Abbott v. Perez, 585 U.S. ___ (2018)

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

_________________

Nos. 17–586 and 17–626

_________________

GREG ABBOTT, GOVERNOR OF TEXAS, ET AL., APPELLANTS

17–586v.

SHANNON PEREZ, ET AL.

GREG ABBOTT, GOVERNOR OF TEXAS, ET AL., APPELLANTS

17–626v.

SHANNON PEREZ, ET AL.

on appeals from the united states district court for the western district of texas

[June 25, 2018]

Justice Alito delivered the opinion of the Court.

Before us for review are orders of a three-judge court in the Western District of Texas effectively directing the State not to conduct this year’s elections using districting plans that the court itself adopted some years earlier. The court developed those plans for use in the 2012 elections pursuant to our directions in Perry v. Perez, 565 U. S. 388 (2012) (per curiam). We instructed the three-judge court to start with the plans adopted by the Texas Legislature in 2011 but to make adjustments as required by the Constitution and the Voting Rights Act. Id., at 392–396. After those plans were used in 2012, the Texas Legislature enacted them (with only minor modifications) in 2013, and the plans were used again in both 2014 and 2016.

Last year, however, the three-judge court reversed its prior analysis and held that some of the districts in those plans are unlawful. After reviewing the repealed 2011 plans, which had never been used, the court found that they were tainted by discriminatory intent and that the 2013 Legislature had not “cured” that “taint.”

We now hold that the three-judge court committed a fundamental legal error. It was the challengers’ burden to show that the 2013 Legislature acted with discriminatory intent when it enacted plans that the court itself had produced. The 2013 Legislature was not obligated to show that it had “cured” the unlawful intent that the court attributed to the 2011 Legislature. Thus, the essential pillar of the three-judge court’s reasoning was critically flawed.

When the congressional and state legislative districts are reviewed under the proper legal standards, all but one of them, we conclude, are lawful.

I

A

The 2010 decennial census revealed that the population of Texas had grown by more than 20% and the State was therefore apportioned four additional seats in the United States House of Representatives. C. J. S. 369a.[1] To accommodate this new allocation and the population changes shown by the census, the Legislature adopted a new congressional districting plan, as well as new districting maps for the two houses of the State Legislature.

Redistricting is never easy, and the task was especially complicated in Texas in 2011. Not only was the Legislature required to draw districts that were substantially equal in population, see Perry, supra, at 391–392; Rey- nolds v. Sims, 377 U. S. 533 (1964); Wesberry v. Sanders, 376 U. S. 1 (1964), and to comply with special state-law districting rules,[2] but federal law imposed complex and delicately balanced requirements regarding the consideration of race.

Then, as now, federal law restricted the use of race in making districting decisions. The Equal Protection Clause forbids “racial gerrymandering,” that is, intentionally assigning citizens to a district on the basis of race without sufficient justification. Shaw v. Reno, 509 U. S. 630, 641 (1993). It also prohibits intentional “vote dilution”—“invidiously . . . minimiz[ing] or cancel[ing] out the voting potential of racial or ethnic minorities.” Mobile v. Bolden, 446 U. S. 55, 66–67 (1980) (plurality opinion).

While the Equal Protection Clause imposes these important restrictions, its application in the field of districting is complicated. For one thing, because a voter’s race sometimes correlates closely with political party preference, see Cooper v. Harris, 581 U. S. ___, ___–___ (2017) (slip op., at 19–20); Easley v. Cromartie, 532 U. S. 234, 243 (2001), it may be very difficult for a court to determine whether a districting decision was based on race or party preference. Here, the three-judge court found that the two factors were virtually indistinguishable.[3]

At the same time that the Equal Protection Clause restricts the consideration of race in the districting process, compliance with the Voting Rights Act of 1965, 79Stat. 437, as amended, 52 U. S. C. §10301 et seq. (VRA), pulls in the opposite direction: It often insists that districts be created precisely because of race. Two provisions of the VRA exert such demands, and in 2011, Texas was subject to both. At that time, Texas was covered by §5 of the VRA[4] and was thus barred from making any districting changes unless it could prove that they did not result in “retrogression” with respect to the ability of racial minorities to elect the candidates of their choice. Alabama Legislative Black Caucus v. Alabama, 575 U. S. ___, ___ (2015) (slip op., at 3). That showing obviously demanded consideration of race.

On top of this, Texas was (and still is) required to comply with §2 of the VRA. A State violates §2 if its districting plan provides “ ‘less opportunity’ ” for racial minorities “ ‘to elect representatives of their choice.’ ” League of United Latin American Citizens v. Perry, 548 U. S. 399, 425 (2006) (LULAC). In a series of cases tracing back to Thornburg v. Gingles, 478 U. S. 30 (1986), we have interpreted this standard to mean that, under certain circumstance, States must draw “opportunity” districts in which minority groups form “effective majorit[ies],” LULAC, supra, at 426.

Since the Equal Protection Clause restricts consideration of race and the VRA demands consideration of race, a legislature attempting to produce a lawful districting plan is vulnerable to “ ‘competing hazards of liability.’ ” Bush v. Vera, 517 U. S. 952, 977 (1996) (plurality opinion). In an effort to harmonize these conflicting demands, we have assumed that compliance with the VRA may justify the consideration of race in a way that would not otherwise be allowed. In technical terms, we have assumed that complying with the VRA is a compelling state interest, see, e.g., Bethune-Hill v. Virginia State Bd. of Elections, 580 U. S. ___, ___ (2017) (slip op., at 13); Shaw v. Hunt, 517 U. S. 899, 915 (1996), and that a State’s consideration of race in making a districting decision is narrowly tailored and thus satisfies strict scrutiny if the State has “ ‘good reasons’ ” for believing that its decision is necessary in order to comply with the VRA. Cooper, supra, at ___ (slip op., at 3).

B

Facing this legal obstacle course, the Texas Legislature in 2011 adopted new districting plans, but those plans were immediately tied up in litigation and were never used. Several plaintiff groups quickly filed challenges in the District Court for the Western District of Texas, arguing that some of the districts in the new plans were racial gerrymanders, some were based on intentional vote dilution, and some had the effect of depriving minorities of the equal opportunity to elect the candidates of their choice. This case was assigned to a three-judge court, as required by 28 U. S. C. §2284(a). (We will call this court “the Texas court” or simply “the District Court.”)

The situation was further complicated by the requirement that Texas obtain preclearance of its new plans. To do this, Texas filed for a declaratory judgment in the District Court for the District of Columbia. See Texas v. United States, 887 F. Supp. 2d 133 (2012). (We will call this court “the D. C. court.”) By early 2012, the D. C. court had not yet issued a decision, and Texas needed usable plans for its rapidly approaching primaries. Accordingly, the Texas court drew up interim plans for that purpose. Perez v. Perry, 835 F. Supp. 2d 209 (2011). In creating those plans, the majority of the Texas court thought that it was not “required to give any deference to the Legislature’s enacted plan.” Id., at 213. Instead, it based its plans on what it called “neutral principles that advance the interest of the collective public good.” Id., at 212.[5]

Texas challenged those court-ordered plans in this Court, and we reversed. Perry v. Perez, 565 U. S. 388 (2012) (per curiam). Noting that “[r]edistricting is ‘primarily the duty and responsibility of the State,’ ” we held that the Texas court should have respected the legislative judgments embodied in the 2011 plans to the extent allowed by the Constitution and the VRA. Id., at 392–399.

We remanded the case with very specific instructions. The Texas court was told to start with the plans adopted by the Legislature but to modify those plans as needed so as “not to incorporate . . . any legal defects.” Id., at 394. With respect to claims under the Constitution or §2 of the VRA, the District Court was told to change a district if the plaintiffs were likely to succeed on the merits of their challenge. Ibid. And with respect to §5 claims, the court was instructed to make whatever changes were needed to obviate any legal claim that was “not insubstantial.”[6] Id., at 395. Thus, our instructions, in an abundance of caution, demanded changes in the challenged 2011 plans without proof that those changes were actually required by either the Constitution or the VRA.

On remand, the Texas court ordered additional briefing and heard two more days of argument. App. 29a, 35a–50a; Order in Civ. No. 11–cv–00360, Doc. No. 616. It issued two opinions, totaling more than 70 pages, and analyzed disputed districts in detail. C. J. S. 367a–423a; H. J. S. 300a–315a. While stressing the preliminary nature of its determinations, see C. J. S. 368a; H. J. S. 314a–315a, the court found that some districts required change and that others were lawful, C. J. S. 367a–423a; H. J. S. 300a–315a. The court then adopted plans for the State’s congressional districts and for both houses of the State Legislature. (The plan for the State Senate is not at issue.)

Both the congressional plan and the plan for the Texas House departed significantly from the State’s 2011 plans. At least 8 of the 36 congressional districts were markedly altered, and 21 districts in the plan for the Texas House were “substantially” changed. H. J. S. 314a; C. J. S. 397a–408a.

In August 2012, the D. C. court denied preclearance of the plans adopted by the Legislature in 2011, see Texas v. United States, supra, so the State conducted the 2012 elections under the interim plans devised by the Texas court. At the same time, Texas filed an appeal in this Court contesting the decision of the D. C. court,[7] but that appeal ultimately died for two reasons.

First, the 2011 plans were repealed. The Texas attorney general urged the Legislature to pass new redistricting plans, C. J. S. 429a, and in his view, the “best way to remedy the violations found by the D. C. court” was to “adopt the [Texas court’s] interim plans as the State’s permanent redistricting maps.” Id., at 432a. Doing so, he said, would “confirm the legislature’s intent” to adopt “a redistricting plan that fully comports with the law.” Id., at 429a.

The Governor called a special session to do just that, and the Legislature complied. One of the legislative sponsors, Senator Seliger, explained that, although “ ‘the Texas Legislature remains confident that the legislatively-drawn maps adopted in 2011 are fair and legal . . . , there remain several outstanding legal questions regarding these maps that undermine the stability and predictability of the electoral process in Texas.’ ” 274 F. Supp. 3d 624, 649, n. 40 (2017). Counsel for one of the plaintiff groups, the Mexican American Legal Defense and Education Fund (MALDEF), testified in favor of the plans. C. J. S. 436a–439a. The 2013 Legislature then repealed the 2011 plans and enacted the Texas court’s interim plans with just a few minor changes. The federal congressional plan was not altered at all, and only small modifications were made to the plan for the Texas House. C. J. S. Findings 231a–232a.

On the day after the Legislature passed the new plans and the day before the Governor signed them, this Court issued its decision in Shelby County v. Holder, 570 U. S. 529 (2013), which invalidated the coverage formula in §4 of the Voting Rights Act. Now no longer subject to §5, Texas obtained a vacatur of the D. C. court’s order on preclearance. 274 F. Supp. 3d, at 634–635, and n. 11.

With the never-effective 2011 plans now repealed and any preclearance issues overcome by events, the State argued in the Texas court that the plaintiffs’ case against the 2011 plans was moot. In September 2013, the Texas court allowed the plaintiffs to amend their complaints to challenge the 2013 plans, but the court held that their challenges to the 2011 plans were still alive, reasoning that the repeal of the 2011 plans represented the “voluntary cessation” of allegedly unconstitutional conduct.[8]

Texas conducted its 2014 and 2016 elections under the plans that had been preliminarily approved by the Texas court and subsequently adopted (with only minor changes) by the Legislature in 2013. But in March and April 2017, after multiple trials, the Texas court issued a pair of rulings on the defunct 2011 plans. The court reaffirmed the conclusions it had reached in 2012 about defects in the 2011 plans, and it went further. Contrary to its earlier decision, it held that Congressional District (CD) 35 is an impermissible racial gerrymander and that CD27 violates §2 of the Voting Rights Act because it has the effect of diluting the electoral opportunities of Latino voters. C. J. S. 181a, 193a–194a. Previously, the court had provided detailed reasons for rejecting the very arguments that it now accepted. Id., at 409a–423a. Similarly, the court held that multiple districts in the plan for the Texas House were the result of intentional vote dilution. These included districts in the counties of Nueces (House District (HD) 32, HD34), Bell (HD54, HD55), and Dallas (HD103, HD104, HD105). H. J. S. 275a–276a.[9]

In August 2017, having ruled on the repealed 2011 plans, the Texas court finally turned its attention to the plans then in effect—i.e., the plans that had been developed by the court, adopted by the Legislature in 2013, and used in both the 2014 and 2016 elections. The court invalidated the districts in those plans that correspond to districts in the 2011 plan that it had just held to be unlawful, i.e., CD27, CD35, HD32, HD34, HD54, HD55, HD103, HD104, and HD105. See 274 F. Supp. 3d 624 (2017) (No. 17–586) and 267 F. Supp. 3d 750 (2017) (No. 17–626).

In reaching these conclusions, the court pointed to the discriminatory intent allegedly harbored by the 2011 Legislature, and it attributed this same intent to the 2013 Legislature because it had failed to “engage in a deliberative process to ensure that the 2013 plans cured any taint from the 2011 plans.” 274 F. Supp. 3d, at 645–652; 267 F. Supp. 3d, at 757. The court saw “no indication that the Legislature looked to see whether any discriminatory taint remained in the plans.” 274 F. Supp. 3d, at 649. And it faulted the State because it “did not accept [findings of the D. C. court] and instead appealed to the Supreme Court.” Ibid. Seeing no evidence that the State had undergone “a change of heart,” the court concluded that the Legislature’s “decision to adopt the [District Court’s] plans” was a “litigation strategy designed to insulate the 2011 or 2013 plans from further challenge, regardless of their legal infirmities.” Id., at 649–650. Finally, summarizing its analysis, the court reiterated that the 2011 Legislature’s “discriminatory taint was not removed by the [2013] Legislature’s enactment of the Court’s interim plans, because the Legislature engaged in no deliberative process to remove any such taint, and in fact intended any such taint to be maintained but be safe from remedy.” Id., at 686.

The Texas court’s decisions about CD35 and all but three of the Texas House districts were based entirely on its finding that the 2013 Legislature had not purged its predecessor’s discriminatory intent. However, the court also held that three districts—CD27, HD32, and HD34—were invalid under §2 of the Voting Rights Act because they had the effect of depriving Latinos of the equal opportunity to elect their candidates of choice. Id., at 682–686; 267 F. Supp. 3d, at 775–783. And the court found independent proof that HD90 was a racial gerrymander. Id., at 788–794.

The court held that violations in all these districts “must be remedied.” 274 F. Supp. 3d, at 686; see also 267 F. Supp. 3d, at 795 (describing State House district violations that “must be remedied”). Mindful that October 1 was the deadline for the Texas Secretary of State to provide voter registration templates to the State’s counties, App. 380a–381a, the court took steps to bring about prompt remedial action. In two orders issued on August 15 and 24, the Texas attorney general was instructed to advise the court, within three days, “whether the Legislature intends to take up redistricting in an effort to cure these violations.” 274 F. Supp. 3d, at 686; 267 F. Supp. 3d, at 795. If the Legislature chose not to do so, the court warned, it would “hold a hearing to consider remedial plans.” Ibid. After the Governor made clear that the State would not act, the court ordered the parties to proceed with a hearing on the congressional plan on September 5, as well as a hearing on the plan for the Texas House on September 6. 274 F. Supp. 3d, at 686; 267 F. Supp. 3d, at 795; App. 134a–136a; Defendants’ Opposed Motion To Stay Order on Plan C235 Pending Appeal or Final Judgment in Civ. No. 11–cv–00360, Doc. 1538, pp. 3–4; Defendants’ Opposed Motion To Stay Order on Plan H358 Pending Appeal or Final Judgment, Doc. No. 1550, pp. 4–5.

