League of United Latin American Citizens v. Perry

announced the judgment of the Court and delivered the opinion of the Court with respect to Parts II-A and III, an opinion with respect to Parts I and IV, in *409which The Chief Justice and Justice Alito join, an opinion with respect to Parts II-B and II-C, and an opinion with respect to Part II-D, in which Justice Souter and Justice Ginsburg join.

These four consolidated cases are appeals from a judgment entered by the United States District Court for the Eastern District of Texas. Convened as a three-judge court under 28 U. S. C. § 2284, the court heard appellants’ constitutional and statutory challenges to a 2003 enactment of the Texas State Legislature that drew new district lines for the 32 seats Texas holds in the United States House of Representatives. (Though appellants do not join each other as to all claims, for the sake of convenience we refer to appellants collectively.) In 2004 the court entered judgment for appellees and issued detailed findings of fact and conclusions of law. Session v. Perry, 298 F. Supp. 2d 451 (per curiam). This Court vacated that decision and remanded for consideration in light of Vieth v. Jubelirer, 541 U. S. 267 (2004). 543 U. S. 941 (2004). The District Court reexamined appellants’ political gerrymandering claims and, in a second careful opinion, again held for the defendants. Henderson v. Perry, 399 F. Supp. 2d 756 (2005). These appeals followed, and we noted probable jurisdiction. 546 U. S. 1074 (2005).

Appellants contend the new plan is an unconstitutional partisan gerrymander and that the redistricting statewide violates § 2 of the Voting Rights Act of 1965, 79 Stat. 437, as amended, 42 U. S. C. § 1973. Appellants also contend that the use of race and polities in drawing lines of specific districts violates the First Amendment and the Equal Protection Clause of the Fourteenth Amendment. The three-judge panel, consisting of Circuit Judge Higginbotham and District Judges Ward and Rosenthal, brought considerable experience and expertise to the instant action, based on their knowledge of the State’s people, history, and geography. Judges Higginbotham and Ward, moreover, had served on the three-judge court that drew the plan the Texas Legisla*410ture replaced in 2003, so they were intimately familiar with the history and intricacies of the eases.

We affirm the District Court’s dispositions on the statewide political gerrymandering claims and the Voting Rights Act claim against District 24. We reverse and remand on the Voting Rights Act claim with respect to District 23. Because we do not reach appellants’ race-based equal protection claim or the political gerrymandering claim as to District 23, we vacate the judgment of the District Court on these claims.

I

To set out a proper framework for the cases, we first recount the history of the litigation and recent districting in Texas. An appropriate starting point is not the reapportionment in 2000 but the one from the census in 1990.

The 1990 census resulted in a 30-seat congressional delegation fori Texas, an increase of 3 seats over the 27 representatives allotted to the State in the decade before. See Bush v. Vera, 517 U. S. 952, 956-957 (1996). In 1991 the Texas Legislature drew new district lines. At the time, the Democratic Party controlled both houses in the state legislature, the governorship, and 19 of the State’s 27 seats in Congress. Yet change appeared to be on the horizon. In the previous 30 years the Democratic Party’s post-Reconstruction dominance over the Republican Party had eroded, and by 1990 the Republicans received 47% of the statewide vote, while the Democrats received 51%. Henderson, supra, at 763; Brief for Appellee Perry et al. in No. 05-204 etc., p. 2 (hereinafter Brief for State Appellees).

Faced with a Republican opposition that could be moving toward majority status, the state legislature drew a congressional redistricting plan designed to favor Democratic candidates. Using then-emerging computer technology to draw district lines with artful precision, the legislature enacted a plan later described as the “shrewdest gerrymander of the 1990s.” M. Barone, R. Cohen, & C. Cook, Almanac of Amer*411ican Politics 2002, p. 1448 (2001). See Henderson, supra, at 767, and n. 47. Although the 1991 plan was enacted by the state legislature, Democratic Congressman Martin Frost was acknowledged as its architect. Session, supra, at 482. The 1991 plan “carefully constructs democratic districts ‘with incredibly convoluted lines’ and packs ‘heavily Republican’ suburban areas into just a few districts.” Henderson, supra, at 767, n. 47 (quoting M. Barone & R. Cohen, Almanac of American Politics 2004, p. 1510 (2003) (hereinafter 2004 Almanac)).

