Cooper v. Harris, 581 U.S. ___ (2017)

NOTICE: This opinion is subject to formal revision before publication in the preliminary print of the United States Reports. Readers are requested to notify the Reporter of Decisions, Supreme Court of the United States, Washington, D. C. 20543, of any typographical or other formal errors, in order that corrections may be made before the preliminary print goes to press.

SUPREME COURT OF THE UNITED STATES

_________________

No. 15–1262

_________________

ROY COOPER, GOVERNOR OF NORTH CAROLINA, et al., APPELLANTS v. DAVID HARRIS, et al.

on appeal from the united states district court for the middle district of north carolina

[May 22, 2017]

Justice Kagan delivered the opinion of the Court.

The Constitution entrusts States with the job of designing congressional districts. But it also imposes an important constraint: A State may not use race as the predominant factor in drawing district lines unless it has a compelling reason. In this case, a three-judge District Court ruled that North Carolina officials violated that bar when they created two districts whose voting-age populations were majority black. Applying a deferential standard of review to the factual findings underlying thatdecision, we affirm.

I

A

The Equal Protection Clause of the Fourteenth Amendment limits racial gerrymanders in legislative districting plans. It prevents a State, in the absence of “sufficient justification,” from “separating its citizens into different voting districts on the basis of race.” Bethune-Hill v. Virginia State Bd. of Elections, 580 U. S. ___, ___ (2017) (slip op., at 6) (internal quotation marks and alteration omitted). When a voter sues state officials for drawing such race-based lines, our decisions call for a two-step analysis.

First, the plaintiff must prove that “race was the predominant factor motivating the legislature’s decision to place a significant number of voters within or without a particular district.” Miller v. Johnson, 515 U. S. 900, 916 (1995) . That entails demonstrating that the legislature “subordinated” other factors—compactness, respect for political subdivisions, partisan advantage, what have you—to “racial considerations.” Ibid. The plaintiff may make the required showing through “direct evidence” of legislative intent, “circumstantial evidence of a district’s shape and demographics,” or a mix of both. Ibid.[1]

Second, if racial considerations predominated over others, the design of the district must withstand strict scrutiny. See Bethune-Hill, 580 U. S., at ____ (slip op., at 13). The burden thus shifts to the State to prove that its race-based sorting of voters serves a “compelling interest” and is “narrowly tailored” to that end. Ibid. This Court has long assumed that one compelling interest is complying with operative provisions of the Voting Rights Act of 1965 (VRA or Act), 79Stat. 437, as amended, 52 U. S. C. §10301 et seq. See, e.g., Shaw v. Hunt, 517 U. S. 899, 915 (1996) (Shaw II).

Two provisions of the VRA—§2 and §5—are involved in this case. §§10301, 10304. Section 2 prohibits any “standard, practice, or procedure” that “results in a denial or abridgement of the right . . . to vote on account of race.” §10301(a). We have construed that ban to extend to “vote dilution”—brought about, most relevantly here, by the “dispersal of [a group’s members] into districts in which they constitute an ineffective minority of voters.” Thornburg v. Gingles, 478 U. S. 30 , n. 11 (1986). Section 5, at the time of the districting in dispute, worked through a different mechanism. Before this Court invalidated its coverage formula, see Shelby County v. Holder, 570 U. S. __ (2013), that section required certain jurisdictions (including various North Carolina counties) to pre-clear voting changes with the Department of Justice, so as to forestall “retrogression” in the ability of racial minorities to elect their preferred candidates, Beer v. United States, 425 U. S. 130, 141 (1976) .

When a State invokes the VRA to justify race-based districting, it must show (to meet the “narrow tailoring” requirement) that it had “a strong basis in evidence” for concluding that the statute required its action. Alabama Legislative Black Caucus v. Alabama, 575 U. S. ___, ___ (2015) (slip op., at 22). Or said otherwise, the State must establish that it had “good reasons” to think that it would transgress the Act if it did not draw race-based district lines. Ibid. That “strong basis” (or “good reasons”) standard gives States “breathing room” to adopt reasonable compliance measures that may prove, in perfect hindsight, not to have been needed. Bethune-Hill, 580 U. S., at ___ (slip op., at 16).

A district court’s assessment of a districting plan, in accordance with the two-step inquiry just described, warrants significant deference on appeal to this Court.[2] We of course retain full power to correct a court’s errors of law, at either stage of the analysis. But the court’s findings of fact—most notably, as to whether racial considerations predominated in drawing district lines—are subject to review only for clear error. See Fed. Rule Civ. Proc. 52(a)(6); Easley v. Cromartie, 532 U. S. 234, 242 (2001) (Cromartie II); id., at 259 (Thomas, J., dissenting). Under that standard, we may not reverse just because we “would have decided the [matter] differently.” Anderson v. Bessemer City, 470 U. S. 564, 573 (1985) . A finding that is “plausible” in light of the full record—even if another is equally or more so—must govern. Id., at 574.

B

This case concerns North Carolina’s most recent redrawing of two congressional districts, both of which have long included substantial populations of black voters. In its current incarnation, District 1 is anchored in the northeastern part of the State, with appendages stretching both south and west (the latter into Durham). District 12 begins in the south-central part of the State (where it takes in a large part of Charlotte) and then travels northeast, zig-zagging much of the way to the State’s northern border. (Maps showing the districts are included in an appendix to this opinion.) Both have quite the history before this Court.

We first encountered the two districts, in their 1992 versions, in Shaw v. Reno, 509 U. S. 630 (1993) . There, we held that voters stated an equal protection claim by alleging that Districts 1 and 12 were unwarranted racial gerrymanders. See id., at 642, 649. After a remand to the District Court, the case arrived back at our door. See Shaw II, 517 U. S. 899 . That time, we dismissed the challenge to District 1 for lack of standing, but struck down District 12. The design of that “serpentine” district, we held, was nothing if not race-centric, and could not be justified as a reasonable attempt to comply with the VRA. Id., at 906; see id., at 911–918.

The next year, the State responded with a new districting plan, including a new District 12—and residents of that district brought another lawsuit alleging an impermissible racial gerrymander. A District Court sustained the claim twice, but both times this Court reversed. See Hunt v. Cromartie, 526 U. S. 541 (1999) (Cromartie I); Cromartie II, 532 U. S. 234 . Racial considerations, we held, did not predominate in designing the revised District 12. Rather, that district was the result of a political gerrymander—an effort to engineer, mostly “without regard to race,” a safe Democratic seat. Id., at 245.

The State redrew its congressional districts again in 2001, to account for population changes revealed in the prior year’s census. Under the 2001 map, which went unchallenged in court, neither District 1 nor District 12 had a black voting-age population (called a “BVAP”) that was a majority of the whole: The former had a BVAP of around 48%, the latter a BVAP of around 43%. See App. 312, 503. Nonetheless, in five successive general elections conducted in those reconfigured districts, all the candidates preferred by most African-American voters won their contests—and by some handy margins. In District 1, black voters’ candidates of choice garnered as much as 70% of the total vote, and never less than 59%. See 5 Record 636, 638, 641, 645, 647 (Pls. Exh. 112). And in District 12, those candidates won with 72% of the vote at the high end and 64% at the low. See id., at 637, 640, 643, 646, 650.

Another census, in 2010, necessitated yet another congressional map—(finally) the one at issue in this case. State Senator Robert Rucho and State Representative David Lewis, both Republicans, chaired the two committees jointly responsible for preparing the revamped plan. They hired Dr. Thomas Hofeller, a veteran political mapmaker, to assist them in redrawing district lines. Several hearings, drafts, and revisions later, both chambers of the State’s General Assembly adopted the scheme the three men proposed.

The new map (among other things) significantly altered both District 1 and District 12. The 2010 census had revealed District 1 to be substantially underpopulated: To comply with the Constitution’s one-person-one-vote principle, the State needed to place almost 100,000 new people within the district’s boundaries. See App. 2690; Evenwel v. Abbott, 578 U. S. ___, ___ (2016) (slip op., at 3) (explaining that “[s]tates must draw congressional districts with populations as close to perfect equality as possible”). Rucho, Lewis, and Hofeller chose to take most of those people from heavily black areas of Durham, requiring a finger-like extension of the district’s western line. See Appendix, infra. With that addition, District 1’s BVAP rose from 48.6% to 52.7%. See App. 312–313. District 12, for its part, had no need for significant total-population changes: It was overpopulated by fewer than 3,000 people out of over 730,000. See id., at 1150. Still, Rucho, Lewis, and Hofeller decided to reconfigure the district, further narrowing its already snakelike body while adding areas at either end—most relevantly here, in Guilford County. See Appendix, infra; App. 1164. Those changes appreciably shifted the racial composition of District 12: As the district gained some 35,000 African-Americans of voting age and lost some 50,000 whites of that age, its BVAP increased from 43.8% to 50.7%. See 2 Record 349 (Fourth Affidavit of Dan Frey, Exh. 5); id., at 416 (Exh. 11).

Registered voters in the two districts (David Harris and Christine Bowser, here called “the plaintiffs”) brought this suit against North Carolina officials (collectively, “the State” or “North Carolina”), complaining of impermissible racial gerrymanders. After a bench trial, a three-judge District Court held both districts unconstitutional. All the judges agreed that racial considerations predominated in the design of District 1. See Harris v. McCrory, 159 F. Supp. 3d 600, 611 (MDNC 2016). And in then applying strict scrutiny, all rejected the State’s argument that it had a “strong basis” for thinking that the VRA compelled such a race-based drawing of District 1’s lines. Id., at 623. As for District 12, a majority of the panel held that “race predominated” over all other factors, including partisanship. Id., at 622. And the court explained that the State had failed to put forward any reason, compelling or otherwise, for its attention to race in designing that district. See ibid. Judge Osteen dissented from the conclusion that race, rather than politics, drove District 12’s lines—yet still characterized the majority’s view as “[e]minently reasonable.” Id., at 640.

The State filed a notice of appeal, and we noted probable jurisdiction. McCrory v. Harris, 579 U. S. ___ (2016).

II

We address at the outset North Carolina’s contention that a victory it won in a very similar state-court lawsuit should dictate (or at least influence) our disposition of this case. As the State explains, the North Carolina NAACP and several other civil rights groups challenged Districts 1 and 12 in state court immediately after their enactment, charging that they were unlawful racial gerrymanders. See Brief for Appellants 19–20. By the time the plaintiffs before us filed this action, the state trial court, in Dickson v. Rucho, had rejected those claims—finding that in District 1 the VRA justified the General Assembly’s use of race and that in District 12 race was not a factor at all. See App. 1969. The North Carolina Supreme Court then affirmed that decision by a 4–3 vote, applying the state-court equivalent of clear error review. See Dickson v. Rucho, 368 N. C. 481, 500, 781 S. E. 2d 404, 419 (2015), modified on denial of reh’g, 368 N. C. 673, 789 S. E. 2d 436 (2016), cert. pending, No. 16–24. In this Court, North Carolina makes two related arguments based on the Dickson litigation: first, that the state trial court’s judgment should have barred this case altogether, under familiar principles of claim and issue preclusion; and second, that the state court’s conclusions should cause us to conduct a “searching review” of the decision below, rather than deferring (as usual) to its factual findings. Reply Brief 6.

The State’s preclusion theory rests on an assertion about how the plaintiffs in the two cases are affiliated. As the State acknowledges, one person’s lawsuit generally does not bar another’s, no matter how similar they are in substance. See Taylor v. Sturgell, 553 U. S. 880 –893 (2008) (noting the “deep-rooted historic tradition that everyone should have his own day in court”). But when plaintiffs in two cases have a special relationship, a judgment against one can indeed bind both. See id., at 893–895 (describing six categories of qualifying relationships). The State contends that Harris and Bowser, the plaintiffs here, are members of organizations that were plaintiffs in Dickson. And according to North Carolina, that connection prevents the pair from raising anew the questions that the state court previously resolved against those groups. See Brief for Appellants 20–21.

