the State's congressional districts that complied with the one person, one vote rule. This algorithm did not take race into account, and it sought to respect traditional redistricting objectives such as contiguity and compactness. The Challengers assert that these maps prove that race drove the State's redistricting process because the average District 1 in these simulations contained a higher BVAP than the District 1 in the Enacted Plan.

The Challengers' inference is fawed because Dr. Imai's models failed to consider partisanship. See J. S. A. Supp. 30a (acknowledging that “no race or partisan information was used” (emphasis added)). That is a fatal omission in this

25

case. As noted, race and politics strongly correlate in South Carolina, and Dr. Imai's algorithm produced maps without requiring that District 1 comply with the legislature's asserted aim of ensuring that District 1 remain a relatively safe Republican seat. The effect of Dr. Imai's omission can be seen by looking at the Democratic vote share (measured by the results in the 2020 Presidential election) in the versions of District 1 that his simulations produced. President Biden's vote share in the average District 1 in Dr. Imai's maps was signifcantly higher than his vote share in the version of District 1 in the Enacted Plan. Rebuttal Report of Sean Trende in South Carolina State Conference of the NAACP v. McMaster, No. 3:21–cv–3302 (D SC, Aug. 19, 2022), ECF Doc. 323–33, pp. 5–6. Indeed, Dr. Sean Trende, the State's expert, showed that District 1 would have voted for the Democratic nominee in 2020 in 91% of Dr. Imai's simulations . Ibid. Because Dr. Imai's model fails to track the Page Proof Pending Publication considerations that governed the legislature's redistricting decision, it is irrelevant that the racial makeup of District 1 in his maps differs from that in the version of the district in the Enacted Plan.

It is also noteworthy that Dr. Imai could have easily controlled for partisan preferences just as he controlled for other redistricting factors such as compactness and county splits. He could have generated maps conditioned on District 1's vote share matching or exceeding the Benchmark Plan's Republican tilt. But he did not take that obvious step.

The Challengers seek to excuse their failures to disentangle race and politics by arguing that South Carolina raised a partisan-gerrymandering defense for the frst time during the trial, but this argument rests on the implausible premise that the Challengers were unaware of the legislature's partisan concerns during the mapmaking process. The fact of the matter is that politics pervaded the highly visible mapmaking process from start to fnish. The Republican and

26

Democratic caucuses submitted competing maps, and the Enacted Plan passed the legislature by a margin of 26 to 15 in the Senate and 72 to 33 in the House, with only Democrats voting in opposition. The public hearings and legislative debates are of a piece. For example, Senator Margie Bright Matthews, a black Democrat, said in a foor debate with Senator Campsen that “ `we're not going to get into the racial gerrymandering thing because you and I both know in Charleston it matters not about your race. It is just that you went by how those folks voted.' ” App. 296. For evidence , she recognized that the Enacted Plan also moved into District 6 predominantly white parts of Charleston that skewed Democratic, such as West Ashley. She added, “ `Senator [Campsen], . . . I really appreciate you agreeing with me that our opposition . . . is not about racial [gerrymandering ].' ” Ibid. Instead, she said, it was about “ `packing ' ” the Democratic-voting area of Charleston into District Page Proof Pending Publication 6 “ `to make [District 1] more electable.' ” Ibid. Former Congressman Cunningham, the Democrat who represented District 1 from 2018 to 2020, also criticized the Enacted Plan's District 1 lines as “ `mak[ing] no sense unless, of course, the sole purpose . . . is to make it harder for a Republican to lose.' ” Id., at 295. He added that “the folks in Washington, D. C.,” did not want a repeat of the 2018 election or even the 2020 election where he lost against the Republican nominee by “a single point in one of the closest elections in the entire country.” Ibid. Under these circumstances, it is safe to say that the Challengers were on notice that the State would raise a partisan-gerrymandering defense at trial.

Dr. Imai's conspicuous failure to control for party preference is alone suffcient to discredit any reliance on his report, but his report exhibited another serious faw: it failed to consider “core district retention,” a term that “refers to the proportion of districts that remain when a State transitions from one districting plan to another.” Allen, 599 U. S., at

27

21. The Enacted Plan retains 83% of District 1's core, but the average map produced by Dr. Imai's model scored 69% on the core-district-retention metric—three standard deviations lower. ECF Doc. 323–33, at 5.

Dr. Imai's failure to consider core retention betrays a blinkered view of the redistricting process. Lawmakers do not typically start with a blank slate; rather, they usually begin with the existing map and make alterations to ft various districting goals. Core retention recognizes this reality. Dr. Imai could have controlled for this metric by restricting the core retention in his simulations to at least 83%. His failure to do so here means we cannot rule out core retention as another plausible explanation for the difference between the Enacted Plan and the average Imai simulation.

Dr. Jordan Ragusa. As evidence that race predominated in District 1's design, the District Court also credited a report by Dr. Jordan Ragusa, another expert for the Chal- Page Proof Pending Publication lengers. Unlike Dr. Imai, Dr. Ragusa attempted to disentangle race from politics, but as we will explain, his analysis has at least two serious defects. First, he failed to account for two key mapmaking factors: contiguity and compactness. Second, he used an inferior method of measuring a precinct's partisan leanings.

We begin with the matter of contiguity and compactness. Dr. Ragusa used three separate models, but none of them controlled for these critical districting factors. Two of his models employed the so-called county envelope approach. Using this approach, he frst identifed the fve counties that have at least one precinct that fell within District 1 in the Benchmark Plan. These counties in their entirety constituted the “county envelope.”

Dr. Ragusa employed a method that we will discuss below to control for the partisan preferences of voters in these precincts , and he also controlled for precinct size. He then asked whether a precinct of a given size with a given parti-

28

san breakdown was more or less likely to be included in District 1 depending on its racial demographics, and he reported that districts with a high percentage of black voters were more likely to be excluded.

His remaining model looked only at the precincts that were in District 1 in the Benchmark Plan, and controlling in the same way for size and partisan leaning, he reported that a precinct was more likely to be moved out if it had a high percentage of black voters.

All three of these models exhibit the same faw. Because they did not control for contiguity or compactness, they all assume that a precinct could be moved into or out of District 1 regardless of its distance from the line between that district and District 6. That is highly unrealistic. A simple example illustrates this point in relation to the county envelope approach, as can be seen with a quick look at Figure 1, which we again reproduce below. Page Proof Pending Publication

29

Under Dr. Ragusa's methodology, any precinct in Colleton County could have been moved into District 1, but many precincts in that county are nowhere near District 1's outer boundaries. For example, precincts near the county's northern border with Bamberg County could not have been moved into District 1 without egregiously fouting the State's important interests in contiguity or compactness. And the same problem arises with respect to the question whether a precinct in District 1 in the Benchmark Plan could have been moved into District 6. Precincts in District 1 that are not close to the district line could not have been moved without making District 6 less contiguous or compact.8 We have already rejected a plaintiff's expert report for failing to account for this feature of mapmaking. In Cromartie II, we faulted the plaintiff's expert for failing to consider whether the excluded precincts “were located near Page Proof Pending Publication enough to [the district's] boundaries or each other for the legislature as a practical matter to have drawn [the district's] boundaries to have included them, without sacrifcing other important political goals.” 532 U. S., at 247. The District

8 The dissent excuses Dr. Ragusa's failure to control for contiguity on the ground that a vast majority of the precincts in old District 1 could have been moved into District 6 without violating contiguity. Post, at 94. However, a quick look at the precincts in the counties that fall within District 1 shows that this is plainly untrue. (Links to some of the relevant precinct maps are provided below.) Many precincts would have had to jump over quite a few others in order to join District 6. In addition, the dissent ignores the other objectives that the new map sought to achieve, namely, the unifcation of Beaufort and Berkeley Counties and the division of Charleston between Districts 1 and 6 so that the city would predictably have one Democratic House Member and one Republican House Member. For the voting precincts in Beaufort County, see https:// rfa.sc.gov/sites/default/files/2024-01/Beaufort%20Precincts%202024.pdf. For Berkeley County, see https://rfa.sc.gov/sites/default/files/2022-04/ Berkeley%20Precincts.pdf.

30

Court clearly erred in crediting Dr. Ragusa's models because his approach made that same mistake.

Dr. Ragusa's report also carries less weight because of how he measured a precinct's partisan leanings. Using the results of the 2020 Presidential election, Dr. Ragusa measured partisan tilt by looking at the total votes cast for President Biden, not the net votes for President Biden. This method fails to account for the fact that voter turnout may vary signifcantly from precinct to precinct, and therefore a precinct in which a candidate gets a large number of votes may also be a precinct in which the candidate fails to win a majority . To illustrate this point, consider this simplified example:

Precinct 1 Precinct 2 Total Voting Age 1,250 1,250 Population Page Proof Pending Publication Biden Vote 500400 Trump Vote 250 600 Net Biden Votes 150 -100 Biden Vote % 62% 45% Black Voting Age 250 0 Population Moved from Dis- Yes No trict 1 to District 6

Dr. Ragusa's model considers only the total number of Biden votes in its partisanship analysis. J. S. A. 502a. But legislators aiming to make District 1 a relatively safe Republican seat would be foolish to exclude Precinct 2 merely because it has more Democratic votes than Precinct 1. Instead , they would look at the net Democratic votes and would thus remove Precinct 1, not Precinct 2. Although the use of total votes may be a statistically permissible measure of partisan lean, it is undoubtedly preferable for an expert re-

31

port to rely on net votes when measuring a district's partisan lean. The Challengers seek to defend Dr. Ragusa's report by suggesting that he followed the same methodology as Professor Stephen Ansolabehere, whose report we blessed in Cooper, 581 U. S., at 315, but that is wrong. There are important differences between Dr. Ragusa's methodology and Professor Ansolabehere's,9 and in all events, Professor Ansolabehere 's report played a minor role in Cooper, where the plaintiffs could also point to direct evidence. Here, by contrast , once the District Court's other circumstantial fndings are set aside, the Challengers must rest their entire case on these expert reports. Dr. Ragusa's report, on its own, cannot prove that District 1's lines are “unexplainable on grounds other than race.” Shaw I, 509 U. S., at 644 (internal quotation marks omitted).

Dr. Baodong Liu. Dr. Baodong Liu, another of the Challengers ' experts, submitted a report that purported to show Page Proof Pending Publication that race rather than politics explains District 1's design. Although the District Court did not cite Dr. Liu's report, the Challengers contend that it bolsters the District Court's fndings. Tr. of Oral Arg. 86–87. But Dr. Liu's methodology was plainly fawed. First, his methodology, like Dr. Ragusa's, failed to account for contiguity and compactness. Dr. Liu examined all voters living within the county envelope for District 1 of the

9 Two differences in particular stand out. First, while Dr. Ragusa looked only at Democratic voters to control for partisanship, Professor Ansolabehere looked at both Democratic and Republican voters. 1 App. in Cooper v. Harris, O. T. 2016, No. 15–1262, pp. 334–337. Only after calculating the percentage of black voters moved in each partisan group did Professor Ansolabehere conclude that “race, and not party, had a disproportionate effect on the confguration of ” the congressional districts. Id., at 337. Second, Professor Ansolabehere's analysis operated at the voter level. Id., at 313–314. That enabled him to compare the demographics of the moved voters to the general population in a way that Dr. Ragusa's precinct-level analysis cannot.

32

Enacted Plan to see which voters were more likely to have been excluded. His analysis suggested that black Democrats were more likely to have been excluded than white Democrats.

This methodology was highly unrealistic because it treated each voter as an independent unit that South Carolina could include or exclude from District 1. No mapmaker who respects contiguity and compactness could take such an approach . For example, a mapmaker could not assign a black Republican to one district while moving a black Democrat who lives in the same apartment building to another district. To accurately refect the districting process, an analysis would have to pay attention to whether a voter's neighbors were moved too.

This defect alone is suffcient to preclude reliance on Dr. Liu's report, but that report exhibited another faw: it used inferior data to measure a district's partisan tilt. While the State used voting data from the 2020 Presidential election, Page Proof Pending Publication Dr. Liu relied on data from the 2018 gubernatorial primaries . Data from that gubernatorial primary is less informative because far fewer voters turn out for off-cycle gubernatorial primary elections. The numbers prove the point. In the 2018 elections, a total of about 610,000 votes were cast across both primaries; in the 2020 Presidential election, by contrast, a total of 2.5 million votes were cast.10 Because Dr. Liu examined only a small, highly non-random sample of the regular voting electorate, we cannot say that the same results would hold true if he had applied his methodology to the State's 2020 data.

10 See N. Y. Times, South Carolina Governor Primary Election Results (June 20, 2018), https://www.nytimes.com/elections/results/south-carolinagovernor -primary-election; N. Y. Times, South Carolina Presidential Election Results (Nov. 3, 2020), https://www.nytimes.com/interactive/2020/11/ 03/us/elections/results-south-carolina-president.html; see also App. 135 (testimony of Baodong Liu) (noting that Presidential election years “usually ha[ve] a very high level of voter turnout”).

33

Dr. Moon Duchin. Dr. Moon Duchin, the fnal expert put forward by the Challengers, provided a report assessing whether the Enacted Plan “cracks” black voters among multiple districts in a way that produced “discernible vote dilution .” J. S. A. Supp. 127a. After fnding that the Enacted Plan diluted the black vote, Dr. Duchin concluded that it is “not plausible” that the dilution was a mere “side effect of partisan concerns.” Id., at 175a.

Neither the District Court nor the Challengers cite Dr. Duchin's report to support the racial-predominance fnding, and that is for a good reason. Like Dr. Imai's report, various parts of Dr. Duchin's report did not account for partisanship or core retention. App. 102–103. Moreover, Dr. Duchin 's conclusion was based on an assessment of the map as a whole rather than District 1 in particular. A state-wide analysis cannot show that District 1 was drawn based on race. See Bethune-Hill, 580 U. S., at 191 (“[T]he basic unit of analysis for racial gerrymandering claims . . . is the dis- Page Proof Pending Publication trict”); Alabama Legislative Black Caucus, 575 U. S., at 262–263 (a racial-gerrymandering claim “does not apply to a State considered as an undifferentiated `whole' ”). Given these serious problems, it is no wonder that the challengers cite Dr. Duchin's report only in support of their racial vote- dilution claim. It has no probative force with respect to their racial-gerrymandering claim regarding District 1's boundaries.

To sum up our analysis so far, no direct evidence supports the District Court's fnding that race predominated in the design of District 1 in the Enacted Plan. The circumstantial evidence falls far short of showing that race, not partisan preferences, drove the districting process, and none of the expert reports offered by the Challengers provides any signifcant support for their position.11 11 The dissent, by contrast, would make it virtually impossible to show clear error in a case like this. The dissent agrees that a plaintiff raising a racial-gerrymandering claim bears a “demanding burden.” Post, at 75

34

D

In addition to all this, the District Court also critically erred by failing to draw an adverse inference against the Challengers for not providing a substitute map that shows how the State “could have achieved its legitimate political objectives” in District 1 while producing “signifcantly greater racial balance.” Cromartie II, 532 U. S., at 258. We have repeatedly observed that an alternative map of this sort can go a long way toward helping plaintiffs disentangle race and politics. In Cooper, we expressed “no doubt that an alternative districting plan . . . can serve as key evidence in a race-versus-politics dispute.” 581 U. S., at 317. By showing that a rational legislature, driven only by its professed mapmaking criteria, could have produced a different map with “greater racial balance,” Cromartie II, 532 U. S., at 258, an alternative map can perform the critical task of distinguishing between racial and political motivations when Page Proof Pending Publication race and partisanship are closely entwined. For that reason , we have said that when all plaintiffs can muster is “meager direct evidence of a racial gerrymander” “only [an alter-

(opinion of Kagan, J.). But according to the dissent's view, clear-error review means that this burden vanishes on appeal because a plaintiff's “hardest job should be done” once it prevails before a three-judge district court. Ibid. That misses the point. In assessing whether a fnding is clearly erroneous, it is important to keep in mind the standard of proof that the district court was required to apply. It is hornbook law, after all, that we must ask on appeal whether the “factfnder in the frst instance made a mistake in concluding that a fact had been proven under the applicable standard of proof.” Concrete Pipe & Products of Cal., Inc. v. Construction Laborers Pension Trust for Southern Cal., 508 U. S. 602, 622– 623 (1993) (emphasis added); see also H. Edwards & L. Elliott, Federal Standards of Review 26 (3d ed. 2018) (“[I]n applying the clearly erroneous standard, a reviewing court must take account of the standard of proof informing the trial court's factual fnding”). Once our task is framed properly, we can easily conclude for the reasons that follow that the District Court clearly erred when it found that the Challengers carried their “demanding burden.”