Voters who considered this unfair and unlawful treatment sought to invalidate the 1991 plan as an unconstitutional partisan gerrymander, but to no avail. See Terrazas v. Slagle, 789 F. Supp. 828, 833 (WD Tex. 1992); Terrazas v. Slagle, 821 F. Supp. 1162, 1175 (WD Tex. 1993) (per curiam). The 1991 plan realized the hopes of Democrats and the fears of Republicans with respect to the composition of the Texas congressional delegation. The 1990’s were years of continued growth for the Texas Republican Party, and by the end of the decade it was sweeping elections for statewide office. Nevertheless, despite carrying 59% of the vote in statewide elections in 2000, the Republicans only won 13 congressional seats to the Democrats’ 17. Henderson, supra, at 763.

These events likely were not forgotten by either party when it came time to draw congressional districts in conformance with the 2000 census and to incorporate two additional seats for the Texas delegation. The Republican Party controlled the governorship and the State Senate; it did not yet control the State House of Representatives, however. As so constituted, the legislature was unable to pass a redistricting scheme, resulting in litigation and the necessity of a court-ordered plan to comply with the Constitution’s one-person, one-vote requirement. See Balderas v. Texas, Civ. Action No. 6:01CV158 (ED Tex., Nov. 14, 2001) (per curiam), summarily aff'd, 536 U. S. 919 (2002), App. E to Juris. Statement in No. 05-276, p. 202a (hereinafter Balderas, App. E to *412Juris. Statement). The congressional districting map resulting from the Balderas litigation is known as Plan 1151C.

As we have said, two members of the three-judge court that drew Plan 1151C later served on the three-judge court that issued the judgment now under review. Thus we have the benefit of their candid comments concerning the redistricting approach taken in the Balderas litigation. Conscious that the primary responsibility for drawing congressional districts is given to political branches of government, and hesitant to “und[o] the work of one political party for the benefit of another,” the three-judge Balderas court sought to apply “only ‘neutral’ redistricting standards” when drawing Plan 1151C. Henderson, 399 F. Supp. 2d, at 768. Once the District Court applied these principles—such as placing the two new seats in high-growth areas, following county and voting precinct lines, and avoiding the pairing of incumbents—“the drawing ceased, leaving the map free of further change except to conform it to one-person, one-vote.” Ibid. Under Plan 1151C, the 2002 congressional elections resulted in a 17-to-15 Democratic majority in the Texas delegation, compared to a 59% to 40% Republican majority in votes for statewide office in 2000. Id., at 763-764. Reflecting on the Balderas plan, the District Court in Henderson was candid to acknowledge “[t]he practical effect of this effort was to leave the 1991 Democratic Party gerrymander largely in place as a ‘legal’ plan.” 399 F. Supp. 2d, at 768.

The continuing influence of a court-drawn map that “perpetuated much of [the 1991] gerrymander,” ibid., was not lost on Texas Republicans when, in 2003, they gained control of the State House of Representatives and, thus, both houses of the legislature. The Republicans in the legislature “set out to increase their representation in the congressional delegation.” Session, 298 F. Supp. 2d, at 471. See also id., at 470 (“There is little question but that the single-minded purpose of the Texas Legislature in enacting [a new plan] was to gain partisan advantage”). After a protracted partisan *413struggle, during which Democratic legislators left the State for a time to frustrate quorum requirements, the legislature enacted a new congressional districting map in October 2003. It is called Plan 1374C. The 2004 congressional elections did not disappoint the plan’s drafters. Republicans won 21 seats to the Democrats’ 11, while also obtaining 58% of the vote in statewide races against the Democrats’ 41%. Henderson, supra, at 764.