But North Carolina never satisfied the District Court that the alleged affiliation really existed. When the State argued that its preclusion theory entitled it to summary judgment, Harris and Bowser responded that they were not members of any of the organizations that had brought the Dickson suit. See 3 Record 1577–1582 (Defs. Motion for Summary Judgment); 4 Record 101–106 (Pls. Opposition to Motion for Summary Judgment). The parties’ dueling contentions turned on intricate issues about those groups’ membership policies (e.g., could Harris’s payment of dues to the national NAACP, or Bowser’s financial contribution to the Mecklenburg County NAACP, have made either a member of the state branch?). Because of those unresolved “factual disputes,” the District Court denied North Carolina’s motion for summary judgment. 4 Record 238 (July 29, 2014 Order). And nothing in the subsequent trial supported the State’s assertion about Harris’s and Bowser’s organizational ties: Indeed, the State chose not to present any further evidence relating to the membership issue. Based on the resulting record, the District Court summarily rejected the State’s claim that Harris and Bowser were something other than independent plaintiffs. See 159 F. Supp. 3d, at 609.

That conclusion defeats North Carolina’s attempt to argue for claim or issue preclusion here. We have no basis for assessing the factual assertions underlying the State’s argument any differently than the District Court did. Nothing in the State’s evidence clearly rebuts Harris’s and Bowser’s testimony that they never joined any of the Dickson groups. We need not decide whether the alleged memberships would have supported preclusion if they had been proved. It is enough that the District Court reason-ably thought they had not.

The State’s back-up argument about our standard of review also falls short. The rule that we review a trial court’s factual findings for clear error contains no exception for findings that diverge from those made in another court. See Fed. Rule Civ. Proc. 52(a)(6) (“Findings of fact . . . must not be set aside unless clearly erroneous”); see also Hernandez v. New York, 500 U. S. 352, 369 (1991) (plurality opinion) (applying the same standard to a state court’s findings). Whatever findings are under review receive the benefit of deference, without regard to whether a court in a separate suit has seen the matter differently. So here, we must ask not which court considering Districts 1 and 12 had the better view of the facts, but simply whether the court below’s view is clearly wrong. That does not mean the state court’s decision is wholly irrelevant: It is common sense that, all else equal, a finding is more likely to be plainly wrong if some judges disagree with it. Cf. Glossip v. Gross, 576 U. S. ___, ___ (2015) (slip op., at 17) (noting that we are even less likely to disturb a factual determination when “multiple trial courts have reached the same finding”). But the very premise of clear error review is that there are often “two permissible”—because two “plausible”—“views of the evidence.” Anderson, 470 U. S., at 574; see supra, at 4. Even assuming the state court’s findings capture one such view, the District Court’s assessment may yet represent another. And the permissibility of the District Court’s account is the only question before us.

III

With that out of the way, we turn to the merits of this case, beginning (appropriately enough) with District 1. As noted above, the court below found that race furnished the predominant rationale for that district’s redesign. See supra, at 6–7. And it held that the State’s interest in complying with the VRA could not justify that consideration of race. See supra, at 7. We uphold both conclusions.

A

Uncontested evidence in the record shows that the State’s mapmakers, in considering District 1, purposefully established a racial target: African-Americans should make up no less than a majority of the voting-age population. See 159 F. Supp. 3d, at 611–614. Senator Rucho and Representative Lewis were not coy in expressing that goal. They repeatedly told their colleagues that District 1 had to be majority-minority, so as to comply with the VRA. During a Senate debate, for example, Rucho explained that District 1 “must include a sufficient number of African-Americans” to make it “a majority black district.” App. 689–690. Similarly, Lewis informed the House and Senate redistricting committees that the district must have “a majority black voting age population.” Id., at 610. And that objective was communicated in no uncertain terms to the legislators’ consultant. Dr. Hofeller testified multiple times at trial that Rucho and Lewis instructed him “to draw [District 1] with a [BVAP] in excess of 50 percent.” 159 F. Supp. 3d, at 613; see, e.g., ibid. (“Once again, my instructions [were] that the district had to be drawn at above 50 percent”).

Hofeller followed those directions to the letter, such that the 50%-plus racial target “had a direct and significant impact” on District 1’s configuration. Alabama, 575 U. S., at __ (slip op., at 17). In particular, Hofeller moved the district’s borders to encompass the heavily black parts of Durham (and only those parts), thus taking in tens of thousands of additional African-American voters. That change and similar ones, made (in his words) to ensure that the district’s racial composition would “add[ ] up correctly,” deviated from the districting practices he other-wise would have followed. App. 2802. Hofeller candidly admitted that point: For example, he testified, he sometimes could not respect county or precinct lines as he wished because “the more important thing” was to create a majority-minority district. Id., at 2807; see id., at 2809. The result is a district with stark racial borders: Within the same counties, the portions that fall inside District 1 have black populations two to three times larger than the portions placed in neighboring districts. See Brief for United States as Amicus Curiae 19; cf. Alabama, 575 U. S., at ___–___ (slip op., at 17–18) (relying on similar evidence to find racial predominance).

Faced with this body of evidence—showing an announced racial target that subordinated other districting criteria and produced boundaries amplifying divisions between blacks and whites—the District Court did not clearly err in finding that race predominated in drawing District 1. Indeed, as all three judges recognized, the court could hardly have concluded anything but. See 159 F. Supp. 3d, at 611 (calling District 1 a “textbook example” of race-based districting).[3]

B

The more substantial question is whether District 1 can survive the strict scrutiny applied to racial gerrymanders. As noted earlier, we have long assumed that complying with the VRA is a compelling interest. See supra, at 2. And we have held that race-based districting is narrowly tailored to that objective if a State had “good reasons” for thinking that the Act demanded such steps. See supra, at 3. North Carolina argues that District 1 passes muster under that standard: The General Assembly (so says the State) had “good reasons to believe it needed to draw [District 1] as a majority-minority district to avoid Section 2 liability” for vote dilution. Brief for Appellants 52. We now turn to that defense.

This Court identified, in Thornburg v. Gingles, three threshold conditions for proving vote dilution under §2 of the VRA. See 478 U. S., at 50–51. First, a “minority group” must be “sufficiently large and geographically compact to constitute a majority” in some reasonably configured legislative district. Id., at 50. Second, the minority group must be “politically cohesive.” Id., at 51. And third, a district’s white majority must “vote[ ] sufficiently as a bloc” to usually “defeat the minority’s preferred candidate.” Ibid. Those three showings, we have explained, are needed to establish that “the minority [group] has the potential to elect a representative of its own choice” in a possible district, but that racially polarized voting prevents it from doing so in the district as actually drawn because it is “submerg[ed] in a larger white voting population.” Growe v. Emison, 507 U. S. 25, 40 (1993) . If a State has good reason to think that all the “Gingles preconditions” are met, then so too it has good reason to believe that §2 requires drawing a majority-minority district. See Bush v. Vera, 517 U. S. 952, 978 (1996) (plurality opinion). But if not, then not.

Here, electoral history provided no evidence that a §2 plaintiff could demonstrate the third Gingles prerequisite—effective white bloc-voting.[4] For most of the twenty years prior to the new plan’s adoption, African-Americans had made up less than a majority of District 1’s voters; the district’s BVAP usually hovered between 46% and 48%. See 159 F. Supp. 3d, at 606; App. 312. Yet throughout those two decades, as the District Court noted, District 1 was “an extraordinarily safe district for African-American preferred candidates.” 159 F. Supp. 3d, at 626. In the closest election during that period, African-Americans’ candidate of choice received 59% of the total vote; in other years, the share of the vote garnered by those candidates rose to as much as 70%. See supra, at 5. Those victories (indeed, landslides) occurred because the district’s white population did not “vote[ ] sufficiently as a bloc” to thwart black voters’ preference, Gingles, 478 U. S., at 51; rather, a meaningful number of white voters joined a politically cohesive black community to elect that group’s favored candidate. In the lingo of voting law, District 1 functioned, election year in and election year out, as a “cross-over” district, in which members of the majority help a “large enough” minority to elect its candidate of choice. Bartlett v. Strickland, 556 U. S. 1, 13 (2009) (plurality opinion). When voters act in that way, “[i]t is difficult to see how the majority-bloc-voting requirement could be met”—and hence how §2 liability could be established. Id., at 16. So experience gave the State no reason to think that the VRA required it to ramp up District 1’s BVAP.

The State counters that, in this context, past performance is no guarantee of future results. See Brief for Appellants 57–58; Reply Brief 19–20. Recall here that the State had to redraw its whole congressional map following the 2010 census. See supra, at 5. And in particular, the State had to add nearly 100,000 new people to District 1 to meet the one-person-one-vote standard. See supra, at 6. That meant about 13% of the voters in the new district would never have voted there before. See App. 2690; Reply Brief 20. So, North Carolina contends, the question facing the state mapmakers was not whether the then-existing District 1 violated §2. Rather, the question was whether the future District 1 would do so if drawn without regard to race. And that issue, the State claims, could not be resolved by “focusing myopically on past elections.” Id., at 19.

But that reasoning, taken alone, cannot justify North Carolina’s race-based redesign of District 1. True enough, a legislature undertaking a redistricting must assess whether the new districts it contemplates (not the old ones it sheds) conform to the VRA’s requirements. And true too, an inescapable influx of additional voters into a district may suggest the possibility that its former track record of compliance can continue only if the legislature intentionally adjusts its racial composition. Still, North Carolina too far downplays the significance of a longtime pattern of white crossover voting in the area that would form the core of the redrawn District 1. See Gingles, 478 U. S., at 57 (noting that longtime voting patterns are highly probative of racial polarization). And even more important, North Carolina can point to no meaningful legislative inquiry into what it now rightly identifies as the key issue: whether a new, enlarged District 1, created without a focus on race but however else the State would choose, could lead to §2 liability. The prospect of a significant population increase in a district only raises—it does not answer—the question whether §2 requires deliberate measures to augment the district’s BVAP. (Indeed, such population growth could cut in either direction, depending on who comes into the district.) To have a strong basis in evidence to conclude that §2 demands such race-based steps, the State must carefully evaluate whether a plaintiff could establish the Gingles preconditions—including effective white bloc-voting—in a new district created without those measures. We see nothing in the legislative record that fits that description.[5]

And that absence is no accident: Rucho and Lewis proceeded under a wholly different theory—arising not from Gingles but from Bartlett v. Strickland—of what §2 demanded in drawing District 1. Strickland involved a geographic area in which African-Americans could not form a majority of a reasonably compact district. See 556 U. S., at 8 (plurality opinion). The African-American community, however, was sizable enough to enable the formation of a crossover district, in which a substantial bloc of black voters, if receiving help from some white ones, could elect the candidates of their choice. See supra, at 14. A plurality of this Court, invoking the first Gingles precondition, held that §2 did not require creating that district: When a minority group is not sufficiently large to make up a majority in a reasonably shaped district, §2 simply does not apply. See 556 U. S., at 18–20. Over and over in the legislative record, Rucho and Lewis cited Strickland as mandating a 50%-plus BVAP in District 1. See App. 355–356, 363–364, 472–474, 609–610, 619, 1044. They apparently reasoned that if, as Strickland held, §2 does not require crossover districts (for groups insufficiently large under Gingles), then §2 also cannot be satisfied by crossover districts (for groups in fact meeting Gingles’ size condition). In effect, they concluded, whenever a legislature can draw a majority-minority district, it must do so—even if a crossover district would also allow the minority group to elect its favored candidates. See 1 Tr. 21–22 (counsel’s explanation that “the [S]tate interpreted” Strickland to say that, in order to protect African-Americans’ electoral strength and thus avoid §2 liability, the BVAP in District 1 “need[ed] to be above 50 percent”).