35

native] ma[p] of that kind” can “carry the day.” Cooper, 581 U. S., at 322.

Nor is an alternative map diffcult to produce. Any expert armed with a computer “can easily churn out redistricting maps that control for any number of specifed criteria, including prior voting patterns and political party registration .” Id., at 337 (opinion of Alito, J.). The Challengers enlisted four experts who could have made these maps at little marginal cost. Dr. Imai's simulations generated 20,000 different maps—but none that actually controlled for politics. The evidentiary force of an alternative map, coupled with its easy availability, means that trial courts should draw an adverse inference from a plaintiff's failure to submit one. The adverse inference may be dispositive in many, if not most, cases where the plaintiff lacks direct evidence or some extraordinarily powerful circumstantial evidence such as the “strangely irregular twenty-eight-sided” district lines in Go- Page Proof Pending Publication million v. Lightfoot, 364 U. S. 339, 341 (1960), which betrayed the State's aim of segregating voters on the basis of race with “mathematical” precision, ibid.

The District Court, however, misunderstood our case law when it held that an alternative map is relevant only for the purpose of showing that a remedy is plausible. 649 F. Supp. 3d, at 198–199. Because “a constitutionally compliant plan for [District 1] can be designed without undue diffculty,” the District Court concluded that it was “not necessary for Plaintiffs to present an acceptable alternative map to prevail on their claims.” Id., at 199. That is wrong. A plaintiff's failure to submit an alternative map—precisely because it can be designed with ease—should be interpreted by district courts as an implicit concession that the plaintiff cannot draw a map that undermines the legislature's defense that the districting lines were “based on a permissible, rather than a prohibited, ground.” Cooper, 581 U. S., at 317. The District Court's conclusions are clearly erroneous because it did not follow this basic logic.

36

E

Despite its length, the dissent boils down to six main points. None is valid.

First, the dissent suggests that clear-error review is a perfunctory task, see post, at 75, but that is not so. While district court fndings of fact are generally correct, conscientious district courts sometimes err, and appellants are entitled to meaningful appellate review. Does the dissent really think that all district court fndings on the question of racial discrimination are virtually immune from reversal?

Second, the dissent attacks the proposition that in redistricting cases the “good faith of [the] state legislature must be presumed.” Miller, 515 U. S., at 915. But, as the citation to Justice Kennedy's opinion for the Court in Miller reveals , that presumption is an established feature of our case law.

Third, the dissent claims that our decision is inconsistent Page Proof Pending Publication with Cooper, but the dissent's argument is based on an imaginary version of that opinion. Nothing in Cooper is inconsistent with the venerable rule that a factfnder may draw an adverse inference when a party fails to produce highly probative evidence that it could readily obtain if in fact such evidence exists. See Interstate Circuit, Inc. v. United States, 306 U. S. 208, 226 (1939); see also 2 J. Wigmore, Evidence in Trials at Common Law § 291, pp. 227–229 (J. Chadbourn rev. 1979). “[T]his rule can be traced as far back as 1722” and “has been utilized in scores of modern cases.” International Union, United Auto, Aerospace and Agricultural Implement Workers of Am. (UAW) v. NLRB, 459 F. 2d 1329, 1336 (CADC 1972). The dissent is correct that this inference “pack[s] a wallop” in such cases, post, at 75, but that is only because an adequate alternative map is remarkably easy to produce—as demonstrated by the fact that the Challengers introduced tens of thousands of other maps into the record. Under such circumstances, if a sophisticated plaintiff bringing a racial-gerrymandering claim cannot pro-

37

vide an alternative map, that is most likely because such a map cannot be created. It would be clear error for the factfnder to overlook this shortcoming.

Fourth, the dissent argues that the Challengers were blindsided when the State argued at trial that its map was drawn to achieve a political goal. Post, at 79. But there is ample evidence that the State's aim was well known before trial. See supra, at 25–26. And neither the Challengers nor the dissent can explain why the Challengers' experts, who created thousands of maps that took into account all sorts of variables, supposedly never even tried to create a District 1 that had a higher BVAP while achieving the legislature's political goals. Nor can they explain why, if such a map can be created, the Challengers' experts did not produce one during the trial.

Fifth, the dissent makes much of the fact that Roberts had taken racial demographics into account in drawing maps in Page Proof Pending Publication the past and was aware of the racial makeup of the various districts he created in this case. But there is nothing nefarious about his awareness of the State's racial demographics. Roberts has spent nearly 20 years drawing maps for various state and local initiatives, and it is therefore entirely unsurprising that he exhibited a wealth of knowledge about who lives in which part of the State. Cf. Miller, 515 U. S., at 916 (state redistricting offcials “will . . . almost always be aware of racial demographics” during the districting process). The dissent seeks to undercut Roberts's credibility by labeling him “a veteran consumer of racial data.” Post, at 83. We think it is unfair for the dissent to question his credibility simply because he, like every other expert who has ever worked on a Voting Rights Act case, has had to “consum[e] . . . racial data” to comply with our precedents.

Finally, the dissent thinks that the State must have used racial data because that data, in its view, is more accurate than political data in predicting future votes. Refusing to use the racial data, according to the dissent, would have re-

38

quired the “self-restraint of a monk.” Post, at 86. This jaded view is inconsistent with our case law's longstanding instruction that the “good faith of [the] state legislature must be presumed” in redistricting cases. Miller, 515 U. S., at 915. And in any event, there is little reason to think that it requires much restraint for a mapmaker with a political aim to use data that bears directly on what he is trying to achieve, namely, political data. That is especially so where, as here, the political data, unlike the racial data that the dissent prefers, took into account voter turnout. See supra, at 22–23, and n. 7.

In sum, there is no substance to the dissent's attacks.

IV

The Challengers also raised an independent vote-dilution claim. The District Court held that this claim was governed by the “same fndings of fact and reasoning” that guided its Page Proof Pending Publication racial-gerrymandering analysis, and it thus entered judgment for the Challengers on that ground as well. 649 F. Supp. 3d, at 198. But in light of our conclusion that those fndings were clearly erroneous, that conclusion cannot stand. Moreover, the District Court's analysis did not take into account the differences between vote-dilution and racial- gerrymandering claims.

A racial-gerrymandering claim asks whether race predominated in the drawing of a district “regardless of the motivations ” for the use of race. Shaw I, 509 U. S., at 645. The racial classifcation itself is the relevant harm in that context. A vote-dilution claim is “analytically distinct” from a racial- gerrymandering claim and follows a “different analysis.” Id., at 650, 652. A plaintiff pressing a vote-dilution claim cannot prevail simply by showing that race played a predominant role in the districting process. Rather, such a plaintiff must show that the State “enacted a particular voting scheme as a purposeful device to minimize or cancel out the voting potential of racial or ethnic minorities.” Miller, 515

Justice Thomas, concurring

U. S., at 911 (internal quotation marks omitted). In other words, the plaintiff must show that the State's districting plan “has the purpose and effect” of diluting the minority vote. Shaw I, 509 U. S., at 649 (emphasis added).

In light of these two errors in the District Court's analysis of the Challengers' vote-dilution claim, a remand is appropriate.

* * *

We reverse the judgment of the District Court in part and remand the case in part for further proceedings consistent with this opinion.

It is so ordered.

Justice Thomas, concurring in part. I join all but Part III–C of the Court's opinion. The Court correctly concludes that the judgment below must be reversed under our precedents. Although I fnd the analy- Page Proof Pending Publication sis in Part III–C persuasive, clear-error review is not an invitation for the Court to “sift through volumes of facts” and “argue its interpretation of those facts.” Easley v. Cromartie , 532 U. S. 234, 262 (2001) (Thomas, J., dissenting). The Court's searching review of the expert reports exceeds the proper scope of clear-error review. But, that analysis is not necessary to resolve the case. In Part III–B, the Court explains that the District Court failed to evaluate evidence refecting the correlation between race and politics with the necessary presumption of legislative good faith. Ante, at 19–24. And, in Part III–D, it explains that the District Court failed to properly account for the plaintiffs' failure to produce an alternative map. Ante, at 34–35. Both of those mistakes are reversible legal errors.

I write separately to address whether our voting-rights precedents are faithful to the Constitution. This case is unique because it presents solely constitutional questions. The plaintiffs do not rely on the Voting Rights Act of 1965 for any of their claims. Nor do the South Carolina off-

40

cials invoke the Voting Rights Act as part of their defense . There can be no more propitious occasion to consider the constitutional underpinnings of our voting-rights jurisprudence.

The plaintiffs press two distinct constitutional claims. First, they bring a “racial gerrymandering” claim, alleging that South Carolina drew its new Congressional District 1 to sort black voters based on their race. To prevail on that claim under our precedents, the plaintiffs must show that race was the “predominant factor” in the legislature's approach to drawing the district. Miller v. Johnson, 515 U. S. 900, 916 (1995). Second, they bring a “vote dilution” claim, alleging that South Carolina drew District 1 to intentionally reduce the voting strength of the district's black residents. To prevail on that claim under our precedents, the plaintiffs must show that District 1's design reduces “minority voters' ability, as a group, `to elect the candidate of their choice.' ” Page Proof Pending Publication Shaw v. Reno, 509 U. S. 630, 641 (1993) (quoting Allen v. State Bd. of Elections, 393 U. S. 544, 569 (1969)).

In my view, the Court has no power to decide these types of claims. Drawing political districts is a task for politicians , not federal judges. There are no judicially manageable standards for resolving claims about districting, and, regardless , the Constitution commits those issues exclusively to the political branches.

The Court's insistence on adjudicating these claims has led it to develop doctrines that indulge in race-based reasoning inimical to the Constitution. As we reiterated last Term, “ `[o]ur Constitution is color-blind.' ” Students for Fair Admissions , Inc. v. President and Fellows of Harvard College, 600 U. S. 181, 230 (2023) (quoting Plessy v. Ferguson, 163 U. S. 537, 559 (1896) (Harlan, J., dissenting)). A colorblind Constitution does not require that racial considerations “predominate ” before subjecting them to scrutiny. Nor does it tolerate groupwide judgments about the preferences and be-

41

liefs of racial minorities. It behooves us to abandon our misguided efforts and leave districting to politicians.

I

Determining the proper shape of a district is a political question not suited to resolution by federal courts. The questions presented by districting claims are “ `nonjusticiable ,' or `political questions.' ” Vieth v. Jubelirer, 541 U. S. 267, 277 (2004) (plurality opinion). We have explained that a question is nonjusticiable when there is “ `a lack of judicially discoverable and manageable standards for resolving' ” the issue or “ `a textually demonstrable constitutional commitment of the issue to a coordinate political department.' ” Id., at 277–278 (quoting Baker v. Carr, 369 U. S. 186, 217 (1962)).

In Rucho v. Common Cause, 588 U. S. 684 (2019), we applied those principles to conclude that partisan gerrymander- Page Proof Pending Publication ing claims are nonjusticiable. Partisan gerrymandering claims allege that a political map unduly favors one political party over another. We explained that partisan gerrymandering claims therefore present questions about how to “apportion political power as a matter of fairness,” despite the fact that “[t]here are no legal standards discernible in the Constitution for making such judgments, let alone limited and precise standards that are clear, manageable, and politically neutral.” Id., at 705, 707. We bolstered our conclusion by reference to “the Framers' decision to entrust districting to political entities” in the Elections Clause, Art. I, § 4, cl. 1. Id., at 697, 701. Because courts “have no commission to allocate political power and infuence in the absence of a constitutional directive or legal standards to guide us in the exercise of such authority,” we held that partisan gerrymandering claims are nonjusticiable. Id., at 721.

The same logic demonstrates that racial gerrymandering and vote dilution claims are also nonjusticiable. As with

42

partisan gerrymandering claims, the racial gerrymandering and vote dilution claims in this case lack “judicially discoverable and manageable standards” for their resolution. Vieth, 541 U. S., at 277–278 (plurality opinion) (internal quotation marks omitted). And, they ask us to address an issue—congressional districting—that is textually committed to a coordinate political department, Congress. Id., at 277. As a result , racial gerrymandering and vote dilution claims brought under the Fourteenth and Fifteenth Amendments are nonjusticiable.

A

Racial gerrymandering and vote dilution claims lack “ `judicially discoverable and manageable standards' ” for their resolution. Id., at 277–278 (quoting Baker, 369 U. S., at 217). Both types of claims turn on questions that cannot be answered through the kind of reasoning that constitutes an exercise of the “judicial Power.” Art. III, § 1, cl. 1. I address Page Proof Pending Publication in turn the reasons why each claim is unmanageable.

1

Racial gerrymandering claims ask courts to reverse- engineer the purposes behind a complex and often arbitrary legislative process. The standard developed under our precedents “require[s] the plaintiff to show that race was the `predominant factor motivating the legislature's decision to place a signifcant number of voters within or without a particular district.' ” Ante, at 7 (quoting Miller, 515 U. S., at 916). In other words, “a plaintiff must prove that the legislature subordinated traditional race-neutral districting principles . . . to racial considerations.” Id., at 916. The Court's focus on legislative purpose is unavoidable because “the constitutional violation in racial gerrymandering cases stems from the racial purpose of state action,” not the resulting map. Bethune-Hill v. Virginia State Bd. of Elections , 580 U. S. 178, 189 (2017) (internal quotation marks omitted).

43

Divining legislative purpose is a dubious undertaking in the best of circumstances, but the task is all but impossible in gerrymandering cases. “Electoral districting is a most diffcult subject for legislatures,” a pure “exercise [of] the political judgment necessary to balance competing interests .” Miller, 515 U. S., at 915. We have therefore cautioned courts to “be sensitive to the complex interplay of forces that enter a legislature's redistricting calculus.” Id., at 915–916.

In cases without smoking-gun evidence, the only practical way to prove that a State considered race when drawing districts is to “show that the State's chosen map conficts with traditional redistricting criteria.” 1 Ante, at 8. The Court's racial gerrymandering precedents use the term “ `traditional districting principles' ” to refer to the “competing interests” and “complex . . . forces” involved in drawing districts. Miller, 515 U. S., at 915–916, 919 (quoting Shaw, 509 U. S., at 647). Judging a map's consistency or confict Page Proof Pending Publication with traditional districting principles requires a court to ascertain what kinds of maps should result from the application of those principles.

Determining how a legislature would have drawn district lines in a vacuum is a fool's errand. Indeed, as we have defned them, “traditional districting principles” are simply anything relevant to drawing districts other than race. They include “principles such as compactness, contiguity, and respect for political subdivisions.” Id., at 647. They also include “keeping communities of interest together, and protecting incumbents,” Rucho, 588 U. S., at 706–707, as well as “minimizing change,” Alabama Legislative Black Caucus v. Alabama, 575 U. S. 254, 259 (2015). Today, the Court identifes “the legislature's partisan interests” as a traditional cri-

1 As the Court observes, the most common direct evidence that a State considered race in drawing a districting plan is the State's admission that it considered race in order to comply with our Voting Rights Act precedents . Ante, at 8.

44

terion. Ante, at 24. Even considerations such as a district 's “consistently urban character, ” “common media sources,” and inclusion of “major transportation lines . . . implicate traditional districting principles.” Bush v. Vera, 517 U. S. 952, 966 (1996) (plurality opinion). We have readily acknowledged that “[t]raditional redistricting principles . . . are numerous and malleable,” and that “some . . . are surprisingly ethereal and admit of degrees.” Bethune- Hill, 580 U. S., at 190 (alteration and internal quotation marks omitted).

