with whom Justice Ginsburg and Justice Breyer join, dissenting.
The 1990 census revealed that Texas’ population had grown, over the past decade, almost twice as fast as the population of the country as a whole. As a result, Texas was entitled to elect three additional Representatives to the United States Congress, enlarging its delegation from 27 to 30. Because Texas’ growth was concentrated in south Texas and the cities of Dallas and Houston, the state legislature concluded that the new congressional districts should be carved out of existing districts in those areas. The consequences of the political battle that produced the new map are some of the most oddly shaped congressional districts in the United States.
Today, the Court strikes down three of Texas’ majority-minority districts, concluding, inter alia, that their odd shapes reveal that the State impermissibly relied on predominantly racial reasons when it drew the districts as it did. For two reasons, I believe that the Court errs in striking down those districts.
First, I believe that the Court has misapplied its own tests for racial gerrymandering, both by applying strict scrutiny to all three of these districts, and then by concluding that none can meet that scrutiny. In asking whether strict scrutiny should apply, the Court improperly ignores the “complex interplay” of political and geographical considerations that went into the creation of Texas’ new congressional districts, Miller v. Johnson, 515 U. S. 900, 915-916 (1995), and *1004focuses exclusively on the role that race played in the State’s decisions to adjust the shape of its districts. A quick comparison of the unconstitutional majority-minority districts with three equally bizarre majority-Anglo districts, compare ante, at Appendixes A-C, with infra, at Appendixes A-C, demonstrates that race was not necessarily the predominant factor contorting the district lines. I would follow the fair implications of the District Court’s findings,1 and conclude that Texas’ entire map is a political, not a racial, gerrymander.2 See Part IV, infra.
Even if strict scrutiny applies, I would find these districts constitutional, for each considers race only to the extent necessary to comply with the State’s responsibilities under the Voting Rights Act while achieving other race-neutral political and geographical requirements. The plurality’s finding to the contrary unnecessarily restricts the ability of States to conform their behavior to the Voting Rights Act while simultaneously complying with other race-neutral goals. See Part V, infra.
Second, even if I concluded that these districts failed an appropriate application of this still-developing law to appropriately read facts, I would not uphold the District Court decision. The decisions issued today serve merely to rein*1005force my conviction that the Court has, with its “analytically distinct” jurisprudence of racial gerrymandering, Shaw v. Reno, 509 U. S. 630, 652 (1993) (Shaw I), struck out into a jurisprudential wilderness that lacks a definable constitutional core and threatens to create harms more significant than any suffered by the individual plaintiffs challenging these districts. See Parts VT-VII, infra; Shaw v. Hunt, ante, at 918-919 (Shaw II) (Stevens, J., dissenting). Though we travel ever farther from it with each passing decision, I would return to the well-traveled path that we left in Shaw I.
I
The factors motivating Texas’ redistricting plan are clearly revealed in the results of the 1992 elections. Both before and immediately after the 1990 census, the Democratic Party was in control of the Texas Legislature. Under the new map in 1992, more than two-thirds of the Districts—including each of the new ones—elected Democrats, even though Texas voters are arguably more likely to vote Republican than Democrat.3 Incumbents of both parties were just as successful: 26 of the 27 incumbents were reelected, while each of the three new districts elected a state legislator who had essentially acted as an incumbent in the districting process,4 giving “incumbents” a 97% success rate.
*1006It was not easy for the State to achieve these results while simultaneously guaranteeing that each district enclosed the residence of its incumbent, contained the same number of people, and complied with other federal and state districting requirements. Much of Dallas and Houston, for example, was already represented in Congress by Democrats, and creating new Democratic districts in each city while ensuring politically safe seats for sitting Representatives required significant political gerrymandering. This task was aided by technological and informational advances that allowed the State to adjust lines on the scale of city blocks, thereby guaranteeing twists and turns that would have been essentially impossible in any earlier redistricting.5 “[T]he result of the Legislature’s efforts,” the District Court concluded, was “a *1007crazy-quilt of districts” that bore little resemblance to “the work of public-spirited representatives.” Vera v. Richards, 861 F. Supp. 1304, 1309 (SD Tex. 1994); see, e.g., Appendixes A-D.