That idea, though, is at war with our §2 jurisprudence—Strickland included. Under the State’s view, the third Gingles condition is no condition at all, because even in the absence of effective white bloc-voting, a §2 claim could succeed in a district (like the old District 1) with an under-50% BVAP. But this Court has made clear that unless each of the three Gingles prerequisites is established, “there neither has been a wrong nor can be a remedy.” Growe, 507 U. S., at 41. And Strickland, far from supporting North Carolina’s view, underscored the necessity of demonstrating effective white bloc-voting to prevail in a §2 vote-dilution suit. The plurality explained that “[i]n areas with substantial crossover voting,” §2 plaintiffs would not “be able to establish the third Gingles precondition” and so “majority-minority districts would not be required.” 556 U. S., at 24; see also ibid. (noting that States can “defend against alleged §2 violations by pointing to crossover voting patterns and to effective crossover districts”). Thus, North Carolina’s belief that it was compelled to redraw District 1 (a successful crossover district) as a majority-minority district rested not on a “strong basis in evidence,” but instead on a pure error of law. Alabama, 575 U. S., at ___ (slip op., at 22).

In sum: Although States enjoy leeway to take race-based actions reasonably judged necessary under a proper interpretation of the VRA, that latitude cannot rescue District 1. We by no means “insist that a state legislature, when redistricting, determine precisely what percent minority population [§2 of the VRA] demands.” Ibid. But neither will we approve a racial gerrymander whose necessity is supported by no evidence and whose raison d’être is a legal mistake. Accordingly, we uphold the District Court’s conclusion that North Carolina’s use of race as the predominant factor in designing District 1 does not withstand strict scrutiny.

IV

We now look west to District 12, making its fifth(!) appearance before this Court. This time, the district’s legality turns, and turns solely, on which of two possible reasons predominantly explains its most recent reconfiguration. The plaintiffs contended at trial that the General Assembly chose voters for District 12, as for District 1, because of their race; more particularly, they urged that the Assembly intentionally increased District 12’s BVAP in the name of ensuring preclearance under the VRA’s §5. But North Carolina declined to mount any defense (similar to the one we have just considered for District 1) that §5’s requirements in fact justified race-based changes to District 12—perhaps because §5 could not reasonably be understood to have done so, see n. 10, infra. Instead, the State altogether denied that racial considerations accounted for (or, indeed, played the slightest role in) District 12’s redesign. According to the State’s version of events, Senator Rucho, Representative Lewis, and Dr. Hofeller moved voters in and out of the district as part of a “strictly” political gerrymander, without regard to race. 6 Record 1011. The mapmakers drew their lines, in other words, to “pack” District 12 with Democrats, not African-Americans. After hearing evidence supporting both parties’ accounts, the District Court accepted the plaintiffs’.[6]

Getting to the bottom of a dispute like this one poses special challenges for a trial court. In the more usual case alleging a racial gerrymander—where no one has raised a partisanship defense—the court can make real headway by exploring the challenged district’s conformity to traditional districting principles, such as compactness and respect for county lines. In Shaw II, for example, this Court emphasized the “highly irregular” shape of then-District 12 in concluding that race predominated in its design. 517 U. S., at 905 (internal quotation marks omitted). But such evidence loses much of its value when the State asserts partisanship as a defense, because a bizarre shape—as of the new District 12—can arise from a “political motivation” as well as a racial one. Cromartie I, 526 U. S., at 547, n. 3. And crucially, political and racial reasons are capable of yielding similar oddities in a district’s boundaries. That is because, of course, “racial identification is highly correlated with political affiliation.” Cromartie II, 532 U. S., at 243. As a result of those redistricting realities, a trial court has a formidable task: It must make “a sensitive inquiry” into all “circumstantial and direct evidence of intent” to assess whether the plaintiffs have managed to disentangle race from politics and prove that the former drove a district’s lines. Cromartie I, 526 U. S., at 546 (internal quotation marks omitted).[7]

Our job is different—and generally easier. As described earlier, we review a district court’s finding as to racial predominance only for clear error, except when the court made a legal mistake. See supra, at 3–4. Under that standard of review, we affirm the court’s finding so long as it is “plausible”; we reverse only when “left with the definite and firm conviction that a mistake has been committed.” Anderson, 470 U. S., at 573–574; see supra, at 4. And in deciding which side of that line to come down on, we give singular deference to a trial court’s judgments about the credibility of witnesses. See Fed. Rule Civ. Proc. 52(a)(6). That is proper, we have explained, because the various cues that “bear so heavily on the listener’s understanding of and belief in what is said” are lost on an appellate court later sifting through a paper record. Anderson, 470 U. S., at 575.[8]

In light of those principles, we uphold the District Court’s finding of racial predominance respecting District 12. The evidence offered at trial, including live witness testimony subject to credibility determinations, adequately supports the conclusion that race, not politics, accounted for the district’s reconfiguration. And no error of law infected that judgment: Contrary to North Carolina’s view, the District Court had no call to dismiss this challenge just because the plaintiffs did not proffer an alternative design for District 12 as circumstantial evidence of the legislature’s intent.

A

Begin with some facts and figures, showing how the redistricting of District 12 affected its racial composition. As explained above, District 12 (unlike District 1) was approximately the right size as it was: North Carolina did not—indeed, could not—much change its total population. See supra, at 6. But by further slimming the district and adding a couple of knobs to its snakelike body (including in Guilford County), the General Assembly incorporated tens of thousands of new voters and pushed out tens of thousands of old ones. And those changes followed racial lines: To be specific, the new District 12 had 35,000 more African-Americans of voting age and 50,000 fewer whites of that age. (The difference was made up of voters from other racial categories.) See ibid. Those voter exchanges produced a sizable jump in the district’s BVAP, from 43.8% to 50.7%. See ibid. The Assembly thus turned District 12 (as it did District 1, see supra, at 10–11) into a majority-minority district.

As the plaintiffs pointed out at trial, Rucho and Lewis had publicly stated that racial considerations lay behind District 12’s augmented BVAP. In a release issued along with their draft districting plan, the two legislators ascribed that change to the need to achieve preclearance of the plan under §5 of the VRA. See App. 358. At that time, §5 covered Guilford County and thus prohibited any “retrogression in the [electoral] position of racial minorities” there. Beer, 425 U. S., at 141; see 31 Fed. Reg. 5081 (1966). And part of Guilford County lay within District 12, which meant that the Department of Justice would closely scrutinize that district’s new lines. In light of those facts, Rucho and Lewis wrote: “Because of the presence of Guilford County in the Twelfth District, we have drawn our proposed Twelfth District at a [BVAP] level that is above the percentage of [BVAP] found in the current Twelfth District.” App. 358. According to the two legislators, that race-based “measure w[ould] ensure preclearance of the plan.” Ibid. Thus, the District Court found, Rucho’s and Lewis’s own account “evince[d] intentionality” as to District 12’s racial composition: Because of the VRA, they increased the number of African-Americans. 159 F. Supp. 3d, at 617.

Hofeller confirmed that intent in both deposition testimony and an expert report. Before the redistricting, Hofeller testified, some black residents of Guilford County fell within District 12 while others fell within neighboring District 13. The legislators, he continued, “decided to reunite the black community in Guilford County into the Twelfth.” App. 558; see id., at 530–531. Why? Hofeller responded, in language the District Court emphasized: “[I]n order to be cautious and draw a plan that would pass muster under the Voting Rights Act.” Id., at 558; see 159 F. Supp. 3d, at 619. Likewise, Hofeller’s expert report highlighted the role of the VRA in altering District 12’s lines. “[M]indful that Guilford County was covered” by §5, Hofeller explained, the legislature “determined that it was prudent to reunify [the county’s] African-American community” into District 12. App. 1103. That change caused the district’s compactness to decrease (in expert-speak, it “lowered the Reock Score”), but that was a sacrifice well worth making: It would “avoid the possibility of a [VRA] charge” that would “inhibit[ ] preclearance.” Ibid.

The State’s preclearance submission to the Justice Department indicated a similar determination to concentrate black voters in District 12. “One of the concerns of the Redistricting Chairs,” North Carolina there noted, had to do with the Justice Department’s years-old objection to “a failure by the State to create a second majority minority district” (that is, in addition to District 1). Id., at 478. The submission then went on to explain that after considering alternatives, the redistricters had designed a version of District 12 that would raise its BVAP to 50.7%. Thus, concluded the State, the new District 12 “increases[ ] the African-American community’s ability to elect their candidate of choice.” Id., at 479. In the District Court’s view, that passage once again indicated that making District 12 majority-minority was no “mere coincidence,” but a deliberate attempt to avoid perceived obstacles to preclearance. 159 F. Supp. 3d, at 617.[9]

And still there was more: Perhaps the most dramatic testimony in the trial came when Congressman Mel Watt (who had represented District 12 for some 20 years) recounted a conversation he had with Rucho in 2011 about the district’s future make-up. According to Watt, Rucho said that “his leadership had told him that he had to ramp the minority percentage in [District 12] up to over 50 percent to comply with the Voting Rights Law.” App. 2369; see id., at 2393. And further, that it would then be Rucho’s “job to go and convince the African-American community” that such a racial target “made sense” under the Act. Ibid.; see id., at 2369.[10] The District Court cred-ited Watt’s testimony about the conversation, citing his courtroom demeanor and “consistent recollection” under “probing cross-examination.” 159 F. Supp. 3d, at 617–618.[11] In the court’s view, Watt’s account was of a piece with all the other evidence—including the redistricters’ on-the-nose attainment of a 50% BVAP—indicating that the General Assembly, in the name of VRA compliance, deliberately redrew District 12 as a majority-minority district. See id., at 618.[12]

The State’s contrary story—that politics alone drove decisionmaking—came into the trial mostly through Hofeller’s testimony. Hofeller explained that Rucho and Lewis instructed him, first and foremost, to make the map as a whole “more favorable to Republican candidates.” App. 2682. One agreed-on stratagem in that effort was to pack the historically Democratic District 12 with even more Democratic voters, thus leaving surrounding districts more reliably Republican. See id., at 2682–2683, 2696–2697. To that end, Hofeller recounted, he drew District 12’s new boundaries based on political data—specifically, the voting behavior of precincts in the 2008 Presidential election between Barack Obama and John McCain. See id., at 2701–2702. Indeed, he claimed, he displayed only this data, and no racial data, on his computer screen while mapping the district. See id., at 2721. In part of his testimony, Hofeller further stated that the Obama-McCain election data explained (among other things) his incorporation of the black, but not the white, parts of Guilford County then located in District 13. See id., at 2824. Only after he drew a politics-based line between those adjacent areas, Hofeller testified, did he “check[ ]” the racial data and “f[ind] out” that the resulting configuration of District 12 “did not have a [§5] issue.” Id., at 2822.