Soon after Texas enacted Plan 1374C, appellants challenged it in court, alleging a host of constitutional and statutory violations. Initially, the District Court entered judgment against appellants on all their claims. See Session, 298 F. Supp. 2d, at 457; id., at 515 (Ward, J., concurring in part and dissenting in part). Appellants sought relief here and, after their jurisdictional statements were filed, this Court issued Vieth v. Jubelirer. Our order vacating the District Court judgment and remanding for consideration in light of Vieth was issued just weeks before the 2004 elections. See 543 U. S. 941 (Oct. 18, 2004). On remand, the District Court, believing the scope of its mandate was limited to questions of political gerrymandering, again rejected appellants’ claims. Henderson, 399 F. Supp. 2d, at 777-778. Judge Ward would have granted relief under the theory—presented to the court for the first time on remand—that mid-decennial redistricting violates the one-person, one-vote requirement, but he concluded such an argument was not within the scope of the remand mandate. Id., at 779,784-785 (specially concurring).

II

A

Based on two similar theories that address the mid-decade character of the 2003 redistricting, appellants now argue that Plan 1374C should be invalidated as an unconstitutional partisan gerrymander. In Davis v. Bandemer, 478 U. S. 109 (1986), the Court held that an equal protection challenge to a political gerrymander presents a justiciable case or contro*414versy, id., at 118-127, but there was disagreement over what substantive standard to apply. Compare id., at 127-137 (plurality opinion), with id., at 161-162 (Powell, J., concurring in part and dissenting in part). That disagreement persists. A plurality of the Court in Vieth would have held such challenges to be nonjusticiable political questions, but a majority declined to do so. See 541 U. S., at 306 (Kennedy, J., concurring in judgment); id., at 317 (STEVENS, J., dissenting); id., at 343 (Souter, J., dissenting); id., at 355 (Breyer, J., dissenting). We do not revisit the justiciability holding but do proceed to examine whether appellants’ claims offer the Court a manageable, reliable measure of fairness for determining whether a partisan gerrymander violates the Constitution.

B

Before addressing appellants’ arguments on mid-decade redistricting, it is appropriate to note some basic principles on the roles the States, Congress, and the courts play in determining how congressional districts are to be drawn. Article I of the Constitution provides:

“Section 2. The House of Representatives shall be composed of Members chosen every second Year by the People of the several States ....

“Section 4. The Times, Places and Manner of holding Elections for ... Representatives, shall be prescribed in each State by the Legislature thereof; but the Congress may at any time by Law make or alter such Regulations ... .”

This text, we have explained, “leaves with the States primary responsibility for apportionment of their federal congressional . . . districts.” Growe v. Emison, 507 U. S. 25, 34 (1993); see also Chapman v. Meier, 420 U. S. 1, 27 (1975) (“[Rjeapportionment is primarily the duty and responsibility of the State through its legislature or other body”); Smiley *415v. Holm, 285 U. S. 355, 366-367 (1932) (reapportionment implicated State’s powers under Art. I, § 4). Congress, as the text of the Constitution also provides, may set further requirements, and with respect to districting it has generally required single-member districts. See U. S. Const., Art. I, §4; Pub. L. 90-196, 81 Stat. 581, 2 U. S. C. §2c; Branch v. Smith, 538 U. S. 254, 266-267 (2003). But see id., at 275 (plurality opinion) (multimember districts permitted by 55 Stat. 762, 2 U. S. C. § 2a(c) in limited circumstances). With respect to a mid-decade redistricting to change districts drawn earlier in conformance with a decennial census, the Constitution and Congress state no explicit prohibition.

Although the legislative branch plays the primary role in congressional redistricting, our precedents recognize an important role for the courts when a districting plan violates the Constitution. See, e. g., Wesberry v. Sanders, 376 U. S. 1 (1964). This litigation is an example, as we have discussed. When Texas did not enact a plan to comply with the one-person, one-vote requirement under the 2000 census, the District Court found it necessary to draw a redistricting map on its own. That the federal courts sometimes are required to order legislative redistrieting, however, does not shift the primary locus of responsibility.