It is clear that race also played a role in Texas’ redistricting decisions. According to the 1990 Census, Texas contained 16,986,510 residents, of whom 22.5% were of Hispanic origin, and 11.6% were non-Hispanic African-American. 861 F. Supp., at 1311. Under the pre-1990 districting scheme, Texas’ 27-member delegation included four Hispanics and one African-American. In Harris County, a concentrated Hispanic community was divided among several majority-Anglo districts as well as the majority-minority District 18. In Dallas County, the majority-black community in South Dallas was split down the middle between two majority-Anglo districts. The legislature was well aware, after the 1990 census, that the minority communities in each county were disproportionately responsible for the growth in population that gained three representatives for the State. Given the omnipresence of § 2 of the Voting Rights Act, 42 U. S. C. § 1973, the demographics of the two communities, and the pressure from leaders of the minority communities in those cities, it was not unreasonable — and certainly not invidious discrimination of any sort — for the State to accede to calls for the creation of majority-minority districts in both cities.6
*1008While complying with a multitude of other political and legal requirements, then, Texas created three new majority-minority congressional districts and significantly reconfigured one pre-existing district. The District Court concluded that the- State impermissibly emphasized race over nonracial factors when it drew two of these new districts (District 30 in Dallas and District 29 in Houston) and the reconfigured District 18 in Houston. To determine whether the Court correctly affirms that decision, I begin, as does the plurality, by asking whether “strict scrutiny” should be applied to the State’s consideration of race in the creation of these majority-minority districts.
II
We have traditionally applied strict scrutiny to state action that discriminates on the basis of race. Prior to Shaw I, however, we did so only in cases in which that discrimination harmed an individual or set of individuals because of their race. In contrast, the harm identified in Shaw I and its progeny is much more diffuse. See Shaw II, ante, at 921-925 (Stevens, J., dissenting). Racial gerrymandering of the sort being addressed in these cases is “discrimination” only in the sense that the lines are drawn based on race, not in the sense that harm is imposed on specific persons on account of their race. Ante, at 923-924 (Stevens, J., dissenting).
Aware of this distinction, a majority of this Court has endorsed a position crucial to a proper evaluation of Texas’ congressional districts: Neither the Equal Protection Clause nor any other provision of the Constitution was offended merely because the legislature considered race when it deliberately *1009created three majority-minority districts.7 The plurality’s statement that strict scrutiny “does [not] apply to all cases of intentional creation of majority-minority districts,” ante, at 958, merely caps a long line of discussions, stretching from Shaw I to Shaw II, which have both expressly and implicitly set forth precisely that conclusion.8
*1010The conclusion that race-conscious districting should not always be subject to strict scrutiny merely recognizes that our equal protection jurisprudence can sometimes mislead us with its rigid characterization of suspect classes and levels of scrutiny. As I have previously noted, all equal protection jurisprudence might be described as a form of rational basis scrutiny; we apply “strict scrutiny” more to describe the likelihood of success than the character of the test to be applied. See Cleburne v. Cleburne Living Center, Inc., 473 U. S. 432, 452-453 (1985) (Stevens, J., concurring). Because race has rarely been a legitimate basis for state classifications, and more typically an irrational and invidious ground for discrimination, a “virtually automatic invalidation of racial classifications” has been the natural result of the application of our equal protection jurisprudence. Id., at 453. In certain circumstances, however, when the state action (i) has neither the intent nor effect of harming any particular group, (ii) is not designed to give effect to irrational prejudices held by its citizens but to break them down, and (iii) uses race as a classification because race is “relevant” to the benign goal of the classification, id., at 454, we need not view the action with the typically fatal skepticism that we have used to strike down the most pernicious forms of state behavior. See Wygant v. Jackson Bd. of Ed., 476 U. S. 267, *1011316-317 (1986) (Stevens, J., dissenting); Regents of Univ. of Cal. v. Bakke, 438 U. S. 265, 320 (1978). While the Court insisted in Shaw I that racial classifications of this sort injure the Nation (though not necessarily any particular group) in myriad ways, see 509 U. S., at 647-648, redistricting that complies with the three factors I outline above simply is not the sort of despicable practice that has been taken in the past to exclude minorities from the electoral process. See Shaw II, ante, at 931-933 (Stevens, J., dissenting); Shaw I, 509 U. S., at 682-685 (Souter, J., dissenting); cf., e. g., Gomillion v. Lightfoot, 364 U. S. 339 (1960); Terry v. Adams, 345 U. S. 461 (1953). While any racial classification may risk some stereotyping, the risk of true “discrimination” in this case is extremely tenuous in light of the remedial purpose the classification is intended to achieve and the long history of resistance to giving minorities a full voice in the political process. Given the balancing of subtle harms and strong remedies — a balancing best left to the political process, not to our own well-developed but rigid jurisprudence — the plurality reasonably concludes that race-conscious redistricting is not always a form of “discrimination” to which we should direct our most skeptical eye.