The District Court, however, disbelieved Hofeller’s asserted indifference to the new district’s racial composition. The court recalled Hofeller’s contrary deposition testimony—his statement (repeated in only slightly different words in his expert report) that Rucho and Lewis “decided” to shift African-American voters into District 12 “in order to” ensure preclearance under §5. See 159 F. Supp. 3d, at 619–620; App. 558. And the court explained that even at trial, Hofeller had given testimony that undermined his “blame it on politics” claim. Right after asserting that Rucho and Lewis had told him “[not] to use race” in designing District 12, Hofeller added a qualification: “except perhaps with regard to Guilford County.” Id., at 2791; see id., at 2790. As the District Court understood, that is the kind of “exception” that goes pretty far toward swallowing the rule. District 12 saw a net increase of more than 25,000 black voters in Guilford County, relative to a net gain of fewer than 35,000 across the district: So the newly added parts of that county played a major role in pushing the district’s BVAP over 50%. See id., at 384, 500–502.[13] The District Court came away from Hofeller’s self-contradictory testimony unpersuaded that this decisive influx of black voters was an accident. Whether the racial make-up of the county was displayed on his computer screen or just fixed in his head, the court thought, Hofeller’s denial of race-based districting “r[ang] hollow.” 159 F. Supp. 3d, at 620, n. 8.

Finally, an expert report by Dr. Stephen Ansolabehere lent circumstantial support to the plaintiffs’ race-not-politics case. Ansolabehere looked at the six counties overlapping with District 12—essentially the region from which the mapmakers could have drawn the district’s population. The question he asked was: Who from those counties actually ended up in District 12? The answer he found was: Only 16% of the region’s white registered voters, but 64% of the black ones. See App. 321–322. Ansolabehere next controlled for party registration, but discovered that doing so made essentially no difference: For example, only 18% of the region’s white Democrats wound up in District 12, whereas 65% of the black Democrats did. See id., at 332. The upshot was that, regardless of party, a black voter was three to four times more likely than a white voter to cast his ballot within District 12’s borders. See ibid. Those stark disparities led Ansolabehere to conclude that “race, and not party,” was “the dominant factor” in District 12’s design. Id., at 337.[14] His report, as the District Court held, thus tended to confirm the plaintiffs’ direct evidence of racial predominance. See 159 F. Supp. 3d, at 620–621.

The District Court’s assessment that all this evidence proved racial predominance clears the bar of clear error review. The court emphasized that the districting plan’s own architects had repeatedly described the influx of African-Americans into District 12 as a §5 compliance measure, not a side-effect of political gerrymandering. And those contemporaneous descriptions comported with the court’s credibility determinations about the trial testimony—that Watt told the truth when he recounted Rucho’s resolve to hit a majority-BVAP target; and conversely that Hofeller skirted the truth (especially as to Guilford County) when he claimed to have followed only race-blind criteria in drawing district lines. We cannot disrespect such credibility judgments. See Anderson, 470 U. S., at 575 (A choice to believe “one of two or more witnesses, each of whom has told a coherent and facially plausible story that is not contradicted by extrinsic evidence,” can “virtually never be clear error”). And more generally, we will not take it upon ourselves to weigh the trial evidence as if we were the first to hear it. See id., at 573 (A “reviewing court oversteps” under Rule 52(a) “if it undertakes to duplicate the role of the lower court”). No doubt other interpretations of that evidence were permissible. Maybe we would have evaluated the testimony differently had we presided over the trial; or then again, maybe we would not have. Either way—and it is only this which matters—we are far from having a “definite and firm conviction” that the District Court made a mistake in concluding from the record before it that racial considerations predominated in District 12’s design.

B

The State mounts a final, legal rather than factual, attack on the District Court’s finding of racial predominance. When race and politics are competing explanations of a district’s lines, argues North Carolina, the party challenging the district must introduce a particular kind of circumstantial evidence: “an alternative [map] that achieves the legislature’s political objectives while improving racial balance.” Brief for Appellants 31 (emphasis deleted). That is true, the State says, irrespective of what other evidence is in the case—so even if the plaintiff offers powerful direct proof that the legislature adopted the map it did for racial reasons. See Tr. of Oral Arg. 8. Because the plaintiffs here (as all agree) did not present such a counter-map, North Carolina concludes that they cannot prevail. The dissent echoes that argument. See post, at 6–11.

We have no doubt that an alternative districting plan, of the kind North Carolina describes, can serve as key evidence in a race-versus-politics dispute. One, often highly persuasive way to disprove a State’s contention that politics drove a district’s lines is to show that the legislature had the capacity to accomplish all its partisan goals without moving so many members of a minority group into the district. If you were really sorting by political behavior instead of skin color (so the argument goes) you would have done—or, at least, could just as well have done—this. Such would-have, could-have, and (to round out the set) should-have arguments are a familiar means of undermining a claim that an action was based on a permissible, rather than a prohibited, ground. See, e.g., Miller-El v. Dretke, 545 U. S. 231, 249 (2005) (“If that were the [real] explanation for striking [juror] Warren[,] the prosecutors should have struck [juror] Jenkins” too).

But they are hardly the only means. Suppose that the plaintiff in a dispute like this one introduced scores of leaked emails from state officials instructing their mapmaker to pack as many black voters as possible into a district, or telling him to make sure its BVAP hit 75%. Based on such evidence, a court could find that racial rather than political factors predominated in a district’s design, with or without an alternative map. And so too in cases lacking that kind of smoking gun, as long as the evidence offered satisfies the plaintiff’s burden of proof. In Bush v. Vera, for example, this Court upheld a finding of racial predominance based on “substantial direct evidence of the legislature’s racial motivations”—including credible testimony from political figures and statements made in a §5 preclearance submission—plus circumstantial evidence that redistricters had access to racial, but not political, data at the “block-by-block level” needed to explain their “intricate” designs. See 517 U. S., at 960–963 (plurality opinion). Not a single Member of the Court thought that the absence of a counter-map made any difference. Similarly, it does not matter in this case, where the plaintiffs’ introduction of mostly direct and some circumstantial evidence—documents issued in the redistricting process, testimony of government officials, expert analysis of demographic patterns—gave the District Court a sufficient basis, sans any map, to resolve the race-or-politics question.

A plaintiff’s task, in other words, is simply to persuade the trial court—without any special evidentiary prerequisite—that race (not politics) was the “predominant consideration in deciding to place a significant number of voters within or without a particular district.” Alabama, 575 U. S., at ___ (slip op., at 4) (internal quotation marks omitted); cf. Bethune-Hill, 580 U. S., at ___, ___ (slip op., at 8, 10) (rejecting a similar effort to elevate one form of “persuasive circumstantial evidence” in a dispute respecting racial predominance to a “mandatory precondition” or “threshold requirement” of proof). That burden of proof, we have often held, is “demanding.” E.g., Cromartie II, 532 U. S., at 241. And because that is so, a plaintiff will sometimes need an alternative map, as a practical matter, to make his case. But in no area of our equal protection law have we forced plaintiffs to submit one particular form of proof to prevail. See Arlington Heights v. Metropolitan Housing Development Corp., 429 U. S. 252 –268 (1977) (offering a varied and non-exhaustive list of “subjects of proper inquiry in determining whether racially discriminatory intent existed”). Nor would it make sense to do so here. The Equal Protection Clause prohibits the unjustified drawing of district lines based on race. An alternative map is merely an evidentiary tool to show that such a substantive violation has occurred; neither its presence nor its absence can itself resolve a racial gerrymandering claim.[15]

North Carolina insists, however, that we have already said to the contrary—more particularly, that our decision in Cromartie II imposed a non-negotiable “alternative-map requirement.” Brief for Appellants 31. As the State observes, Cromartie II reversed as clearly erroneous a trial court’s finding that race, rather than politics, predominated in the assignment of voters to an earlier incarnation of District 12. See 532 U. S., at 241; supra, at 5. And as the State emphasizes, a part of our opinion faulted the Cromartie plaintiffs for failing to offer a convincing account of how the legislature could have accomplished its political goals other than through the map it chose. See 532 U. S., at 257–258. We there stated:

“In a case such as this one where majority-minority districts . . . are at issue and where racial identification correlates highly with political affiliation, the party attacking the legislatively drawn boundaries must show at the least that the legislature could have achieved its legitimate political objectives in alternative ways that are comparably consistent with traditional districting principles. That party must also show that those districting alternatives would have brought about significantly greater racial balance.” Id., at 258.

According to North Carolina, that passage alone settles this case, because it makes an alternative map “essential” to a finding that District 12 (a majority-minority district in which race and partisanship are correlated) was a racial gerrymander. Reply Brief 11. Once again, the dissent says the same. See post, at 7.

But the reasoning of Cromartie II belies that reading. The Court’s opinion nowhere attempts to explicate or justify the categorical rule that the State claims to find there. (Certainly the dissent’s current defense of that rule, see post, at 8–11, was nowhere in evidence.) And given the strangeness of that rule—which would treat a mere form of evidence as the very substance of a constitutional claim, see supra, at 29–31—we cannot think that the Court adopted it without any explanation. Still more, the entire thrust of the Cromartie II opinion runs counter to an inflexible counter-map requirement. If the Court had adopted that rule, it would have had no need to weigh each piece of evidence in the case and determine whether, taken together, they were “adequate” to show “the predominance of race in the legislature’s line-drawing process.” 532 U. S., at 243–244. But that is exactly what Cromartie II did, over a span of 20 pages and in exhaustive detail. Item by item, the Court discussed and dismantled the supposed proof, both direct and circumstantial, of race-based redistricting. All that careful analysis would have been superfluous—that dogged effort wasted—if the Court viewed the absence or inadequacy of a single form of evidence as necessarily dooming a gerrymandering claim.

Rightly understood, the passage from Cromartie II had a different and narrower point, arising from and reflecting the evidence offered in that case. The direct evidence of a racial gerrymander, we thought, was extremely weak: We said of one piece that it “says little or nothing about whether race played a predominant role” in drawing district lines; we said of another that it “is less persuasive than the kinds of direct evidence we have found significant in other redistricting cases.” Id., at 253–254 (emphasis deleted). Nor did the report of the plaintiffs’ expert impress us overmuch: In our view, it “offer[ed] little insight into the legislature’s true motive.” Id., at 248. That left a set of arguments of the would-have-could-have variety. For example, the plaintiffs offered several maps purporting to “show how the legislature might have swapped” some mostly black and mostly white precincts to obtain greater racial balance “without harming [the legislature’s] political objective.” Id., at 255 (internal quotation marks omitted). But the Court determined that none of those proposed exchanges would have worked as advertised—essentially, that the plaintiffs’ “you could have redistricted differently” arguments failed on their own terms. See id., at 254–257. Hence emerged the demand quoted above, for maps that would actually show what the plaintiffs’ had not. In a case like Cromartie II—that is, one in which the plaintiffs had meager direct evidence of a racial gerrymander and needed to rely on evidence of forgone alternatives—only maps of that kind could carry the day. Id., at 258.

But this case is most unlike Cromartie II, even though it involves the same electoral district some twenty years on. This case turned not on the possibility of creating more optimally constructed districts, but on direct evidence of the General Assembly’s intent in creating the actual District 12, including many hours of trial testimony subject to credibility determinations. That evidence, the District Court plausibly found, itself satisfied the plaintiffs’ burden of debunking North Carolina’s “it was really politics” defense; there was no need for an alternative map to do the same job. And we pay our precedents no respect when we extend them far beyond the circumstances for which they were designed.

V

Applying a clear error standard, we uphold the District Court’s conclusions that racial considerations predominated in designing both District 1 and District 12. For District 12, that is all we must do, because North Carolina has made no attempt to justify race-based districting there. For District 1, we further uphold the District Court’s decision that §2 of the VRA gave North Carolina no good reason to reshuffle voters because of their race. We accordingly affirm the judgment of the District Court.

It is so ordered.

Justice Gorsuch took no part in the consideration or decision of this case.

Congressional District 1 (Enacted 2011)

Congressional District 12 (Enacted 2011)

SUPREME COURT OF THE UNITED STATES

_________________

No. 15–1262

_________________

ROY COOPER, GOVERNOR OF NORTH CAROLINA, et al., APPELLANTS v. DAVID HARRIS, et al.

on appeal from the united states district court for the middle district of north carolina

[May 22, 2017]

Justice Thomas, concurring.