Texas applied for stays of both orders, but the District Court denied the applications. App. 134a–136a. Texas then asked this Court to stay the orders, and we granted that relief. After receiving jurisdictional statements, we postponed consideration of jurisdiction and set the cases for consolidated argument. 583 U. S. ___ (2018).

II

Before reaching the merits of these appeals, we must assure ourselves that we have jurisdiction to review the orders at issue. Appellants claim that the orders amount to injunctions and are therefore appealable to this Court under 28 U. S. C. §1253. Appellees disagree, contending that the orders do not qualify as injunctions. We hold that we have jurisdiction because the orders were effectively injunctions in that they barred Texas from using the districting plans now in effect to conduct this year’s elections.

A

The Judiciary Act of 1789, 1Stat. 73, “established the general principle that only final decisions of the federal district courts would be reviewable on appeal.” Carson v. American Brands, Inc., 450 U. S. 79, 83 (1981) (emphasis deleted). But because “rigid application of this principle was found to create undue hardship in some cases,” Congress created exceptions. Ibid. Two are relevant here. We have jurisdiction under 28 U. S. C. §1253 to hear an appeal from an order of a three-judge district court “granting or denying . . . an interlocutory or permanent injunction.” Similarly, §1292(a)(1) gives the courts of appeals jurisdiction over “[i]nterlocutory orders of the district courts” “granting, continuing, modifying, refusing or dissolving injunctions,” “except where a direct review may be had in the Supreme Court.”

The orders in these cases fall within §1253. To be sure, the District Court did not call its orders “injunctions”—in fact, it disclaimed the term, App. 134a–136a—but the label attached to an order is not dispositive. We have previously made clear that where an order has the “practical effect” of granting or denying an injunction, it should be treated as such for purposes of appellate jurisdiction. Carson, supra, at 83; see also Gulfstream Aerospace Corp. v. Mayacamas Corp., 485 U. S. 271, 287–288 (1988). We applied this test in Carson, holding that an order that declined to enter a consent decree prohibiting certain conduct could be appealed under §1292(a)(1) because it was the practical equivalent of an order denying an injunction and threatened serious and perhaps irreparable harm if not immediately reviewed. 450 U. S., at 83–84, 86–90.

This “practical effect” rule serves a valuable purpose. If an interlocutory injunction is improperly granted or denied, much harm can occur before the final decision in the district court. Lawful and important conduct may be barred, and unlawful and harmful conduct may be allowed to continue. Recognizing this, Congress authorized interlocutory appellate review of such orders. But if the availability of interlocutory review depended on the district court’s use of the term “injunction” or some other particular language, Congress’s scheme could be frustrated. The harms that Congress wanted to avoid could occur so long as the district court was careful about its terminology. The “practical effect” inquiry prevents such manipulation.

In analogous contexts, we have not allowed district courts to “shield [their] orders from appellate review” by avoiding the label “injunction.” Sampson v. Murray, 415 U. S. 61, 87 (1974). For instance, in Sampson, we held that an order labeled a temporary restraining order (which is not appealable under §1292(a)(1)) should be treated as a “preliminary injunction” (which is appealable) since the order had the same practical effect as a preliminary injunction. Id., at 86–88.

Appellees and the dissent contend that the “practical effect” approach should be confined to §1292(a)(1), but we see no good reason why it should not apply to §1253 as well. Appellees note that we “narrowly constru[e]” §1253, Goldstein v. Cox, 396 U. S. 471, 478 (1970), but we also construe §1292(a)(1) “narrowly,” Carson, supra, at 84. In addition, the relevant language in the two provisions is nearly identical;[10] both provisions serve the same purpose; and we have previously called them “analogous.” Goldstein, supra, at 475.

The provisions are also textually interlocked. Section 1292(a)(1) does not apply where “direct review may be had in the Supreme Court,” i.e., where §1253 applies. If the “practical effects” test applied under §1292(a)(1) but not §1253, the consequences would be unfortunate and strange. We would have to identify the magic language needed for an order to qualify as an order granting or denying an injunction, and that standard would hardly constitute the sort of “[s]imple” rule that the dissent prizes. See post, at 14 (opinion of Sotomayor, J.). Then, having developed that standard, we would have to apply it in any case in which a party took an appeal to us from an order of a three-judge court that clearly had the practical effect of an injunction. If we concluded that the magic-words test was not met, the order would appear to be appealable to one of the courts of appeals under §1292(a)(1). In the language of that provision, the order would be an “orde[r] of [a] district cour[t] of the United States . . . granting [an] injunctio[n].” And because this Court would lack jurisdiction under §1253, the appeal would not fall within §1292(1)’s exception for cases “where a direct review may be had in the Supreme Court.” Having taken pains to provide for review in this Court, and not in the courts of appeals, of three-judge court orders granting injunctions Congress surely did not intend to produce that result.[11]

Appellees argue that an order denying an injunction (the situation in Carson) and an order granting an injunction (the situation here) should be treated differently, Brief for Appellees in No. 17–586, p. 27, but they offer no convincing reason for doing so. No authority supports their argument. The language of §§1253 and 1292(a)(1) makes no such distinction, and we have stated that the “practical effect” analysis applies to the “granting or denying” of injunctions. Gulfstream, supra, at 287–288.

In addition, appellees’ suggested distinction would put appellate courts in an awkward position. Suppose that a district court granted an injunction that was narrower than the one requested by the moving party. Would an appellate court (whether this Court or a court of appeals) have jurisdiction to rule on only part of that decision? Suppose the appellate court concluded that the district court was correct in refusing to give the movant all the injunctive relief it sought because the movant’s entire claim was doomed to fail. Would the appellate court be limited to holding only that the lower court properly denied the relief that was withheld? The rule advocated by the appellees would needlessly complicate appellate review.[12]

Finally, appellees point in passing to Rule 65(d) of the Federal Rules of Civil Procedure, which requires that an injunction “state its terms specifically” and “describe in reasonable detail . . . the act or acts restrained or required.” Rules 65(d)(1)(B), (C); see Brief for Appellees in No. 17–586, at 27. But as explained in Gunn v. University Comm. to End War in Viet Nam, 399 U. S. 383, 389, n. 4 (1970), we have never suggested that a failure to meet the specificity requirements of Rule 65(d) would “deprive the Court of jurisdiction under §1253.”

A contrary holding would be perverse. Rule 65(b) protects the party against which an injunction is issued by requiring clear notice as to what that party must do or refrain from doing. Where a vague injunction does not comply with Rule 65(b), the aggrieved party has a particularly strong need for appellate review. It would be odd to hold that there can be no appeal in such a circumstance.

For these reasons, we hold that we have jurisdiction under §1253 to hear an appeal from an order that has the same practical effect as one granting or denying an injunction.

B

With these principles settled, we conclude that the orders in these cases qualify as interlocutory injunctions under §1253. The text of the orders and the context in which they were issued make this clear.

The orders are unequivocal that the current legislative plans “violate §2 and the Fourteenth Amendment” and that these violations “must be remedied.” 274 F. Supp. 3d, at 686; see also, e.g., 267 F. Supp. 3d, at 795 (“[V]iolations found by this Court in its Order on [the State House plan] now require a remedy”); ibid. (“In Bell County, the intentional discrimination previously found by the Court must be remedied”); ibid. (“In Dallas County, the intentional discrimination previously found by the Court must be remedied”).

We do not suggest that this language alone is sufficient to show that the orders had the practical effect of enjoining use of the current plans in this year’s elections, but the court did not stop with these pronouncements. As we have noted, the orders required the Texas attorney general to inform the court within three days whether the Legislature would remedy the violations, and the orders stated that if the Legislature did not intend to adopt new plans, the court would hold remedial hearings.

The short time given the Legislature to respond is strong evidence that the three-judge court did not intend to allow the elections to go ahead under the plans it had just condemned. The Legislature was not in session, so in order to take up the task of redistricting, the Governor would have been required to convene a special session—which is no small matter. And, when the Governor declined to call a special session, the court moved ahead with its scheduled hearings and invited the parties to continue preparing for them even after this Court administratively stayed the August 15 order.

The import of these actions is unmistakable: The court intended to have new plans ready for use in this year’s elections. Nothing in the record even hints that the court contemplated the possibility of allowing the elections to proceed under the 2013 plans.

What is more, Texas had reason to believe that it would risk deleterious consequences if it defied the court and attempted to conduct the elections under the plans that the court had found to be based on intentional racial discrimination. In the very orders at issue, the court inferred discriminatory intent from Texas’s choice to appeal the D. C. court’s preclearance decision rather than immediately taking steps to bring its plans into compliance with that decision. 274 F. Supp. 3d, at 649; see Part III, infra. Reading such an order, Texas had reason to fear that if it tried to conduct elections under plans that the court had found to be racially discriminatory, the court would infer an evil motive and perhaps subject the State once again to the strictures of preclearance under §3(c) of the Voting Rights Act.[13] This is a remedy that the plaintiffs hoped to obtain, see, e.g., App. 177a, and that the District Court seemed inclined to consider, see C. J. S. 122a–123a (declining to declare moot the challenges to the long-since-repealed 2011 plans because “there remains the possibility of declaratory and equitable relief under §3(c)”).

Contending that the orders here do not qualify under §1253, appellees analogize this case to Gunn, 399 U. S. 383, but there is no relevant similarity. In Gunn, anti-war protesters were charged with violating a Texas “disturbing-the-peace statute,” id., at 384, and they challenged the constitutionality of the statute in federal court. After the state charges were dismissed, the District Court issued a “discursive” opinion “expressing the view that [the statute was] constitutionally invalid.” Id., at 386–387. But the court then refrained from going any further, “pending the next session, special or general, of the Texas legislature, at which time the State of Texas may, if it so desires, enact such disturbing-the-peace statute as will meet constitutional requirements.” University Comm. to End War in Viet Nam v. Gunn, 289 F. Supp. 469, 475 (WD Tex. 1968). The defendants appealed to this Court, and at the time of our decision two years later, neither the Legislature nor the District Court had taken any further action. We therefore held that we lacked jurisdiction under §1253. The District Court order in that case did not have the same practical effect as an injunction. Indeed, it had no practical effect whatsoever and is thus entirely different from the orders now before us.[14]

Appellees suggest that appellate jurisdiction is lacking in this case because we do not know at this point “what a remedy would entail, who it would affect, and when it would be implemented.” Brief for Appellees in No. 17–586, at 27. The dissent makes a similar argument with respect to two of the Texas House districts. Post, at 13–14.[15] But the issue here is whether this year’s elections can be held under the plans enacted by the Legislature, not whether any particular remedies would have ulti- mately been ordered by the District Court.

Appellees and the dissent also fret that this Court will be inundated with redistricting appeals if we accept jurisdiction here, Brief for Appellees in No. 17–626, p. 34; post, at 15–16, and n. 8, but there is no reason to fear such a flood. Because §1253 expressly authorizes “interlocutory” appeals, there is no question that there can be more than one appeal in a case challenging a redistricting plan. District courts sometimes expressly enjoin the use of districting plans before moving on to the remedial phase. See, e.g., Whitford v. Gill, No. 3:15–cv–421, Doc. No. 190 (WD Wis., Feb. 22, 2017); Harris v. McCrory, No. 1:13–cv–949, Doc. No. 143 (MDNC, Feb. 5, 2016). But appeals from such orders have not overwhelmed our docket. Our holding here will affect only a small category of additional cases.[16]

It should go without saying that our decision does not mean that a State can always appeal a district court order holding a redistricting plan unlawful. A finding on liability cannot be appealed unless an injunction is granted or denied, and in some cases a district court may see no need for interlocutory relief. If a plan is found to be unlawful long before the next scheduled election, a court may defer any injunctive relief until the case is completed. And if a plan is found to be unlawful very close to the election date, the only reasonable option may be to use the plan one last time.

We appreciate our obligation to heed the limits of our jurisdiction, and we reiterate that §1253 must be strictly construed. But it also must be sensibly construed, and here the District Court’s orders, for all intents and purposes, constituted injunctions barring the State from conducting this year’s elections pursuant to a statute enacted by the Legislature. Unless that statute is unconstitutional, this would seriously and irreparably harm[17] the State, and only an interlocutory appeal can protect that State interest. See Carson, supra, at 89–90. As a result, §1253 provides jurisdiction.

III

We now turn to the merits of the appeal. The primary question is whether the Texas court erred when it required the State to show that the 2013 Legislature somehow purged the “taint” that the court attributed to the defunct and never-used plans enacted by a prior legislature in 2011.

A

Whenever a challenger claims that a state law was enacted with discriminatory intent, the burden of proof lies with the challenger, not the State. Reno v. Bossier Parish School Bd., 520 U. S. 471, 481 (1997). This rule takes on special significance in districting cases.

Redistricting “is primarily the duty and responsibility of the State,” and “[f]ederal-court review of districting legislation represents a serious intrusion on the most vital of local functions.” Miller v. Johnson, 515 U. S. 900, 915 (1995) (internal quotation marks omitted). “[I]n assessing the sufficiency of a challenge to a districting plan,” a court “must be sensitive to the complex interplay of forces that enter a legislature’s redistricting calculus.” Id., at 915–916. And the “good faith of [the] state legislature must be presumed.” Id., at 915.

The allocation of the burden of proof and the presumption of legislative good faith are not changed by a finding of past discrimination. “[P]ast discrimination cannot, in the manner of original sin, condemn governmental action that is not itself unlawful.” Mobile, 446 U. S., at 74 (plurality opinion). The “ultimate question remains whether a discriminatory intent has been proved in a given case.” Ibid. The “historical background” of a legislative enactment is “one evidentiary source” relevant to the question of intent. Arlington Heights v. Metropolitan Housing Development Corp., 429 U. S. 252, 267 (1977). But we have never suggested that past discrimination flips the evidentiary burden on its head.

Neither the District Court nor appellees have pointed to any authority that would justify shifting the burden. The appellees rely primarily on Hunter v. Underwood, 471 U. S. 222 (1985), but that case addressed a very different situation. Hunter involved an equal protection challenge to an article of the Alabama Constitution adopted in 1901 at a constitutional convention avowedly dedicated to the establishment of white supremacy. Id., at 228–230. The article disenfranchised anyone convicted of any crime on a long list that included many minor offenses. Id., at 226–227. The court below found that the article had been adopted with discriminatory intent, and this Court accepted that conclusion. Id., at 229. The article was never repealed, but over the years, the list of disqualifying offenses had been pruned, and the State argued that what remained was facially constitutional. Id., at 232–233. This Court rejected that argument because the amendments did not alter the intent with which the article, including the parts that remained, had been adopted. Id., at 233. But the Court specifically declined to address the question whether the then-existing version would have been valid if “[re]enacted today.” Ibid.