III
While the Court has agreed that race can, to a point, govern the drawing of district lines, it nonetheless suggests that at a certain point, when the State uses race “too much,” illegitimate racial stereotypes threaten to overrun and contaminate an otherwise legitimate redistricting process. In Miller, the Court concluded that this point was reached when “race for its own sake, and not other districting principles, was the . .. dominant and controlling rationale” behind the shape of the district. 515 U. S., at 913. For strict scrutiny to apply, therefore, the plaintiff must demonstrate that “the legislature subordinated traditional race-neutral dis-tricting principles, including but not limited to compactness, *1012contiguity, [and] respect for political subdivisions ... to racial considerations.” Id., at 916; see also id., at 928 (O’Connor, J., concurring) (strict scrutiny should be applied only if State emphasized race in “substantial disregard” for traditional districting principles); ante, at 962 (opinion of O’Connor, J.).
Of course, determining the “predominant” motive of the Texas Legislature, ante, at 959 (citing Miller, 515 U. S., at 916), is not a simple matter.9 The members of that body *1013faced many unrelenting pressures when they negotiated the creation of the contested districts. They had to ensure that there was no deviation in population from district to district.10 They reasonably believed that they had to create districts that would comply with the Voting Rights Act. See supra, at 1007. If the redistricting legislation was to be enacted, they had to secure the support of incumbent Congressmen of both parties by drawing districts that would ensure their election. And all of these desires had to be achieved within a single contiguous district. Every time a district line was shifted from one place to another, each of these considerations was implicated, and additional, compensating shifts were necessary to ensure that all competing goals were simultaneously accomplished. In such a constrained environment, there will rarely be one “dominant and controlling” influence. Nowhere is this better illustrated *1014than in Dallas’ District 30 where, at the very least, it is clear that race was not such an overriding factor.
IV
The plurality lists several considerations which, when taken in combination, lead it to conclude that race, and no other cause, was the predominant factor influencing District 30’s configuration. First, there is the shape itself. Second, there is evidence that the districts were intentionally drawn with consciousness of race in an effort to comply with the Voting Rights Act. Third, the plurality dismisses two race-neutral considerations (communities of interest and incumbency protection) that appellants advanced as race-neutral considerations that led to the odd shape of the districts. Finally, the plurality concludes that race was impermissibly used as a proxy for political affiliation during the course of redistricting. In my opinion, an appropriate reading of the record demonstrates that none of these factors — either singly or in combination — suggests that racial considerations “subordinated” race-neutral districting principles. I discuss each in turn.
Bizarre Shape
As noted, supra, at 1003, and n. 6, Texas’ Legislature concluded that it would add a new district to Dallas County that would incorporate the rapidly growing minority communities in South Dallas. To do so, the new district would have to fit into the existing districts: Before redistricting, most of southern Dallas County (including the African-American communities in South Dallas) was divided between Districts 5 and 24, represented by Democratic Representatives Bryant and Frost, respectively. The middle of the northern section of the county was divided between Districts 3 and 26, both represented by Republicans.
Then-State Senator Johnson began the redistricting process by proposing a compact, Democratic, majority-minority *1015district encompassing all of South Dallas. See App. 139; 861 F. Supp., at 1321, n. 22. Representatives Bryant and Frost objected, however, because the proposed district included not only Johnson’s residence, but their own homes, located within only 10 miles of each other on opposite sides of the city. Furthermore, Johnson’s plan transferred many of Frost and Bryant’s most reliable Democratic supporters into the proposed district. Rather than acquiesce to the creation of this compact majority-minority district, Frost and Bryant insisted that the new district avoid both their own homes and many of the communities that had been loyal to them. Johnson’s plan was, therefore, “quickly abandoned.” Ibid.