I join the opinion of the Court because it correctly applies our precedents under the Constitution and the Voting Rights Act of 1965 (VRA), 52 U. S. C. §10301 et seq. I write briefly to explain the additional grounds on which I would affirm the three-judge District Court and to note my agreement, in particular, with the Court’s clear-error analysis.

As to District 1, I think North Carolina’s concession that it created the district as a majority-black district is by itself sufficient to trigger strict scrutiny. See Brief for Appellants 44; see also, e.g., Bethune-Hill v. Virginia State Bd. of Elections, 580 U. S. ___, ___–___ (2017) (slip op., at 1–2) (Thomas, J., concurring in judgment in part and dissenting in part). I also think that North Carolina cannot satisfy strict scrutiny based on its efforts to comply with §2 of the VRA. See ante, at 12. In my view, §2 does not apply to redistricting and therefore cannot justify a racial gerrymander. See Holder v. Hall, 512 U. S. 874 –923 (1994) (Thomas, J., concurring in judgment).

As to District 12, I agree with the Court that the District Court did not clearly err when it determined that race was North Carolina’s predominant motive in drawing the district. See ante, at 21. This is the same conclusion I reached when we last reviewed District 12. Easley v. Cromartie, 532 U. S. 234, 267 (2001) (Cromartie II) (dissenting opinion). The Court reached the contrary conclusion in Cromartie II only by misapplying our deferential standard for reviewing factual findings. See id., at 259–262. Today’s decision does not repeat Cromartie II’s error, and indeed it confines that case to its particular facts. It thus represents a welcome course correction to this Court’s application of the clear-error standard.

SUPREME COURT OF THE UNITED STATES

_________________

No. 15–1262

_________________

ROY COOPER, GOVERNOR OF NORTH CAROLINA, et al., APPELLANTS v. DAVID HARRIS, et al.

on appeal from the united states district court for the middle district of north carolina

[May 22, 2017]

Justice Alito, with whom The Chief Justice and Justice Kennedy join, concurring in the judgment in part and dissenting in part.

A precedent of this Court should not be treated like a disposable household item—say, a paper plate or napkin—to be used once and then tossed in the trash. But that is what the Court does today in its decision regarding North Carolina’s 12th Congressional District: The Court junks a rule adopted in a prior, remarkably similar challenge to this very same congressional district.

In Easley v. Cromartie, 532 U. S. 234 (2001) (Cromartie II ), the Court considered the constitutionality of the version of District 12 that was adopted in 1997. Id., at 238. That district had the same basic shape as the district now before us, and the challengers argued that the legislature’s predominant reason for adopting this configuration was race. Ibid. The State responded that its motive was not race but politics. Id., at 241. Its objective, the State insisted, was to create a district in which the Democratic candidate would win. See ibid.; Brief for State Appellants in Easley v. Cromartie, O. T. 2000, Nos. 99–1864, 99–1865, p. 25. Rejecting that explanation, a three-judge court found that the legislature’s predominant motive was racial, specifically to pack African-Americans into District 12. See Cromartie v. Hunt, 133 F. Supp. 2d 407, 420 (EDNC 2000). But this Court held that this finding of fact was clearly erroneous. Cromartie II, 532 U. S., at 256.

A critical factor in our analysis was the failure of those challenging the district to come forward with an alternative redistricting map that served the legislature’s political objective as well as the challenged version without producing the same racial effects. Noting that race and party affiliation in North Carolina were “highly corre-lated,” id., at 243, we laid down this rule:

“In a case such as this one . . . , the party attacking the legislatively drawn boundaries must show at the least that the legislature could have achieved its legitimate political objectives in alternative ways that are comparably consistent with traditional districting principles. That party must also show that those districting alternatives would have brought about significantly greater racial balance. Appellees failed to make any such showing here.” Id., at 258.

Now, District 12 is back before us. After the 2010 census, the North Carolina Legislature, with the Republicans in the majority, drew the present version of District 12. The challengers contend that this version violates equal protection because the predominant motive of the legislature was racial: to pack the district with African-American voters. The legislature responds that its objective was political: to pack the district with Democrats and thus to increase the chances of Republican candidates in neighboring districts.

You might think that the Cromartie II rule would be equally applicable in this case, which does not differ in any relevant particular, but the majority executes a stunning about-face. Now, the challengers’ failure to produce an alternative map that meets the Cromartie II test is inconsequential. It simply “does not matter.” Ante, at 30.

This is not the treatment of precedent that state legislatures have the right to expect from this Court. The failure to produce an alternative map doomed the challengers in Cromartie II, and the same should be true now. Partisan gerrymandering is always unsavory, but that is not the issue here. The issue is whether District 12 was drawn predominantly because of race. The record shows that it was not.[1]

I

Under the Constitution, state legislatures have “the initial power to draw districts for federal elections.” Vieth v. Jubelirer, 541 U. S. 267, 275 (2004) (plurality opinion).[2] This power, of course, must be exercised in conformity with the Fourteenth Amendment’s Equal Protection Clause. And because the Equal Protection Clause’s “central mandate is racial neutrality in governmental decisionmaking,” Miller v. Johnson, 515 U. S. 900, 904 (1995) , “effort[s] to separate voters into different districts on the basis of race” must satisfy the rigors of strict scrutiny. Shaw v. Reno, 509 U. S. 630, 649, 653 (1993) (Shaw I ).

We have stressed, however, that courts are obligated to “exercise extraordinary caution in adjudicating claims that a State has drawn district lines on the basis of race.” Miller, 515 U. S., at 916. “Federal-court review of districting legislation represents a serious intrusion on the most vital of local functions,” and “the good faith of a state legislature must be presumed.” Id., at 915. A legislature will “almost always be aware of racial demographics” during redistricting, but evidence of such awareness does not show that the legislature violated equal protection. Id., at 916. Instead, the Court has held, “[r]ace must not simply have been a motivation for the drawing of a majority-minority district, but the predominant factor motivating the legislature’s districting decision.” Cromartie II, 532 U. S., at 241 (citation and internal quotation marks omitted; emphasis in original).

This evidentiary burden “is a demanding one.” Ibid. (internal quotation marks omitted). Thus, although “[t]he legislature’s motivation is . . . a factual question,” Hunt v. Cromartie, 526 U. S. 541, 549 (1999) (Cromartie I ), an appellate court conducting clear-error review must always keep in mind the heavy evidentiary obligation borne by those challenging a districting plan. See Cromartie II, supra, at 241, 257. Recognizing “the intrusive potential of judicial intervention into the legislative realm,” Miller, supra, at 916, we have warned that courts must be very cautious about imputing a racial motive to a State’s redistricting plan.

II

That caution “is especially appropriate . . . where the State has articulated a legitimate political explanation for its districting decision, and the voting population is one in which race and political affiliation are highly correlated.” Cromartie II, 532 U. S., at 242. We have repeatedly acknowledged the problem of distinguishing between racial and political motivations in the redistricting context. See id., at 242, 257–258; Cromartie I, supra, at 551–552; Bush v. Vera, 517 U. S. 952 –968 (1996) (plurality opinion).

The problem arises from the confluence of two factors. The first is the status under the Constitution of partisan gerrymandering. As we have acknowledged, “[p]olitics and political considerations are inseparable from districting and apportionment,” Gaffney v. Cummings, 412 U. S. 735, 753 (1973) , and it is well known that state legislative majorities very often attempt to gain an electoral advantage through that process. See Davis v. Bandemer, 478 U. S. 109, 129 (1986) . Partisan gerrymandering dates back to the founding, see Vieth, supra, at 274–276 (plural-ity opinion), and while some might find it distasteful, “[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 and even if the State were conscious of that fact.” Cromartie I, supra, at 551 (emphasis in original); Vera, supra, at 964 (plurality opinion).

The second factor is that “racial identification is highly correlated with political affiliation” in many jurisdictions. Cromartie II, 532 U. S., at 243 (describing correlation in North Carolina). This phenomenon makes it difficult to distinguish between political and race-based decisionmaking. If around 90% of African-American voters cast their ballots for the Democratic candidate, as they have in recent elections,[3] a plan that packs Democratic voters will look very much like a plans that packs African-American voters. “[A] legislature may, by placing reliable Demo-cratic precincts within a district without regard to race, end up with a district containing more heavily African–American precincts, but the reasons would be political rather than racial.” Id., at 245.

A

We addressed this knotty problem in Cromartie II, which, as noted, came to us after the District Court had held a trial and found as a fact that the legislature’s predominant reason for drawing District 12 was race, not politics. Id., at 239–241. Our review for clear error in that case did not exhibit the same diffidence as today’s decision. We carefully examined each piece of direct and circumstantial evidence on which the District Court had relied and conceded that this evidence provided support for the court’s finding. Id., at 257. Then, at the end of our opinion, we stated:

“We can put the matter more generally as follows: In a case such as this one where majority-minority districts (or the approximate equivalent) are at issue and where racial identification correlates highly with political affiliation, the party attacking the legislatively drawn boundaries must show at the least that the legislature could have achieved its legitimate political objectives in alternative ways that are comparably consistent with traditional districting principles. That party must also show that those districting alternatives would have brought about significantly greater racial balance.” Id., at 258.

Because the plaintiffs had “failed to make any such showing,” we held that the District Court had clearly erred in finding that race predominated in drawing District 12. Ibid.

Cromartie II plainly meant to establish a rule for use in a broad class of cases and not a rule to be employed one time only. We stated that we were “put[ting] the matter more generally” and were describing what must be shown in cases “where majority-minority districts (or the approximate equivalent) are at issue and where racial identification correlates highly with political affiliation.” Ibid. We identified who would carry the burden of the new rule (“the party attacking the legislatively drawn boundaries”) and what that party must show (that “the legislature could have achieved its legitimate political objectives in alternative ways that are comparably consistent with traditional districting principles” while achieving “significantly greater racial balance”). Ibid. And we reversed the finding of racial predominance due to the plaintiffs’ failure to carry the burden established by this evidentiary rule. Ibid.

Here, too, the plaintiffs failed to carry that burden. In this case, as in Cromartie II, the plaintiffs allege a racial gerrymander, and the State’s defense is that political motives explain District 12’s boundaries. In such a case, Cromartie II instructed, plaintiffs must submit an alternative redistricting map demonstrating that the legislature could have achieved its political goals without the racial effects giving rise to the racial gerrymandering allegation. But in spite of this instruction, plaintiffs in this case failed to submit such a map.[4] See Brief for Appellees 31–36. Based on what we said in Cromartie II about the same type of claim involving the same congressional district, reversal should be a foregone conclusion. It turns out, however, that the Cromartie II rule was good for one use only. Even in a case involving the very same district, it is tossed aside.

B

The alternative-map requirement deserves better. It is a logical response to the difficult problem of distinguishing between racial and political motivations when race and political party preference closely correlate.

This is a problem with serious institutional and federalism implications. When a federal court says that race was a legislature’s predominant purpose in drawing a district, it accuses the legislature of “offensive and demeaning” conduct. Miller, 515 U. S., at 912. Indeed, we have said that racial gerrymanders “bea[r] an uncomfortable re-semblance to political apartheid.” Shaw I, 509 U. S., at 647. That is a grave accusation to level against a state legislature.

In addition, “[f]ederal-court review of districting legislation represents a serious intrusion on the most vital of local functions” because “[i]t is well settled that reapportionment is primarily the duty and responsibility of the State.” Miller, supra, at 915 (internal quotation marks omitted); see also Cromartie II, 532 U. S., at 242. When a federal court finds that race predominated in the redistricting process, it inserts itself into that process. That is appropriate—indeed, constitutionally required—if the legislature truly did draw district boundaries on the basis of race. But if a court mistakes a political gerrymander for a racial gerrymander, it illegitimately invades a traditional domain of state authority, usurping the role of a State’s elected representatives. This does violence to both the proper role of the Judiciary and the powers reserved to the States under the Constitution.