To evaluate whether a map aligns with traditional districting principles, a court must “rank the relative importance of those . . . criteria.” Rucho, 588 U. S., at 708. Without such a ranking, it is impossible to say what kinds of maps the principles should yield. But, that analysis ensnarls courts in a political thicket. Traditional districting principles often confict with one another, and there is no princi- Page Proof Pending Publication pled way for judges to resolve those conficts. Consider the question whether the principles of contiguity and compactness can justify a map that retains a relatively small part of the old district's core. See ante, at 24, 26–27. Or, consider whether the principle of keeping communities of interest together can justify uniting one community at the cost of splitting another between several districts, or healing partially an existing split at the cost of introducing a new one. See Allen v. Milligan, 599 U. S. 1, 57, 61 (2023) (Thomas, J., dissenting). These questions do not ask for legal answers, only political compromises. Judicial resolution of racial gerrymandering claims thus requires precisely the kind of “inconsistent, illogical, and ad hoc” decisionmaking that we have said is beyond the judicial power. Vieth, 541 U. S., at 278 (plurality opinion).

Evaluating compliance with traditional districting principles is further complicated by the fact that many decisions are equally consistent with both a good-faith application of those principles and with common gerrymandering tech-

45

niques. A legislature seeking to gerrymander a district will often proceed by “packing” or “cracking” groups of minority voters. “Packing” means concentrating minority voters in a single district to reduce their infuence in surrounding districts . “Cracking” means splitting a group of minority voters between multiple districts to avoid strong minority infuence in any one district. But, in areas where “political groups . . . tend to cluster (as is the case with Democratic voters in cities),” apparent packing or cracking can simply refect “adherence to compactness and respect for political subdivision lines” or “the traditional criterion of incumbency protection.” Id., at 290, 298. This case exemplifes the problem—the majority observes that Dr. Moon Duchin's report failed to “account for” the traditional districting principles of “partisanship or core retention” in “assessing whether the Enacted Plan `cracks' black voters among multiple districts .” Ante, at 33. The difference between illegitimate Page Proof Pending Publication packing and the legitimate pursuit of compactness is too often in the eye of the beholder.

Perhaps the most serious obstacle to evaluating whether a map is consistent with traditional districting principles is the fact that race and politics are, at present, highly correlated in American society. Racial gerrymandering is constitutionally suspect, but “a jurisdiction may engage in constitutional political gerrymandering.” Rucho, 588 U. S., at 701 (internal quotation marks omitted). So, even if a court is able to navigate all the complications I have identifed so far, it must still contend with the reality that “political and racial reasons are capable of yielding similar oddities in a district's boundaries.” Cooper v. Harris, 581 U. S. 285, 308 (2017). To that end, “when the State asserts partisanship as a defense ,” plaintiffs must meet the “formidable task” of “disentangl [ing] race from politics and prov[ing] that the former drove a district's lines. ” Ibid. Courts are not well equipped to evaluate whether plaintiffs succeed in disentangling race and politics.

46

As the Court observes, roughly 90% of black voters in South Carolina supported the Democratic candidate in the last Presidential election. Ante, at 9, and n. 2. When nearly all black voters support Democrats, an effort to strategically sort Democratic voters can be indistinguishable from an effort to strategically sort black voters. In this case, all Democratic-leaning maps presented during the districting process featured a black share of the voting-age population of 21% or higher, and all Republican-leaning maps featured a black voter share of 17% or lower. Ante, at 20. The dispute in this case therefore focuses on whether that correlation refected a racial purpose, or merely refected the result of a political purpose.

The majority's reasoning highlights the diffculties inherent in disentangling race and politics. Its explanation of why the expert evidence was insuffcient does not rest on the application of legal principles, but on the likely errors it fnds Page Proof Pending Publication in the experts' statistical models after a “foray into the minutiae of the record.” Cromartie, 532 U. S., at 262 (opinion of Thomas, J.). The majority discounts four separate expert reports based on methodological concerns. One report is insuffcient because it fails to model partisanship. Ante, at 24– 26. Another “carries less weight” because it measures partisanship through the wrong statistical method. Ante, at 30. And, another cannot be relied upon because it measures partisanship with the wrong election data. Ante, at 32. The dissent accuses the Court of “play[ing] armchair statistician.” Post, at 96–97 (opinion of Kagan, J.). But, the dissent's defense of the expert reports includes an exercise in armchair cartography. The dissent justifes the experts' assumption that the legislature could move any precinct in District 1 to District 6 by explaining that District 1 is thin, coastal, and shares a long border with District 6. Post, at 93–95. It supports its hunch with two zoomed-out maps that include no information about precinct size or location. Post, at 100, appendix to opinion of Kagan, J. This type of back-and-forth is the

47

inevitable result of our voting-rights doctrine. One worries that the Court will soon begin drawing its own sample maps and performing in-house regression analyses.

A system in which only specialized experts can discern the existence of a constitutional injury is intolerable, and strongly suggests that the racial gerrymandering injury is not amenable to judicial resolution. We should resist the temptation to reduce the Fourteenth Amendment to a battle of expert witnesses. Our gerrymandering misadventures demonstrate that these claims lack judicially manageable standards.

2

As I have long maintained, vote dilution claims are also “not readily subjected to any judicially manageable standards . ” Holder v. Hall, 512 U. S. 874, 901–902 (1994) (Thomas, J., concurring in judgment). To prove vote dilution as a constitutional claim, our precedents require plain- Page Proof Pending Publication tiffs to show that the design of a district reduces “minority voters' ability, as a group, to elect the candidate of their choice.” Shaw, 509 U. S., at 641 (internal quotation marks omitted). The same consideration is used for vote dilution claims brought under § 2 of the Voting Rights Act. See Allen, 599 U. S., at 13 (explaining that § 2 “borrow[s] language from a Fourteenth Amendment [vote dilution] case”).

To assess whether a legislature has diluted a minority's vote, “the critical question . . . is: `Diluted relative to what benchmark?' ” Id., at 50 (opinion of Thomas, J.) (quoting Gonzalez v. Aurora, 535 F. 3d 594, 598 (CA7 2008) (opinion of Easterbrook, C. J.)). Despite repeated efforts in our Voting Rights Act cases, the Court has “never succeeded” in formulating “an objective and workable method of identifying the undiluted benchmark.” 599 U. S., at 69 (opinion of Thomas, J.). The Court's failure is not surprising because the task is futile. The Constitution does not offer “a theory for defning effective participation in representative government.” Holder, 512 U. S., at 897 (opinion of Thomas, J.).

48

Choosing among theories of effective representation depends on particular voters' objectives and preferred political strategies, not principles of constitutional law. Are a minority 's votes “more `effective' when they provide infuence over a greater number of seats, or control over a lesser number of seats”? Id., at 899. Are minority voters “ `represented' only when they choose a delegate who will mirror their views in the legislative halls,” or does the “practical infuence” of a small group of potential swing voters also amount to effective representation? Id., at 900. Only minority voters themselves can answer these questions. No “theory of the `effective' vote” is “inherent in the concept of representative democracy itself.” Id., at 899. So, when our precedents ask a court to determine if a minority's vote is diluted, they are “actually ask[ing]” the court “ `to choose among competing bases of representation—ultimately, really, among competing theories of political philosophy.' ” Id., at 897 (quoting Baker, Page Proof Pending Publication 369 U. S., at 300 (Frankfurter, J., dissenting)). The Constitution expresses no view on such issues, and they are not amenable to judicial resolution.

In practice, this Court has endorsed a theory of representation that distributes legislative seats in direct proportion to racial demographics. “[T]he `lack of any better alternative ' identifed in our case law” and the “intuitive appeal” of “direct proportionality” make a racial proportionality standard irresistible. Allen, 599 U. S., at 72 (opinion of Thomas, J.) (quoting Holder, 512 U. S., at 937 (opinion of Thomas, J.)). As a result, there is a “near-perfect correlation between [courts'] proportionality fndings and [vote dilution] liability results.” 599 U. S., at 72 (citing 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)). A proportionality approach is easy to apply, but it is “radically inconsistent with the [Reconstruction ] Amendments' command that government treat citizens

49

as individuals and their `goal of a political system in which race no longer matters.' ” 599 U. S., at 82 (quoting Shaw, 509 U. S., at 657).

I continue to believe that “[t]he matters the Court has set out to resolve in vote dilution cases are . . . not questions of law,” and that “they are not readily subjected to any judicially manageable standards.” Holder, 512 U. S., at 901–902 (opinion of Thomas, J.). The Court's determination to nonetheless adjudicate these cases has yielded an unconstitutional practice of distributing political power based on race.

B

Racial gerrymandering and vote dilution claims—at a minimum , those challenging congressional districts—are nonjusticiable for an additional reason: The Elections Clause makes a “textually demonstrable constitutional commitment” of the power to oversee congressional districting to “a coordinate political department,” Congress. Vieth, 541 U. S., at 277 Page Proof Pending Publication (plurality opinion) (internal quotation marks omitted). And, no other constitutional provision overcomes that commitment to Congress. The Constitution contemplates no role for the federal courts in the districting process.

1

Although States have the initial duty to draw district lines, the Elections Clause commits exclusive supervisory authority over the States' drawing of congressional districts to Congress—not federal courts. It provides: “The Times, Places and Manner of holding Elections for Senators and Representatives, shall be prescribed in each State by the Legislature thereof; but the Congress may at any time by Law make or alter such Regulations, except as to the Places of chusing Senators.” Art. I, § 4, cl. 1. The frst part of the Clause “imposes a duty upon” state legislatures to “prescribe the details necessary to hold congressional elections.” U. S. Term Limits, Inc. v. Thornton, 514 U. S. 779, 862 (1995) (Thomas, J., dissenting). The second part “grants power

50

exclusively to Congress” to police the state legislatures' performance of their duty. Id., at 864. Critically, the Clause leaves the Judiciary out of the districting process entirely.

The Clause's assignment of roles is comprehensive. For example, a state legislature's responsibility over congressional elections “ `transcends any limitations sought to be imposed by the people of a State' ” through other state actors; the state legislature is the exclusive state authority. Moore v. Harper, 600 U. S. 1, 58 (2023) (Thomas, J., dissenting) (quoting Leser v. Garnett, 258 U. S. 130, 137 (1922)). In a similar vein, the Clause makes Congress the exclusive federal authority over States' efforts to draw congressional districts , to the exclusion of courts.

The historical record compels this interpretation of the Elections Clause's text. Gerrymandering and vote dilution are not new phenomena. The founding generation was familiar with political districting problems from the American Page Proof Pending Publication colonial experience. See Vieth, 541 U. S., at 274 (plurality opinion) (collecting examples). But, the Framers nowhere suggested the federal courts as a potential solution to those problems. Instead, they relied on congressional oversight. The Framers' considered choice of a nonjudicial remedy is highly relevant context to the interpretation of the Elections Clause. See New York State Rife & Pistol Assn., Inc. v. Bruen, 597 U. S. 1, 26–27 (2022).

Because the Elections Clause attracted considerable criticism during the ratifcation debates, ample contemporaneous discussion sheds light on the original understanding of the Clause. As a delegate to the Virginia ratifying convention observed, Congress's power to regulate the time, place, and manner of elections drew objections that “echoed from one end of the continent to the other.” 3 Debates on the Constitution 9 (J. Elliot ed. 1836) (Elliot's Debates). Opponents of ratifcation attacked the Clause as a radical expansion of national power and a grave danger to liberty. Patrick Henry argued: “What can be more defective than the clause

51

concerning the elections? The control given to Congress over the time, place, and manner of holding elections, will totally destroy the end of suffrage.” Id., at 60.

Defenses of the Elections Clause demonstrate that it was designed at least in part as a way to address abusive districting . To be sure, proponents of ratifcation primarily justifed the Clause as a “constitutional remedy for th[e] evil” presented by the possibility that “the states [might] neglect to appoint representatives” to the new Federal Government. 2 id., at 326 (statement of John Jay). But, other defenses of the Elections Clause resonate with modern concerns about gerrymandering and vote dilution.

Some proponents of ratifcation championed the Clause as necessary “for securing to the people their equal rights of election.” Id., at 26. A delegate to the Massachusetts ratifying convention cautioned that “a state legislature . . . in times of popular commotion, and when faction and party Page Proof Pending Publication spirit run high, . . . might make an unequal and partial division of the states into districts for the election of representatives .” Id., at 27. In such a situation, he explained, “the people can have no remedy” except for that created by the Elections Clause: the “controlling power” by which Congress may “preserve and restore to the people their equal and sacred rights of election.” Ibid. And, James Madison raised similar arguments at the Constitutional Convention. See 2 Records of the Federal Convention of 1787, pp. 240–241 (M. Farrand ed. 1911).

It was Congress, not the courts, that the Founders contemplated would provide recourse against state intrusions on voting rights through the districting process. Even when listing all entities that could possibly regulate congressional elections, the founding generation did not consider the federal courts. To support his assertion that “the discretionary power over elections ought to exist somewhere,” Alexander Hamilton posited that “there were only three ways in which this power could have been reasonably organized; that

52

it must either have been lodged wholly in the National Legislature , or wholly in the State Legislatures, or primarily in the latter, and ultimately in the former.” The Federalist No. 59, p. 326 (E. Scott ed. 1898) (emphasis added). A delegate made the same observation at the Massachusetts ratifying convention: “The power . . . to regulate the elections of our federal representatives must be lodged somewhere. I know of but two bodies wherein it can be lodged—the legislatures of the several states, and the general Congress.” 2 Elliot's Debates 24.

The Elections Clause's text and history therefore point to the same conclusion: The Clause commits supervisory authority over congressional districting to Congress alone. “At no point” during the drafting or ratifcation of the Constitution “was there a suggestion that the federal courts had a role to play” in resolving “electoral districting problems.” Rucho, 588 U. S., at 699. Even when the debate touched on how political districting could affect the voting rights of indi- Page Proof Pending Publication viduals, it was understood that any remedy related to districting would come from Congress, not federal courts.2

2

None of the Constitution's other provisions undercuts or countermands the Elections Clause's clear mandate for Congress to supervise the States' districting efforts. The Court 2 Congress has, at times, wielded its power under the Elections Clause to impose compactness and contiguity requirements for congressional districts . See, e.g., Apportionment Act of 1842, ch. 47, 5 Stat. 491; Apportionment Act of 1911, ch. 5, 37 Stat. 13. More recently, in the Uniform Congressional District Act of 1967, Congress required the States to use single-member congressional districts instead of at-large elections. See Pub. L. 90–196, 81 Stat. 581, 2 U. S. C. § 2c. And, Congress created a system for addressing a State's failure to properly redistrict following a decennial census. See § 2a(c). Some Elections Clause legislation may give rise to justiciable controversies regarding the application of federal statutes. Cf. Wood v. Broom, 287 U. S. 1, 8 (1932). But, constitutional districting claims are not justiciable in and of themselves.

53

has viewed the Fourteenth and Fifteenth Amendments as the source of its authority to entertain challenges to districts . But, the Reconstruction Amendments are perfectly consistent with Congress's exclusive authority to oversee congressional districting.

Our decisions primarily identify the Equal Protection Clause as the textual basis for judicial resolution of districting claims. See Shaw, 509 U. S., at 642; Davis v. Bandemer, 478 U. S. 109, 151 (1986) (O'Connor, J., concurring in judgment ) (asserting that, in contrast to political gerrymandering , “the greater warrant the Equal Protection Clause gives the federal courts to intervene for protection against racial discrimination . . . render[s] racial gerrymandering claims justiciable”). That conclusion does not comport with the text of the Equal Protection Clause or the structure of the Reconstruction Amendments.