To accommodate the incumbents’ desires, District 30 required geographical adjustments that had telling effects on its shape. First, two notches carefully avoiding the residences of and neighborhoods surrounding Frost and Bryant were carved out of District 30’s side. See Appendix D, infra.11 Furthermore, Frost and Bryant retained several communities — many majority-black — along the southern and eastern sides of the proposed district. See generally 861 F. Supp., at 1321-1322.12
*1016Had these communities been retained by District 30, it would have been much more compact. By giving up these voters to Frost and Bryant, however, District 30 was forced to seek out population and Democratic voters elsewhere. The Democratic incumbents had blocked its way to the south and east; north (and, to a lesser extent, west) was the only way it could go.13
It would not have helped the prospects of a Democratic candidate in the new District 30 had it simply plowed directly north to pick up additional population. Immediately north of the city of Dallas are the “Park Cities,” which include a population that has voted strongly Republican throughout recent elections. See State’s Exhs. 9A and 9B (depicting one index of political affiliation in 1990 and 1992 elections). Rather than dilute the Democratic vote (and threaten the Republican incumbents) in this manner, District 30 skirted these communities on the west, and then curved east, picking up communities on either side of the region’s major interstate freeways.14
As the process of extracting Democratic voters out of the core of the Republican districts in North Dallas progressed, the distinction between Democratic and Republican voters moved from the precinct level (the smallest level at which political affiliation data was immediately available in the re*1017districting programs) down to the smaller census block level (the smallest level at which demographic and socioeconomic data was available).15 In an effort to further identify which census blocks were likely to support their candidacy, the incumbents used not only census data, but their own long experience as local representatives as well as the experiences of staffers and supporters. See 3 Tr. 177-179, 181-182 (describing methods, such as simply driving through neighborhoods, that staff members and candidates for office used to develop block-specific information regarding the likely political affiliation of voters).16
In addition, although information about political affiliation was not available at the block level through the computer program, legislators and staffers were able to get relatively precise information about voter preferences through a system, developed by the Democratic Party, that allowed candidates to determine in which party primary voters had participated. Id., at 179-180. By examining this information, legislators were able to further fine-tune district lines to include likely supporters and exclude those who would prob*1018ably support their opponents. Cf. Gaffney v. Cummings, 412 U. S. 735, 753 (1973) (“[W]hen [political profiles are] overlaid on a census map, it requires no special genius to recognize the political consequences of drawing a district line along one street rather than another”).17
The careful gerrymandering conducted by the Texas Legislature under the watchful eye of Johnson and her staff was a success not only on a districtwide level (Johnson was elected with over 70% of the vote in both 1992 and 1994), but on a precinct level. While the pre-1990 precincts in the heavily Republican North Dallas gave little reason for a Democratic incumbent to hope for much support, see State’s Exh. 9B (maps of Dallas and Collin Counties with 1990 election index results showing only a few Democrat-leaning precincts in North Dallas), the gerrymandering that occurred in 1991 resulted in smaller precincts that, by all indications, gathered concentrations of Democratic voters into District 30 while leaving concentrations of Republican voters in surrounding Districts 3 and 26. See State’s Exh. 9A (maps of Dallas and Collin Counties with 1992 election index results showing many more Democrat-leaning precincts in the North Dallas sections of District 30).
Presumably relying on Shaw J’s statement that “a reapportionment plan may be so highly irregular that, on its face, it rationally cannot be understood as anything other than an effort to ‘segregat[e] . . . voters’ on the basis of race,” 509 *1019U. S., at 646-647, the plurality offers mathematical proof that District 30 is one of the most bizarre districts in the Nation, see ante, at 960, and relates the now-obligatory florid description of the district’s shape, ante, at 965-966; see also ante, at 973-974 (describing District 29). As the maps appended to this opinion demonstrate, neither District 30 nor the Houston districts have a monopoly on either of these characteristics. Three other majority-white districts are ranked along with the majority-minority districts as among the oddest in the Nation. See Pildes & Niemi, Expressive Harms, “Bizarre Districts,” and Voting Rights: Evaluating Election-District Appearances After Shaw v. Reno, 92 Mich. L. Rev. 483, 565 (1993). Perhaps the clearest example of partisan gerrymandering outside of the context of majority-minority districts is District 6, a majority-Anglo district represented by a Republican.18
*1020For every geographic atrocity committed by District 30, District 6 commits its own and more. District 30 split precincts to gerrymander Democratic voters out of Republican precincts; District 6 did the same. See State’s Exh. 9B (Tar-rant County, showing District 6 cuts). District 30 travels down a riverbed; District 6 follows the boundaries of a lake. District 30 combines various unrelated communities of interest within Dallas and its suburbs; District 6 combines rural, urban, and suburban communities. District 30 sends tentacles nearly 20 miles out from its core; District 6 is a tentacle, hundreds of miles long (as the candidate walks), and it has no core.