“Legislative bodies should not leave their reapportionment tasks to the federal courts; but when those with legislative responsibilities do not respond, or the imminence of a state election makes it impractical for them to do so, it becomes the ‘unwelcome obligation’ of the federal court to devise and impose a reapportionment plan pending later legislative action.” Wise v. Lipscomb, 437 U. S. 535, 540 (1978) (principal opinion) (quoting Connor v. Finch, 431 U. S. 407, 415 (1977)).

Quite apart from the risk of acting without a legislature’s expertise, and quite apart from the difficulties a court faces in drawing a map that is fair and rational, see id., at 414-415, *416the obligation placed upon the Federal Judiciary is unwelcome because drawing lines for congressional districts is one of the most significant acts a State can perform to ensure citizen participation in republican self-governance. That Congress is the federal body explicitly given constitutional power over elections is also a noteworthy statement of preference for the democratic process. As the Constitution vests redistricting responsibilities foremost in the legislatures of the States and in Congress, a lawful, legislatively enacted plan should be preferable to one drawn by' the courts.

It should follow, too, that if a legislature acts to replace a court-drawn plan with one of its own design, no presumption of impropriety should attach to the legislative decision to act. As the District Court noted here, Session, 298 F. Supp. 2d, at 460-461, our decisions have assumed that state legislatures are free to replace court-mandated remedial plans by enacting redistricting plans of their own. See, e. g., Upham v. Seamon, 456 U. S. 37, 44 (1982) (per curiam); Wise, supra, at 540 (principal opinion) (quoting Connor, supra, at 415); Burns v. Richardson, 384 U. S. 73, 85 (1966); Reynolds v. Sims, 377 U. S. 533, 587 (1964). Underlying this principle is the assumption that to prefer a court-drawn plan to a legislature’s replacement would be contrary to the ordinary and proper operation of the political process. Judicial respect for legislative plans, however, cannot justify legislative reliance on improper criteria for districting determinations. With these considerations in mind, I now turn to consider appellants’ challenges to the new redistricting plan.

C

Appellants claim that Plan 1374C, enacted by the Texas Legislature- in 2003, is an unconstitutional political gerrymander. A decision, they claim, to effect mid-decennial redistricting, when solely motivated by partisan objectives, violates equal protection and the First Amendment because it *417serves no legitimate public purpose and burdens one group because of its political opinions and affiliation. The mid-decennial nature of the redistricting, appellants say, reveals the legislature’s sole motivation. Unlike Vieth, where the legislature acted in the context of a required decennial redistricting, the Texas Legislature voluntarily replaced a plan that itself was designed to comply with new census data. Because Texas had “no constitutional obligation to act at all” in 2003, Brief for Appellant Jackson et al. in No. 05-276, p. 26, it is hardly surprising, according to appellants, that the District Court found “[t]here is little question but that the single-minded purpose of the Texas Legislature in enacting Plan 1374C was to gain partisan advantage” for the Republican majority over the Democratic minority, Session, supra, at 470.

A rule, or perhaps a presumption, of invalidity when a mid-decade redistricting plan is adopted solely for partisan motivations is a salutary one, in appellants’ view, for then courts need not inquire about, nor parties prove, the discriminatory effects of partisan gerrymandering—a matter that has proved elusive since Bandemer. See Vieth, 541 U. S., at 281 (plurality opinion); Bandemer, 478 U. S., at 127 (plurality opinion). Adding to the test’s simplicity is that it does not quibble with the drawing of individual district lines but challenges the decision to redistrict at all.

For a number of reasons, appellants’ case for adopting their test is not convincing. To begin with, the state appellees dispute the assertion that partisan gain was the “sole” motivation for the decision to replace Plan 1151C. There is some merit to that criticism, for the pejorative label overlooks indications that partisan motives did not dictate the plan in its entirety. The legislature does seem to have decided to redistrict with the sole purpose of achieving a Republican congressional majority, but partisan aims did not guide every line it drew. As the District Court found, the contours of some contested district lines were drawn based *418on more mundane and local interests. Session, supra, at 472-473. The state appellees also contend, and appellants do not contest, that a number of line-drawing requests by Democratic state legislators were honored. Brief for State Appellees 34.