The text of the Equal Protection Clause makes it an unlikely source for claims about political districting. The Page Proof Pending Publication Equal Protection Clause provides that “[n]o State shall . . . deny to any person within its jurisdiction the equal protection of the laws.” Amdt. 14, § 1. The Clause's “focus on `protection' ” suggests that it imposes only “ `a duty on each state to protect all persons and property within its jurisdiction from violence and to enforce their rights through the court system,' ” not a “prohibit[ion on] discriminatory legislative classifcations.” United States v. Vaello Madero, 596 U. S. 159, 178–179, n. 4 (2022) (Thomas, J., concurring) (quoting C. Green, The Original Sense of the (Equal) Protection Clause: Pre-Enactment History, 19 Geo. Mason U. Civ. Rights L. J. 1, 3 (2008)). So understood, the Equal Protection Clause has no obvious bearing on districting.3

3 Other Clauses in § 1 of the Fourteenth Amendment fare no better.

The Privileges or Immunities Clause provides that “[n]o State shall make or enforce any law which shall abridge the privileges or immunities of citizens of the United States.” It “grants `United States citizens a certain collection of rights—i.e., privileges or immunities—attributable to that

54

Reading the Equal Protection Clause—or anything else in § 1 of the Fourteenth Amendment—to invite judicial involvement in disputes over voting rights also ignores the fact that another part of the Fourteenth Amendment deals directly with those rights. Section 2 provides that “when the right to vote . . . is denied” to a State's voting-age male citizens “or in any way abridged,” the State's apportionment of congressional representatives “shall be reduced in the proportion ” of the denial of the franchise. Congress alone can provide that remedy through its power to apportion representatives among the States. See Art. I, § 2, cl. 3. Federal courts are therefore unable to enforce § 2. See Saunders v. Wilkins, 152 F. 2d 235 (CA4) (1945), cert. denied, 328 U. S. 870 (1946). The express provision of a nonjudicial remedy for voting-rights violations in § 2 counsels against reading § 1 to allow judicial remedies implicitly in those same voting- rights disputes. Cf. Reynolds v. Sims, 377 U. S. 533, 594 (1964) (Harlan, J., dissenting). Page Proof Pending Publication Reading the Equal Protection Clause to support claims for racial gerrymandering or vote dilution also makes the existence of the Fifteenth Amendment unexplainable. If § 1 of the Fourteenth Amendment allows for such fulsome protection of the franchise by federal courts, it is hard to see why status.' ” Ramos v. Louisiana, 590 U. S. 83, 138 (2020) (Thomas, J., concurring in judgment) (quoting McDonald v. Chicago, 561 U. S. 742, 808 (2010) (Thomas, J., concurring in part and concurring in judgment)). And, the Citizenship Clause provides that “[a]ll persons born or naturalized in the United States . . . are citizens of the United States and of the State wherein they reside.” It likely “guarantees citizens equal treatment . . . with respect to civil rights.” Vaello Madero, 596 U. S., at 179 (opinion of Thomas, J.). It is questionable whether the terms “privileges and immunities” and “civil rights” were understood by the generation that ratifed the Fourteenth Amendment “to extend to political rights, such as voting.” J. Harrison, Reconstructing the Privileges or Immunities Clause, 101 Yale L. J. 1385, 1417 (1992).

The Due Process Clause, of course, is a nonstarter as a source for substantive rights. See Dobbs v. Jackson Women's Health Organization, 597 U. S. 215, 330–336 (2022) (Thomas, J., concurring).

55

“Congress and the States still found it necessary to adopt the Fifteenth Amendment—which protects `[t]he right of citizens of the United States to vote'—two years after the Fourteenth Amendment's passage.” McDonald, 561 U. S., at 852 (opinion of Thomas, J.).

Nor can the Fifteenth Amendment justify racial gerrymandering or vote dilution claims in its own right. The Fifteenth Amendment is the primary constitutional protection for the voting rights of racial minorities. But, the Fifteenth Amendment “address[es] only matters relating to access to the ballot.” Holder, 512 U. S., at 930 (opinion of Thomas, J.). “[I]ts protections [are] satisfed as long as members of racial minorities [can] ` “register and vote without hindrance .” ' ” Id., at 921 (quoting Mobile v. Bolden, 446 U. S. 55, 65 (1980) (plurality opinion)). The Court's decision in Gomillion v. Lightfoot, 364 U. S. 339 (1960)—a Fifteenth Amendment case often cited as a touchstone of our racial gerrymandering jurisprudence—is consistent with this un- Page Proof Pending Publication derstanding. Gomillion involved only a claim “that the boundaries of a city had been drawn to prevent blacks from voting in municipal elections altogether,” not a claim about the way minority voters were distributed between two districts. Holder, 512 U. S., at 920, n. 20 (opinion of Thomas, J.).

At this juncture, I see no directive in the Reconstruction Amendments for courts to police the lines between political districts. Instead, the Elections Clause assigns the responsibility for supervising the States' drawing of congressional districts solely to Congress.

* * *

Racial gerrymandering and vote dilution claims lack judicially manageable standards for their resolution. And, they confict with the Constitution's textual commitment of congressional districting issues to the state legislatures and Congress. They therefore present nonjusticiable political

56

questions. The Court should extricate itself from this business and return political districting to the political branches, where it belongs.

II

When an institution strays from its competencies, one does not expect good results. This Court's efforts in the districting feld are no exception. The underlying nonjusticiability of racial gerrymandering and vote dilution claims leads us to distort our doctrines in numerous ways. The standard that the Court uses to resolve racial gerrymandering claims betrays the colorblind promise of the Fourteenth Amendment by endorsing the notion that some racial classifcations are benign. The standard that the Court uses to resolve vote dilution claims invariably falls back on racial stereotypes. And, the remedy commonly ordered in redistricting cases—a judicially imposed map—ignores the normal limits on federal equity power. Taken together, the Court's misconceived Page Proof Pending Publication doctrines leave the States in an unenviable position.

A

The racial predominance standard for racial gerrymandering claims is plainly inconsistent with the fact that “ `[o]ur Constitution is color-blind.' ” Harvard College, 600 U. S., at 230 (quoting Plessy, 163 U. S., at 559 (opinion of Harlan, J.)). Ordinarily, any governmental consideration of race—even as a second-order consideration—triggers strict scrutiny. For example, using race merely as a “tip” or a “plus” factor in college admissions does not excuse a university from satisfying strict scrutiny. 600 U. S., at 195–196, 213 (internal quotation marks omitted).

Our voting-rights precedents diverge from this rule by subjecting an alleged racial gerrymander to strict scrutiny only if “race was the `predominant factor motivating the legislature 's decision to place a signifcant number of voters within or without a particular district.' ” Ante, at 7 (quoting Miller, 515 U. S., at 916) (emphasis added). A “predomi-

57

nance” requirement conficts with the classifcation-based harm that racial gerrymandering claims purport to address. The constitutional injury underlying a racial gerrymandering claim is the legislature's mere use of a racial classifcation in drawing its map. See Bethune-Hill, 580 U. S., at 189. That injury exists whether race is a legislature's frst or last consideration in drawing districts. “Racial classifcations of any sort pose the risk of lasting harm to our society.” Shaw, 509 U. S., at 657 (emphasis added). “They reinforce the belief . . . that individuals should be judged by the color of their skin” and “balkanize us into competing racial factions .” Ibid. All racial classifcations are inherently suspect , whether predominant or not.

The Court developed the racial predominance standard with concerns about the justiciability of gerrymandering claims in mind. The Court initially formulated the predominance standard while observing that “[f]ederal-court review Page Proof Pending Publication of districting legislation represents a serious intrusion on the most vital of local functions,” and stressing the need to allow States “discretion to exercise the political judgment necessary to balance competing interests.” Miller, 515 U. S., at 915. And, after describing the predominance standard, the Court cautioned that federal courts must consider the problem of racial gerrymandering in light of “the intrusive potential of judicial intervention into the legislative realm.” Id., at 916. These concerns about intruding on the political process should have been a clear sign to retreat. Instead, the Court forged ahead to adopt a constitutionally suspect compromise.

The racial predominance standard does not even purport to be consistent with the colorblind Constitution. The Miller Court simply borrowed that standard from the District Court's fawed opinion below. The Court endorsed the District Court's decision “to require strict scrutiny whenever race is the `overriding, predominant force' in the redistricting process.” Id., at 909, 917 (quoting Johnson v.

58

Miller, 864 F. Supp. 1354, 1372 (SD Ga. 1994)). But, the District Court's opinion could not have been a stronger rejection of our colorblind Constitution. It acknowledged that the racial predominance standard allowed legislatures to “intentionally consider race in redistricting—and even alter the occasional line in keeping with that consideration—without incurring constitutional review.” Id., at 1373. But, the District Court reasoned, “[b]oth the Supreme Court and Congress have already admitted that the Constitution is not genuinely `color-blind.' ” Id., at 1374. This provenance underscores the inconsistency of the racial predominance standard with our colorblind Constitution.

Any use of race in drawing political districts—no matter how minor—must be justifed by a compelling interest. The Court's insistence on hearing nonjusticiable districting claims leads it to disregard that principle in favor of a distorted standard that legitimizes racial classifcations. If the Court is truly concerned about intruding on the political Page Proof Pending Publication process, it should acknowledge that districting is a political question and vacate the feld.

B

The Court's standard for vote dilution claims is similarly fawed, because it requires judges to engage in racial stereotyping . As I have explained, the Constitution does not defne a baseline of effective representation by which to evaluate the dilution of a vote. Supra, at 47–49. The Court has purported to fll that gap by looking to “minority voters' ability , as a group, `to elect the candidate of their choice.' ” Shaw, 509 U. S., at 641 (quoting Allen, 393 U. S., at 569). Simply put, the lack of a manageable vote dilution standard has led the Court to fall back on generalized expectations about members of minority groups.

“Our constitutional history does not tolerate [the] choice” to treat as “the touchstone of an individual's identity . . . the color of their skin.” Harvard College, 600 U. S., at 231. It

59

therefore does not permit courts to make judgments about what candidate “minority voters as a group” would choose. That assessment requires a court to assume that “members of racial and ethnic groups must all think alike on important matters of public policy.” Holder, 512 U. S., at 903 (opinion of Thomas, J.). And, it requires a court to construct a caricature of the racial group to determine—in the abstract— the attributes that defne “the candidate of its choice.” The Constitution does not indulge the belief that members of racial minorities “always (or even consistently) express some characteristic minority viewpoint on any issue.” Harvard College, 600 U. S., at 219 (internal quotation marks omitted).

The racial stereotyping encouraged by our vote dilution precedents is pronounced here. To establish vote dilution, the plaintiffs point to the District Court's observation that recent elections in the district “ `were close, with less than one percent separating the candidates,' so increasing the dis- Page Proof Pending Publication trict's Black population to 20% `would produce a “toss up” district' ” instead of a Republican one. Brief for Appellees 64. But, that reasoning simply equates the ability of black South Carolinians to elect the candidate of their choice with their ability to elect a Democrat—an exercise in racial stereotyping . The mere fact that “members of a racial group tend to prefer the same candidates” is not license to treat that correlation as an absolute truth. Holder, 512 U. S., at 904 (opinion of Thomas, J.). Plaintiffs make no effort to explore whether the affnity of the district's black population toward the Democratic Party “might be the product of similar socioeconomic interests rather than some other factor related to race.” Ibid. They instead proceed on the “working assumption that racial groups can be conceived of largely as political interest groups.” Id., at 905. The Constitution forbids such an assumption.

The plaintiffs' stereotyping does not stop there. They contend that their vote dilution claim also fnds support in an expert report evaluating the ability of black South Caro-

60

linians to elect the candidate of their choice. That expert based her conclusion on the results of “elections with Black candidates on the ballot.” Brief for Appellees 64. The plaintiffs' argument therefore assumes that the “candidate of choice” for black voters is simply a black candidate. But, the stereotyping is worse than that. In 2016, South Carolina reelected Republican Tim Scott to the United States Senate; Scott is the frst black senator from the South since Reconstruction. The plaintiffs and their expert nonetheless decided that this race was not “considered probative for Black electoral opportunity.” Supp. App. to Juris. Statement 174a. Plaintiffs' argument therefore combines two stereotypes by assuming that black South Carolinians can be properly represented only by a black Democrat.

Such stereotyping is, of course, not limited to this case or black voters. For example, a District Court recently concluded that Hispanic voters in a majority-Hispanic district Page Proof Pending Publication lacked an opportunity to elect the candidate of their choice, even though the district elected a Hispanic Republican. Soto Palmer v. Hobbs, 686 F. Supp. 3d 1213, 1224–1225, 1230– 1231, 1234–1235 (WD Wash. 2023). The court later purported to correct the lack of Hispanic opportunity by imposing a remedial map that made the district “substantially more Democratic,” but slightly less Hispanic. Soto Palmer, 2024 WL 1138939, *2, *5 (Mar. 15, 2024). In short, the court concluded that securing the rights of Hispanic voters required replacing some of those voters with non-Hispanic Democrats. That dismissive attitude toward non- Democratic members of minority groups exemplifes the tendency of the Court's race-obsessed jurisprudence to “balkanize us into competing racial factions.” Shaw, 509 U. S., at 657. The Court should correct course now before it inficts further damage.

The vote dilution analysis in this case inevitably reduces black Charlestonians to partisan pawns and racial tokens. The analysis is demeaning to the courts asked to perform it,

61

to say nothing of the black voters that it stereotypes. “The assumptions upon which our vote dilution decisions have been based should be repugnant to any nation that strives for the ideal of a color-blind Constitution.” Holder, 512 U. S., at 905–906 (opinion of Thomas, J.).

C

The Court's insistence on adjudicating racial gerrymandering and vote dilution claims has also tempted it to ignore constitutional limits on its remedial powers. Ultimately, the only remedy for the constitutional injuries caused by an illegally drawn map is a new map. But, federal courts lack “the power to create remedies previously unknown to equity jurisprudence .” Grupo Mexicano de Desarrollo, S. A. v. Alliance Bond Fund, Inc., 527 U. S. 308, 332 (1999). And, there is no “indication that the Framers had ever heard of courts” playing any role in resolving electoral districting problems. Page Proof Pending Publication Rucho, 588 U. S., at 699. The power to redraw a States' electoral districts therefore exceeds “the jurisdiction in equity exercised by the High Court of Chancery in England at the time of the adoption of the Constitution and the enactment of the original Judiciary Act.” Grupo Mexicano, 527 U. S., at 318 (internal quotation marks omitted).

The Court once recognized its limited equitable powers in this area. We previously acknowledged that “[o]f course no court can affrmatively re-map [a State's] districts so as to bring them more in conformity with the standards of fairness for a representative system. At best we could only declare the existing electoral system invalid.” Colegrove v. Green, 328 U. S. 549, 553 (1946) (plurality opinion); see also Baker, 369 U. S., at 328 (Frankfurter, J., dissenting) (“Surely a Federal District Court could not itself remap the State”).

The view of equity required to justify a judicial map- drawing power emerged only in the 1950s. The Court's “impatience with the pace of desegregation” caused by resistance to Brown v. Board of Education, 347 U. S. 483 (1954),

62

“led us to approve . . . extraordinary remedial measures,” Missouri v. Jenkins, 515 U. S. 70, 125 (1995) (Thomas, J., concurring). In the follow-on case to Brown, the Court considered “the manner in which relief [was] to be accorded” for vindication of “the fundamental principle that racial discrimination in public education is unconstitutional.” Brown v. Board of Education, 349 U. S. 294, 298 (1955) (Brown II). In doing so, the Court took a boundless view of equitable remedies, describing equity as being “characterized by a practical fexibility in shaping its remedies and by a facility for adjusting and reconciling public and private needs.” Id., at 300 (footnote omitted). That understanding may have justifed temporary measures to “overcome the widespread resistance to the dictates of the Constitution” prevalent at that time, but, as a general matter, “[s]uch extravagant uses of judicial power are at odds with the history and tradition of the equity power and the Framers' design.” Jenkins, 515 Page Proof Pending Publication U. S., at 125–126 (opinion of Thomas, J.). Federal courts have the power to grant only the equitable relief “traditionally accorded by courts of equity,” not the fexible power to invent whatever new remedies may seem useful at the time. Grupo Mexicano, 527 U. S., at 319.