There is a final, often-unstated danger where race and politics correlate: that the federal courts will be transformed into weapons of political warfare. Unless courts “exercise extraordinary caution” in distinguishing race-based redistricting from politics-based redistricting, Miller, supra, at 916, they will invite the losers in the redistricting process to seek to obtain in court what they could not achieve in the political arena. If the majority party draws districts to favor itself, the minority party can deny the majority its political victory by prevailing on a racial gerrymandering claim. Even if the minority party loses in court, it can exact a heavy price by using the judicial process to engage in political trench warfare for years on end.

Although I do not imply that this is what occurred here, this case does reflect what litigation of this sort can look like. This is the fifth time that North Carolina’s 12th Congressional District has come before this Court since 1993, and we have almost reached a new redistricting cycle without any certainty as to the constitutionality of North Carolina’s current redistricting map. Given these dangers, Cromartie II was justified in crafting an evidentiary rule to prevent false positives.[5]

C

The majority nevertheless absolves the challengers of their failure to submit an alternative map. It argues that an alternative map cannot be “the only means” of proving racial predominance, and it concludes from this that an alternative map “does not matter in this case.” Ante, at 29–30 (emphasis in original). But even if there are cases in which a plaintiff could prove a racial gerrymandering claim without an alternative map, they would be exceptional ones in which the evidence of racial predominance is overwhelming. This most definitely is not one of those cases, see Part III–C, infra, and the plaintiffs’ failure to produce an alternative map mandates reversal. Moreover, even in an exceptional case, the absence of such a map would still be strong evidence that a district’s boundaries were determined by politics rather than race.[6] The absence of a map would “matter.” Cf. ante, at 30.

The majority questions the legitimacy of the alternative-map requirement, ante, at 29–31, and n. 15, but the rule is a sound one. It rests on familiar principles regarding the allocation of the burdens of production and persuasion and the assessment of evidence. First, in accordance with the general rule in civil cases, plaintiffs in a case like this bear the burden of proving that the legislature’s motive was unconstitutional. Second, what must be shown is not simply that race played a part in the districting process but that it played the predominant role. Third, a party challenging a districting plan must overcome the strong presumption that the plan was drawn for constitutionally permissible reasons. Miller, supra, at 915. Fourth, when those responsible for adopting a challenged plan contend that the plan was devised for partisan political ends, they are making an admission that may not sit well with voters, so the explanation should not be lightly dismissed. Cf. Fed. Rule Evid. 804(b)(3). And finally, the Cromartie II rule takes into account the difficulty of proving a negative.

For challengers like those in the present case, producing a map that meets the Cromartie II test should not be hard if the predominant reason for a challenged plan really was race and not politics. Plaintiffs mounting a challenge to a districting plan are almost always sophisticated litigants who have the assistance of experts, and that is certainly true in the present case. Today, an expert with a computer can easily churn out redistricting maps that control for any number of specified criteria, including prior voting patterns and political party registration. Therefore, if it is indeed possible to find a map that meets the Cromartie II test, it should not be too hard for the challengers to do so. The State, on the other hand, cannot prove that no map meeting the Cromartie II test can be drawn. Even if a State submits, say, 100 alternative maps that fail the test, that would not prove that no such map could pass it. The relative ease with which the opposing parties can gather evidence is a familiar consideration in allocating the burden of production. See 1 C. Mueller & L. Kirkpatrick, Federal Evidence §63, p. 316 (2d ed. 1994); 21 C. Wright & K. Graham, Federal Practice and Procedure §5122, pp. 556–557 (1977).

III

Even if we set aside the challengers’ failure to submit an alternative map, the District Court’s finding that race predominated in the drawing of District 12 is clearly erroneous. The State offered strong and coherent evidence that politics, not race, was the legislature’s predominant aim, and the evidence supporting the District Court’s contrary finding is weak and manifestly inadequate in light of the high evidentiary standard that our cases require challengers to meet in order to prove racial predominance.[7]

My analysis will proceed in three steps. First, I will discuss what the legislature’s mapmaker did and why this approach is entirely consistent with his stated political objectives. Then, I will explain why this approach inevitably had the racial effect to which the challengers object. Finally, I will address the evidence of racial predominance on which the majority relies and show why it is inadequate to sustain the District Court’s judgment.

A

In order to understand the mapmaker’s approach, the first element to be kept in mind is that the basic shape of District 12 was legitimately taken as a given. When a new census requires redistricting, it is a common practice to start with the plan used in the prior map and to change the boundaries of the prior districts only as needed to comply with the one-person, one-vote mandate and to achieve other desired ends. This approach honors settled expectations and, if the prior plan survived legal challenge, minimizes the risk that the new plan will be overturned. And that is the approach taken by the veteran mapmaker in the present case, Dr. Thomas Hofeller. App. 523 (“the normal starting point is always from the existing districts”).

Dr. Hofeller began with the prior version of District 12 even though that version had a strange, serpentine shape. Cromartie I, 526 U. S., at 544; App. 1163. That design has a long history. It was first adopted in 1992, and subsequent redistricting plans have built on the 1992 plan. Ibid. In Cromartie II, we sustained the constitutionality of the 1997 version of District 12, which featured the same basic shape. See 532 U. S., at 258. And retention of this same basic shape is not challenged in this case.[8]

Using the prior design as his starting point, Dr. Hofeller assumed that District 12 would remain a “strong Democratic distric[t].” App. 521. He stated that he drew “the [overall redistricting] plan to . . . have an increased number of competitive districts for GOP candidates,” id., at 520, and that he therefore moved more Democratic voters into District 12 in order to “increase Republican opportunities in the surrounding districts,” id., at 1606.

Under the map now before us, District 12 is bordered by four districts.[9] Running counterclockwise, they are: District 5 to the northwest; District 9 to the southwest; District 8 to the southeast; and District 6 to the northeast. See Appendix, ante. According to Dr. Hofeller, the aim was to make these four districts—considered as a whole—more secure for Republicans. App. 1606, 2696.

To do this, Dr. Hofeller set out in search of pockets of Democratic voters that could be moved into District 12 from areas adjoining or very close to District 12’s prior boundaries. Of the six counties through which District 12 passes, the three most heavily Democratic (and also the most populous) are Forsyth, Guilford, and Mecklenburg, which contain the major population centers of Winston-Salem, Greensboro, and Charlotte, respectively. See 7 Record 480–482; App. 1141. As a measure of voting preferences, Dr. Hofeller used the results of the then-most recent Presidential election, i.e., the election of 2008. Id., at 1149, 2697, 2721–2722. In that election, these three counties voted strongly for the Democratic candidate, then-Senator Barack Obama, while the other three counties, Cabarrus, Davidson, and Rowan, all voted for the Republican candidate, Senator John McCain. See 4 Record 1341–1342.

Two of the three Democratic counties, Forsyth and Guilford, are located at the northern end of District 12, while the other Democratic county, Mecklenburg, is on the southern end. See Appendix, ante. The middle of the district (often called the “corridor”) passes through the three more Republican-friendly counties—Cabarrus, Davidson, and Rowan. Ibid. Thus, if a mapmaker sat down to increase the proportion of Democrats in District 12 and to reduce the proportion in neighboring districts, the most obvious way to do that was to pull additional Democrats into the district from the north and south (the most populous and heavily Democratic counties) while shifting Republican voters out of the corridor.

That, in essence, is what Dr. Hofeller did—as the majority acknowledges. Ante, at 6 (Dr. Hofeller “narrow[ed District 12’s] already snakelike body while adding areas at either end”); App. 1150 (Table 1), 1163. Dr. Hofeller testified that he sought to shift parts of Mecklenburg County out of Districts 8 and 9 (in order to reduce the percentage of Democrats in these two districts) and that this required him to increase the coverage of Mecklenburg County in District 12. Id., at 1142–1143, 1607, 2753.

Dr. Hofeller testified that he also had political plans for the current map’s District 6, which differed substantially from the version in the prior map. Dr. Hofeller wanted to improve the Republicans’ prospects in this new district by minimizing its coverage of Guilford County’s Democratic population. Id., at 1143, 1607, 2693, 2697, 2752. That also meant increasing the population of Guilford County Democrats in District 12. Id., at 1143, 1607, 2697.

This influx of Democratic voters from the two most populous counties in District 12 required shedding voters elsewhere in order to comply with this Court’s mandate of one-person, one-vote, see Kirkpatrick v. Preisler, 394 U. S. 526 –531 (1969),[10] and the population removed had to be added to a bordering district. App. 523. Parts of Davidson and Rowan Counties were therefore shifted to District 5, id., at 1143, 1150 (Table 1), but Dr. Hofeller testified that this would not have been sufficient to satisfy the one-person, one-vote standard, so he also had to move voters from heavily Democratic Forsyth County into District 5, id., at 1143, 2697, 2752–2753. Doing so did not undermine his political objective, he explained, because District 5 “was stronger [for Republicans] to begin with and could take those [Forsyth] Democratic precincts” without endangering Republican chances in the district. Id., at 2753; see also id., at 2697. The end result was that, under the new map now at issue, the three major counties in the north and south constitute a larger percentage of District 12’s total population, while the corridor lost population. See id., at 1150 (Table 1), 2149 (Finding 187).

A comparison of the 2008 Presidential election vote under the old and new versions of the districts shows the effect of Dr. Hofeller’s map. District 8 (which, of the four districts bordering District 12 under the 2011 map, was the most Democratic district) saw a drop of almost 11% in the Democratic vote under the new map. See 2 Record 354, 421. District 9 saw a drop in the percentage of registered Democrats, id., at 350, 417, although the vote percentage for the Democratic Presidential candidate remained essentially the same (increasing by 0.39%). Id., at 354, 421. District 5, which was heavily Republican under the prior map and was redrawn to absorb Democrats from Forsyth County, saw about a 7-point swing in favor of the Democratic candidate, but it remained a strong Republican district. Ibid. New District 6 is less susceptible to comparison because its boundaries are completely different from the district bearing that number under the old plan, but the new District 6 was solidly Republican, with a Republican Presidential vote percentage of nearly 56%. Ibid. As stated by the state court that considered and rejected the same constitutional challenge now before us:

“By increasing the number of Democratic voters in the 2011 Twelfth Congressional District located in Mecklenburg and Guilford Counties, the 2011 Congressional Plan created other districts that were more competitive for Republican candidates as compared to the 2001 versions of these districts . . . .” App. 2150 (Finding 191).

The results of subsequent congressional elections show that Dr. Hofeller’s plan achieved its goal. In 2010, prior to the adoption of the current plan, Democrats won 7 of the 13 districts, including District 8.[11] But by 2016, Republicans controlled 10 of the 13 districts, including District 8, and all the Republican candidates for the House of Representatives won their races with at least 56% of the vote.[12] In accordance with the map’s design, the only Democratic seats remaining after 2016 were in Districts 1, 4, and 12. Id., at 521.

In sum, there is strong evidence in the record to sup-port Dr. Hofeller’s testimony that the changes made to the 2001 map were designed to maximize Republican opportunities.

B

I now turn to the connection between the mapmaker’s strategy and the effect on the percentage of African-Americans in District 12.