Evaluating the legality of acts arising out of mixed motives can be complex, and affixing a single label to those acts can be hazardous, even when the actor is an individual performing a discrete act. See, e. g., Hartman v. Moore, 547 U. S. 250, 259-260 (2006). When the actor is a legislature and the act is a composite of manifold choices, the task can be even more daunting. Appellants’ attempt to separate the legislature’s sole motive for discarding Plan 1151C from the complex of choices it made while drawing the lines of Plan 1374C seeks to avoid that difficulty. We should be skeptical, however, of a claim that seeks to invalidate a statute based on a legislature’s unlawful motive but does so without reference to the content of the legislation enacted.

Even setting this skepticism aside, a successful claim attempting to identify unconstitutional acts of partisan gerrymandering must do what appellants’ sole-motivation theory explicitly disavows: show a burden, as measured by a reliable standard, on the complainants’ representational rights. For this reason, a majority of the Court rejected a test proposed in Vieth that is markedly similar to the one appellants present today. Compare 541 U. S., at 336 (Stevens, J., dissenting) (“Just as race can be a factor in, but cannot dictate the outcome of, the districting process, so too can partisanship be a permissible consideration in drawing district lines, so long as it does not predominate”), and id., at 338 (“[A]n acceptable rational basis can be neither purely personal nor purely partisan”), with id., at 292-295 (plurality opinion), and id., at 307-308 (Kennedy, J., concurring in judgment).

The sole-intent standard offered here is no more compelling when it is linked to the circumstance that Plan 1374C is mid-decennial legislation. The text and structure of the *419Constitution and our case law indicate there is nothing inherently suspect about a legislature’s decision to replace mid-decade a court-ordered plan with one of its own. And even if there were, the fact of mid-decade redistricting alone is no sure indication of unlawful political gerrymanders. Under appellants’ theory, a highly effective partisan gerrymander that coincided with decennial redistricting would receive less scrutiny than a bumbling, yet solely partisan, mid-decade redistricting. More concretely, the test would leave untouched the 1991 Texas redistricting, which entrenched a party on the verge of minority status, while striking down the 2003 redistricting plan, which resulted in the majority Republican Party capturing a larger share of the seats. A test that treats these two similarly effective power plays in such different ways does not have the reliability appellants ascribe to it.

Furthermore, compared to the map challenged in Vieth, which led to a Republican majority in the congressional delegation despite a Democratic majority in the statewide vote, Plan 1374C can be seen as making the party balance more congruent to statewide party power. To be sure, there is no constitutional requirement of proportional representation, and equating a party’s statewide share of the vote with its portion of the congressional delegation is a rough measure at best. Nevertheless, a congressional plan that more closely reflects the distribution of state party power seems a less likely vehicle for partisan discrimination than one that entrenches an electoral minority. See Gaffney v. Cummings, 412 U. S. 735, 754 (1973). By this measure, Plan 1374C can be seen as fairer than the plan that survived in Vieth and the two previous Texas plans—all three of which would pass the modified sole-intent test that Plan 1374C would fail.

A brief for one of the amici proposes a symmetry standard that would measure partisan bias by “comparing] how both parties would fare hypothetically if they each (in turn) had received a given percentage of the vote.” Brief for Gary *420King et al. 5. Under that standard the measure of a map’s bias is the extent to which a majority party would fare better than the minority party, should their respective shares of the vote reverse. Amici’s, proposed standard does not compensate for appellants’ failure to provide a reliable measure of fairness. The existence or degree of asymmetry may in large part depend on conjecture about where possible vote-switchers will reside. Even assuming a court could choose reliably among different models of shifting voter preferences, we are wary of adopting a constitutional standard that invalidates a map based on unfair results that would occur in a hypothetical state of affairs. Presumably such a challenge could be litigated if and when the feared inequity arose. Cf. Abbott Laboratories v. Gardner, 387 U. S. 136, 148 (1967). More fundamentally, the counterfactual plaintiff would face the same problem as the present, actual appellants: providing a standard for deciding how much partisan dominance is too much. Without altogether discounting its utility in redistricting planning and litigation, I would conclude asymmetry alone is not a reliable measure of unconstitutional partisanship.