Redistricting remedies rest on the same questionable understanding of equitable power. No court has explained where the power to draw a replacement map comes from, but all now assume it may be exercised as a matter of course. The most consideration this Court has given to the question, if it can be called consideration, was in Reynolds v. Sims, 377 U. S. 533. In that case, the Court foreswore any attempt to “consider . . . the diffcult question of the proper remedial devices which federal courts should utilize in state legislative apportionment cases,” but nonetheless upheld, as an act of “proper judicial restraint,” the District Court “ordering its own temporary reapportionment plan.” Id., at 585–586. The Court's only support for that conclusion was the naked statement in Justice Douglas's Baker concurrence that “ `any

63

relief accorded can be fashioned in the light of well-known principles of equity.' ” Reynolds, 377 U. S., at 585 (quoting 369 U. S., at 250). Douglas's statement is an obvious fallback to the “practical fexibility” extolled as a “traditional attribut[e] of equity power” in Brown II. 349 U. S., at 300. The explanation is wholly inadequate; the Court has never attempted to ground the map-drawing power in “the jurisdiction in equity exercised by the High Court of Chancery in England” in 1789. Grupo Mexicano, 527 U. S., at 318 (internal quotation marks omitted).

The lack of a historically grounded map-drawing remedy is an enormous problem for districting claims, because no historically supportable remedy can correct an improperly drawn district. The most promising option is “[t]he negative injunction remedy against state offcials countenanced in Ex parte Young,” a “standard tool of equity that federal courts have authority to entertain under their traditional eq- Page Proof Pending Publication uitable jurisdiction.” Whole Woman's Health v. Jackson, 595 U. S. 30, 53 (2021) (Thomas, J., concurring in part and dissenting in part) (citation and internal quotation marks omitted); see also Ex parte Young, 209 U. S. 123 (1908). The Court has “long held that federal courts may in some circumstances grant injunctive relief against state offcers who are violating, or planning to violate, federal law.” Armstrong v. Exceptional Child Center, Inc., 575 U. S. 320, 326 (2015).

But, a negative-injunction remedy does not actually redress racial gerrymandering or vote dilution, for two reasons . First, it is not apparent that an Ex parte Young injunction can prevent a state election offcial from conducting an election under an unconstitutional map, or force him to draw a new map. Such an injunction “permits a party to assert in equity a defense that would otherwise have been available in the State's enforcement proceedings at law,” and it “extends no further than permitting private parties in some circumstances to prevent state offcials from bringing an action to enforce a state law that is contrary to federal

64

law.” Whole Woman's Health, 595 U. S., at 53 (opinion of Thomas, J.) (alteration and internal quotation marks omitted ). It is thus not clear that such an injunction could stop an election. Second, even if it is possible to enjoin state officials from conducting an election, it is questionable whether that remedy is ever “equitable.” Our system of government depends on regular elections; putting elections indefnitely on hold may do more harm than good. Cf. Baker, 369 U. S., at 327 (opinion of Frankfurter, J.) (“An injunction restraining a general election unless the legislature reapportions would paralyze the critical centers of a State's political system and threaten political dislocation whose consequences are not foreseeable”). Ultimately, to remedy racial gerrymandering or vote dilution, someone must draw a new map. I can fnd no explanation why that “someone” can be a federal court.

D

Page Proof The Court's attemptsPending to adjudicate thePublication impossible have put the States in an untenable position. We have hesitated to subject States to the “ ` “competing hazards of liability” ' ” that arise from the fact that the Constitution “restricts consideration of race and the [Voting Rights Act] demands consideration of race.” Abbott v. Perez, 585 U. S. 579, 587 (2018) (quoting Vera, 517 U. S., at 977 (plurality opinion)). But, the lack of manageable standards for districting claims and the unfortunate trajectory of the Court's Voting Rights Act precedents combine to make it impossible for States to navigate these hazards.

Last Term, the Court held that the Voting Rights Act required Alabama to draw a map that would give black Alabamians a majority in two of the State's seven congressional districts. Because black Alabamians make up less than twosevenths of the State's population, such a map could result only from an obsessive focus on race in the map-drawing process. See Allen, 599 U. S., at 56 (opinion of Thomas, J.). For example, one of the plaintiffs' experts used a race-

65

neutral algorithm to generate 2 million random maps; not a single map yielded two majority-black districts. Id., at 58– 59. In this case, however, South Carolina faced a real risk of constitutional liability based on allegations that it considered race too heavily in drawing a district that was 17% black instead of 20%.

In fact, the Court recently granted emergency relief after a State failed to thread the impossible needle created by our voting-rights precedents. Voters in Louisiana challenged the State's 2022 congressional map, arguing that “Louisiana was required under the Voting Rights Act to create a second black-majority district.” Robinson v. Ardoin, 86 F. 4th 574, 585 (CA5 2023). The Fifth Circuit concluded that the plaintiffs were likely to succeed on their Voting Rights Act claim. Louisiana argued that, under the Voting Rights Act, “the possibility of drawing a majority-minority district does not require the drawing of the district,” but the court pointed to Page Proof Pending Publication our decision in Allen to reject that contention. 86 F. 4th, at 599. Louisiana then held a special legislative session and adopted a new map that “established a second majority- Black congressional district to resolve the [Voting Rights Act] litigation.” Callais, v. Landry, ––– F. Supp. 3d –––, 2024 WL 1903930, *1 (WD La., Apr. 30, 2024). The result? A different group of voters brought constitutional gerrymandering and vote dilution claims against the State. Id., at *6–*7. That suit was also successful. A District Court found that race predominated in Louisiana's process of adding the second majority-minority district, and enjoined the use of the new map. Id., at *17, *24. After the State argued that the proximity of the District Court's order to important election deadlines would cause “election chaos,” Emergency Application in No. 23A1002, p. 19, we stayed the order, Order in No. 23A1002, 601 U. S. ––– (2024) (citing Purcell v. Gonzalez, 549 U. S. 1 (2006) (per curiam)).

As these cases make clear, this Court's jurisprudence puts States in a lose-lose situation. Taken together, our prece-

Justice Kagan, dissenting

dents stand for the rule that States must consider race just enough in drawing districts. And, what “just enough” means depends on a federal court's answers to judicially unanswerable questions about the proper way to apply the State's traditional districting principles, or about the groupwide preferences of racial minorities in the State. There is no density of minority voters that this Court's jurisprudence cannot turn into a constitutional controversy. We have extracted years of litigation from every districting cycle, with little to show for it. The Court's involvement in congressional districting is unjustifed and counterproductive.

* * *

“When, under our direction, federal courts are engaged in methodically carving the country into racially designated electoral districts, it is imperative that we stop to reconsider whether the course we have charted for the Nation is the Page Proof Pending Publication one” required by the Constitution. Holder, 512 U. S., at 945 (opinion of Thomas, J.). The Constitution provides courts no power to draw districts, let alone any standards by which they can attempt to do so. And, it does not authorize courts to engage in the race-based reasoning that has come to dominate our voting-rights precedents. It is well past time for the Court to return these political issues where they belong —the political branches.

Justice Kagan, with whom Justice Sotomayor and Justice Jackson join, dissenting.

This voting case, as the Court acknowledges, turns on a quintessential factual dispute: Did South Carolina rely on racial data to reconfgure the State's Congressional District 1? The parties here agree that the South Carolina Legislature wanted to make District 1 more Republican. They further agree that in pursuit of that aim, the legislature moved nearly 200,000 people into or out of the district. What the parties disagree about is how the people expelled from the

67

district were chosen. The State contends that its mapmakers looked exclusively at data from the last election and targeted people who had voted Democratic. If that is true, the State's actions (however unsavory and undemocratic) are immune from federal constitutional challenge. The Challengers , though, offer a different account. They say that the mapmakers, not content with what the election data revealed , also reviewed and heavily relied on racial data—thus exploiting the well-known correlation between race and voting behavior. And if that is true, the Challengers have a good constitutional claim, because the Equal Protection Clause forbids basing election districts mainly on race in order to achieve partisan aims. See Cooper v. Harris, 581 U. S. 285, 291, and n. 1, 308, n. 7 (2017); Miller v. Johnson, 515 U. S. 900, 914 (1995). So the key question again: In drawing District 1, did the mapmakers consider voting data alone, or did they also closely attend and respond to which residents Page Proof Pending Publication were Black and which were White?

A three-judge District Court undertook to resolve that factual dispute. And the court, over nearly a year, did everything one could ask to carry out its charge. After overseeing broad discovery, the court held a 9-day trial, featuring some two dozen witnesses and hundreds of exhibits. It evaluated evidence about South Carolina geography and politics . It heard frst-hand testimony about the redistricting process. And it considered the views of statistical experts on how the State's new district lines could—and could not— have come about. In the end, the court had to decide between two starkly different stories, backed by opposing bodies of evidence. One side you know from having read the majority opinion: The state offcials repeatedly denied using race in choosing the people kicked out of District 1, insisting that they based their decisions on political data alone. The other side you have not yet heard, except in the sketchiest of terms. It is that the State's mapmakers were experienced and skilled in the use of racial data to draw electoral

68

maps; that they confgured their mapmaking software to show how any change made to the district would affect its racial composition; that the racial make-up they landed on was precisely what they needed, to the decimal point, to achieve their partisan goals; and that their politics-only story could not account, as a statistical matter, for their large-scale exclusion of African-American citizens. Faced with that proof, all three judges agreed: The Challengers' version of events was the more credible. The court, to put the matter bluntly, did not believe the state offcials. It thought they had gerrymandered District 1 by race.

In reviewing those conclusions, the majority goes seriously wrong. Factfnding about electoral districting, as about other matters, is reversible “only for clear error.” Cooper, 591 U. S., at 293. This Court must give a district court's view of events “signifcant deference,” which means we must uphold it so long as it is “plausible.” Ibid. Under Page Proof Pending Publication that standard, South Carolina should now have to redraw District 1. As I'll detail, the Challengers introduced more than enough evidence of racial gerrymandering to support the District Court's judgment. The majority's attempt to explain its contrary result fails at every turn. The majority picks and chooses evidence to its liking; ignores or minimizes less convenient proof; disdains the panel's judgments about witness credibility; and makes a series of mistakes about expert opinions. The majority declares that it knows better than the District Court what happened in a South Carolina map-drawing room to produce District 1. But the proof is in the pudding: On page after page, the majority's opinion betrays its distance from, and lack of familiarity with, the events and evidence central to this case.

Yet there is worse: The majority cannot begin to justify its ruling on the facts without in two ways reworking the law—each to impede racial-gerrymandering cases generally. First, the majority, though ostensibly using the clear-error standard, effectively inverts it whenever a trial court rules

69

against a redistricting State. In the majority's version, all the deference that should go to the court's factual fndings for the plaintiffs instead goes to the losing defendant, because it is presumed to act in good faith. See ante, at 10. So the wrong side gets the beneft of the doubt: Any “possibility ” that favors the State is treated as “dispositive.” Ante, at 20. Second, the majority invents a new rule of evidence to burden plaintiffs in racial-gerrymandering cases. As of today, courts must draw an adverse inference against those plaintiffs when they do not submit a so-called alternative map—no matter how much proof of a constitutional violation they otherwise present. See ante, at 34–35. Such micro-management of a plaintiff's case is elsewhere unheard of in constitutional litigation. But as with its upside-down application of clear-error review, the majority is intent on changing the usual rules when it comes to addressing racial- gerrymandering claims. Page Proof Pending Publication To be fair, we have seen all this once before—except that it was in a dissent. Just seven years ago, this Court decided another racial-gerrymandering case, strikingly similar to this one. In Cooper v. Harris, the Court rejected the State's request for an alternative-map requirement; the dissent vehemently objected. See 581 U. S., at 318; id., at 334– 337 (Alito, J., dissenting). The Court applied normal clear- error review, deferring to all plausible trial court fndings. See id., at 293. The dissent, invoking a presumption of good faith, instead deferred to all plausible arguments of the losing State defendant. See id., at 357 (Alito, J., dissenting). Today, for all practical purposes, the Cooper dissent becomes the law.

Perhaps most dispiriting is what lies behind the Court's new approach—its special rules to specially disadvantage suits to remedy race-based redistricting. The Cooper dissent thought plaintiffs would use racial-gerrymandering actions as “weapons of political warfare.” Id., at 335 (Alito, J., dissenting). And it lamented that courts fnding gerry-

70

manders were “accus[ing]” States of “offensive and demeaning conduct.” Id., at 334 (internal quotation marks omitted ). So the problem was more with challenging racial gerrymanders than with putting them into place. Today, that view becomes central to the majority opinion. See ante, at 11. The suspicion, and indeed derision, of suits brought to stop racial gerrymanders are self-evident; the intent to insulate States from those suits no less so. But consider what this altered perspective misses. That a State may in fact have engaged in such “offensive and demeaning” conduct. That it may have sorted citizens by their race with respect to the most fundamental of all their political rights. That it may have done so for no reason other than to achieve partisan gain. And here, that a three-judge court unanimously found all this to have occurred.

The proper response to this case is not to throw up novel roadblocks enabling South Carolina to continue dividing citizens along racial lines. It is to respect the plausible—no, Page Proof Pending Publication the more than plausible—fndings of the District Court that the State engaged in race-based districting. And to tell the State that it must redraw District 1, this time without targeting African-American citizens.

I

Begin with the law, and more particularly the usual standard of review. This Court all the time recites the words: “only for clear error.” Cooper, 581 U. S., at 293, 309. And those words always mean (or anyway, always meant) the same thing. Under the clear-error standard, a lower court's factual fndings “warrant[ ] signifcant deference.” Id., at 293. We do not rubber stamp those fndings, but we affrm them so long as they are “plausible” in light of the full record. Anderson v. Bessemer City, 470 U. S. 564, 574 (1985). And that is so even if, left to our own devices, we “would have decided the [matter] differently.” Id., at 573. We can reverse only when “left with the defnite and frm conviction

71

that a mistake has been committed.” Ibid. And nowhere is that high bar higher than when witness credibility is at issue. A trial court's judgment about whether a witness is telling the truth is entitled to “singular deference.” Cooper, 581 U. S., at 309.

The reasons for thus deferring to trial court factfnding are equally well-settled. Trial courts are the judiciary's factfnding specialists. They live with a case for months or years, supervising discovery, ruling on the admission of expert opinions, and watching how the evidence unfolds. They preside over the trial and see the live witnesses (24 in this case) up close. They can observe “the variations in demeanor and tone” that “bear so heavily” on credibility judgments. Anderson, 470 U. S., at 575. They know the ins and outs of often massive records. (This case boasts, for example, a 2,122-page trial transcript, a 1,694-page compilation of key deposition testimony, and (as one judge re- Page Proof Pending Publication marked) too many exhibits to ft in the courtroom. No. 3:21– cv–3302 (D SC), ECF Doc. 503, p. 23.) Chances are, then, that a trial court will do better factfnding than an appellate court parachuting in at the last moment. The clear-error standard is a recognition of comparative competence. And it is a forced dose of humility—a virtue which sometimes doesn't come naturally to appellate courts. Apply that last point to this Court in particular. The clear-error standard tells us that when we disagree with a trial court's view of the facts, we are the ones likely to be wrong. So we should make triple sure that we are correcting, not creating, an error before we reverse.

Cooper illustrates how the ordinary clear-error standard works in districting litigation. The question there, as here, was whether a state legislature chose voters for a congressional district based on their race, or instead based on their past political choices. The three-judge District Court found that race accounted for the new district lines. On review, we decided the evidence “adequately support[ed]” that con-

72

clusion. 581 U. S., at 309. As that phrasing suggests, we nowhere claimed the court was actually right. To the contrary , we observed that in this “thoroughly two-sided case,” both views of the evidence were “plausible” and “permissible ,” and we declined to choose between them. Id., at 299, 307, n. 6; see id., at 316–317 (“Maybe we would have evaluated the testimony differently had we presided over the trial; or then again, maybe we would not have”). Our decision followed from the deference we thought owed to the District Court. Under clear-error review, we noted, “we will not take it upon ourselves to weigh the trial evidence as if we were the frst to hear it.” Id., at 316. Because the District Court's view was “plausible in light of the full record,” it “must govern”—even if another were “equally or more so.” Id., at 293 (internal quotation marks omitted).