As we recognized in Cromartie II, political party preference and race are highly correlated in North Carolina generally and in the area of Congressional District 12 in particular. App. 2022 (state trial court finding that “racial identification correlates highly with political affiliation” in North Carolina). The challenger’s expert, Dr. Stephen Ansolabehere, corroborated this important point. Dr. Ansolabehere calculated the statewide correlation between race and voting in 2008[13] and found a correlation of 0.8, which is “very high.” Id., at 342, 352 (Table 1). See also J. Levin, J. Fox, & D. Forde, Elementary Statistics in Social Research 370 (12th ed. 2014); R. Witte & J. Witte, Statistics 138 (10th ed. 2015).

In the area of District 12, the correlation is even higher. There, Dr. Ansolabehere found that the correlation “approach[ed] 1,” App. 342, that is, almost complete overlap. These black Democrats also constitute a supermajority of Democrats in the area covered by the district. Under the 2001 version of District 12—which was drawn by Democrats and was never challenged as a racial gerrymander—black registered voters constituted 71.44% of Democrats in the district. 2 Record 350; see also App. 2145 (Finding 173).[14] What this means is that a mapmaker seeking to pull Democrats into District 12 would unavoidably pull in a very large percentage of African-Americans.

The distribution of Democratic voters magnified this effect. Dr. Hofeller’s plan required the identification of areas of Democratic strength that were near District 12’s prior boundaries. Dr. Hofeller prepared maps showing the distribution of Democratic voters by precinct,[15] see id., at 1148–1149, 1176–1177, 1181, and those maps show that these voters were highly concentrated around the major urban areas of Winston-Salem (in Forsyth County), Greensboro (in Guilford County), and Charlotte (in Mecklenberg County). Dr. Ansolabehere, the challengers’ expert, prepared maps showing the distribution of black registered voters in these same counties, see id., at 322–328; 1 Record 128–133, and a comparison of these two sets of maps reveals that the clusters of Democratic voters generally overlap with those of registered black voters. In other words, the population of nearby Democrats who could be moved into District 12 was heavily black.

The upshot is that, so long as the legislature chose to retain the basic shape of District 12 and to increase the number of Democrats in the district, it was inevitable that the Democrats brought in would be disproportionately black.

None of this should come as a surprise. After all, when the basic shape of District 12 was created after the 1990 census, the express goal of the North Carolina Legislature was to create a majority-minority district. See Shaw I, 509 U. S., at 633–636. It has its unusual shape because it was originally designed to capture pockets of black voters. See Shaw v. Hunt, 517 U. S. 899 –906 (1996) (Shaw II). Although the legislature has modified the district since then, see Cromartie I, 526 U. S., at 544 (describing changes from the 1991 version to the 1997 version), “it retains its basic ‘snakelike’ shape and continues to track Interstate 85.” Ibid.; 1 Record 35 (Appellees’ Complaint) (“Congressional District 12 has existed in roughly its current form since 1992, when it was drawn as a majority African-American district . . . ”); see also App. 1163 (showing the 1997, 2001, and 2011 versions of District 12). The original design of the district was devised to ensure a high concentration of black voters, and as long as the basic design is retained (as it has been), one would expect that to continue.

While plaintiffs failed to offer any alternative map, Dr. Hofeller produced a map showing what District 12 would have looked like if his computer was programmed simply to maximize the Democratic vote percentage in the district, while still abiding by the requirement of one-person, one-vote. Id., at 1148. The result was a version of District 12 that is very similar to the version approved by the North Carolina Legislature. See id., at 1175; id., at 1615–1618. Indeed, this maximum-Democratic plan had a black voting age population of 50.73%, which is actually higher than District 12’s black voting age population of 50.66%. Id., at 1154 (Table 5).

Thus, the increase in the black voting age population of District 12 is easily explained by a coherent (and generally successful) political strategy. Cromartie II, 532 U. S., at 245 (“[A] legislature may, by placing reliable Democratic precincts within a district without regard to race, end up with a district containing more heavily African-American precincts, but the reasons would be political rather than racial”).

Amazingly, a reader of the majority opinion (and the opinion of the District Court) would remain almost entirely ignorant of the legislature’s political strategy and the relationship between that strategy and the racial composition of District 12.[16] The majority’s analysis is like Hamlet without the prince.[17]

C

The majority focuses almost all its attention on a few references to race by those responsible for the drafting and adoption of the redistricting plan. But the majority reads far too much into these references. First, what the plaintiffs had to prove was not simply that race played some role in the districting process but that it was the legislature’s predominant consideration. Second, as I have explained, a court must exercise “extraordinary caution” before finding that a state legislature’s predominant reason for a districting plan was racial. Miller, 515 U. S., at 916. This means that comments should not be taken out of context and given the most sinister possible meaning. Third, the findings of the state courts in a virtually identical challenge to District 12 are entitled to respectful consideration. A North Carolina trial court, after hearing much the same evidence as the court below, found that the legislature’s predominant motive was political, not racial. That decision was affirmed by the North Carolina Supreme Court. Dickson v. Rucho, 367 N. C. 542, 766 S. E. 2d 238 (2014), vacated and remanded, 575 U. S. ___, aff ’d on remand, 368 N. C. 481, 781 S. E. 2d 404 (2015), cert. pending, No. 16–24. Even if the judgment in the state case does not bar the present case under the doctrine of res judicata, see ante, at 7–9, the state-court finding illustrates the thinness of the plaintiffs’ proof.

Finally, it must be kept in mind that references to race by those responsible for drawing or adopting a redistricting plan are not necessarily evidence that the plan was adopted for improper racial reasons. Under our precedents, it is unconstitutional for the government to consider race in almost any context, and therefore any mention of race by the decisionmakers may be cause for suspicion. We have said, however, that that is not so in the redistricting context. For one thing, a State like North Caro-lina that was either wholly or partially within the coverage of §5 of the Voting Rights Act of 1965 could not redistrict without heeding that provision’s prohibition against racial retrogression, see 52 U. S. C. §10304(b); Alabama Legislative Black Caucus v. Alabama, 575 U. S. ___, ___–___ (2015) (slip op., at 3–4), and therefore race had to be kept in mind. In addition, all legislatures must also take into account the possibility of a challenge under §2 of the Voting Rights Act claiming that a plan illegally dilutes the voting strength of a minority community. See League of United Latin American Citizens v. Perry, 548 U. S. 399, 425 (2006) . If a State ultimately concludes that it must take race into account in order to comply with the Voting Rights Act, it must show that it had a “ ‘strong basis in evidence’ in support of the (race-based) choice that it has made.” Alabama Legislative Black Caucus, supra, at ___ (slip op., at 22). But those involved in the redistricting process may legitimately make statements about Voting Rights Act compliance before deciding that the Act does not provide a need for race-based districting. And it is understandable for such individuals to explain that a race-neutral plan happens to satisfy the criteria on which Voting Rights Act challengers might insist. In short, because of the Voting Rights Act, consideration and discussion of the racial effects of a plan may be expected.

1

The June 17, 2011, Statement

I begin with a piece of evidence that the majority does not mention, namely, the very first item cited by the District Court in support of its racial-predominance finding. This evidence consisted of a June 17, 2001, statement by Senator Rucho and Representative Lewis, the state legislators who took the lead in the adoption of the current map. In that statement, Rucho and Lewis referred to “constructing [Voting Rights Act] majority black districts.” App. 1025. Seizing upon the use of the plural term “districts,” the court below seemed to think that it had found a smoking gun. Harris v. McCrory, 159 F. Supp. 3d 600, 616 (MDNC 2016). The State had insisted that its plan drew only one majority-minority congressional district, District 1, but since the June 17 statement “clearly refers to multiple districts that are now majority minority,” ibid., the court below viewed the statement as telling evidence that an additional congressional district, presumably District 12, had been intentionally designed to be a majority-minority district and was thus based on race.

There is a glaring problem with this analysis: The June 17 statement was about state legislative districts, not federal congressional districts. See App. 1024–1033. The United States, as amicus curiae in support of plaintiffs, concedes that the District Court made a mistake by relying on the June 17 statement. Brief for United States 27, n. 13. The majority, by contrast, tries to ignore this error. But the District Court gave the June 17 statement pride-of-place in its opinion, mentioning it first in its analysis, and the District Court seemed to think that this evidence was particularly significant, stating that the reference to multiple districts was not “the result of happenstance, a mere slip of the pen.” 159 F. Supp. 3d, at 616. The District Court’s error shows a troubling lack of precision.

2

The §5 Preclearance Request

Under §5 of the Voting Rights Act, North Carolina requested preclearance from the Department of Justice shortly after the Legislature approved the new congressional plan. Id., at 608. In its preclearance application, the State noted that “[o]ne of the concerns of the Redistricting Chairs was that in 1992, the Justice Department had objected to the 1991 Congressional Plan because of a failure by the State to create a second majority minority district.” App. 478. The application says that the Redistricting Chairs “sought input from Congressman [Mel] Watt[, the African-American incumbent who represented District 12,] regarding options for re-drawing his district,” and that after this consultation, “the Chairs had the impression that Congressman Watt would oppose any redrawing of the Twelfth District . . . as originally contemplated by the 1992 Justice Department objection.” Ibid. The Chairs drew District 12 “[b]ased in part on this input from Congressman Watt.” Id., at 478–479. Two sentences later in the same paragraph, the application observed that the black voting age population for District 12 went up from 43.77% to 50.66% and that therefore the district “maintains, and in fact increases, the African-American community’s ability to elect their candidate of choice in District 12.” Id., at 479.

According to the majority, this statement shows a “determination to concentrate black voters in District 12.” Ante, at 23. In fact, it shows no such thing. The statement explains that Senator Rucho and Representative Lewis decided not to construct District 12 as a majority-minority district—as the 1992 Justice Department had demanded—“[b]ased in part on” the input they received from Congressman Watt, whom they thought “would oppose” drawing the district “as originally contemplated by the 1992 Justice Department objection.” App. 478–479. If anything, this document cuts against a finding of racial predominance.

The statement’s matter-of-fact reference to the increase in District 12’s black voting age population hardly shows that the legislature altered District 12 for the purpose of causing this increase. An entirely natural interpretation is that the Redistricting Chairs simply reported this fact so that it would be before the Justice Department in the event that the Department had renewed Voting Rights Act concerns. Only by reading a great deal between the lines and adopting the most sinister possible interpretation can the statement be viewed as pointed evidence of a predominantly racial motive.

3

The Mel Watt Testimony

In both the District Court and the state trial court, Congressman Watt testified that, while the redistricting plan was being developed, Senator Rucho invited him to his home to discuss the new boundaries of District 12. Id., at 2368–2369, 1343–1344. According to Congressman Watt, Senator Rucho said that the Republican leadership wanted him to “ramp the 12th Congressional District up to over 50 percent black” because “they believed it was required . . . by the Voting Rights Act.” Id., at 1344, 2369, 2393. In the state proceedings, Senator Rucho denied making any such statement, id., at 1703, and another state legislator present at the meeting, Representative Ruth Samuelson, gave similar testimony, id., at 1698. Neither Senator Rucho nor Representative Samuelson testified in federal court (although their state court testimony was made part of the federal record). See id., at 2847. But the District Court credited Congressman Watt’s testimony based on its assessment of his demeanor and the consistency of his recollection, 159 F. Supp. 3d, at 617–618, and I accept that credibility finding for purposes of our review.[18]

But even assuming that Congressman Watt’s recollection was completely accurate, all that his testimony shows is that legislative leaders at one point in the process thought that they had to draw District 12 as a majority-minority district in order to comply with the Voting Rights Act; it does not show that they actually did draw District 12 with the goal of creating a majority-minority district. And as explained in the discussion of the preclearance request above, Senator Rucho and Representative Lewis stated that they ultimately turned away from the creation of a majority-minority district after consulting with Congressman Watt. “Based in part on this input from Congressman Watt,” they said they decided not to draw the district as the 1992 Department of Justice had suggested—that is, as a majority-minority district. App. 478–479.