In these cases, we do not confront a situation like the one in Hunter. Nor is this a case in which a law originally enacted with discriminatory intent is later reenacted by a different legislature. The 2013 Texas Legislature did not reenact the plan previously passed by its 2011 predecessor. Nor did it use criteria that arguably carried forward the effects of any discriminatory intent on the part of the 2011 Legislature. Instead, it enacted, with only very small changes, plans that had been developed by the Texas court pursuant to instructions from this Court “not to incorporate . . . any legal defects.” Perry, 565 U. S., at 394.

Under these circumstances, there can be no doubt about what matters: It is the intent of the 2013 Legislature. And it was the plaintiffs’ burden to overcome the presumption of legislative good faith and show that the 2013 Legislature acted with invidious intent.

The Texas court contravened these basic principles. Instead of holding the plaintiffs to their burden of overcoming the presumption of good faith and proving discriminatory intent, it reversed the burden of proof. It imposed on the State the obligation of proving that the 2013 Legislature had experienced a true “change of heart” and had “engage[d] in a deliberative process to ensure that the 2013 plans cured any taint from the 2011 plans.” 274 F. Supp. 3d, at 649.

The Texas court’s references to the need to “cure” the earlier Legislature’s “taint” cannot be dismissed as stray comments. On the contrary, they were central to the court’s analysis. The court referred repeatedly to the 2013 Legislature’s duty to expiate its predecessor’s bad intent, and when the court summarized its analysis, it drove the point home. It stated: “The discriminatory taint [from the 2011 plans] was not removed by the Legislature’s enactment of the Court’s interim plans, because the Legislature engaged in no deliberative process to remove any such taint, and in fact intended any such taint to be maintained but be safe from remedy.” Id., at 686.[18]

The dissent labors to explain away all these references to the 2013’s Legislature’s supposed duty to purge its predecessor’s allegedly discriminatory intent, but the dissent loses track of its own argument and characterizes the District Court’s reasoning exactly as we have. Indeed, the dissent criticizes us on page 33 of its opinion for saying precisely the same thing that it said 11 pages earlier. On page 33, the dissent states:

“[T]he majority quotes the orders as requiring proof that the Legislature “ ‘engage[d] in a deliberative process to ensure that the 2013 plans cured any taint from the 2011 plans.’ ” But the District Court did not put the burden on Texas to make that affirmative showing.” Post, at 33–34 (quoting supra, at 23, in turn quoting 274 F. Supp. 3d, at 649; citations omitted).

But earlier, the dissent itself describes the District Court’s analysis as follows:

“Despite knowing of the discrimination in its 2011 maps, ‘the Legislature did not engage in a deliberative process to ensure that the 2013 plans cured any taint from the 2011 plans.’ ” Post, at 22 (quoting 274 F. Supp. 3d, at 649).

And this is not just a single slip of the pen. The dissent writes that the District Court was required “to assess how the 2013 Legislature addressed the known discrimination that motivated” the districts approved by that Court in 2012. Post, at 31. The dissent quotes the District Court’s statement that “ ‘there is no indication that the Legislature looked to see whether any discriminatory taint remained in the plans.’ ” Post, at 24 (quoting 274 F. Supp. 3d, at 649). And there is also this: “Texas was just ‘not truly interested in fixing any remaining discrimination in [its 2011 maps].’ ” Post, at 22 (quoting 274 F. Supp. 3d, at 651, n. 45). The District Court’s true mode of analysis is so obvious that the dissent cannot help but repeat it. And that approach was fundamentally flawed and demands reversal.

While a district court’s finding of fact on the question of discriminatory intent is reviewed for clear error, see Cromartie, 532 U. S., at 242, whether the court applied the correct burden of proof is a question of law subject to plenary review, U. S. Bank N. A. v. Village at Lakeridge, LLC, 583 U. S. ___, ___ (2018) (slip op., at 5); Highmark Inc. v. Allcare Health Management System, Inc., 572 U. S. ___, ___ (2014) (slip op., at 4). And when a finding of fact is based on the application of an incorrect burden of proof, the finding cannot stand. Bose Corp. v. Consumers Union of United States, Inc., 466 U. S. 485, 501 (1984) (“An appellate cour[t has] power to correct errors of law, including those that . . . infect . . . a finding of fact that is predicated on a misunderstanding of the governing rule of law”).

B

In holding that the District Court disregarded the presumption of legislative good faith and improperly reversed the burden of proof, we do not suggest either that the intent of the 2011 Legislature is irrelevant or that the plans enacted in 2013 are unassailable because they were previously adopted on an interim basis by the Texas court. Rather, both the intent of the 2011 Legislature and the court’s adoption of the interim plans are relevant to the extent that they naturally give rise to—or tend to refute—inferences regarding the intent of the 2013 Legislature. They must be weighed together with any other direct and circumstantial evidence of that Legislature’s intent. But when all the relevant evidence in the record is taken into account, it is plainly insufficient to prove that the 2013 Legislature acted in bad faith and engaged in intentional discrimination.[19] See, e.g., Ricci v. DeStefano, 557 U. S. 557, 585 (2009); McCleskey v. Zant, 499 U. S. 467, 497 (1991). There is thus no need for any further prolongation of this already protracted litigation.

The only direct evidence brought to our attention suggests that the 2013 Legislature’s intent was legitimate. It wanted to bring the litigation about the State’s districting plans to an end as expeditiously as possible. The attorney general advised the Legislature that the best way to do this was to adopt the interim, court-issued plans. The sponsor of the 2013 plans voiced the same objective, and the Legislature then adopted the court-approved plans.

On its face, this explanation of the Legislature’s intent is entirely reasonable and certainly legitimate. The Legislature had reason to know that any new plans it devised were likely to be attacked by one group of plaintiffs or another. (The plaintiffs’ conflicting positions with regard to some of the districts in the plans now before us bear this out.) Litigating districting cases is expensive and time consuming, and until the districts to be used in the next election are firmly established, a degree of uncertainty clouds the electoral process. Wishing to minimize these effects is understandable and proper.

The court below discounted this direct evidence, but its reasons for doing so are not sound. The court stated that the “strategy” of the 2013 Legislature was to “insulate [the plans] from further challenge, regardless of [the plans’] legal infirmities.” 274 F. Supp. 3d, at 650; see also id., at 651, n. 45. But there is no evidence that the Legislature’s aim was to gain acceptance of plans that it knew were unlawful.[20] Indeed, there is no evidence that the Legislature thought that the plans were invalid—and as we will explain, the Legislature had sound reasons to believe just the opposite.[21]

The District Court found it significant that the Legislature must have realized that enacting the interim plans would not “end the litigation,” because it knew that at least some plaintiffs would pursue their challenges anyway. Id., at 651, n. 45. But even if, as seems likely, the Legislature did not think that all the plaintiffs would immediately abandon all their claims, it does not follow that the Legislature was insincere in stating that it adopted the court-approved plan with the aim of bringing the litigation to a close. It was reasonable for the Legislature to think that approving the court-approved plans might at least reduce objections and thus simplify and expedite the conclusion of the litigation.[22] That MALDEF, counsel for one of the plaintiff groups, testified in favor of the plans is evidence that the Legislature’s objective was reasonable. C. J. S. 436a–439a.

Not only does the direct evidence suggest that the 2013 Legislature lacked discriminatory intent, but the circumstantial evidence points overwhelmingly to the same conclusion. Consider the situation when the Legislature adopted the court-approved interim plans. First, the Texas court had adopted those plans, and no one would claim that the court acted with invidious intent when it did so. Second, the Texas court approved those plans only after reviewing them and modifying them as required to comply with our instructions. Not one of the judges on that court expressed the view that the plans were unlawful. Third, we had directed the Texas court to make changes in response to any claims under the Equal Protection Clause and §2 of the Voting Rights Act if those claims were merely likely to prevail. Perry, 565 U. S., at 394. And the Texas court was told to accommodate any claim under §5 of the VRA unless it was “insubstantial.” Id., at 395. Fourth, the Texas court had made a careful analysis of all the claims, had provided a detailed examination of individual districts, and had modified many districts. Its work was anything but slapdash. All these facts gave the Legislature good reason to believe that the court-approved interim plans were legally sound.

Is there any evidence from which a contrary inference can reasonably be drawn? Appellees stress the preliminary nature of the Texas court’s approval of the interim plans, and as we have said, that fact is relevant. But in light of our instructions to the Texas court and the care with which the interim plans were developed, the court’s approval still gave the Legislature a sound basis for thinking that the interim plans satisfied all legal requirements.

The court below and the dissent infer bad faith because the Legislature “pushed the redistricting bills through quickly in a special session.” 274 F. Supp. 3d, at 649. But we do not see how the brevity of the legislative process can give rise to an inference of bad faith—and certainly not an inference that is strong enough to overcome the presumption of legislative good faith (a concept to which the dissent pays only the briefest lip service, post, at 21). The “special session” was necessary because the regular session had ended. As explained, the Legislature had good reason to believe that the interim plans were sound, and the adoption of those already-completed plans did not require a prolonged process. After all, part of the reason for adopting those plans was to avoid the time and expense of starting from scratch and leaving the electoral process in limbo while that occurred.[23]

The District Court and the dissent also err when they charge that Representative Darby, the chair of the Texas House Redistricting Committee at the time in question, “ ‘willfully ignored those who pointed out deficiencies’ ” in the plans. Post, at 23 (quoting 274 F. Supp. 3d, at 651, n. 45). This accusation is not only misleading, it misses the point. The Legislature adopted the interim plans in large part because they had the preliminary approval of the District Court, and Darby was open about the fact that he wanted to minimize amendments to the plans for that reason. See, e.g., Joint Exh. 17.3, pp. S1–S2. That Darby generally hoped to minimize amendments—so that the plans would remain legally compliant—hardly shows that he, or the Legislature, acted with discriminatory intent. In any event, it is misleading to characterize this attitude as “willfu[l] ignor[ance].” The record shows that, al- though Darby hoped to minimize amendments, he did not categorically refuse to consider changes. This is illustrated by his support for an amendment to HD90, which was offered by the then-incumbent, Democrat Lon Burnam, precisely because it fixed an objection raised by the Mexican-American Legal Caucus (MALC) that the district’s Latino population was too low. 267 F. Supp. 3d, at 790.[24]

The Texas court faulted the 2013 Legislature for failing to take into account the problems with the 2011 plans that the D. C. court identified in denying preclearance, ibid., but the basis for that criticism is hard to understand. One of the 2013 Legislature’s principal reasons for adopting the court-approved plans was to fix the problems identified by the D. C. court. The attorney general advised the Legislature to adopt the interim plans because he thought that was the “best way to remedy the violations found by the D. C. court.” C. J. S. 432a. Chairman Darby similarly stated that the 2013 plans fixed the errors found by the D. C. court, Tr. 1498, 1584–1585 (July 14, 2017), as did Senator Seliger, Joint Exh. 26.2, p. A–5.

There is nothing to suggest that the Legislature proceeded in bad faith—or even that it acted unreasonably—in pursuing this strategy. Recall that we instructed the Texas court, in developing the interim plans, to remedy any §5 claim that was “not insubstantial.” Perry, 565 U. S., at 395. And that is just what the interim plans, which the Legislature later enacted, attempted to do. For instance, the D. C. court held that the congressional plan had one too few “ability to elect” districts for Latinos, largely because of changes to CD23, Texas, 887 F. Supp. 2d, at 156–159; the interim plan (and, by extension, the 2013 plan) amended CD23, C. J. S. 397a–399a. Similarly, in the plan for the Texas House, the D. C. court found §5 retrogression with respect to HD35, HD117, and HD149, Texas, supra, at 167–175, and all of those districts were changed in the 2013 plans, H. J. S. 305a–307a, 312a.

Although the D. C. court found that the 2011 Legislature acted with discriminatory intent in framing the congressional plan, that finding was based on evidence about districts that the interim plan later changed. The D. C. court was concerned about the intent reflected in the drawing of CDs 9, 18, and 30, but all those districts were amended by the Texas court. Texas, supra, at 159–160; C. J. S. 406a–408a. With respect to the plan for the Texas House, the D. C. court made no intent findings, but its areas of concern were generally addressed by the Texas court and the 2013 plans. Compare Texas, supra, at 178 (noting evidence of unlawful intent in HD117), with H. J. S. 307a (amending HD117).[25]

It is indicative of the District Court’s mistaken approach that it inferred bad faith from Texas’s decision to take an appeal to this Court from the D. C. court’s decision denying preclearance. See 274 F. Supp. 3d, at 649 (“Defendants did not accept [these findings] and instead appealed to the Supreme Court”). Congress gave the State the right to appeal, and no bad motive can be inferred from its decision to make use of this right—unless of course the State had no reasonable grounds for appeal. Before our decision in Shelby County mooted Texas’s appeal to this Court from the D. C. court’s preclearance decision, Texas filed a jurisdictional statement claiming that the D. C. court made numerous errors, but the Texas court made no attempt to show that Texas’s arguments were frivolous.

As a final note, appellees assert that the 2013 Legislature should have either defended the 2011 plans in litigation or gone back to the drawing board and devised en- tirely new plans, Brief for Appellees in No. 17–626, at 45, but there is no reason why the Legislature’s options should be limited in this way. It was entirely permissible for the Legislature to favor a legitimate option that promised to simplify and reduce the burden of litigation. That the Legislature chose this course is not proof of discriminatory intent.

IV

Once the Texas court’s intent finding is reversed, there remain only four districts that were invalidated on alternative grounds. For three of these districts, the District Court relied on the “effects” test of §2. We reverse as to each of these, but we affirm the District Court’s final holding that HD90 is a racial gerrymander.

A

To make out a §2 “effects” claim, a plaintiff must establish the three so-called “Gingles factors.” These are (1) a geographically compact minority population sufficient to constitute a majority in a single-member district, (2) political cohesion among the members of the minority group, and (3) bloc voting by the majority to defeat the minority’s preferred candidate. Gingles, 478 U. S., at 48–51; LULAC, 548 U. S., at 425. If a plaintiff makes that showing, it must then go on to prove that, under the totality of the circumstances, the district lines dilute the votes of the members of the minority group. Id., at 425–426.

1

The Texas court held that CD27 violates §2 of the VRA because it has the effect of diluting the votes of Latino voters in Nueces County. C. J. S. 191a. CD27 is anchored in Nueces County (home to Corpus Christi) and follows the Gulf of Mexico to the northeast before taking a turn inland to the northwest in the direction of Austin. Nueces County contains a Latino population of roughly 200,000 (a little less than one-third the size of an ideal Texas congressional district), and the court held that the Nueces County Latinos should have been included in a Latino opportunity district, rather than CD27, which is not such a district. The court found that an area centered on Nueces County satisfies the Gingles factors and that, under the totality of the circumstances, the placement of the Nueces County Latinos in CD27 deprives them of the equal opportunity to elect candidates of their choice. C. J. S. 181a–195a.