Today's decision could not be more different. To be sure, the majority recites the clear-error standard. See ante, at Page Proof Pending Publication 18. But from then on, the majority ignores it—no, worse, does the opposite of what the standard commands. It is not just that the majority refuses to defer to the District Court's fndings in favor of the Challengers. It is that the majority defers to the assertions of the State defendants—the side that lost below. Invoking a “presumption of legislative good faith,” the majority insists that “when confronted with evidence that could plausibly support multiple conclusions,” a court must “draw the inference that cuts” in the State's favor. Ante, at 10. So over and over the majority puts its thumb on the scale against the District Court. Each time it takes up a piece of evidence, the majority declares that there is a “possibility” of seeing it the State's way. Ante, at 20, 24. And that possibility is “dispositive”; because of it, the State's version of the facts must control. Ante, at 20; see also, e. g., ante, at 10, 23, 27 (similarly awarding points to the State because its claims were “plausible,” even if the Challengers' were more so). In effect, the majority's demand for deference to the State overrides clear-error re-

73

view's call for deference to the trial court. If the District Court wants deference, it had better just rule for the State.

That approach conficts with this Court's precedent. Indeed , it has only ever appeared in the Cooper . . . dissent. There too, Justice Alito argued for reversing the trial court's view of evidence because it was not “the only plausible interpretation.” 581 U. S., at 357. There too, he called for accepting the State's contrary view because the evidence could “as easily be understood” that way. Ibid.; see id., at 345, 350, 352, 358–359. The Cooper Court noticed—and disapproved . The dissent, it said, “repeatedly fips the appropriate standard of review,” to give the State rather than the trial court deference. Id., at 309, n. 8. But that move refected “an elemental error”: There is no “super-charged, pro-State presumption on appeal, trumping clear error review .” Ibid. Of course clear-error review takes into account the standard of proof in the trial court. See ante, at Page Proof Pending Publication 33–34, n. 11. But that standard is not transformed because of the good-faith presumption. In our precedents, that presumption tells a court not to assume a districting plan is fawed or to limit the State's opportunities to defend it. See Abbott v. Perez, 585 U. S. 579, 603 (2018) (the presumption requires a plan's challengers to bear the burden of proof); Hunt v. Cromartie, 526 U. S. 541, 553 (1999) (the presumption may suggest sending a case to trial, rather than rejecting a plan on summary judgment). And the presumption reminds a court that it is a serious matter to fnd a State in breach of the Constitution. See Miller, 515 U. S., at 915. But that is all. Nothing in our decisions suggests that a trial court must resolve every plausibly disputed factual issue for the State (as if we could hardly imagine offcials violating the law). And still less do our decisions suggest that the trial court's factual fndings are deprived of deference on appeal. To the contrary, as Cooper stated, clear- error review of those fndings proceeds just as usual, unaffected by the presumption. See 581 U. S., at 309, n. 8; see

74

also Miller, 515 U. S., at 915 (good faith is presumed “until a claimant makes a showing” of “race-based decisionmaking” (emphasis added)).

The majority's deeper reasons for specially indulging the State also clash with this Court's decisions. In the majority 's view, claims of racial gerrymanders are often “weapons of political warfare,” using courts for illegitimate ends. Ante, at 11. And when courts vindicate those claims, they “accus[e]” States of “offensive and demeaning conduct,” bearing “an uncomfortable resemblance to political apartheid ,” ibid.—an apparently intolerable insult even when justifed . Those sentiments, again, come straight out of the dissent in Cooper. See 581 U. S., at 334–335. The Court there took a different view, more refective of our precedents. See id., at 319, n. 15. Time and again, this Court has noted the important role suits like this one play in stopping the unlawful race-based division of citizens into electoral dis- Page Proof Pending Publication tricts. See, e. g., Bethune-Hill v. Virginia State Bd. of Elections , 580 U. S. 178, 187 (2017). For sorting of that kind does occur—sometimes (as here) to serve partisan goals, occasionally just to suppress the political infuence of minority voters. See Cooper, 581 U. S., at 319, n. 15. And when it does, the Court has held, it requires a judicial response. See, e. g., Shaw v. Reno, 509 U. S. 630, 649 (1993). If calling out a racial gerrymander “accus[es]” a State of a grave wrong, then so be it. This Court is not supposed to be so fearful of telling discriminators, including States, to stop discriminating . In other recent decisions, the Court has prided itself on halting race-based decision-making wherever it arises—even though serving far more commendable goals than partisan advantage. See, e. g., Students for Fair Admissions , Inc. v. President and Fellows of Harvard College, 600 U. S. 181, 213–214 (2023). It is not the ordinary thing to agonize so much about giving “offens[e]” to a discriminating State. Ante, at 11.

75

And it is not the right thing either. In adopting its novel credit-the-losing-State approach, the majority thwarts efforts to undo a pernicious kind of race-based discrimination. See Shaw, 509 U. S., at 643 (recognizing racial gerrymanders as “odious”). True enough, as the majority highlights, that the judicial system fails when a State is wrongly found to have gerrymandered a district. But the system fails as badly or worse when a State that has gerrymandered a district gets away with it. This Court has prohibited race- based gerrymanders for a reason: They divide citizens on racial lines to engineer the results of elections (without the justifcation of protecting minority voters' rights). And litigation to remedy that harm is already none too easy. Because of the complex political context, this Court has required challengers of electoral maps to show that race was not just a single but the “predominant” factor in moving voters between districts. Bethune-Hill, 580 U. S., at 187. Page Proof Pending Publication That is, and is meant to be, a demanding burden. But once plaintiffs have met it to a three-judge district court's satisfaction , their hardest job should be done. They should not have to face an upside-down form of clear-error review, in which this Court reverses if it decides there is a “possibility” of seeing the evidence the State's way. Ante, at 20. The principal effect of that novel rule will be to defeat valid voting-discrimination claims.

And the majority is not yet done putting uncommon burdens on gerrymandered plaintiffs. From now on, those plaintiffs will also be subject to an “adverse inference” unless they present a specifc form of evidence—an “alternative map” that would “achieve[ the State's] legitimate political objectives” while “producing signifcantly greater racial balance .” Ante, at 34–35 (internal quotation marks omitted). And that inference gives every sign of packing a wallop. The majority labels it “dispositive in many, if not most, cases,” except when the plaintiff presents (1) direct evidence

76

of a gerrymander (say, an email admitting to the targeting of Black voters) or (2) “some extraordinarily powerful circumstantial evidence such as the strangely irregular twenty- eight-sided district lines” in Gomillion v. Lightfoot, 364 U. S. 339 (1960). Ante, at 35 (internal quotation marks omitted). Think about that last category, as the majority frames it. The majority must go back 65 years, to the most grotesque racial gerrymander in the U. S. Reports, to fnd a case based on circumstantial evidence that could have survived its adverse inference. How better to make the point: The majority 's new evidentiary rule is meant to scuttle gerrymandering cases. Odd that the majority fails to mention a seemingly pertinent fact: Cooper expressly rejected a similar demand that a plaintiff alleging a gerrymander submit an alternative map. In that case, North Carolina argued that “[w]hen race and politics are competing explanations of a district's lines,” the challenger must introduce “an alternative map that achieves Page Proof Pending Publication the legislature's political objectives while improving racial balance. ” 581 U. S., at 317 (alterations omitted). The Cooper dissent agreed. See id., at 332–337. The Cooper Court did not. See id., at 317–322. The Court freely acknowledged that such a map could be good evidence of a racial gerrymander. See id., at 317. So too, it recognized “as a practical matter” that a plaintiff with an otherwise weak case would not prevail without a map. Id., at 319.1 1 The example Cooper gave was Easley v. Cromartie, 532 U. S. 234 (2001). The plaintiffs' direct evidence there, Cooper noted, was “meager” and “weak.” 581 U. S., at 321–322. Cromartie described it as saying “little or nothing” about the role race had played in drawing district lines. 532 U. S., at 253. And the additional, circumstantial evidence did not fll the gap, because it too “offer[ed] little insight” into the basis of the legislature 's mapmaking. Id., at 248. In that evidentiary vacuum, Cooper explained , an alternative map was needed to “carry the day.” 581 U. S., at 322. Not because, as today's majority decides, there is something special about that form of evidence. Just because in Cromartie there was basi-

77

But we could not have been more adamant in rebuffng the State's proposed requirement. “[I]n no area of our equal protection law,” we reasoned, “have we forced plaintiffs to submit one particular form of proof.” Ibid. And we were not about to start. A “plaintiff's task” in a gerrymander case, we stated, “is simply to persuade the trial court—without any special evidentiary prerequisite”—that race was the predominant factor in redistricting voters. Id., at 318. Like all other submissions in a gerrymandering case—the “testimony of government offcials,” proof about the data available to mapmakers, and “expert analysis”—“[a]n alternative map is merely an evidentiary tool.” Id., at 318–319. So “neither [a map's] presence nor its absence can itself resolve a racial gerrymandering claim.” Id., at 319. The majority cannot evade Cooper's force by casting today 's holding as an “adverse inference” rule rather than a simple requirement. First, there is precious little difference Page Proof Pending Publication between the two. Given the apparent strength of the majority 's adverse inference, few litigants will feel free to proceed without commissioning alternative maps. The majority 's inference is effectively a requirement, whether or not it goes by that label. And anyway, Cooper's reasoning easily encompasses—which is to say forbids—the majority's new inference rule. The point in Cooper was to treat maps equivalently to—rather than “elevate” them above—other forms of evidence. Id., at 318. So if the plaintiff's non-map evidence supports a claim, the Court stated, the absence of a map “does not matter.” Ibid. The Cooper dissent well understood the point. No less than three times, the dissent quoted the Court's “does not matter” line, arguing vociferously that a map's absence should matter, if not in all cases, at least in all but “exceptional ones.” Id., at 336; see id., at

cally nothing else. As I'll soon show, that is far from true in this case. See infra, at 80–98.

78

329, 359. The dissent lost that battle, but now succeeds in overturning the essence of Cooper's map ruling.

The majority-née-dissent's reasons for elevating maps above other evidence have not improved since Cooper held to the contrary. The majority states that maps can serve as a good way to undermine a State's “it was all politics” defense . See ante, at 34–35. No argument there: The Cooper Court also said as much. 581 U. S., at 317. But it went on to say that maps “are hardly the only means” of attacking such a defense—as this case well shows. Id., at 318; see infra, at 80–98. The majority also insists that plaintiffs can “easily churn out” alternative maps at “little marginal cost.” Ante, at 35 (quoting, of course, the Cooper dissent). Maybe or maybe not; either way, the Cooper Court said, the matter is irrelevant: We have no “warrant to demand” that plaintiffs jump through “evidentiary hoops” of our creation, “whether the exercise would cost a hundred dollars or a million, a week's more time or a year's,” if they can otherwise prove Page Proof Pending Publication that race predominated in drawing district lines. 581 U. S., at 319, n. 15.2 Finally, the majority suggests that all plaintiffs with serious gerrymandering cases should have known to produce an alternative map. See ante, at 10. But that assertion requires airbrushing Cooper out of our caselaw. What plaintiffs should have known after Cooper was that they could but need not submit an alternative map. The majority today punishes the Challengers for thinking that this Court would be good to its word.

2 And that view is in no way an outlier. Note that the majority must go back almost a century to fnd a decision in which this Court drew an adverse inference against a civil litigant for failure to offer a certain form of evidence. See ante, at 36 (citing Interstate Circuit, Inc. v. United States, 306 U. S. 208, 226 (1939)). And even that decision merely applied an inference in a particular case; it did not create a rule to cover a whole category of suits, as the majority does today. Nor did that old decision relate to a constitutional claim. As far as I know, today's decision is the frst to impose a rule defeating claims of that type merely because plaintiffs chose not to offer one form of evidence, and instead relied on others.

79

In any event, the Challengers had an understandable reason for not offering the kind of map the majority demands. The point of such a map, as the majority explains, is to help fgure out whether race or politics accounts for districting lines. See ante, at 34–35. That function becomes important—so a map makes sense—only if a State in fact defends its plan as arising from political considerations. At trial, South Carolina indeed adopted that defense. But it was not clear beforehand, when the plaintiffs were developing their evidence for trial, that the State would do so. The plain fact is, politicians don't like admitting to partisan gerrymanders : They often deny them as aggressively as they draw them. That is because “[e]xcessive partisanship in districting ” is–and is thought by voters to be—“incompatible with democratic principles.” Rucho v. Common Cause, 588 U. S. 684, 718 (2019). So it is scarcely surprising that, during legislative debate, the districting plan's sponsor responded to Page Proof Pending Publication charges of a partisan gerrymander by asserting “that's really not the case.” J. S. A. Supp. 286a.3 Or that during pretrial proceedings key State witnesses continued to deny partisan motives. Luke Rankin, the Republican chair of the Senate Judiciary Committee, testifed in discovery that it was not “a goal of [his] to make” District 1 “more reliably republican.” Id., at 425a. Likewise, a Republican member of the House Redistricting Committee testifed that he “never considered partisan gain as a goal” of redistricting, and “never” heard “anyone else” admit that goal either. Id., at 409a–410a. And the Senate Redistricting Subcommittee's counsel swore

3 The majority does not help its cause by noting that two Democratic members of the legislature described the districting plan as a partisan gerrymander. See ante, at 26. Even as a districting plan's proponents deny partisan gerrymandering, a plan's opponents often allege it. (And both for the same reason—because voters don't like excessive partisan manipulation of district lines.) That Democrats were attacking the plan as a partisan gerrymander hardly shows that Republicans were likely to defend it in that way.

80

that there was “no effort” to make District 1 “more Republican leaning.” Id., at 392a. So the Challengers, prior to trial, were not on notice of a partisanship defense. The State, to be sure, changed tack in the end: A strong case made by plaintiffs can powerfully concentrate a defendant's mind. But by that time, the Challengers' mapmaker (Dr. Kosuke Imai) had completed his work, and the trial had begun.

Even before looking at the trial evidence, the majority thus places the Challengers in a deep hole. Although this Court recently disclaimed any need for an alternative map, the majority today draws an adverse inference from such a map's absence. And contrary to settled practice, the majority decrees that, even on clear-error review of a ruling for the Challengers, the State will emerge victorious if its version of events is so much as possible. Combine those two facets of the majority's approach, and the trial evidence fades into insignifcance. A legal twist here and a legal bend Page Proof Pending Publication there ensure that the majority need show no respect for the three-judge District Court's well-considered factual fndings.

II

Normal clear-error review would lead to a different outcome . The District Court faced a factual question: Did the State rely signifcantly on racial data in drawing its new District 1? Based on the mountains of evidence presented, the court decided that the State had done so. That fnding was reasonable, and deserves to be affrmed.

As the majority explains, this case concerns changes that South Carolina made in its most recent redistricting to Congressional District 1. See ante, at 11–17. Under the preexisting map, District 1 was a thin strip of land stretching along the Atlantic Coast. See Appendix, infra, at 100, Figure 1 (2011 Congressional Map). It was bordered to the northwest by District 6, the State's only majority-Black district . See ibid.; J. S. A. 429a. After the 2020 census, South

81

Carolina had to redraw both those districts to comply with the Constitution's one-person, one-vote requirement. District 1 was overpopulated by about 88,000 people, and District 6 was underpopulated by about 85,000. The State chose, though, not to make a one-way transfer of residents from the overpopulated to the underpopulated district. To unite two counties, the State frst moved around 53,000 residents from (the underpopulated) District 6 into (the overpopulated ) District 1. That shift, of course, exacerbated the problem: The State now needed to transfer some 140,000 residents in the opposite direction. It did so mainly by moving a large chunk of Charleston County from District 1 to District 6.