In the absence of any other workable test for judging partisan gerrymanders, one effect of appellants’ focus on mid-decade redistricting could be to encourage partisan excess at the outset of the decade, when a legislature redistricts pursuant to its decennial constitutional duty and is then immune from the charge of sole motivation. If mid-decade redistricting were barred or at least subject to close judicial oversight, opposition legislators would also have every incentive to prevent passage of a legislative plan and try their luck with a court that might give them a better deal than negotiation with their political rivals. See Henderson, 399 F. Supp. 2d, at 776-777.

D

Appellants’ second political gerrymandering theory is that mid-decade redistricting for exclusively partisan purposes *421violates the one-person, one-vote requirement. They observe that population variances in legislative districts are tolerated only if they “are unavoidable despite a good-faith effort to achieve absolute equality, or for which justification is shown.” Karcher v. Daggett, 462 U. S. 725, 730 (1983) (quoting Kirkpatrick v. Preisler, 394 U. S. 526, 531 (1969); internal quotation marks omitted). Working from this unchallenged premise, appellants contend that, because the population of Texas has shifted since the 2000 census, the 2003 redistrieting, which relied on that census, created unlawful interdistrict population variances.

To distinguish the variances in Plan 1374C from those of ordinary, 3-year-old districting plans or belatedly drawn court-ordered plans, appellants again rely on the voluntary, mid-decade nature of the redistricting and its partisan motivation. Appellants do not contend that a decennial redistricting plan would violate equal representation three or five years into the decade if the State’s population had shifted substantially. As they must, they concede that States operate under the legal fiction that their plans are constitutionally apportioned throughout the decade, a presumption that is necessary to avoid constant redistricting, with accompanying costs and instability. See Georgia v. Ashcroft, 539 U. S. 461, 488, n. 2 (2003); Reynolds, 377 U. S., at 583. Appellants agree that a plan implemented by a court in 2001 using 2000 population data also enjoys the benefit of the so-called legal fiction, presumably because belated court-drawn plans promote other important interests, such as ensuring a plan complies with the Constitution and voting rights legislation.

In appellants’ view, however, this fiction should not provide a safe harbor for a legislature that enacts a voluntary, mid-decade plan overriding a legal court-drawn plan, thus “ ‘unnecessarily’ ” creating population variance “when there was no legal compulsion” to do so. Brief for Appellant Travis County et al. in No. 05-254, p. 18. This is particularly so, appellants say, when a legislature acts because of an *422exclusively partisan motivation. Under appellants’ theory this improper motive at the outset seems enough to condemn the map for violating the equal-population principle. For this reason, appellants believe that the State cannot justify under Karcher v. Daggett the population variances in Plan 1374C because they are the product of partisan bias and the desire to eliminate all competitive districts.

As the District Court noted, this is a test that turns not on whether a redistricting furthers equal-population principles but rather on the justification for redrawing a plan in the first place. Henderson, supra, at 776. In that respect appellants’ approach merely restates the question whether it was permissible for the Texas Legislature to redraw the districting map. Appellants’ answer, which mirrors their attack on mid-decennial redistricting solely motivated by partisan considerations, is unsatisfactory for reasons we have already discussed.

Appellants also contend that the legislature intentionally sought to manipulate population variances when it enacted Plan 1374C. There is, however, no District Court finding to that effect, and appellants present no specific evidence to support this serious allegation of bad faith. Because appellants have not demonstrated that the legislature’s decision to enact Plan 1374C constitutes a violation of the equal-population requirement, we find unavailing their subsidiary reliance on Larios v. Cox, 300 F. Supp. 2d 1320 (ND Ga.) (per curiam), summarily aff'd, 542 U. S. 947 (2004). In Larios, the District Court reviewed the Georgia Legislature’s decennial redistricting of its State Senate and House of Representatives districts and found deviations from the equal-population requirement. The District Court then held the objectives of the drafters, which included partisan interests along with regionalist bias and inconsistent incumbent protection, did not justify those deviations. 300 F. Supp. 2d, at 1351-1352.. The Larios holding and its examination of the legislature’s motivations were relevant only in response to *423an equal-population violation, something appellants have not established here. Even in addressing political motivation as a justification for an equal-population violation, moreover, Latios does not give clear guidance. The panel explained it “need not resolve the issue of whether or when partisan advantage alone may justify deviations in population” because the plans were “plainly unlawful” and any partisan motivations were “bound up inextricably” with other clearly rejected objectives. Id., at 1352.