The problem with this holding is that plaintiffs could not establish a violation of §2 of the VRA without showing that there is a “ ‘possibility of creating more than the existing number of reasonably compact’ ” opportunity districts. LULAC, supra, at 430. And as the Texas court itself found, the geography and demographics of south and west Texas do not permit the creation of any more than the seven Latino opportunity districts that exist under the current plan. 274 F. Supp. 3d, at 684, and n. 85.

Attempting to get around this problem, the Texas court relied on our decision in LULAC, but it misapplied our holding. In LULAC, we held that the State should have created six proper Latino opportunity districts but instead drew only five. 548 U. S., at 435. Although the State claimed that the plan actually included a sixth opportunity district, that district failed to satisfy the Gingles factors. 548 U. S., at 430. We held that a “State’s creation of an opportunity district for those without a §2 right offers no excuse for its failure to provide an opportunity district for those with a §2 right.” Ibid.

Here, the Texas court concluded that Texas committed the same violation as in LULAC: It created “an opportu-nity district for those without a §2 right” (the Latinos in CD35), while failing to create such a district “for those with a §2 right” (the Latinos of Nueces County). Ibid. This holding is based on a flawed analysis of CD35.

CD35 lies to the north of CD27 and runs along I–35 from San Antonio up to Austin, the center of Travis County. In the District Court’s view, the Latinos of CD35 do not have a §2 right because one of the Gingles factors, majority bloc voting, is not present. The Court reached this conclusion because the non-Latino voters of Travis County tend to favor the same candidates as the great majority of Latinos. There are two serious problems with the District Court’s analysis.

First, the Court took the wrong approach in evaluating the presence of majority bloc voting in CD35. The Court looked at only one, small part of the district, the portion that falls within Travis County. 274 F. Supp. 3d, at 683; C. J. S. 175a–176a. But Travis County makes up only 21% of the district. We have made clear that redistricting analysis must take place at the district level. Bethune-Hill, 580 U. S., at ___ (slip op., at 12). In failing to perform that district-level analysis, the District Court went astray.

Second, here, unlike in LULAC, the 2013 Legislature had “good reasons” to believe that the district at issue (here CD35) was a viable Latino opportunity district that satisfied the Gingles factors. CD35 was based on a concept proposed by MALDEF, C. J. S. Findings 315a–316a, and the Latino Redistricting Task Force (a plaintiff group) argued that the district is mandated by §2. C. J. S. 174a. The only Gingles factor disputed by the court was majority bloc voting, and there is ample evidence that this factor is met. Indeed, the court found that majority bloc voting exists throughout the State. C. J. S. Findings 467a. In addition, the District Court extensively analyzed CD35 in 2012 and determined that it was likely not a racial gerrymander and that even if it was, it likely satisfied strict scrutiny. C. J. S. 415a. In other words, the 2013 Legislature justifiably thought that it had placed a viable opportunity district along the I–35 corridor.

2

The District Court similarly erred in holding that HD32 and HD34 violate §2. These districts make up the entirety of Nueces County, which has a population that is almost exactly equal to twice the population of an ideal Texas House district. (It can fit 2.0295 ideal districts. H. J. S. Findings 91a.) In 2010, Latinos made up approximately 56% of the voting age population of the county. Ibid. The 2013 plan created two districts that lie wholly within the county; one, HD34, is a Latino opportunity district, but the other, HD32, is not. 267 F. Supp. 3d, at 767.

Findings made by the court below show that these two districts do not violate §2 of the Voting Rights Act. Under Gingles, the ultimate question is whether a districting decision dilutes the votes of minority voters, see LULAC, supra, at 425–426, and it is hard to see how this standard could be met if the alternative to the districting decision at issue would not enhance the ability of minority voters to elect the candidates of their choice.

The only plaintiff that pressed a §2 claim with respect to HD32 and HD34 was MALC, 267 F. Supp. 3d, at 767, and as the District Court recognized, that group’s own expert determined that it was not possible to divide Nueces County into more than one performing Latino district. In his analysis, the expert relied on Nueces County election returns for statewide elections between 2010 and 2016. Id., at 775–776. Based on this data, he calculated that when both HD32 and HD34 were maintained as Latino-majority districts, one performed for Latinos in only 7 out of 35 relevant elections, and the other did so in none of the 35 elections. Ibid. In order to create two performing districts in that area, it was necessary, he found, to break county lines in multiple places, id., at 778, but the District Court held that “breaking the County Line Rule” in the Texas Constitution, see Art. III, §26, to “remove Anglos and incorporate even more Hispanics to improve electoral outcomes goes beyond what §2 requires.” 267 F. Supp. 3d, at 783. So if Texas could not create two performing districts in Nueces County and did not have to break county lines, the logical result is that Texas did not dilute the Latino vote.

The court refused to accept this conclusion, but its reasons for doing so cannot stand up. As an initial matter, the court thought that the two districts would have to be redrawn based on its finding regarding the intent of the 2013 Legislature,[26] and it therefore deferred a final decision on the §2 issue and advised the plaintiffs to consider at the remedial phase of the case whether they preferred to have two districts that might not perform or just one safe district. Id., at 783. The court’s decision cannot be sustained on this ground, since its finding of discriminatory intent is erroneous.

The only other reason provided by the court was the observation that MALC “failed to show” that two majority-Latino districts in Nueces County would not perform. Id., at 782. This observation twisted the burden of proof beyond recognition. It suggested that a plaintiff might succeed on its §2 claim because its expert failed to show that the necessary factual basis for the claim could not be established.[27] Courts cannot find §2 effects violations on the basis of uncertainty. In any event, if even the District Court remains unsure how to draw these districts to comply with §2 (after six years of litigation, almost a dozen trials, and numerous opinions), the Legislature surely had the “ ‘broad discretion’ ” to comply as it reasonably saw fit in 2013, LULAC, supra, at 429.

The dissent charges us with ignoring the District Court’s “ ‘intensely local appraisal” ’ of Nueces County, post, at 44, but almost none of the “findings” that the District Court made with respect to HD32 and HD34 referred to present local conditions, and none cast any significant light on the question whether another opportunity district is possible at the present time. For instance, what the dissent describes as Texas’s “long ‘history of voting-related discrimination,’ ” ibid., in no way undermines—or even has any logical bearing on—the conclusions reached by MALC’s expert about whether Latino voters would have a real opportunity to elect the candidates of their choice if the county was divided into two districts with narrow majorities of Latino citizens of voting age. The same is true with respect to the District Court’s findings regarding racially polarized voting in the county and Latinos’ “continuing pattern of disadvantage relative to” non-Latinos. 267 F. Supp. 3d, at 779 (internal quotation marks omitted). Perhaps recognizing as much, both the District Court and the dissent point to the anticipated future growth in the percentage of eligible voters of Latino descent, but the districts now at issue would not necessar-ily be used beyond 2020, after which time the 2020 census would likely require redistricting once again.

B

HD90 is a district in Tarrant County that, unlike the other districts at issue in this appeal, was not copied from the District Court’s interim plans. Instead, the 2013 Legislature substantially modified the district developed by the District Court, and the District Court held that the 2013 Legislature’s creation is an invalid racial gerrymander. 267 F. Supp. 3d, at 794.

In drawing HD90, the Legislature was pulled in opposite directions by competing groups. In 2011, the Legislature, responding to pressure from MALDEF, increased the Latino population of the district in an effort to make it a Latino opportunity district. H. J. S. Findings 258a–262a. In the process of doing so, the Legislature moved the community of Como, which is predominantly African-American, out of the district. But Como residents and the member of the Texas House who represented the district, Lon Burnam, objected, and in 2013, the Legislature moved Como back into the district. 267 F. Supp. 3d, at 788–789. That change was opposed by MALC because it decreased the Latino population below 50%. App. 398a–399a. So the Legislature moved Latinos into the district to bring the Latino population back above 50%. 267 F. Supp. 3d, at 789–790.

In light of these maneuvers, Texas does not dispute that race was the predominant factor in the design of HD90, but it argues that this was permissible because it had “ ‘good reasons to believe’ ” that this was necessary to satisfy §2 of the Voting Rights Act.” Bethune-Hill, 580 U. S., at ___ (slip op., at 14).

Texas offers two pieces of evidence to support its claim. The first—that one of the plaintiffs, MALC, demanded as much—is insufficient. A group that wants a State to create a district with a particular design may come to have an overly expansive understanding of what §2 demands. So one group’s demands alone cannot be enough.

The other item of evidence consists of the results of the Democratic primaries in 2012 and 2014. In 2012, Representative Burnham, who was not the Latino candidate of choice, narrowly defeated a Latino challenger by 159 votes. And in 2014, the present representative, Ramon Romero, Jr., beat Burnam by 110 votes. See Brief for Appellants 70. These election returns may be suggestive, but standing alone, they were not enough to give the State good reason to conclude that it had to alter the district’s lines solely on the basis of race. And putting these two evidentiary items together helps, but it is simply too thin a reed to support the drastic decision to draw lines in this way.

We have previously rejected proffers of evidence that were at least as strong as Texas’s here. For example, in Cooper, 581 U. S., at ___ (slip op., at 11), we analyzed North Carolina’s justification for deliberately moving “African-American voters” into a district to “ensure . . . the district’s racial composition” in the face of its expansion in size. North Carolina argued that its race-based decisions were necessary to comply with §2, but the State could point to “no meaningful legislative inquiry” into “whether a new, enlarged” district, “created without a focus on race, . . . could lead to §2 liability.” Id., at ___ (slip op., at 15). North Carolina pointed to two expert reports on “voting patterns throughout the State,” but we rejected that evidence as insufficient. Id., at ___, n. 5 (slip op., at 15, n. 5). Here, Texas has pointed to no actual “legislative inquiry” that would establish the need for its manipulation of the racial makeup of the district.

By contrast, where we have accepted a State’s “good reasons” for using race in drawing district lines, the State made a strong showing of a pre-enactment analysis with justifiable conclusions. In Bethune-Hill, the State established that the primary mapdrawer “discussed the district with incumbents from other majority-minority districts[,] . . . considered turnout rates, the results of the recent contested primary and general elections,” and the district’s large prison population. 580 U. S., at ___ (slip op., at 15). The State established that it had performed a “functional analysis,” and acted to achieve an “informed bipartisan consensus.” Id., at ___ (slip op., at 14). Texas’s showing here is not equivalent.

Perhaps Texas could have made a stronger showing, but it is the State’s burden to prove narrow tailoring, and it did not do so on the record before us. We hold that HD90 is an impermissible racial gerrymander. On remand, the District Court will have to consider what if any remedy is appropriate at this time.

*  *  *

Except with respect to one Texas House district, we hold that the court below erred in effectively enjoining the use of the districting maps adopted by the Legislature in 2013. We therefore reverse with respect to No. 17–586; reverse in part and affirm in part with respect to No. 17–626; and remand for proceedings consistent with this opinion.

It is so ordered.

SUPREME COURT OF THE UNITED STATES

_________________

Nos. 17–586 and 17–626

_________________

GREG ABBOTT, GOVERNOR OF TEXAS, ET AL., APPELLANTS

17–586v.

SHANNON PEREZ, ET AL.

GREG ABBOTT, GOVERNOR OF TEXAS, ET AL., APPELLANTS

17–626v.

SHANNON PEREZ, ET AL.

on appeals from the united states district court for the western district of texas

[June 25, 2018]

Justice Thomas, with whom Justice Gorsuch joins, concurring.

I adhere to my view that §2 of the Voting Rights Act of 1965 does not apply to redistricting. See Cooper v. Harris, 581 U. S. ___, ___ (2017) (concurring opinion) (slip op., at 1) (citing Holder v. Hall, 512 U. S. 874, 922–923 (1994) (Thomas, J., concurring in judgment)). Thus, §2 cannot provide a basis for invalidating any district, and it cannot provide a justification for the racial gerrymander in House District 90. Because the Court correctly applies our precedents and reaches the same conclusion, I join its opinion in full.

SUPREME COURT OF THE UNITED STATES

_________________

Nos. 17–586 and 17–626

_________________

GREG ABBOTT, GOVERNOR OF TEXAS, ET AL., APPELLANTS

17–586v.

SHANNON PEREZ, ET AL.

GREG ABBOTT, GOVERNOR OF TEXAS, ET AL., APPELLANTS

17–626v.

SHANNON PEREZ, ET AL.

on appeals from the united states district court for the western district of texas

[June 25, 2018]

Justice Sotomayor, with whom Justice Ginsburg, Justice Breyer, and Justice Kagan join, dissenting.

The Court today goes out of its way to permit the State of Texas to use maps that the three-judge District Court unanimously found were adopted for the purpose of preserving the racial discrimination that tainted its previous maps. In reaching its desired result, the majority commits three fundamental errors along the way.

First, the majority disregards the strict limits of our appellate jurisdiction and reads into the District Court orders a nonexistent injunction to justify its premature intervention. Second, the majority indulges Texas’ distorted reading of the District Court’s meticulous orders, mistakenly faulting the court for supposedly shifting the burden of proof to the State to show that it cured the taint of past discrimination, all the while ignoring the clear language and unambiguous factual findings of the orders below. Third, the majority elides the standard of review that guides our resolution of the factual disputes in these appeals—indeed, mentioning it only in passing—and selectively parses through the facts. As a result of these errors, Texas is guaranteed continued use of much of its discriminatory maps.

This disregard of both precedent and fact comes at serious costs to our democracy. It means that, after years of litigation and undeniable proof of intentional discrimination, minority voters in Texas—despite constituting a majority of the population within the State—will continue to be underrepresented in the political process. Those voters must return to the polls in 2018 and 2020 with the knowledge that their ability to exercise meaningfully their right to vote has been burdened by the manipulation of district lines specifically designed to target their communities and minimize their political will. The fundamental right to vote is too precious to be disregarded in this manner. I dissent.

I

A

The first obstacle the majority faces in its quest to intervene in these cases is jurisdictional. The statute that governs our jurisdiction over these appeals is 28 U. S. C. §1253, which provides that “any party may appeal to the Supreme Court from an order granting or denying . . . an interlocutory or permanent injunction in any civil action, suit or proceeding required by any Act of Congress to be heard and determined by a district court of three judges.” Unlike the more typical certiorari process, for cases falling within §1253, appellate review in this Court is mandatory. That is why, until today, this Court has repeatedly recognized and adhered to a “long-established rule” requiring “strict construction” of this jurisdictional statute “to protect our appellate docket.” Stainback v. Mo Hock Ke Lok Po, 336 U. S. 368, 375, 378 (1949); see, e.g., Gonzalez v. Automatic Employees Credit Union, 419 U. S. 90, 98 (1974) (noting that “only a narrow construction” of our jurisdiction under §1253 “is consonant with the overriding policy, historically encouraged by Congress, of minimizing the mandatory docket of this Court in the interests of sound judicial administration”); Gunn v. University Comm. to End War in Viet Nam, 399 U. S. 383, 387 (1970) (similar); Goldstein v. Cox, 396 U. S. 471, 477–478 (1970) (rejecting a construction of §1253 that would “involve an expansion of [our] mandatory appellate jurisdiction,” even where the statutory text “is subject to [that] construction,” in light of “canon of construction” requiring that §1253 be “narrowly construed”); Phillips v. United States, 312 U. S. 246, 248–250 (1941) (explaining that §1253 is an “exceptional procedure” and that “inasmuch as this procedure . . . brings direct review of a district court to this Court, any loose construction . . . would defeat the purposes of Congress . . . to keep within narrow confines our appellate docket”).