And here is the rub—the thing that created this case. The part of the county that the legislature moved out of District 1 was disproportionately Black, and by a lot. The mapmakers targeted several heavily Black neighborhoods in Page Proof Pending Publication North Charleston, while leaving many heavily White neighborhoods alone. See id., at 261a–262a. And no matter how you slice the numbers, the effects were stark. More than 60% of Black Charleston County residents previously in District 1 were relocated to District 6. 649 F. Supp. 3d 177, 189 (SC 2023). Of the 11 precincts with the largest Black populations, 10 were gone. Ibid. Overall, the proportion of African Americans in the excised part of the county (23.8%) was more than twice as high as in the remaining part (10.3%). See id., at 190; Supp. App. 153a. The upshot was that 79% of Charleston County's Black population now found itself in District 6, whereas only 53% had been there before. See 649 F. Supp. 3d, at 190, and n. 9. As the State's main mapmaker—and star witness—acknowledged, the new lines created a “tremendous [racial] disparity” in comparison to the old districting plan. J. S. A. 262a; 649 F. Supp. 3d, at 189.

The question at trial was how that disparity had come about. By that time, the State had adopted its politics-only defense. It argued, as the majority says, that the point of

82

redrawing District 1 was to “enhance[ ] the Republican advantage ” there—i. e., to make sure a Democratic candidate could not win. Ante, at 14. But that claim, even if true, would not be enough for the State to prevail. As this Court has held, a State cannot divide voters by race to achieve political ends. See Miller, 515 U. S., at 914. “[T]he sorting of voters on the grounds of their race” is a constitutional problem “even if race is meant to function as a proxy” for political affliation. Cooper, 581 U. S., at 309, n. 7; see id., at 291, and n. 1. So the critical issue was not whether the State's ultimate aim was political or racial (though the majority often phrases it that way, see, e. g., ante, at 6, 9–10, 21– 22). Instead, the issue was whether the State had advanced its partisan objective primarily by racial means. The Challengers maintained that it had. They said the State's mapmakers had consciously removed Black citizens from District 1 on the ( justifed) assumption that doing so would turn the district redder. The State, by contrast, denied in any way Page Proof Pending Publication using race to draw District 1's lines. According to its account , the disproportionate removal of African Americans from District 1 was just an accidental byproduct of political sorting—more specifcally, of ejecting precincts that had strongly supported then-candidate Biden in the 2020 election .4 Faced with those competing stories, the District Court had to decide which to credit.

The court's decision to credit the Challengers, as I'll next show, was not clear error—indeed, far from it. There was 4 A notable feature of this case is that the State chose to litigate it in categorical terms, claiming that the new district lines were based only on political data and not at all on racial data. The State did not need to go that far. In a gerrymandering case, a defendant can prevail by arguing that although race played some role in redistricting, it was not the “predominant factor.” Miller, 515 U. S., at 916. The State's eschewal of that more moderate assertion turned the factual issue about what its mapmakers did into a binary choice. I therefore mainly address it in those terms, though the Challengers' evidence was powerful enough to support a fnding of gerrymandering even had the State put predominance at issue.

83

of course evidence pointing in each direction; like Cooper, this was a “two-sided case.” 581 U. S., at 307, n. 6. But the Challengers made a weighty showing that the mapmakers relied substantially on racial data in moving voters around. The mapmakers had the incentive to do so, given the limits of the political information in their possession. They had the ability to do so—both access to data and experience using it. And direct testimony showed that the mapmakers had in fact continually examined racial data during the line- drawing process. The map yielded by that process hit on the dot the Black voting percentage that state offcials knew they needed to achieve their partisan goal. And when statistics experts reviewed the map, they found that the State's politics-only story could not explain the redistricting's extreme racial disparity. In dismissing that strong case, the majority cherry-picks evidence, ignores credibility fndings, misunderstands expert views, and substitutes its own statis- Page Proof Pending Publication tical theories. Its opinion gives not a whit of respect to the District Court's factual fndings, thus defying the demands of clear-error review.

A

Start with the State's chief mapmaker. William Roberts, as the majority notes, was a “nonpartisan staffer with 20 years of experience” drawing maps for Republicans and Democrats alike. Ante, at 13. He was good at what he did—expert, “helpful,” and “precise.” J. S. A. 74a, 254a. And also this—he was a veteran consumer of racial data. On cross-examination, Roberts testifed as follows:

Q: I think I heard the number of 75 to a hundred localities you've worked in over the past 20 years? A: Yes. . . . Q: Before this redistricting cycle, you always looked at race data in the 75 to a hundred districts you worked in, correct? A: Yes. . . .

84

Q: Indeed, . . . you provided guidance to localities that they should be looking at BVAP [Black Voting-Age Population ] in drawing lines, correct? A: That's correct.

Id., at 204a–205a. The point of looking at BVAP, according to the mapmaker's testimony, was not to suppress the Black vote. Rather, Roberts stated that he did so to achieve a panoply of lawful districting goals—like assessing Voting Rights Act compliance and “help[ing] the general public understand the race of voters getting moved in and out.” Id., at 206a; see id., at 205a. Whatever the particular purpose, he consulted racial data constantly. Now as you know from the majority, Roberts denied doing so in the redistricting at issue here. See ante, at 19. But when asked “so in your 20 years of redistricting, this was the only time [that] you didn't look at race?,” Roberts answered “That's correct.” J. S. A. 207a. Page Proof Pending Publication True to his persistent practice (if not to his this-case-only denial), Roberts confgured maproom computers to show how every line-drawing decision would affect the new District 1's racial make-up. In other words, as a mapmaker moved a district line this way or that, he could immediately see the resulting change in the district's BVAP. Displaying racial data in that way was not an unavoidable feature of the mapmaking software. As one staffer explained: “[Y]ou could confgure” the computer setup “in a multitude of ways.” ECF Doc. 462–9, at 114. You could make it so that new BVAP numbers appeared on your screen “while you manipulated geography”—but “there [was] no requirement that you ha[d] to set it up that way.” Ibid. The mapmakers had to choose to display racial data. And here is the key thing: They did. A Senate staffer who often sat with Roberts in the maproom explained that not only “political data” but also “demographic data”—specifcally, “race” and “voting age population by race”—was “visible” on computer screens “[a]

85

lot of the time.” ECF Doc. 462–4, at 40. And on cross- examination, Roberts admitted that to be true:

Q: So BVAP was visible on the screen while you were drawing maps? A: Yeah. It was in the statistics window at the bottom of the screen. Q: So, you could see BVAP as you were making changes in real time as you were drawing lines? A: We could see the statistics update after a change was made. Q: So, if you moved a district line, you could see if the BVAP went up or down, right? A: You could see on the statistics what the overall district BVAP would be.

J. S. A. 207a; see J. S. A. Supp. 402a (another staffer acknowledging : “Was I aware of, while I was drawing, what the Page Proof Pending Publication racial makeup of what I was drawing was? Yes”).

So Roberts's testimony presented a puzzle. As the majority highlights, Roberts consistently denied relying on racial data. See, e. g., ante, at 19, 22. But racial data, according to both him and others, was easily accessible—in fact, was usually visible—on his computer while the line-drawing was going on. And he never explained why it was there. Why confgure a computer to tell you, at every stage of the mapmaking process, how the slightest change in a district line would affect Black voting-age population if you weren't tracking and manipulating Black voting-age population? Roberts had no answer.

But there was an obvious reason for attending so closely to racial data, as even the majority acknowledges: One surefre way of making a South Carolina district more Republican is to make it less Black. See ante, at 20. The difference between a “Republican tilt” and a “Democratic tilt” in District 1, notes the majority, is the difference between a 17% BVAP and a 21% BVAP. Ibid. That is because in recent

86

statewide elections, more than 90% of Black South Carolina voters—and usually more than 95%—have supported the Democratic candidate. See J. S. A. Supp. 82a. In South Carolina, to remove a Black voter from a congressional district is pretty nearly to remove a future Democratic vote. That is no secret. So it is small wonder that racial data was conspicuously displayed on Roberts's computer. And then small wonder that the District Court found Roberts to have used that data to draw district lines. See 649 F. Supp. 3d, at 191. More doubt would properly have attached to the opposite fnding—that Roberts put this hugely relevant data on his screen only to ignore it as he worked to make District 1 more Republican. That would have taken the self- restraint of a monk.

Especially so because using only the political data at hand would not have done the job as well. “Why,” the majority asks, “would Roberts have used racial data” when he had Page Proof Pending Publication access to sub-precinct-level voting data from the 2020 election ? Ante, at 22–23; see ante, at 38. The question is apparently meant to be rhetorical; but the trial record provides a ready answer—and one more than suffcient on clear-error review. One of the Challengers' experts testifed that “[t]he 2020 election data” was “not a good” measure of partisan tilt—neither so “accurate” nor so “reliable.” App. 135. And racial data, another expert suggested, served the mapmakers' goal better. See id., at 112. The single-sentence explanation is this: In South Carolina, a Black voter is more likely to vote for a Democrat in the next election than is someone who voted for a Democrat in the last election. That is because White voting preferences in the State are not as “stable” as Black voting preferences. Ibid. A White voter “might vote for a Democrat in one election ” only to vote “for a Republican in another.” Ibid. So to remove a past Democratic voter (as contrasted with a Black voter) is not necessarily to remove a future Democratic

87

vote.5 And the gap only widens for past presidential voters , like those who participated in the 2020 election. In presidential elections, one expert explained, more people than usual switch party lines to “vote for the candidate”—a trend that then-President Trump's candidacy may have further amplifed. Id., at 135; see J. S. A. 382a. Given all that, the South Carolina mapmakers' racial data was peculiarly predictive: The single best thing Roberts and his staff could do to increase the future Republican vote in District 1 was to exclude a Black voter. That fact would not have meant they looked at racial data alone; they also had the 2020 election data on their computers. But the racial data offered a potent tool for ensuring that District 1 would vote for a Republican in coming elections.6 And strong evidence showed, as the District Court found, that the mapmakers wielded this tool—that they used their racial data to meet the BVAP level needed to achieve their partisan goal. Recall the large turnover of voters in Dis- Page Proof Pending Publication trict 1. See supra, at 81. Some 53,000 people were moved into, and 140,000 people were moved out of, the district (which wound up with 730,000 total). Yet the district's racial balance did not budge. The district began with a 16.6% 5 The same variability occurs the other way around. In other words, a White voter might vote for a Republican in one election only to vote for a Democrat in another. So to retain a past Republican voter in a district is not necessarily to retain a future Republican vote.

6 In arguing to the contrary—that the political data was superior to, and would have removed any incentive to use, racial data—the majority emphasizes that only the political data “accounted for voter turnout.” Ante, at 22–23, and n. 7, 38. But as one of the Challengers' experts explained , that fact is a double-edged sword, because turnout in presidential elections is highly unrepresentative of turnout in off-year ones. See App. 135. And still more important, the mapmakers did not have to make a choice between using political data alone and racial data alone. They could get whatever turnout (or other) information the political data provided even as they used the racial data as an especially reliable and accurate measure of individual voting behavior.

88

BVAP. See J. S. A. 430a. That number went up with the 53,000-person addition, because almost 40% of the new residents were Black. See id., at 439a. So what did the mapmakers do? As noted earlier, they removed from District 1 over 60% of Black Charleston County residents, by excising a part of the county more than twice as Black (23.8%) as the part they kept in (10.3%). See 649 F. Supp. 3d, at 189–190; Supp. App. 153a; supra, at 81. That brought the district's BVAP right back down to 16.7%—again below the 17% required to create the desired Republican tilt. See J. S. A. 452a; 649 F. Supp. 3d, at 188. In the majority's description, what happened was of no particular note—just that the District 's BVAP “stayed more or less constant.” Ante, at 20. But consider: With approximately a quarter of District 1's population moving in or out, the district's BVAP shifted by . . . one-tenth of one percentage point. The District Court observed that uncanny stability, knowing that racial data Page Proof Pending Publication was at the mapmakers' fngertips. See 649 F. Supp. 3d, at 191. And the court, as addressed shortly, had heard statistical experts deny that the racially disparate districting could have come about through political sorting. See infra, at 91– 98. So it was no large step—and hardly clear error—for the court to conclude that the mapmakers had gerrymandered Charleston County to achieve “a target of 17%” BVAP. 649 F. Supp. 3d, at 193.

As against all that, what does the majority offer? Only a series of self-serving denials. The sum and substance of the State's case came from the testimony of Roberts and State Senator George Campsen, who was the redistricting plan's sponsor. Yes, the new map, Roberts conceded, had a “tremendous ” racial skew. J. S. A. 262a. But Roberts and Campsen maintained that they had never sorted by race— never used their (constantly accessible) racial data to draw district lines. Both insisted that they had looked only to voting results from the 2020 election to ensure their partisan goal. The majority buys it—hook, line, and sinker. Indeed,

89

the majority relies on nothing else. It treats Roberts's and Campsen's account as a “fact of the matter,” rather than a vigorously contested assertion. Cooper, 581 U. S., at 307, n. 6; see, e. g., ante, at 13–15. The majority trusts the two State witnesses, and believes what they said.

The problem is that the three judges who sat on the District Court did not. And they are the ones entitled to make credibility judgments. See supra, at 71; Cooper, 581 U. S., at 309 (“[W]e give singular deference to a trial court's judgments about the credibility of witnesses”). That is for an obvious reason: They were there. They could assess every aspect of a witness's testimony, including demeanor, tone of voice, and facial expression. They could see when the witness was at ease and when he stumbled. And after taking account of all those cues, the three judges all reached the same conclusion about Roberts and Campsen. They thought that those two witnesses were not telling the truth. Page Proof Pending Publication The panel was especially disbelieving of Roberts, if almost in spite of itself. The court (contra the majority) well understood what the presumption of good faith required. The judges were predisposed, as the majority has to acknowledge , to think that this “good man,” who had for so long been a fxture on the South Carolina political scene, would play it straight. Ante, at 13, and n. 5 (citing J. S. A. 74a–75a, 254a, 263a, 421a). But in the end, the court felt compelled to fnd that Roberts's old habit of relying on race died hard. To the panel, the mapmaker's tale did not hang together. He said he did not consider race in drawing lines; but he could recite “off the top of his head” the racial breakdown of particular precincts in District 1. 649 F. Supp. 3d, at 191. Those “highly accurate” estimates, the court noted, refected Roberts 's obvious knowledge of “the racial demographics of the state down to the individual precinct level.” Ibid., n. 12. And Roberts never did—never could—explain why he put so much racial data on his computer screen if not to look at it as he drew district lines. Especially given the surrounding

90

evidence, the court found, Roberts's “claim that he did not consider race” in excluding voters from District 1 “rings hollow .” Id., at 191 (internal quotation marks omitted). On normal clear-error review, that credibility judgment would control.

And so too for Campsen, who obfuscated at every turn. At trial, Campsen reversed his own deposition testimony about whether state senators knew the racial makeup of their districts. (First they knew, then he couldn't possibly speak for them.) See J. S. A. 377a–378a. He answered as simple a question as whether “race and party are correlated in South Carolina” this way:

“Yes—well, yes and no. I guess that's fuid. It is fuid, but yes. . . . Well, it's not in every instance, but generally African Americans tend to vote higher, you know, more—you can look at the polls—when you look at the numbers after the fact—I didn't look at them drawing Page Proof Pending Publication the map—but you see that in the numbers.” Id., at 381a.

And he contradicted common knowledge—as well as the State's own defense—when he point-blank denied that sorting people based on their voting behavior could result in racial disparities. See id., at 383a (“Q: You would agree with me that if you . . . focus on partisan numbers, there's a risk that you might disproportionately impact Black voters in drawing lines, right? A: No, I'm not going to agree with that”). Would you buy what this man was selling? As the contradictions, non-answers, and evasions mounted, the District Court quite reasonably decided that it could not.