The existence of the equally bizarre majority-white District 6 makes the plurality’s discussion of District 30’s odd shape largely irrelevant. If anything, the similarities between Districts 6 and 30 suggest that it is more likely than not that the incumbency considerations that led to the mutation of District 6 were the same considerations that forced District 30 to twist and turn its way through North Dallas.19
*1021The political, rather than the racial, nature of District 30’s gerrymander is even more starkly highlighted by comparing it with the districts struck down in Shaw II and Miller. District 30’s black population is, for instance, far more concentrated than the minority population in North Carolina’s District 12. And in Miller, the Court made it clear that the odd shape of Georgia’s Eleventh District was the result of a conscious effort to increase its proportion of minority populations: It was, the Court found, “ ‘exceedingly obvious’ from the shape of the Eleventh District, together with the racial demographics, that the drawing of narrow land bridges to incorporate within the district outlying appendages containing nearly 80% of the district’s total black population was a deliberate attempt to bring black populations into the district.” Miller, 515 U. S., at 917 (emphasis added; citation omitted).
District 30 is the precise demographic converse of the district struck down in Miller. District 30, for example, has a compact core in South Dallas which contains 50% of the district population and nearly 70% of the district’s total black population. Cf. ibid. Unlike the appendages to Georgia’s District 11, the tentacles stretching north and west *1022from District 30 add progressively less in the way of population, and, more important for purposes of this inquiry, they actually reduce the proportional share of minorities in the district. See State’s Exh. 33.
For example: The worst offender, in the trained eye of the Court, may be the northern arm of the district that winds around the Park Cities and then up into Collin County. But that arm, which contains 22% of the population, is only 21% black, ibid. — a proportion essentially identical to the proportion of African-Americans in Dallas County as a whole.20
The plurality is certainly correct in pointing out that District 30’s outlying reaches encompass some communities with high concentrations of minorities.21 It is implausible *1023to suggest, however, that an effort to “segregate” voters drove District 30 to collect those populations. After all, even the District Court noted that African-American voters immediately adjacent to the core of District 30 were intentionally excluded, from the district “in order to protect incumbents.” 861 F. Supp., at 1339 (emphasis added). Forced into Republican territory to collect Democratic votes, the district intentionally picked up some minority communities (though far more majority-white communities). If it had not, the goal of creating a majority-black district would have been sacrificed to incumbency protection (the very sort of “predominance” of race over race-neutral factors that the plurality discredits). But unlike Georgia’s District 11 and North Carolina’s District 12, the reason that the district was there in the first place was not to collect minority communities, but to collect population — preferably Democrats. It would, therefore, be fanciful to assert that the “several appendages” to District 30 were “drawn for the obvious,” let alone the predominant, “purpose of putting black populations into the district.” Miller, 515 U. S., at 910.22
In sum, a fair analysis of the shape of District 30, like the equally bizarre shape of District 6, belies the notion that its shape was determined by racial considerations.
*1024 Intent
Perhaps conscious that noncompact congressional districts are the rule rather than the exception in Texas, the plurality suggests, ante, at 960-961, 969-970, that the real key is the direct evidence, particularly in the form of Texas’ § 5 Voting Rights Act submissions and the person of then-State Senator Johnson, that the State expressed an intent to create these districts with a given “minimum percentage of the favored minority.” 861 F. Supp., at 1309. Even if it were appropriate to rest this test of dominance on an examination of the subjective motivation of individual legislators,23 or on *1025testimony given in a legal proceeding designed to prove a conflicting conclusion,24 this information does little more than confirm that the State believed it necessary to comply with the Voting Rights Act. Given its reasonable understanding of its legal responsibilities, see supra, at 1007, the legislature acted to ensure that its goal of creating a majority-black district in Dallas County was not undermined by the changes made to accommodate District 30 to other, race-neutral dis-tricting principles. As the plurality admits, see ante, at 958, the intent to create majority-minority districts does not in itself trigger strict scrutiny; these admissions prove nothing more than that. See also Shaw II, ante, at 930-932 (Stevens, J., dissenting).