In line with that command, this Court has held that a ruling on the merits will not suffice to invoke our mandatory appellate jurisdiction in the absence of an order granting or denying an injunction. In fact, even if a three-judge district court unequivocally indicates that a state law must be enjoined as it stands, we have required more before accepting mandatory review. For example, the Court in Gunn found no jurisdiction where the three-judge District Court held that a Texas disturbing-the-peace statute was “ ‘impermissibly and unconstitutionally broad,’ ” concluded that the plaintiffs were “ ‘entitled to their declaratory judgment to that effect, and to injunctive relief against the enforcement of [the statute] as now worded, insofar as it may affect the rights guaranteed under the First Amendment,’ ” and stayed the mandate to allow the State to, “ ‘if it so desires, enact such disturbing-the-peace statute as will meet constitutional requirements.’ ” 399 U. S., at 386. Despite the District Court’s resolution of the merits and its clear indication that, unless amended, the disturbing-the-peace statute would be enjoined, this Court dismissed an appeal from the State for want of jurisdiction, concluding that the District Court merely wrote a “rather discursive per curiam opinion” and “there was no order of any kind either granting or denying an injunction—interlocutory or permanent.” Id., at 387. The Court explained that, in addition to the congressional command to “ ‘keep within narrow confines our appellate docket,’ ” other “policy considerations” counseled limiting “our power of review,” including “that until a district court issues an injunction, or enters an order denying one, it is simply not possible to know with any certainty what the court has decided.” Id., at 387–388. Those considerations, the Court thought, were “conspicuously evident” in that case, where the opinion did not specify, for instance, exactly what was to be enjoined or against whom the injunction would run. Id., at 388.

Similarly, Whitcomb v. Chavis, 403 U. S. 124 (1971), concerned a redistricting challenge in which a three-judge District Court held that “a redistricting of [the challenged county was] necessitated” and “that the evidence adduced . . . and the additional apportionment requirements set forth by the Supreme Court call[ed] for a redistricting of the entire state as to both houses of the General Assembly,” Chavis v. Whitcomb, 305 F. Supp. 1364, 1391 (SD Ind. 1969). Recognizing “that the federal judiciary functions within a system of federalism which entrusts the responsibility of legislative apportionment and districting primarily to the state legislature,” the District Court afforded the Governor “a reasonable opportunity to call a Special Session of the General Assembly of the State of Indiana so that it may enact legislation to redistrict the State and reapportion the legislative seats in the General Assembly in accordance with federal constitutional requirements and in compliance with [its] opinion.” Id., at 1392. The District Court gave the State a little over two months to enact new statutes “to remedy the improper districting and malapportionment.” Ibid. When the Governor appealed from that order, this Court dismissed for want of jurisdiction because “at [the] time no judgment had been entered and no injunction had been granted or denied.” 403 U. S., at 138, n. 19. The findings of liability on the merits and the unequivocal indication that the redistricting and malapportionment violations had to be remedied were not enough.

B

Straightforward application of this precedent compels the conclusion that this Court lacks jurisdiction over these appeals. Here, Texas appeals from two orders entered by the three-judge District Court on August 15 and 24, 2017. Those orders concern the constitutional and statutory challenges to Texas’ State House and federal congressional redistricting plans, enacted by the Texas Legislature (hereinafter Legislature) in 2013 (hereinafter the 2013 maps). As relevant here, the orders concerned Texas House districts in Bell County (HD54 and HD55), Dallas County (HD103, HD104, and HD105), Nueces County (HD32 and HD34), and Tarrant County (HD90), as well federal congressional districts encompassing Nueces County (CD27) and parts of Travis County (CD35). The District Court concluded that plaintiffs had proved intentional discrimination as to HD54, HD55, HD103, HD104, HD105, HD32, HD34, and CD27.[1] It also concluded that plaintiffs had proved a “results” violation under §2 of the Voting Rights Act as to HD32, HD34, and CD27,[2] and had established a racial gerrymandering claim as to HD90 and CD35.[3]

Having ruled on the challengers’ statutory and constitutional claims, the District Court stated that all but one of the “violations must be remedied by either the Texas Legislature or [the District] Court.” 274 F. Supp. 3d 624, 686 (WD Tex. 2017); see also 267 F. Supp. 3d 750, 795 (WD Tex. 2017).[4] With respect to the §2 results violation concerning HD32 and HD34, however, the District Court noted that it had yet to decide “whether §2 requires a remedy for this results violation.” Id., at 783, 795. The District Court then ordered “the [Texas] Office of the Attorney General to file a written advisory within three business days stating whether the Legislature intends to take up redistricting in an effort to cure these violations and, if so, when the matter will be considered.” 274 F. Supp. 3d, at 686; see also 267 F. Supp. 3d, at 795. The court went on: “If the Legislature does not intend to take up redistricting, the [District] Court will hold a hearing to consider remedial plans” on September 5 and 6, 2017, respecting the congressional and Texas House districts. 274 F. Supp. 3d, at 686–687; see also 267 F. Supp. 3d, at 795. “In preparation for the hearing[s],” the District Court ordered the parties to confer and to “take immediate steps to consult with their experts and mapdrawers and prepare” maps to present at those hearings. 274 F. Supp. 3d, at 687; 267 F. Supp. 3d, at 795.

The District Court went no further. Though there had been a determination on the merits that Texas violated both the Equal Protection Clause and §2 of the Voting Rights Act with respect to a number of districts in the 2013 maps, the District Court did not enjoin use of the 2013 maps for the upcoming 2018 elections. For instance, with respect to the congressional map, the District Court explained that its order “only partially addresse[d]” the challengers’ claims, as it had “bifurcated the remedial phase” from the merits phase. 274 F. Supp. 3d, at 687. Importantly, in denying Texas’ motions for a stay, the District Court took care to make abundantly clear the scope of its orders: “Although the [District] Court found violations [in the congressional and Texas House maps], the [District] Court has not enjoined [their] use for any upcoming elections.” App. 134a–136a.

That is the end of the inquiry under our precedent, as our past cases are directly on point. Like in Gunn and Whitcomb, the District Court issued a ruling on the merits against the State. Like in Gunn and Whitcomb, the District Court was clear that those violations required a remedy. Like in Gunn and Whitcomb, the District Court stayed its hand and did not enter an injunction, instead allowing the State an opportunity to remedy the violations. Therefore, like in Gunn and Whitcomb, this Court lacks jurisdiction under §1253 because there is “no order of any kind either granting or denying an injunction—interlocutory or permanent.” Gunn, 399 U. S., at 387.[5]

C

1

Despite this precedent, the majority nonetheless concludes that our intervention at this early stage is not only authorized, but mandatory. None of the justifications that the majority offers for deviating from our precedent is persuasive.

The majority justifies its jurisdictional overreach by holding that §1253 mandates appellate review in this Court if a three-judge district court order “has the ‘practical effect’ of granting or denying an injunction.” Ante, at 12. It reasons that the Court has “previously made clear that where an order has the ‘practical effect’ of granting or denying an injunction, it should be treated as such for purposes of appellate jurisdiction.” Ibid. That reasoning, however, has no application here. Whereas this Court has applied the “practical effect” rule in the context of the courts of appeals’ appellate jurisdiction under 28 U. S. C. §1292(a)(1), it has never applied it to questions of its own mandatory appellate docket under §1253. That explains why the only cases the majority can round up to support its position concern jurisdiction under §1292(a)(1). Ante, at 12 (citing Carson v. American Brands, Inc., 450 U. S. 79, 83–84 (1981), and Gulfstream Aerospace Corp. v. Mayacamas Corp., 485 U. S. 271, 287–288 (1988)).

This distinction matters a great deal. Courts of appeals generally have jurisdiction over direct appeals from the district courts. See 15A C. Wright, A. Miller, & E. Cooper, Federal Practice and Procedure §3901, p. 13 (3d ed. 2005) (“Courts of appeals jurisdiction extends to nearly every action that might be taken by a district court”). In contrast, exercising mandatory review over direct appeals in this Court is a truly “exceptional procedure,” Phillips, 312 U. S., at 248, in no small part due to our “necessarily finite docket,” 16B Wright, Federal Practice and Procedure §4003, at 19. Reading §1253 broadly risks transforming that exceptional procedure into a routine matter, when our precedent commands a strict construction precisely so that we can “ ‘keep within narrow confines our appellate docket.’ ” Goldstein, 396 U. S., at 478.

Brushing that distinction aside, the majority contends that “we also construe Section 1292(a)(1) ‘narrowly,’ ” and have referred to the statutes as “ ‘analogous.’ ” Ante, at 13. True, but that is no response to the jurisdictional obstacle of §1253. The command from our precedent is not simply one to undertake the same narrow interpretation as we do for §1292(a)(1). Rather, our “long-established rule” requires “strict construction” of §1253, Stainback, 336 U. S., at 378, so that even where the statutory text could be read to expand our mandatory appellate docket, this Court will not adopt that reading if a narrower construction is available, Goldstein, 396 U. S., at 477–478. That “strict construction” rule exists for a purpose specific to this Court: to protect our “carefully limited appellate jurisdiction.” Board of Regents of Univ. of Tex. System v. New Left Ed. Project, 404 U. S. 541, 543 (1972). Unlike the courts of appeals, which hear cases on mandatory jurisdiction regularly, this Court hears cases on mandatory jurisdiction only rarely. The majority nowhere grapples with that vital contextual distinction between §1253 and §1292(a)(1). Nor does the majority acknowledge that, in interpreting §1253, this Court has itself recognized that distinction, noting that “this Court above all others must limit its review of interlocutory orders.” Goldstein, 396 U. S., at 478 (emphasis added).

2

Looking to escape that pitfall in its reasoning, the majority turns to the text of the two jurisdictional statutes. But the text provides no refuge for its position. The majority first states that “the relevant language in the two provisions is nearly identical.” Ante, at 13. But whereas §1253 provides for appeal “from an order granting or denying . . . an interlocutory or permanent injunction,” §1292(a)(1) provides for appeal from “[i]nterlocutory orders . . . granting, continuing, modifying, refusing or dissolving injunctions, or refusing to dissolve or modify injunctions.” It is a stretch, to say the least, to characterize these provisions as “nearly identical.” Ante, at 13.

Next, the majority contends that §1253 and §1292(a)(1) are “textually interlocked,” ante, at 14, in that §1292(a)(1) provides for appeal to the courts of appeals, “except where a direct review may be had in the Supreme Court.” In its view, this demonstrates that the “practical effect” rule must apply under §1253. The majority reasons that “the consequences would be unfortunate and strange” otherwise, imagining that an order from a three-judge district court that had the practical effect of an injunction but did not invoke §1253 jurisdiction would “appear to be appealable to one of the courts of appeals” in light of the “except” clause, a result “Congress surely did not intend” given that it took “pains to provide for review in this Court, and not in the courts of appeals, of three-judge court orders granting injunctions.” Ante, at 14.

This reasoning rests on a mistaken premise. Congress did not provide for review of every three-judge court order in this Court. It provided for review of only certain narrow categories of orders, i.e., those granting or denying an injunction. There is nothing “unfortunate” or “strange” about the proposition that orders from a three-judge court that do not fall within these narrow categories of actions made directly appealable to this Court can be appealed only to the courts of appeals. In fact, this Court itself has recognized as much. See, e.g., Rockefeller v. Catholic Medical Center of Brooklyn & Queens, Inc., 397 U. S. 820 (1970) (per curiam) (“The judgment appealed from does not include an order granting or denying an interlocutory or permanent injunction and is therefore not appealable to this Court under 28 U. S. C. §1253.  The judgment of the District Court is vacated and the case is remanded to that court so that it may enter a fresh decree from which timely appeal may be taken to the Court of Appeals” (citation omitted)); see also Mitchell v. Donovan, 398 U. S. 427, 431–432 (1970) (per curiam) (concluding that “this Court lacks jurisdiction of the appeal” under §1253 and directing “the District Court [to] enter a fresh order . . . thus affording the appellants an opportunity to take a timely appeal to the Court of Appeals”).[6] And to the extent a party prematurely appeals to the court of appeals an order that would otherwise fall within §1253, e.g., if Texas had appealed the August 15 and 24 orders to the Court of Appeals for the Fifth Circuit, that court surely will be more than capable of identifying as much and instructing the party to wait for an actual injunction before bringing an appeal to this Court.

3

The majority attempts to bolster its jurisdictional conclusion with a passing reference to the “valuable purpose” served by the “practical effect” rule, i.e., preventing district courts from manipulating proceedings by avoiding labeling their orders as “injunctions.” Ante, at 12–13. Notably, the majority cites no evidence for the proposition that district courts are engaging in any kind of manipulation. Nor is there any indication that the District Court here attempted to manipulate the proceedings by shielding its orders from appellate review. Instead, the District Court carefully adhered to a common practice in cases implicating important state interests, staying its hand as to the remedy to allow the State an opportunity to act, as happened in Gunn and Whitcomb.

More important, the majority ignores the “valuable purposes” served by the longstanding rule requiring strict construction of §1253. Not only does it comply with the congressional command to “ ‘keep within narrow confines our appellate docket,’ ” but without strict enforcement of the requirement that an order grant or deny an injunction, “it is simply not possible to know with any certainty what the court has decided.” Gunn, 399 U. S., at 387–388. Such clarity “is absolutely vital in a case where a federal court is asked to nullify a law duly enacted by a sovereign State.” Id., at 389. Orders coming to this Court on direct appeal under the “practical effect” rule will more often than not lack that clarity.

In these cases, for instance, what does the majority read the “practical effect” of the orders to have been with respect to HD32 and HD34? The District Court held that the challengers had “not proven that §2 requires breaking the County Line Rule” in the Texas Constitution, Art. III, but that “§2 could require” drawing two majority-HCVAP[7] districts. 267 F. Supp. 3d, at 783, 795. Does the majority read that to mean that the §2 results violation could potentially go without a remedy? If so, there would have been no obstacle to use of the 2013 maps for those districts even after a remedial phase. Or does the majority read that to mean that the challengers still had more to show before the District Court “would” redraw the districts that §2 “could” require to be redrawn? And what is the effect of the conclusion respecting the County Line Rule on the potential remedy for the intentional vote dilution holding as to HD32 and HD34? The majority conveniently avoids confronting this lack of clarity by ignoring the relevant record, instead stating without explanation that it believes “it clear that the District Court effectively enjoined use of these districts as currently configured.” Ante, at 19, n. 15. But it cannot escape the reality that its rule will “needlessly complicate appellate review,” ante, at 15, given that “it is simply not possible [absent an injunction] to know with any certainty what the court has decided,” Gunn, 399 U. S., at 388.