Put all this together, and the Challengers offered—even before getting to their statistical studies—a more than plausible case of racial gerrymandering. They showed that the exclusion of voters from District 1 was racially disproportionate —not by a little but by a lot. They showed that the State's star mapmaker had always—always—before

91

considered race in drawing district lines. They showed why he would want to do so here, to create a reliable Republican tilt. They showed that the mapmaker confgured his computer to exhibit in real time how every adjustment of a district line affected the district's racial make-up. And they showed that after moving nearly 200,000 residents this way and that, the mapmaker managed to land on the exact BVAP fgure he knew would ensure his political goal. Now it is true that the State, when confronted with this evidence, did not confess error, as the majority comes close to demanding. Its offcials, as you might expect, adamantly disputed the charge of racial discrimination. But they could not keep their story straight or make it believable to three judges. The more the offcials talked, the more the court became convinced that, to create a red District 1, they had divided citizens by race. And that, again, was even before the statisticians took center stage.

B

Page Proof Pending Publication Once the statisticians did so, the Challengers' case was clinched—at the least, from a clear-error perspective. Consider how much the controverted issue lent itself to statistical evidence. That issue began with a simple fact: The part of Charleston County that the mapmakers excised from District 1 was (vastly) disproportionately Black. The dispute was about what caused that disparity. Statistical evidence showing that it could have arisen from political sorting would signifcantly beneft the State's defense. Conversely, statistical evidence showing that the racial disparity could not have arisen in that way would signifcantly beneft the Challengers' case. So you might think that the trial would feature a war of statistical experts, each presenting their own multivariate regressions. But you would be wrong. The Challengers did their part, but the State failed to respond in kind. Rather than submit its own statistical studies , the State devoted all its efforts to trying to pick apart the Challengers'. It thus anticipated today's majority,

92

which (given the unbalanced record) can do nothing more than search for holes, however minute, in the Challengers' expert evidence. But two separate studies emerge unscathed , and with signifcant probative force—fully suffcient on clear-error review to justify the District Court's conclusion . Each analysis was designed to answer the critical question: whether Charleston County was split as it was based on its residents' race. And each found that it was. Even controlling for political preference, Black voters were more likely than White voters to be removed from District 1.7 Dr. Jordan Ragusa's regression found that race, separate and apart from partisanship, was “an important factor in the design of the 1st district.” J. S. A. 509a; see 649 F. Supp. 3d, at 192. Ragusa looked at the size, racial demographics, and partisan composition of each precinct in the old District 1. (His measure of partisanship was the vote count for then- candidate Biden in the 2020 election, which mirrored the po- Page Proof Pending Publication litical data the State's mapmakers possessed.) By controlling for all three of those variables, Ragusa explained, he could “statistically disentangle the effect of each factor.” J. S. A. 505a. And when he did so, Ragusa determined that “the decision to move a [precinct] out of [District 1] was highly correlated to the number of African American voters” in the precinct. 649 F. Supp. 3d, at 192; see J. S. A. 508a– 509a, 514a. If, for example, a precinct had 100 to 500 Black voters, “the chance of [its] being moved out” of District 1 was “no greater than 20%.” 649 F. Supp. 3d, at 192. But 7 Two other studies on which the majority expends much effort, see ante, at 24–27, 33, had only a tenuous connection to the race-versus-politics question. Dr. Moon Duchin's analysis was offered primarily to support the Challengers' independent vote-dilution claim. And Dr. Kosuke Imai's report was designed to address a different defense the State could have raised—that traditional districting principles accounted for District 1's lines. Those two studies are therefore irrelevant. They do not help the Challengers on the disputed issue. But neither does the majority score any points for saying as much.

93

as the number climbed, so did the likelihood: When a district had 1,500 Black voters, the probability of exclusion reached 60%. See ibid. And on top of that analysis, Ragusa directly compared the effects of partisanship and race on the exclusion decision. He found that the mapmakers removed 41% of precincts with more than 1,000 Biden voters, but 62% of precincts with more than 1,000 Black voters. See J. S. A. Supp. 14a. That comparison showed that “the racial composition of a precinct was a stronger predictor of whether it was removed” from District 1 “than its partisan composition .” Ibid.; see 649 F. Supp. 3d, at 192.

A second expert, Dr. Baodong Liu, reinforced Ragusa's conclusions about the signifcance of race, using a complementary methodology and data set. Liu evaluated the different likelihoods that White Democrats and Black Democrats would wind up outside or inside District 1. Based on demographic data and vote tabulations from the 2018 Demo- Page Proof Pending Publication cratic primary, Liu frst found that Black Democrats were moved out of District 1 disproportionately to White Democrats . Whereas 26% of Black Democrats in the district were excluded, only 19% of White Democrats were; so the rate at which Black Democrats were excluded was more than onethird higher. See J. S. A. Supp. 94a. And then Liu sliced his data another way, which confrmed his results. Replicating a methodology that this Court approved in Cooper, see 581 U. S., at 315, Liu looked at Democratic voters in all the counties that at least partly overlapped with District 1. Which of those voters, Liu asked, actually wound up in District 1 and which did not? Once again, the answer showed a signifcant racial disproportion. Whereas 69% of White Democrats in the region were placed in the new District 1, only 51% of Black Democrats were put there. J. S. A. Supp. 100a.

The majority's primary objection to Ragusa's and Liu's studies—that they did not “control for contiguity or compactness ,” ante, at 28, 31—is woefully misplaced. The gripe is

94

that the experts assumed “unrealistic[ally]” that any precinct , no matter where located, could be moved. Ante, at 28. If the experts had thought about geography, the majority suggests, they might have found that Black Democrats were disproportionately relocated because they lived in precincts closer to a district boundary. The argument is reprised from Cooper—but (what a surprise) only from the dissent. See 581 U. S., at 358. And the reason the objection got nowhere in Cooper applies once again. The relevant district in Cooper was super-thin, so that the lion's share of precincts within it were close enough to a boundary line to be easily moved. See id., at 326. And so too here. Recall that the only issue under review is whether the State improperly moved Black voters from District 1 to District 6—because that is the only gerrymander the District Court found. Now turn to the map of South Carolina's old districts in this opinion 's Appendix. District 1 was a narrow strip on the Atlantic coast; District 6 ran along its whole length. Nearly ev- Page Proof Pending Publication eryone within District 1 lived close to the border line; so nearly everyone could have been sent to District 6, consistent with contiguity and compactness. That is true even of people who lived on the beach. Under the State's districting guidelines, “[c]ontiguity by water is suffcient,” so the mapmakers could—and in fact did—split the new District 1's land area by pulling District 6 all the way to the water. J. S. A. 541a; see Appendix, infra, at 100, Figure 2 (Inset to 2022 Congressional Map). The upshot is that precinct location did not meaningfully constrain the State's choice of which voters to move from District 1 to District 6. And so the Challengers ' experts were not required to pretend that it did.8

8 None of that is to say, as the majority seems to think I say, that all or nearly all District 1 precincts touch the District 1-District 6 line. See ante, at 29, n. 8. Some of the district's precincts are indeed several precincts away from the border. But that fact in no way revives the majority's objection to the expert reports. Because of District 1's thinness , almost all of its 300 precincts could (contra the majority) “[ ]realisti-

95

That is why the majority, to support its contiguity theory, must use a “simple example” of zero relevance to this case. Ante, at 28. Says the majority: District 6 “precincts near [Colleton C]ounty's northern border with Bamberg County could not have been moved into District 1 without egregiously fouting the State's important interests in contiguity or compactness.” Ante, at 29. That is true: As the map shows, District 6 is fat, and the precincts the majority mentions are far away from the District 1-District 6 line. See Appendix, infra, at 100, Figure 1. But of course this case has nothing to do with those outermost District 6 precincts, or even with the closer-in District 6 precincts that could have been moved into District 1. The sole issue here, again, is whether the State disproportionately selected heavily African-American precincts to move out of District 1. When it gets around to that issue, the majority says: “[T]he same problem” as in its example “arises with respect to the question whether a precinct in District 1 . . . could have been Page Proof Pending Publication moved into District 6.” Ante, at 29. But that is not true, for self-evident reasons. As just described—and shown on the map—the old District 1 was thin, and the great bulk of its precincts were close to the District 1-District 6 line. See Appendix, infra, at 100, Figure 1. So they could have been moved “without egregiously flouting ”—actually, without fouting at all—“the State's important interests in contiguity or compactness.” Ante, at 29. The majority's inapt comparison is revelatory in one sense only: It shows why appellate courts are supposed to use a clear-error standard—to make sure we are fxing, not introducing, mistakes.

c[ally]” have been moved, either alone or with a few others, to District 6. Ante, at 28. (And so what if with a few others?: The State generally moved precincts around in clumps.) In other words, the State's preference for contiguity and compactness left almost all precincts on the table as candidates for removal. The choice of which of those precincts to move must therefore have been explained by other variables, as the Challengers' experts concluded.

96

The majority's other main criticism, aimed solely at Ragusa , is original to this Court: It was never raised or considered below (or, as far as I know, in other voting suits). The objection relates to the way Ragusa measured each precinct 's partisan tilt. He asked how many 2020 Biden voters lived in a precinct relative to its voting-age population. So, for example, a 1,250-person precinct with 700 Biden voters would count as much more Democratic than the same-sized precinct with 350 Biden voters. The majority says that measure may be “statistically permissible”—but still is not good enough. Ante, at 30. In the majority's view, Ragusa should have “account[ed] for” potential variance in precinct turnout by looking to the Biden net vote instead of the Biden total vote. Ante, at 30–31. Now I'll admit: I'm not a statistician . I can see what the majority is saying, but my inclination would be to seek out other opinions—including from Ragusa himself—about the net-vote approach, and whether it Page Proof Pending Publication would matter. The problem is I can't do that here. The theory is the majority's brainchild, absent from the District Court's proceedings. The State never asked Ragusa about it, before or during trial. The State's own expert did not bring it up. The State did not raise it in briefng below. And most important: Nothing in the trial record suggests that adopting the net-vote measure would have made a real difference. The majority, to show you why it might, offers what it calls a “simplifed” example. Ante, at 30. For simplifed read “fctional”—meaning, not refective of any actual precinct's vote. And for simplifed, also read “unrepresentative ”? To take just one example: Maybe there are some, but I doubt there are many, precincts in which 1,100 of 1,250 voting-age people make it to the polls. See ibid. A number of things about precinct composition and turnout would need to be true for the net-vote/total-vote distinction to make a signifcant difference to Ragusa's analysis —and we know none of them. Sure, it's fun to play

97

armchair statistician. But it's irresponsible to reverse a trial court's decision—on clear-error review—based on such hypothesizing.

A couple of fnal attacks fare no better. The majority faults Liu for testing partisan tilt in District 1 with data from the 2018 gubernatorial primaries, rather than the 2020 presidential election. The majority confdently declares that because an off-year primary has a lower turnout, the “[d]ata from [it] is less informative.” Ante, at 32. Liu's explanation is deemed unworthy of mention. It was that the higher turnout of a presidential election, along with its greater focus on individual candidates, makes it a poorer measure of a district's year-in, year-out partisan tilt. See App. 135. The State's own expert did not contest that view, so the majority's skepticism again fnds no support in the trial record. And even if 2020 data is better than 2018 data—it might be—what is better than either is both. That Page Proof Pending Publication is what the Challengers had: Ragusa's study based on 2020 data and Liu's based on 2018 data, each showing a racial gerrymander.

Much the same thing is true as to a more obscure methodological issue the majority raises (again, needless to say, sua sponte): whether statistical analysis should “operate[ ] at the voter level” or the precinct level. Ante, at 31, n. 9. Here, the majority cannot get its attack-line consistent. First the majority claims that Ragusa's testimony was worse than the expert's in Cooper because Ragusa's relied on “precinct-level analysis” rather than looking at individual voters. Ibid. But within a page the majority asserts that Liu's study was “highly unrealistic” because he “treated each voter as an independent unit” rather than considering “neighbors” together. Ante, at 32. So an expert challenging a gerrymander can't win either way. But put that aside; the key thing, once more, is that the Challengers had not one but two types of analysis working in their favor.

98

However a statistician looked at the data—whether voterlevel or precinct-level—he reached the same conclusion: that the State's mapmakers targeted Black voters.

And the State offered little by way of rebuttal. It, too, had an expert witness. And that witness, Sean Trende, took a couple of shots at Ragusa's methods. See ECF Doc. 510, at 46–52. But he did not offer the most relevant kind of evidence—a counter-analysis showing that partisanship subsumed race in the design of District 1. Trende had access to all the same data Ragusa did. He even had access to Ragusa's computer code, so that he would not have needed to start from scratch. See id., at 58. He could just have rerun the code after fxing whatever variables he thought wrong. What should one make of Trende's failure to do so? If I were adopting the majority's methods, I would draw an “adverse inference” from the decision not to submit such “easily churn[ed] out” evidence. Ante, at 34–35. Surely it must count as an “implicit concession” by the State that the Page Proof Pending Publication statistical analysis, even with the desired fxes, would keep showing evidence of a racial gerrymander? Ante, at 35. But I don't need to create a novel adverse inference to make the critical point. It was hardly clear error for the District Court to credit the Challengers' statistical evidence about race's predominant role when the State presented no similar evidence to support its partisanship theory. The majority's contrary view—that the State's nothing necessarily beat the Challengers' something—is one more tell that it has left the proper review standard way behind.

III

In every way, the majority today stacks the deck against the Challengers. They must lose, the majority says, because the State had a “possible” story to tell about not considering race—even if the opposite story was the more credible . Ante, at 20. And they must lose again, the majority says, because they failed to offer a particular form of proof—

99

which they did not know would be relevant and which this Court recently told plaintiffs was not required. It does not matter that the Challengers offered extensive evidence, including expert statistical analyses, that the State's districting plan was the product of racial sorting. It does not matter that the State, by way of response, offered little more than strained and awkward denials. It does not matter that three judges—entitled to respect for their factual fndings— thought that those denials were not believable, and did not put a dent in the plaintiffs' proof. When racial classifcations in voting are at issue, the majority says, every doubt must be resolved in favor of the State, lest (heaven forfend) it be “accus[ed]” of “offensive and demeaning” conduct. Ante, at 11.

What a message to send to state legislators and mapmakers about racial gerrymandering. For reasons I've addressed , those actors will often have an incentive to use race Page Proof Pending Publication as a proxy to achieve partisan ends. See supra, at 85–87. And occasionally they might want to straight-up suppress the electoral infuence of minority voters. See Cooper, 581 U. S., at 319, n. 15. Go right ahead, this Court says to States today. Go ahead, though you have no recognized justifcation for using race, such as to comply with statutes ensuring equal voting rights. Go ahead, though you are (at best) using race as a short-cut to bring about partisan gains—to elect more Republicans in one case, more Democrats in another . It will be easy enough to cover your tracks in the end: Just raise a “possibility” of non-race-based decision- making, and it will be “dispositive.” Ante, at 20. And so this “odious” practice of sorting citizens, built on racial generalizations and exploiting racial divisions, will continue. Shaw, 509 U. S., at 643. In the electoral sphere especially, where “ugly patterns of pervasive racial discrimination” have so long governed, we should demand better—of ourselves , of our political representatives, and most of all of this Court. Id., at 639. Respectfully, I dissent.

Appendix

APPENDIX

Figure 1. 2011 Congressional Map (adapted from ECF Doc. 323–1, p. 2) Page Proof Pending Publication

Figure 2. 2022 Congressional Map (adapted from J. S. A. Supp. 306a)

Reporter's Note

The attached opinion has been revised to refect the usual publication and citation style of the United States Reports. The revised pagination makes available the offcial United States Reports citation in advance of publication. The syllabus has been prepared by the Reporter of Decisions Page Proof Pending Publication for the convenience of the reader and constitutes no part of the opinion of the Court. A list of counsel who argued or fled briefs in this case, and who were members of the bar of this Court at the time this case was argued, has been inserted following the syllabus. Other revisions may include adjustments to formatting, captions, citation form, and any errant punctuation. The following additional edits were made:

None