In sum, we disagree with appellants’ view that a legislature’s decision to override a valid, court-drawn plan mid-decade is sufficiently suspect to give shape to a reliable standard for identifying unconstitutional political gerrymanders. We conclude that appellants have established no legally impermissible use of political classifications. For this reason, they state no claim on which relief may be granted for their statewide challenge.

Ill

Plan 1374C made changes to district lines in south and west Texas that appellants challenge as violations of §2 of the Voting Rights Act and the Equal Protection Clause of the Fourteenth Amendment. The most significant changes occurred to District 23, which—both before and after the redistricting—covers a large land area in west Texas, and to District 25, which earlier included Houston but now includes a different area, a north-south strip from Austin to the Rio Grande Valley.

After the 2002 election, it became apparent that District 23 as then drawn had an increasingly powerful Latino population that threatened to oust the incumbent Republican, Henry Bonilla. Before the 2003 redistricting, the Latino share of the citizen voting-age population was 57.5%, and Bonilla’s support among Latinos had dropped with each successive election since 1996. Session, 298 F. Supp. 2d, at 488-489. In 2002, Bonilla captured only 8% of the Latino vote, *424ibid., and 51.5% of the overall vote. Faced with this loss of voter support, the legislature acted to protect Bonilla’s incumbency by changing the lines—and hence the population mix—of the district. To begin with, the new plan divided Webb County and the city of Laredo, on the Mexican border, that formed the county’s population base. Webb County, which is 94% Latino, had previously rested entirely within District 23; under the new plan, nearly 100,000 people were shifted into neighboring District 28. Id., at 489. The rest of the county, approximately 93,000 people, remained in District 23. To replace the numbers District 23 lost, the State added voters in counties comprising a largely Anglo, Republican area in central Texas. Id., at 488. In the newly drawn district, the Latino share of the citizen voting-age population dropped to 46%, though the Latino share of the total voting-age population remained just over 50%. Id., at 489.

These changes required adjustments elsewhere, of course, so the State inserted a third district between the two districts to the east of District 23, and extended all three of them farther north. New District 25 is a long, narrow strip that winds its way from McAllen and the Mexican-border towns in the south to Austin, in the center of the State and 300 miles away. Id., at 502. In between it includes seven full counties, but 77% of its population resides in split counties at the northern and southern ends. Of this 77%, roughly half reside in Hidalgo County, which includes McAllen, and half are in Travis County, which includes parts of Austin. Ibid. The Latinos in District 25, comprising 55% of the district’s citizen voting-age population, are also mostly divided between the two distant areas, north and south. Id., at 499. The Latino communities at the opposite ends of District 25 have divergent “needs and interests,” id., at 502, owing to. “differences in socio-economic status, education, employment, health, and other characteristics,” id., at 512;

The District Court summed up the purposes underlying the redistricting in south and west Texas: “The change to *425Congressional District 23 served the dual goal of increasing Republican seats in general and protecting Bonilla’s incumbency in particular, with the additional political nuance that Bonilla would be reelected in a district that had a majority of Latino voting age population—although clearly not a majority of citizen voting age population and certainly not an effective voting majority.” Id., at 497. The goal in creating District 25 was just as clear: “[t]o avoid retrogression under § 5” of the Voting Rights Act given the reduced Latino voting strength in District 23. Id., at 489.

A

The question we address is whether Plan 1374C violates § 2 of the Voting Rights Act. A State violates § 2

“if, based on the totality of circumstances, it is shown that the political processes leading to nomination or election in the State or political subdivision are not equally open to participation by members of [a racial group] in that its members have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice.” 42 U.S. C. § 1973(b).

The Court has identified three threshold conditions for establishing a §2 violation: (1) the racial g