I do not disagree that “lack of specificity in an injunctive order would [not] alone deprive the Court of jurisdiction under §1253.” Id., at 389, n. 4; see also ante, at 16 (quoting Gunn). “But the absence of any semblance of effort by the District Court to comply with [the specificity required of injunctive orders under the Federal Rules] makes clear that the court did not think its [orders] constituted an order granting an injunction.” Gunn, 399 U. S., at 389. If any doubt remained as to the effect of the orders here, moreover, the District Court explicitly assured the parties that, even though it had found violations, it was not enjoining use of the 2013 maps for the upcoming elections. App. 134a–136a.

Finally, it is axiomatic that “administrative simplicity is a major virtue in a jurisdictional statute.” Hertz Corp. v. Friend, 559 U. S. 77, 94 (2010).

“Complex jurisdictional tests complicate a case . . . . Complex tests produce appeals and reversals, [and] encourage gamesmanship . . . . Judicial resources too are at stake [as] courts benefit from straightforward rules under which they can readily assure themselves of their power to hear a case. Simple jurisdictional rules also promote greater predictability.” Ibid. (citations omitted).

Simple is thus the name of the game when it comes to jurisdictional rules. The rule in the majority opinion is anything but. Although the majority claims that a mere “finding on liability cannot be appealed unless an injunction is granted or denied,” ante, at 20, the rule it embraces today makes it hard to understand when a finding on liability would not be read, as the majority does here, as having the “practical effect” of an injunction. It is a worrisome prospect that, after today, whenever a three-judge district court expresses that a statutory or constitutional violation must be remedied, the party held liable will straightaway file an appeal in this Court and assert jurisdiction under §1253, even where the district court is clear that no injunction has issued.[8]

The majority opinion purports to add a limit by distinguishing between unappealable orders that find a plan “unlawful long before the next scheduled election” or “very close to the election date,” and those (presumably) appealable orders that are entered neither “long before” nor “very close” to the next election. Ante, at 20.[9] What does that even mean? The orders at issue here were entered about 15 months before the 2018 elections, and according to the majority fall within the not “long before” but not “very close” appealable range. Why this is so, however, the majority never says. Without any definitions for its boundary posts, courts will be left to wonder: What about orders entered 17 or 18 months before an election? Are those considered “long before” so they would be unappealable? And are orders entered 14, 13, or 12 months before the election similarly unappealable because they were entered “very close” to the election date? And what does the majority mean by “the election date”? Does that include primaries? What about registration deadlines, or ballot-printing deadlines? It is not uncommon for there to be, at any given time, multiple impending deadlines relating to an upcoming election. Thinking through the many variations of jurisdictional disputes that will arise over the years following this novel reading of §1253 should be enough to stop the majority from rewriting our long established jurisprudence in this area.

After today, our mandatory appellate docket will be flooded by unhappy litigants in three-judge district court cases, demanding our review. Given the lack of predict- ability, the rule will incentivize appeals and “encourage gamesmanship.” Hertz Corp., 559 U. S., at 94. The Court will no doubt regret the day it opened its courthouse doors to such time-consuming and needless manipulation of its docket.

D

Even if the majority were correct to import the “practical effect” rule into the §1253 context, moreover, that would still not justify the Court’s premature intervention in these appeals for at least two reasons. First, while taking from Carson the “practical effect” rule it likes, the majority gives short shrift to the second half of that case, in which the Court was explicit that “[u]nless a litigant can show that an interlocutory order . . . might have a ‘serious, perhaps irreparable, consequence,’ and that the order can be ‘effectually challenged’ only by immediate appeal, the general congressional policy against piecemeal review will preclude interlocutory appeal.” 450 U. S., at 84. Texas has made no showing of a “serious, perhaps irreparable consequence” requiring our immediate intervention, nor has Texas shown that the orders could not be “effectually challenged” after the remedial stage was completed. In fact, when Texas sought a stay of those orders before this Court, the 2018 elections were more than a year away. For the majority, however, it is enough that the District Court found the Texas redistricting maps to be in violation of federal law. Ante, at 20–21. That cursory application of Carson, in particular whether the injunctions the majority reads into the August 15 and 24 orders could be “effectually challenged” absent immediate appeal to this Court, deprives that limit to our jurisdiction of much of its meaning when assessing Texas’ request for our intervention in these cases. Nothing in our precedent supports that truncated approach. And in any event, if Texas wanted review of the orders after any injunction was entered by the District Court, it could have asked this Court for an emergency stay.

Second, the August 15 and 24 orders at issue here sim- ply did not have the “practical effect” of enjoining Texas’ use of the 2013 maps. The majority thinks otherwise in part because the District Court noted that the violations “ ‘must be remedied.’ ” Ante, at 16. In addition, the major- ity believes that “Texas had reason to fear that if it tried to conduct elections under plans that the court had found to be ‘racially discriminatory,’ the court would infer an evil motive and perhaps subject the State once again to the strictures of preclearance under §3(c) of the Voting Rights Act.” Ante, at 17–18. But the majority forgets that the District Court made explicit that “[a]lthough [it] found violations [in the 2013 maps], [it] ha[d] not enjoined [their] use for any upcoming elections.” App. 134a–136a. That the District Court requested the Texas attorney general to advise it, within “three business days,” whether “the Legislature intends to take up redistricting in an effort to cure [the] violations,” 274 F. Supp. 3d, at 686; 267 F. Supp. 3d, at 795, does not undermine that unequivocal statement. Nothing in that language indicates that the District Court required the Legislature to “redraw both maps immediately” or else “the court would do so itself.” Brief for Appellants 20 (emphasis in original). Instead, recognizing “that the federal judiciary functions within a system of federalism which entrusts the responsibility of legislative . . . districting primarily to the state legislature,” Whitcomb, 305 F. Supp., at 1392, the District Court gave Texas an opportunity to involve its Legislature and asked for a simple statement of intent so that the court could manage its docket accordingly. This request for a statement of intent, which was necessary for the District Court to manage its own docket, does not transform the orders into injunctions.

As to the second point, if Texas had any “fear” regarding the use of its maps, despite having been explicitly told that the maps were not enjoined, that would still not be enough. This Court recognized in Gunn that the State in that case, faced with the order declaring its statute unconstitutional, “would no doubt hesitate long before disregarding it.” 399 U. S., at 390. That hesitation was not enough in Gunn to magically transform an order into an injunction for purposes of §1253, and nothing about these cases justifies the majority taking out its wand today. Whatever “fear” Texas had does not transform the August 15 and 24 orders into injunctions. And absent an injunction, this Court lacks jurisdiction over these appeals. The cases should thus be dismissed.

II

Having rewritten the limits of §1253, the majority moves to the merits. There again the Court goes astray. It asserts that the District Court legally erred when it purportedly shifted the burden of proof and “required the State to show that the 2013 Legislature somehow purged the ‘taint’ that the court attributed to the defunct and never-used plans enacted by a prior legislature in 2011.” Ante, at 21. But that holding ignores the substantial amount of evidence of Texas’ discriminatory intent, and indulges Texas’ warped reading of the legal analysis and factual record below.[10]

A

Before delving into the content of the August 15 and 24 orders, a quick recap of the rather convoluted history of these cases is useful. In 2011, the Texas Legislature redrew its electoral districts. Various plaintiff groups challenged the 2011 maps under §2 of the Voting Rights Act and the Fourteenth Amendment, and those lawsuits were consolidated before the three-judge District Court below pursuant to 28 U. S. C. §2284(a). Because Texas then was subject to preclearance under §5 of the Voting Rights Act, the 2011 maps did not take effect immediately, and Texas filed a declaratory action in the District Court for the District of Columbia to obtain preclearance.

“Faced with impending election deadlines and un-precleared plans that could not be used in the [2012] election, [the District] Court was faced with the ‘unwelcome obligation’ of implementing interim plans so that the primaries could proceed.” 274 F. Supp. 3d, at 632. In January 2012, this Court vacated the first iteration of those interim maps in Perry v. Perez, 565 U. S. 388, 394–395 (2012) (per curiam), finding that the District Court failed to afford sufficient deference to the Legislature. In February 2012, the District Court issued more deferential interim plans, but noted that its analysis had been expedited and curtailed, and that it had only made preliminary conclusions that might be revised on full consideration. C. J. S. 367a–424a; H. J. S. 300a–315a.

In August 2012, the D. C. District Court denied preclearance of the 2011 maps. Texas v. United States, 887 F. Supp. 2d 133 (2012). It concluded that the federal congressional map had “retrogressive effect” and “was enacted with discriminatory intent,” id., at 159, 161, and that the State House map was retrogressive and that “the full record strongly suggests that the retrogressive effect . . . may not have been accidental,” id., at 178. Texas appealed, and the case was eventually dismissed following Shelby County v. Holder, 570 U. S. 529 (2013) (holding unconstitutional the formula used to subject States to the preclearance requirement).

In June 2013, the Texas Governor called a special legislative session, and that same month the Legislature adopted the 2012 interim maps as the permanent maps for the State. The Legislature made small changes to the maps, including redrawing the lines in HD90, but the districts at issue in these appeals all remained materially unchanged from the 2011 maps.

The District Court in these cases denied Texas’ motion to dismiss the challenges to the 2011 maps, and the challengers amended their complaints to assert claims respecting the 2013 maps. In April and May 2017, the District Court held that districts in Texas’ 2011 maps violated §2 and the Fourteenth Amendment. The August 15 and 24 orders respecting the 2013 maps followed.

B

The majority believes that, in analyzing the 2013 maps, the District Court erroneously “attributed [the] same [discriminatory] intent [harbored by the 2011 Legislature] to the 2013 Legislature” and required the 2013 Legislature to purge that taint. Ante, at 9–10. The District Court did no such thing. It engaged in a painstaking analysis of discriminatory intent under Arlington Heights v. Metropolitan Housing Development Corp., 429 U. S. 252 (1977), which is critical to understanding why, as explained in Part II–D, infra, the District Court did not improperly presume that the Legislature acted with discriminatory intent.

Under Arlington Heights, “in determining whether racially discriminatory intent existed,” this Court considers “circumstantial and direct evidence” of: (1) the discriminatory “impact of the official action,” (2) the “historical background,” (3) the “specific sequence of events leading up to the challenged decision,” (4) departures from procedures or substance, and (5) the “legislative or administrative history,” including any “contemporary statements” of the lawmakers. 429 U. S., at 266–268. Although this analysis must start from a strong “presumption of good faith,” Miller v. Johnson, 515 U. S. 900, 916 (1995), a court must not overlook the relevant facts. This Court reviews the “findings of fact” made by the District Court, including those respecting legislative motivations, “only for clear error.” Cooper v. Harris, 581 U. S. ___, ___–___ (2017) (slip op., at 3–4); see also Anderson v. Bessemer City, 470 U. S. 564, 573 (1985). The Court therefore “may not reverse just because we ‘would have decided the [matter] differently. . . . A finding that is ‘plausible’ in light of the full record—even if another is equally or more so—must govern.” Harris, 581 U. S., at ___ (slip op., at 4).

The District Court followed the guidance in Arlington Heights virtually to a tee, and its factual findings are more than “plausible” in light of the record. To start, there is no question as to the discriminatory impact of the 2013 plans, as the “specific portions of the 2011 plans that [the District Court] found to be discriminatory or unconstitutional racial gerrymanders continue unchanged in the 2013 plans, their harmful effects ‘continu[ing] to this day.’ ” 274 F. Supp. 3d, at 649 (alteration in original). Texas, more- over, has a long “history of discrimination” against minority voters. Id., at 648, n. 37. “In the last four decades, Texas has found itself in court every redistricting cycle, and each time it has lost.” Texas, 887 F. Supp. 2d, at 161.

There is also ample evidence that the 2013 Legislature knew of the discrimination that tainted its 2011 maps. “The 2013 plans were enacted by a substantially similar Legislature with the same leadership only two years after the original enactment.” 274 F. Supp. 3d, at 648, n. 37. The Legislature was also well aware that “the D. C. court concluded that [its 2011] maps were tainted by evidence of discriminatory purpose,” H. J. S. 443a, and despite the District Court having warned of the potential that the Voting Rights Act may require further changes to the maps, “the Legislature continued its steadfast refusal to consider [that] possibility,” 274 F. Supp. 3d, at 649.

Turning to deliberative process—on which the majority is singularly focused, to the exclusion of the rest of the factors analyzed in the orders below, see Part II–D, infra—the District Court concluded that Texas was just “not truly interested in fixing any remaining discrimination in the [maps].” 274 F. Supp. 3d, at 651, n. 45. Despite knowing of the discrimination in its 2011 maps, “the Legislature did not engage in a deliberative process to ensure that the 2013 plans cured any taint from the 2011 plans.”[11] Id., at 649. For instance, Representative Darby, a member of the redistricting committee, “kept stating that he wanted to be informed of legal deficiencies so he could fix them,” but “he did not himself seek to have the plan evaluated for deficiencies and he willfully ignored those who pointed out deficiencies, continuing to emphasize that he had thought ‘from the start’ that the interim plans were fully legal.” Id., at 651, n. 45.[12] The Legislature made no substantive changes to the challenged districts that were the subject of the 2011 complaints, and “there is no indication that the Legislature looked to see whether any discriminatory taint remained in the plans.” Id., at 649. In fact, the only substantive change that the Legislature made to the maps was to add more discrimination in the form of a new racially gerrymandered HD90, as the majority concedes. Ante, at 39–41.

The absence of a true deliberative process was coupled with a troubling sequence of events leading to the enactment of the 2013 maps. Specifically, “the Legislature pushed the redistricting bills through quickly in a special session,” 274 F. Supp. 3d, at 649, despite months earlier having been urged by the Texas attorney general to take on redistricting during the regular session, id., at 634; see also H. J. S. 440a. By pushing the bills through a special session, the Legislature did not have to comply with “a two-thirds rule in the Senate or a calendar rule in the House,” 274 F. Supp. 3d, at 649, n. 38, and it avoided the “full public notice and hearing” that would have allowed “ ‘meaningful input’ from all Texans, including the minor- ity community,” H. J. S. 444a. In addition, “necessary resources were not allocated to support a true deliberative process.” 274 F. Supp. 3d, at 649. For instance, the House committee “did not have counsel when the session started.” Ibid., n. 39.