This account is fully consistent with Congressman Watt’s testimony about his meeting with Senator Rucho. Congressman Watt noted that Senator Rucho was uncomfortable with the notion of increasing the black voting age population, id., at 2369, 2393, and Congressman Watt testified that he told Senator Rucho that he was opposed to the idea, id., at 1345, 2369, 2393. So it makes sense that Senator Rucho was dissuaded from taking that course by Congressman Watt’s reaction. And Dr. Hofeller consistently testified that he was never asked to meet a particular black voting age population target, see Part III–C–5, infra, and that the only data displayed on his screen when he drew District 12 was political data. See infra, at 27, n. 19. Thus, Congressman Watt’s testimony, even if taken at face value, is entirely consistent with what the preclearance request recounts: After initially contemplating the possibility of drawing District 12 as a majority-minority district, the legislative leadership met with Congressman Watt, who convinced them not to do so.

4

Dr. Hofeller’s Statements About Guilford County

Under the prior map, both Guilford County and the Greensboro African-American community were divided between the 12th and 13th Districts. This had been done, Dr. Hofeller explained, “to make both the Old 12th and 13th Districts strongly Democratic.” App. 1103; see also id., at 555, 2821; 1 Record 132–133 (showing racial demographics of Guilford precincts under 2001 and 2011 maps). But the Republican legislature wanted to make the area surrounding District 12 more Republican. The new map eliminated the old 13th District and created a new district bearing that number farther to the east. The territory to the north of Greensboro that had previously been in the 13th District was placed in a new district, District 6, which was constructed to be a Republican-friendly district, and the new map moved more of the Greensboro area into the new District 12. This move was entirely consistent with the legislature’s stated goal of concentrating Democrats in the 12th District and mak-ing the surrounding districts hospitable to Republican candidates.

Dr. Hofeller testified that the placement of the Greensboro African-American community in the 12th District was the result of this political strategy. He stated that the portion of Guilford County absorbed by District 12 “wasn’t moved into CD 12 because it had a substantial black population. It was moved into CD 12 because it had a substantial Democratic political voting record . . . .” App. 2824. And Dr. Hofeller maintained that he was never instructed to draw District 12 as a majority-minority district or to increase the district’s black voting age population. See, e.g., id., at 520, 556–558, 1099, 1603–1604, 2682–2683, 2789. Instead, he testified that political considerations determined the boundaries of District 12 and that the only data displayed on his computer screen when he drew the challenged map was voting data from the 2008 Presidential election.[19] Id., at 1149, 2697, 2721–2722.

Dr. Hofeller acknowledged, however, that there had been concern about the possibility of a Voting Rights Act challenge to this treatment of the Greensboro African-American community. Guilford County was covered by §5 of the Voting Rights Act, and as noted, §5 prohibits retrogression. Under the old map, the Guilford County African-American community was split between the old District 13 and District 12, and in both of those districts, black voters were able to elect the candidates of their choice by allying with white Democratic voters. Under the new map, however, if the Greensboro black community had been split between District 12 and the new Republican-friendly District 6, the black voters in the latter district would be unlikely to elect the candidate of their choice. Placing the African-American community in District 12 avoided this consequence. Even Congressman Watt conceded that there were potential §5 concerns relating to the black community in Guilford County. Id., at 2387–2388.

The thrust of many of Dr. Hofeller’s statements about the treatment of Guilford County was that the reuniting of the Greensboro black community in District 12 was nothing more than a welcome byproduct of his political strat-egy. He testified that he first drew the district based on political considerations and then checked to ensure that Guilford County’s black population was not fractured. Id., at 2822 (“[W]hen we checked it, we found that we did not have an issue in Guilford County with fracturing the black . . . community”); see also id., at 556, 2821, 2823. This testimony is entirely innocuous.

There is no doubt, however, that Dr. Hofeller also made a few statements that may be read to imply that concern about Voting Rights Act litigation was part of the motivation for the treatment of Guilford County. He testified at trial that he “was instructed [not] to use race in any form except perhaps with regard to Guilford County.” Id., at 2791 (emphasis added). See id., at 1103 (the legislature “determined that it was prudent to reunify the African-American community in Guilford County”); id., at 558 (“[I]t was decided to reunite the black community in Guilford County into the Twelfth”).

These statements by Dr. Hofeller convinced the District Court that the drawing of District 12 was not a “purely . . . politically driven affair.” 159 F. Supp. 3d, at 619. But in order to prevail, the plaintiffs had to show much more—that race was the predominant reason for the drawing of District 12, and these few bits of testimony fall far short of that showing.

Our decision in Cromartie II illustrates this point. In that case, the legislature’s mapmaker made a statement that is remarkably similar to Dr. Hofeller’s. Gerry Cohen, the “legislative staff member responsible for drafting districting plans,” reported: “ ‘I have moved Greensboro Black community into the 12th, and now need to take [about] 60,000 out of the 12th. I await your direction on this.’ ” 532 U. S., at 254. This admission did not persuade the Court that the legislature’s predominant motive was racial. The majority ignores this obvious parallel with Cromartie II.

Moreover, in an attempt to magnify the importance of the treatment of Guilford County, the majority plays games with statistics. It states that “District 12 saw a net increase of more than 25,000 black voters in Guilford County, relative to a net gain of fewer than 35,000 across the district: So the newly added parts of that county played a major role in pushing the district’s BVAP over 50%.” Ante, at 26.

This is highly misleading. First, since the black voting age population of District 12 is just barely over 50%—specifically, 50.66%—almost any decision that increased the number of voting age blacks in District 12 could be said to have “played a major role in pushing the district’s BVAP over 50%.”

Second, the majority provides the total number of voting age blacks added to District 12 from Guilford County (approximately 25,000) alongside the total number of voting age blacks added to the district (approximately 35,000), and this has the effect of making Guilford County look like it is the overwhelming contributor to the district’s net increase in black voting age population. In truth, Mecklenburg County was by far the greatest contributor of voting age blacks to District 12 in both absolute terms (approximately 147,000) and in terms of new voting age blacks (approximately 37,000). See App. 384, 500–502. Indeed, if what matters to the majority is how much individual counties increased District 12’s black voting age population percentage, Davidson County deserves attention as well, since the portion of the county within District 12 lost over 26,000 more voting age whites than blacks. Ibid. That is greater than the net number of voting age blacks added to the district by Guilford County or Mecklenburg County. Ibid. As with so much in the majority opinion, the issue here is more nuanced—and much more favorable to the State—than the majority would have it seem.

5

The July 1, 2011, Statement

For reasons similar to those just explained, the majority makes far too much of a statement issued by Senator Rucho and Representative Lewis on July 1, 2011, when the new districting plan was proposed. Particularly in light of Dr. Hofeller’s later testimony about the legislature’s partisan objectives, it is apparent that this statement does not paint an entirely reliable picture of the legislature’s aims. The statement begins with this proclamation: “From the beginning, our goal has remained the same: the development of fair and legal congressional and legislative districts,” id., at 353, and the statement seriously downplays the role of politics in the map-drawing process, acknowledging only that “we have not been ignorant of the partisan impacts of the districts we have created,” id., at 361.

The statement discusses the treatment of Guilford County in a section with the heading “Compliance with the Voting Rights Act.” Id., at 355–358. In that section, Rucho and Lewis state: “Because of the presence of Guilford County in the Twelfth District, we have drawn our proposed Twelfth District at a black voting age level that is above the percentage of black voting age population found in the current Twelfth District. We believe that this measure will ensure preclearance of the plan.” Id., at 358.

The majority and the District Court interpret this passage to say that Rucho and Lewis decided to move black voters from Guilford County into District 12 in order to ward off Voting Rights Act liability. Ante, at 22 (“Because of the VRA, [Rucho and Lewis] increased the number of African-Americans” in District 12 (citing 159 F. Supp. 3d, at 617; emphasis in original)). But that is hardly the only plausible interpretation. The statement could just as easily be understood as “an explanation by [the] legislature that because they chose to add Guilford County back into CD 12, the district ended up with an increased ability to elect African-American candidates, rather than the legislature explaining that they chose to add Guilford County back into CD 12 because of the [racial] results that addition created.” Id., at 635 (Osteen, J., concurring in part and dissenting in part) (emphasis in original). And because we are obligated to presume the good faith of the North Carolina Legislature, this latter interpretation is the appropriate one.

But even if one adopts the majority’s interpretation, it adds little to the analysis. The majority’s close and incriminating reading of a statement issued to win public support for the new plan may represent poetic justice: Having attempted to blur the partisan aim of the new District 12, the legislature is hoisted on its own petard. But poetic justice is not the type of justice that we are supposed to dispense. This statement is some evidence that race played a role in the drawing of District 12, but it is a mistake to give this political statement too much weight.

Again, we made precisely this point in Cromartie II. There, the “legislative redistricting leader,” then-Senator Roy Cooper, testified before a legislative committee that the proposed plan “ ‘provides for . . . racial and partisan balance.’ ” 532 U. S., at 253 (emphasis added). The District Court read the statement literally and concluded that the district had been drawn with a racial objective. Ibid. But this Court dismissed the statement, reasoning that although “the phrase shows that the legislature considered race, along with other partisan and geographic considerations; . . . it says little or nothing about whether race played a predominant role comparatively speaking.” Ibid.

What was good in Cromartie II should also be good here.

6

Dr. Ansolabehere’s Testimony

Finally, the majority cites Dr. Ansolabehere’s testimony that black registered voters in the counties covered by District 12 were more likely to be drawn into District 12 than white registered voters and that black registered Democrats were more likely to be pulled in than white registered Democrats. Ante, at 26–27.

There is an obvious flaw in Dr. Ansolabehere’s analysis. He assumed that, if race was not the driving force behind the drawing of District 12, “white and black registered voters would have approximately the same likelihood of inclusion in a given Congressional District.” App. 2597 (internal quotation marks omitted). But that would be true only if black and white voters were evenly distributed throughout the region, and his own maps showed that this was not so. See id., at 322–328; 1 Record 128–133. Black voters were concentrated in the cities located at the north and south ends of the district and constituted a supermajority of Democrats in the area covered by District 12. See Part III–B, supra. As long as the basic shape of the district was retained, moving Democrats from areas outside but close to the old district boundaries naturally picked up far more black Democrats than white Democrats.

This explanation eluded Dr. Ansolabehere because he refused to consider either the implications of the political strategy that the legislature claimed to have pursued or the effects of the changes to District 12 on the surrounding districts. App. 2578–2582. The result was a distorted—and largely useless—analysis.

IV

Reviewing the evidence outlined above,[20] two themes emerge. First, District 12’s borders and racial composition are readily explained by political considerations and the effects of the legislature’s political strategy on the demographics of District 12. Second, the majority largely ignores this explanation, as did the court below, and instead adopts the most damning interpretation of all available evidence.

Both of these analytical maneuvers violate our clearly established precedent. Our cases say that we must “ ‘exercise extraordinary caution’ ” “ ‘where the State has articulated a legitimate political explanation for its districting decision,’ ” Cromartie II, supra, at 242 (emphasis deleted); the majority ignores that political explanation. Our cases say that “the good faith of a state legislature must be presumed,” Miller, 515 U. S., at 915; the majority presumes the opposite. And Cromartie II held that plaintiffs in a case like this are obligated to produce a map showing that the legislature could have achieved its political objectives without the racial effect seen in the challenged plan; here, the majority junks that rule and says that the plaintiffs’ failure to produce such a map simply “does not matter.” Ante, at 30.

The judgment below regarding District 12 should be reversed, and I therefore respectfully dissent.