Nor can Texas credibly claim to have understood the 2012 interim orders as having endorsed the legality of its maps so that adopting them would resolve the challengers’ complaints. In its 2012 interim orders, “the [District] Court clearly warned that its preliminary conclusions . . . were not based on a full examination of the record or the governing law and were subject to revision” “given the severe time constraints . . . at the time” the orders were adopted. Id., at 650. The District Court also explained that the “claims presented . . . involve difficult and unsettled legal issues as well as numerous factual disputes.” C. J. S. 367a. During the redistricting hearings, chief legislative counsel for the Texas Legislative Council in 2013, Jeff Archer, advised the Legislature that the District Court “ ‘had not made full determinations, . . . had not made fact findings on every issue, had not thoroughly analyzed all the evidence’ ” and had “ ‘made it explicitly clear that this was an interim plan to address basically first impression of voting rights issues.’ ” 274 F. Supp. 3d, at 650 (alterations in original); see also App. 441a–442a (testimony that interim plans were “impromptu” and “preliminary” and that the District Court “disclaimed making final determinations”). Archer explained that although the Legislature had “ ‘put to bed’ ” challenges regarding “ ‘those issues that the [District] Court identified so far,’ ” it had not “ ‘put the rest to bed.’ ” 274 F. Supp. 3d, at 651, n. 45; see also App. 446a–447a (advising that, “on a realistic level,” the Legislature had not “removed legal challenges” and that adopting the interim maps “in no way would inoculate the plans”).

There was substantial evidence that the 2013 Legislature instead adopted the interim plans as part of a “strategy [that] involved adopting the interim maps, however flawed,” to insulate (and thus continue to benefit from) the discriminatory taint of its 2011 maps. 274 F. Supp. 3d, at 651. Texas hoped that, by adopting the 2012 interim maps, the challengers “would have no remedy, and [the Legislature] would maintain the benefit of such discrimination or unconstitutional effects.” Ibid. That strategy originated with the Texas attorney general, who was responsible for defending the State in the redistricting challenges. Id., at 650, and n. 41. He advised the Legislature that adopting the interim plans was the “ ‘best way to avoid further intervention from federal judges’ ” and to “ ‘insulate [Texas’] redistricting plans from further legal challenge.’ ” Id., at 650 (emphasis added); see also H. J. S. 443a. The Texas attorney general also drafted the “legislative fact findings accompanying the plans, before the Legislature had engaged in any fact findings on the bills,” stating that the 2012 interim plans “complied ‘with all federal and state constitutional provisions or laws applicable to redistricting plans.’ ” 274 F. Supp. 3d, at 650, n. 41 (emphasis added). That the legislative factfindings were predrafted by the attorney defending Texas in these redistricting challenges—purporting to conclude that the 2012 interim plans complied with the law, when in fact the evidence showed that the Legislature did not engage in a true deliberative process or meaningfully consider evidence of the legality of the plans so that it could have endorsed such factfindings—demonstrates that the adoption of the interim plans was a mere pretext to insulate the discriminatory benefits of the 2011 plans. That explains why legislators thought that removal of those fact findings would “ ‘gu[t] the bill.’ ” Ibid.

In the end, having presided over years of litigation and seeing firsthand all of the evidence, the District Court thought it clear that Texas’ “strategy involved adopting the interim maps, however flawed” so that the challengers “would have no remedy, and [Texas] would maintain the benefit of such discrimination and unconstitutional effects.” Id., at 651. It is hard to imagine what a more thorough consideration of the Arlington Heights factors in these cases would have looked like. Review of the District Court’s thorough inquiry leads to the inescapable conclusion that it did not err—let alone clearly err—in concluding that the “Legislature in 2013 intentionally furthered and continued the existing discrimination in the plans.” 274 F. Supp. 3d, at 652.

C

In contrast to that thorough Arlington Heights inquiry, the majority engages in a cursory analysis of the record to justify its conclusion that the evidence “overwhelmingly” shows that Texas acted with legitimate intent. Ante, at 28. Two critical things are conspicuously missing from its analysis: first, consideration of the actual factual record (or most of it, anyway),[13] and second, meaningful consideration of the limits of our review of facts on these appeals.[14]

The majority first makes reference to the fact that the Texas attorney general “advised the Legislature that the best way to [end the redistricting litigation] was to adopt the interim, court-issued plans,” a position repeated by the sponsor of the plans. Ante, at 26. And in its view, it was reasonable for the Legislature to believe that adopting the interim plans “might at least reduce objections and thus simplify and expedite the conclusion of the litigation.” Ante, at 28. The majority also states that “there is no evidence that the Legislature thought that the plans were invalid.” Ante, at 27. In reaching those findings, however, the majority ignores all of the evidence in the record that demonstrates that the Legislature was aware of (and ignored) the infirmities in the maps, that it knew that adopting the interim plans would not resolve the litigation concerning the disputed districts, and that it nevertheless moved forward with the bills as a strategy to “insulate” the discriminatory maps from further judicial scrutiny and perpetuate the discrimination embedded in the 2012 interim maps. See Part II–B, supra.

Instead of engaging with the factual record, the majority opinion sets out its own view of “the situation when the Legislature adopted the court-approved interim plans.” Ante, at 28. Under that view, “the Legislature [had] good reason to believe that the court-approved interim plans were legally sound,” particularly in light of our remand instructions in Perry, 565 U. S. 388. Ante, at 28–29. The majority nowhere considers, however, the evidence regarding what the Legislature actually had before it concerning the effect of the interim orders, including the explicit cautionary statements in the orders and the repeated warnings of the chief legislative counsel that the interim plans were preliminary, incomplete, and impromptu.[15] See Part II–B, supra.

The majority finds little significance in the fact that the Legislature “ ‘pushed the redistricting bills through quickly in a special session,’ ” reasoning that a special session was needed “because the regular session had ended.” Ante, at 29. That of course ignores the evidence that the Legislature disregarded requests by the Texas attorney general, months earlier, to take up redistricting during the regular session, that proceeding through a special session permitted the Legislature to circumvent procedures that would have ensured full and adequate consideration, and that resources were not sufficiently allocated to permit considered review of the plans. See Part II–B, supra.

Finally, the majority sees nothing wrong with the fact that the Legislature failed “to take into account the problems with the 2011 plans that the D. C. court identified in denying preclearance.” Ante, at 30. It maintains that the purpose of adopting the interim plans was to “fix the problems identified by the D. C. court,” and reasons that the interim maps did just that by modifying any problematic districts. Ibid. But of course the finding of discriminatory intent rested not only on what happened with particular districts. Rather, the evidence suggested that discriminatory motive permeated the entire 2011 redistricting process, as the D. C. court considered that “Texas has found itself in court every redistricting cycle [in the last four decades], and each time it has lost”; that “Black and Hispanic members of Congress testified at trial that they were excluded completely from the process of drafting new maps, while the preferences of Anglo members were frequently solicited and honored”; that the redistricting committees “released a joint congressional redistricting proposal for the public to view only after the start of a special legislative session, and each provided only seventy-two hours’ notice before the sole public hearing on the proposed plan in each committee”; that minority members of the Texas Legislature “raised concerns regarding their exclusion from the drafting process and their inability to influence the plan”; and that the Legislature departed from normal procedure in the “failure to release a redistricting proposal during the regular session, the limited time for review, and the failure to provide counsel with the necessary election data to evaluate [Voting Rights Act] compliance.” 887 F. Supp. 2d, at 161. The majority also ignores the findings of retrogression concerning the previous version of CD25, which of course are relevant to the challengers’ claims about CD27 and CD35 in this litigation and were not addressed in the 2012 interim plans. See Part III–A, infra. That the 2012 interim maps addressed some of the deficiencies identified by the D. C. court in the preclearance litigation does not mean that the Legislature in 2013 was free to wholly disregard the significance of other evidence of discrimination that tainted its 2011 maps and were entrenched in the 2012 interim maps.

Even had the majority not ignored the factual record, it still would be wrong in concluding that the District Court erred in finding that the 2013 Legislature acted with the intent to further and benefit from the discrimination in the 2011 maps. In light of the record before this Court, the finding of invidious intent is at least more than “ ‘plausible’ ” and thus “must govern.” Harris, 581 U. S., at ___ (slip op., at 4). The majority might think that it has a “better view of the facts” than the District Court did, but “the very premise of clear error review is that there are often ‘two permissible’—because two ‘plausible’—‘views of the evidence.’ ” Id., at ___–___ (slip op., at 9–10).

D

The majority resists the weight of all this evidence of invidious intent not only by disregarding most of it and ignoring the clear-error posture, but also by endorsing Texas’ distorted characterizations of the intent analysis in the orders below. Specifically, the majority accepts Texas’ argument that the District Court “reversed the burden of proof” and “imposed on the State the obligation of proving that the 2013 Legislature had experienced a true ‘change of heart’ and had ‘engage[d] in a deliberative process to ensure that the 2013 plans cured any taint from the 2011 plans.’ ” Ante, at 23 (alteration in original). The District Court did no such thing, and only a selective reading of the orders below could support Texas’ position.

It is worth noting, as a preliminary matter, that the majority does not question the relevance of historical discrimination in assessing present discriminatory intent. Indeed, the majority leaves undisturbed the longstanding principle recognized in Arlington Heights that the “ ‘historical background’ of a legislative enactment is ‘one evidentiary source’ relevant to the question of intent.” Ante, at 22 (quoting Arlington Heights, 429 U. S., at 267). With respect to these cases, the majority explicitly acknowledges that, in evaluating whether the 2013 Legislature acted with discriminatory purpose, “the intent of the 2011 Legislature . . . [is] relevant” and “must be weighed together with any other direct and circumstantial evidence” bearing on intent. Ante, at 26.

If consideration of this “ ‘historical background’ ” factor means anything in the context of assessing intent of the 2013 Legislature, it at a minimum required the District Court to assess how the 2013 Legislature addressed the known discrimination that motivated the drawing of the district lines that the Legislature was adopting, unchanged, from the 2011 maps. Therefore, the findings as to whether the 2013 Legislature engaged in a good-faith effort to address any known discrimination that tainted its 2011 plans were entirely apposite, so long as the District Court “weighed [this factor] together with any other direct and circumstantial evidence” bearing on the intent question, and so long as the burden remained on the challengers to establish invidious intent. Ante, at 26.

The majority faults the District Court for not adequately engaging in that weighing and giving too “central” a focus to the historical factor in its intent analysis. Ante, at 23; see also ante, at 23–24, n. 18. That alleged “central” focus, the majority contends, led the District Court to shift the burden of proof on the intent inquiry away from the challengers, instead requiring Texas to show that the Legislature cured its past transgressions. Ante, at 23. Those conclusions can only be supported if, as Texas and the majority have done, one engages in a highly selective reading of the District Court orders.

To begin, entirely absent from the majority opinion is any reference to the portions of the District Court orders that unequivocally confirm its understanding that the burden remained on the challengers to show that the 2013 Legislature acted with invidious intent. The District Court was explicit that the challengers bore the burden to “establish their claim by showing that the Legislature adopted the plans with a discriminatory purpose, maintained the district lines with a discriminatory purpose, or intentionally furthered preexisting intentional discrimination.” 274 F. Supp. 3d, at 646; see also id., at 645 (discussing Circuit precedent regarding the showing needed for “a plaintiff [to] meet the purpose standard”).[16]

Even when it does look at the actual language of the orders, the majority picks the few phrases that it believes support its argument, choosing to disregard the rest. For instance, the majority quotes the District Court order as having required Texas to show that the 2013 Legislature had a “ ‘change of heart.’ ” Ante, at 23 (quoting 274 F. Supp. 3d, at 649). When that sentence is read in full, however, it is evident that the District Court was not imposing a “duty to expiate” the bad intent of the previous Legislature, as the majority contends, ante, at 23, but instead was describing what the weighing of the direct and circumstantial evidence revealed about the motivations of the 2013 Legislature: “The decision to adopt the interim plans was not a change of heart concerning the validity of [the challengers’] claims . . . —it was a litigation strategy designed to insulate the 2011 or 2013 plans from further challenge, regardless of their legal infirmities.” 274 F. Supp. 3d, at 649–650.

Likewise, the majority quotes the orders as requiring proof that the Legislature “ ‘engage[d] in a deliberative process to ensure that the 2013 plans cured any taint from the 2011 plans.’ ” Ante, at 23 (quoting 274 F. Supp. 3d, at 649). But the District Court did not put the burden on Texas to make that affirmative showing. Instead, that partial quote is lifted from a sentence in which the District Court, having held a trial on these factual issues, concluded that the challengers had met their burden to show that “the Legislature did not engage in a deliberative process,” which it supported later in that paragraph with findings that the Legislature “pushed the redistricting bills through quickly in a special session” without allocating the “necessary resources . . . to support a true deliberative process.” 274 F. Supp. 3d, at 649.

The majority finally asserts that the District Court “drove the point home” when it “summarized its analysis” as follows: “ ‘The discriminatory taint [from the 2011 plans] was not removed by the Legislature’s enactment of the Court’s interim plans, because the Legislature engaged in no deliberative process to remove any such taint, and in fact intended any such taint to be maintained but safe from remedy.’ ” Ante, at 23 (quoting 274 F. Supp. 3d, at 686). The majority no doubt hopes that the reader will focus on the portion of the sentence in which the District Court concludes that the discriminatory taint found in the 2011 maps “ ‘was not removed’ ” by the enactment of the interim maps “ ‘because the Legislature engaged in no deliberative process to remove any such taint.’ ” Ante, at 23 (quoting 274 F. Supp. 3d, at 686).[17] But the majority ignores the import of the remaining part of the sentence, in which the District Court held that the Legislature “in fact intended any such taint to be maintained but be safe from remedy.” 274 F. Supp. 3d, at 652; see also id., at 686. The majority also conveniently leaves out the sentence that immediately follows: “The Legislature in 2013 intentionally furthered and continued the existing discrimination in the plans.” Id., at 652. When read in full and in context, it is clear that the District Court remained focused on the evidence proving the intent of the 2013 Legislature to shield its plans from a remedy and thus further the discrimination, rather than simply presuming invidious intent from the failure to remove the taint, as the majority claims.

In selectively reviewing the record below, the majority attempts to shield itself from the otherwise unavoidable conclusion that the District Court did not err. If forced to acknowledge the true scope of the legal analysis in the orders below, the majority would find itself without support for its insistence that the District Court was singu- larly focused on whether the Legislature “removed” past taint. And then the majority would have to contend with the thorough analysis of the Arlington Heights factors, Part II–B, supra, that led the District Court to conclude that the 2013 Legislature acted with invidious intent.

III

The majority fares no better in its district-by-district analysis. In line with the theme underlying the rest of its analysis, the majority opinion overlooks the factual record and mischaracterizes the bulk of the analysis in the orders below in concluding that the District Court erred in finding a §2 results violation as to CD27, HD32, and HD34. I first address CD27, and then turn to HD32 and HD34.

A

1

To put in context the objections to the District Court’s conclusion regarding CD27, a brief review of the District Court’s factual findings as to that district is necessary. Before 2011, CD27 was a Latino opportunity district, i.e., a majority-HCVAP district with an opportunity to elect a Hispanic-preferred candidate. When the Legislature reconfigured the district in 2013, it moved Nueces County, a majority-HCVAP county, into a new Anglo-majority district to protect an incumbent “who was not the candidate of choice of those Latino voters” and likely would have been “ousted” by them absent the redistricting. C. J. S. 181a, 191a. The District Court found that the “placement of Nueces County Hispanics in an Anglo-majority district ensures that the Anglo majority usually will defeat the minority-preferred candidate, given the racially polarized voting in the area.” Id., at 189a–190a. It also found that “the political processes are not equally open to Hispanics” in Texas as a result of its “history of official discrimination touching on the right of Hispanics to register, vote, and otherwise to participate in the democratic process [that] is well documented,” and that “Latinos bear the effects of past discrimination in areas such as education and employment/income, which hinder their ability to participate effectively in the political process.” Id., at 190a–191a. Given those findings, the District Court concluded that the newly constituted CD27 “has the effect of diluting Nueces County Hispanic voters’ electoral opportunity.” Id., at 191a.

Texas nevertheless contended (and maintains here) that no §2 results violation existed because only “seven compact Latino opportunity districts could be drawn in South/West Texas,” id., at 181a, and that all seven districts already existed under its maps. To explain how it counted to seven, Texas pointed to the creation of CD35 as a supposed new Latino opportunity district that joined Travis County Hispanics with Hispanics in San Antonio. The District Court agreed that only seven such districts could be drawn in the area, but rejected Texas’ invocation of CD35 as a defense. The District Court concluded that because Travis County “[did] not have Anglo bloc voting,” 274 F. Supp. 3d, at 683, §2 did not require the placement of Travis County Hispanics in an opportunity district, C. J. S. 176a; see also Thornburg v. Gingles, 478 U. S. 30, 51 (1986). The District Court found that Texas had moved Travis County Hispanics from their pre-2011 district, CD25, to the newly constituted CD35, not to comply with §2, but “to use race as a tool for partisan goals . . . to intentionally destroy an existing district with significant minority population (both African American and Hispanic) that consistently elected a Democrat (CD25).” 274 F. Supp. 3d, at 683. Thus, it concluded that “CD35 was an impermissible racial gerrymander because race predominated in its creation without furthering a compelling state interest.” Ibid.

Importantly, the District Court concluded that, without CD35, Texas could have drawn one more Latino opportunity district in South/West Texas that included Nueces County Hispanics. C. J. S. 181a; see also id., at 190a (“Plaintiffs have thus shown that a district could be drawn in which Hispanics, including Nueces County Hispanics, are sufficiently numerous and geographically compact to constitute a majority HCVAP”); id., at 192a (“Numerous maps also demonstrated that accommodating the §2 rights of all or most Nueces County Hispanic voters would not compromise the §2 rights of any other voters, and in fact including it substantially accommodates the §2 rights of Hispanic voters in South/West Texas”). Indeed, “[p]lans were submitted during the legislative session and during this litigation that showed that seven compact districts could be drawn that included all or most Nueces County Hispanic voters but not Travis County voters.” Id., at 181a, n. 47.

2

Nothing in the record or the parties’ briefs suggests that the District Court clearly erred in these findings of fact, which unambiguously support its conclusion that there is a §2 results violation with respect to CD27. Nevertheless, the majority offers two reasons for reversing that conclusion. First, the majority contends that the District Court erred because “in evaluating the presence of majority bloc voting in CD35,” it “looked at only one, small part of the district, the portion that falls within Travis County.” Ante, at 34. It cites to Bethune-Hill v. Virginia State Bd. of Elections, 580 U. S. ___, ___ (2017) (slip op., at 12), an equal protection racial gerrymandering case, for the proposition “that redistricting analysis must take place at the district level.” Ante, at 34. According to the majority, then, the District Court should have looked at the existence of majority bloc voting in CD35 as a whole after the 2011 redistricting.

But the majority confuses the relevant inquiry, as well as the relevant timeline. The particular §2 question here does not concern the status of Travis County Latinos in the newly constituted CD35 after the 2011 redistricting. Rather, it concerns the status of Travis County Latinos in the old CD25, prior to the 2011 redistricting. That is because the challengers’ §2 claim concerns the choices before the Legislature at the time of the 2011 redistricting, when it was deciding which Latinos in South/West Texas to place in the new opportunity district to be created in that area of the State. The Legislature chose to include Travis County Latinos in an opportunity district at the expense of the Nueces County Latinos, who were instead moved into a majority-Anglo district. So the question is whether, knowing that Nueces County Latinos indisputably had a §2 right, the Legislature’s choice was nevertheless justified because the Travis County Latinos also had a §2 right that needed to be accommodated. In other words, did the Legislature actually create a new §2 opportunity district for persons with a §2 right, or did it simply move people without a §2 right into a new district and just call it an opportunity district? To answer that question, the status of Travis County Latinos in 2011 is the only thing that matters, and the District Court thus correctly focused its inquiry on whether bloc voting existed in Travis County prior to the 2011 redistricting, such that Travis County Latinos could be found to have a §2 right. Whether the newly constituted CD35 now qualifies as a §2 opportunity district—an inquiry that would, as the majority suggests, call for district-wide consideration—is beside the point.

Second, the majority reasons that “the 2013 Legislature had ‘good reasons’ to believe that [CD35] was a viable Latino opportunity district that satisfied the Gingles factors.” Ante, at 35. For this, the majority cites to the fact that the district “was based on a concept proposed by MALDEF” and that one group of plaintiffs “argued that the district [was] mandated by §2,” and vaguely suggests that, contrary to the District Court’s finding, “there is ample evidence” of majority bloc voting in CD35. Ibid.[18]

The majority forgets, yet again, that we review factual findings for clear error. Harris, 581 U. S., at ___–___ (slip op., at 3–4). Indeed, its analysis is too cursory even for de novo review. The majority does not meaningfully engage with the full factual record below. Instead, it looks only to the handful of favorable facts cited in Texas’ briefs. Compare Brief for Appellants 46 with ante, at 35. Had the majority considered the full record, it could only have found that the District Court cited ample evidence in support of its conclusion that the Legislature had no basis for believing that §2 required its drawing of CD35. In fact, the District Court noted that Texas in 2011 “actually asserted that CD35 is not required by §2,” C. J. S. 174a, n. 40, that the main plan architect testified that he was not sure whether §2 required drawing the district, and that testimony at trial showed that the district was drawn because, on paper, it would fulfill the requirement of being majority-HCVAP while providing Democrats only one new district, and “not because all of the Gingles factors were satisfied,” id., at 178a–179a, n. 45. The District Court also concluded that “there is no evidence that any member of the Legislature . . . had any basis in evidence for believing that CD35 was required by §2 other than its HCVAP-majority status.” Ibid.

Had the majority properly framed the inquiry and ap-plied the clear-error standard to the full factual record, it could not convincingly dispute the existence of a §2 results violation as to CD27. Texas diluted the voting strength of Nueces County Latinos by transforming a minority-opportunity district into a majority-Anglo district. The State cannot defend that result by pointing to CD35, because its “creation of an opportunity district for [Travis County Latinos] without a §2 right offers no excuse for its failure to provide an opportunity district for [Nueces County Latinos] with a §2 right.” League of United Latin American Citizens v. Perry, 548 U. S. 399, 430 (2006) (LULAC).[19]

B

1

I turn now to HD32 and HD34. Before the 2011 redistricting, Nueces County had within it two Latino opportunity districts and part of one Anglo-represented district. 267 F. Supp. 3d, at 767. Due to slower population growth reflected in the 2010 census, however, Nueces County was entitled to have within it only two districts. Accordingly, during the 2011 redistricting, the Legislature opted to “eliminate one of the Latino opportunity districts . . . and draw two districts wholly within Nueces County—one strongly Latino (HD34) and one a safe Anglo Republican seat (HD32) to protect [an] incumbent.” Ibid. “Based on an analysis of the Gingles requirements and the totality of the circumstances,” however, the District Court found that the Legislature could have drawn two compact minority districts in Nueces County. Id., at 780. Namely, the evidence demonstrated that it was possible to draw a map with “two districts with greater than 50% HCVAP,” that “Latinos in Nueces County are highly cohesive, and that Anglos vote as a block usually to defeat minority preferred candidates.” Id., at 777–778.

The District Court then considered two proposed configurations for those districts: one with two HCVAP-majority districts located wholly within Nueces County, and another that required breaking the County Line Rule. Id., at 777. The challengers preferred the latter configuration because, according to their expert, “an exogenous election index” revealed that the two HCVAP-majority districts wholly within Nueces County did “not perform suffi- ciently.” Id., at 778. The District Court did not accept that expert’s assessment at face value. Instead, it explained that “an exogenous election index alone will not determine opportunity,” and so evaluated the expert testing and ample other evidence and ultimately concluded that the challengers had “not adequately demonstrated that they lack equal opportunity in [an alternative] configuration . . . such that a county line break is necessary.” Id., at 778, 781. Thus, although it found that “two HCVAP-districts could have been drawn that would provide Hispanics with equal electoral opportunity, and that §2 could require those two districts,” because §2 did not require the challengers’ requested remedy (i.e., breaking the County Line Rule), the District Court had to “consider whether §2 requires a remedy” and directed the challengers to “consider their preferred configuration for the remedy stage” that was to follow (before Texas prematurely appealed). Id., at 783.

2

The majority purports to accept these factual findings, and contends that they “show that [HD32 and HD34] do not violate §2.” Ante, at 35. Specifically, the majority points to the fact that the challengers’ “own expert determined that it was not possible to divide Nueces County into more than one performing Latino district” without breaking the County Line Rule, a remedy the District Court concluded was not required by §2. Ante, at 36 (emphasis in original). “So if Texas could not create two performing districts in Nueces County and did not have to break county lines,” the majority reasons, “the logical result is that Texas did not dilute the Latino vote.” Ibid. (emphasis in original). In its view, a districting decision cannot be said to dilute the votes of minority voters “if the alternative to the districting decision at issue would not enhance the ability of minority voters to elect the candidates of their choice.” Ibid.

At bottom, then, the majority rests its conclusion on one aspect of the challengers’ expert evidence, i.e., that it was not possible to place within Nueces County more than one performing Latino district without breaking county lines. The majority acknowledges the District Court’s finding that the challengers had “ ‘failed to show’  that two majority-Latino districts in Nueces County would not perform,” but waves away that finding by concluding that the District Court “twisted the burden of proof beyond recognition” by “suggest[ing] that a plaintiff might succeed on its §2 claim because its expert failed to show that the necessary factual basis for the claim could not be established.” Ante, at 37. That conclusion is only possible because the majority closes its eyes to significant evidence in the record and misrepresents the District Court’s conclusion about the potential for creating two performing Latino-majority districts in Nueces County.

The majority, of course, is right on one thing: The District Court recognized that the challengers’ expert opined that the two HCVAP-majority districts would not perform based on the results of an exogenous election index. See ante, at 36. But the majority ignores that the District Court rejected that expert’s conclusion because “the results of an exogenous election index alone will not determine opportunity,” as “[s]uch indices often do not mirror endogenous election performance.” 267 F. Supp. 3d, at 778. Instead of “just relying on an exogenous election index to measure opportunity,” the District Court “conduct[ed] an intensely local appraisal to determine whether real electoral opportunity exists.” Ibid.

That “intensely local appraisal” resulted in a lengthy analysis that considered, among other facts: that Texas had a long “history of voting-related discrimination”; that “racially polarized voting exist[s] in Nueces County and its house district elections, the level is high, and the high degree of Anglo bloc voting plays a role in the defeat of Hispanic candidates”; “that Hispanics, including in Nueces County, suffer a ‘continuing pattern of disadvantage’ relative to non-Hispanics”; that population growth in the county “was [driven by] Hispanic growth” and that the “HCVAP continues to climb”; that the districts “include demographic distributions strongly favoring Hispanic voters,” and that the “numbers translate into a significant advantage in house district elections”; and that data analysis showed that “performance for Latinos increased significantly in presidential election years,” which “indicates that the districts provide potential to elect.” Id., at 778–782.[20]

The District Court’s focus on the history of the county as well as its potential performance going forward was an important point of departure from the challengers’ expert, who considered only the former. See LULAC, 548 U. S., at 442 (noting “a significant distinction” in analysis of what district performance “ ‘had been’ ” compared to “how it would operate today . . . given the growing Latino political power in the district”). The District Court also found the expert’s analysis lacking in other key respects. Namely, the District Court noted that one of the majority-HCVAP districts “provides opportunity, at least in presidential election years”; that “[m]ost of the elections in [the exogenous election] index did not involve a Latino Democrat candidate”; and that the expert “only looked at statewide races and no county races,” even though it was “conceiv- able that, in competitive local races with Latino candidates, Hispanic voters would mobilize in significantly higher numbers.” 267 F. Supp. 3d, at 781 (emphasis in original).

Based on this review of the evidence, the District Court concluded “that Hispanics have equal opportunity in two districts drawn wholly within Nueces County (or at least [the challengers] failed to show that they do not).” Id., at 782. It further explained that, whereas the “evidence shows that two HCVAP-districts could have been drawn that would provide Hispanics with equal electoral opportunity, . . . the evidence does not show that the Legislature was required to break the County Line Rule to draw what [the challengers] consider to be ‘effective’ districts.” Id., at 783.

When read in the context of the full analysis just detailed, it is clear that the District Court was not “twist[ing] the burden of proof,” ante, at 37, when it observed that the challengers “failed to show that” the two HCVAP-majority districts drawn wholly within Nueces County would not perform. That statement plainly refers to the challengers’ failure to rebut the finding that the two districts wholly within Nueces County provided equal electoral opportunity to Hispanics, as they needed to do to show that §2 required breaking the County Line Rule. If anything is “twisted . . . beyond recognition,” ibid., it is the majority opinion’s description of the District Court’s findings. For while relying on a reference to what the challengers’ expert opined, the majority wholly ignores the District Court’s lengthy discussion rejecting that opinion on the basis of other evidence in the record.[21]

This Court has been clear that “the ultimate right of §2 is equality of opportunity.” Johnson v. De Grandy, 512 U. S. 997, 1014, n. 11 (1994). The District Court found that two HCVAP-majority districts drawn wholly within Nueces County provided such “equality of opportunity,” and its findings of fact are not clearly erroneous. Only by selectively reading the factual record and ignoring the relevant analysis of those facts can the majority escape the §2 results violation that flows from those findings.

IV

The Equal Protection Clause of the Fourteenth Amendment and §2 of the Voting Rights Act secure for all voters in our country, regardless of race, the right to equal participation in our political processes. Those guarantees mean little, however, if courts do not remain vigilant in curbing States’ efforts to undermine the ability of minority voters to meaningfully exercise that right. For although we have made progress, “voting discrimination still exists; no one doubts that.” Shelby County, 570 U. S., at 536.

The Court today does great damage to that right of equal opportunity. Not because it denies the existence of that right, but because it refuses its enforcement. The Court intervenes when no intervention is authorized and blinds itself to the overwhelming factual record below. It does all of this to allow Texas to use electoral maps that, in design and effect, burden the rights of minority voters to exercise that most precious right that is “preservative of all rights.” Yick Wo v. Hopkins, 118 U. S. 356, 370 (1886); see Husted v. A. Philip Randolph Institute, 584 U. S. ___, ___ (2018) (Sotomayor, J., dissenting) (slip op., at 5) (“Our democracy rests on the ability of all individuals, regardless of race, income, or status, to exercise their right to vote”). Because our duty is to safeguard that fundamental right, I dissent.