Nonracial Factors: Community
In an effort to provide a definitive explanation for the odd shape of the district, the State emphasized two factors: The *1026presence of communities of interest tying together the populations of the district, and the role of incumbency protection. The District Court and the plurality improperly dismissed these considerations as ultimately irrelevant to the shape of the districts.
First, the appellants presented testimony that the districts were drawn to align with certain communities of interest, such as land use, family demographics, and transportation corridors. See 861 F. Supp., at 1322-1328. Although the District Court recognized that these community characteristics amounted to accurate descriptions of District 30, id., at 1323, it dismissed them as irrelevant to the districting process, concluding that there was no evidence that “the Legislature had these particular ‘communities of interest’ in mind when drawing the boundaries of District 30.” Ibid. The plurality concludes that appellants present no reason to displace that conclusion. Ante, at 966-967.
I do not understand why we should require such evidence ever to exist. It is entirely reasonable for the legislature to rely on the experience of its members when drawing particular boundaries rather than on clearly identifiable “evidence” presented by demographers and political scientists. Most of these representatives have been members of their communities for years. Unless the Court intends to interfere in state political processes even more than it has already expressed an intent to do, I presume that it does not intend to require States to create a comprehensive administrative record in support of their redistricting process. State legislators should be able to rely on their own experience, not only prepared reports. To the extent that the presence of obvious communities of interest among members of a district explicitly or implicitly guided the shape of District 30, it amounts to an entirely legitimate nonracial consideration.25
*1027 Nonracial Factors: Incumbency
The plurality admits that the appellants “present a . . . substantial case for their claim that incumbency protection rivaled race in determining the district’s shape.” Ante, at 967. Every individual who participated in the redistricting process knew that incumbency protection was a critical factor in producing the bizarre lines and, as the plurality points out, ante, at 963-964, even the District Court recognized that this nearly exclusive focus on the creation of “safe” districts for incumbents was intimately related to the bizarre shape of district lines throughout the State.
“[I]n Texas in 1991, many incumbent protection boundaries sabotaged traditional redistricting principles as they routinely divided counties, cities, neighborhoods, and regions. For the sake of maintaining or winning seats in the House of Representatives, Congressmen or would-be Congressmen shed hostile groups and potential opponents by fencing them out of their districts. The Legislature obligingly carved out districts of apparent supporters of incumbents, . . . and then added appendages to connect their residences to those districts. The final result seems not one in which the people select their representatives, but in which the representatives *1028have selected the people.” 861 F. Supp., at 1834 (citations and footnotes omitted).
See also id., at 1335, n. 43. Despite this overwhelming evidence that incumbency protection was the critical motivating factor in the creation of the bizarre Texas districts, the District Court reached the stunning conclusion that because the process was so “different in degree” from the “generalized, and legitimate, goal of incumbent and seniority protection” that this Court has previously recognized, it could not serve as a legitimate explanation for the bizarre boundaries of the congressional districts. Id., at 1334-1335. In dismissing incumbency protection once and for all, the District Court stated that “[i]ncumbent protection is a valid state interest only to the extent that it is not a pretext for unconstitutional racial gerrymandering.” Id., at 1336.
It is difficult to know where to begin to attack the misper-ceptions reflected in these conclusions,26 and the plurality’s failure to do so seriously taints its evaluation of the relative importance of nonracial considerations in the creation of District 30. The initial problem, of course, is that under the Court’s threshold test as set forth in Miller, one must consider the role of incumbency protection before determining whether there is an “unconstitutional racial gerrymander.” And because the ultimate focus in these gerrymandering cases is the claim that race was the “dominant and controlling rationale in drawing [the] district lines,” 515 U. S., at 913, a court must, in applying that test, consider a State’s claim that a given race-neutral rationale controlled the creation of those lines. See id., at 916 (“Where [compactness, contiguity,] or other race-neutral considerations are the basis for redistricting legislation, and are not subordinated to race, a State can ‘defeat a claim that a district has been *1029gerrymandered on racial lines’ ”). Although a court may not like the State’s explanation, that is no excuse for ignoring it.
If some independent bar prevented the use of that race-neutral criterion, then the District Court might be in a position to object to the State’s use of it. We have, however, affirmed that a State has an interest in incumbency protection, see, e. g., ante, at 964-965 (opinion of O’Connor, J.); White v. Weiser, 412 U. S. 783, 791, 797 (1973), and also assured States that the Constitution does not require compactness, contiguity, or respect for political borders, see Shaw I, 509 U. S., at 647. While egregious political gerrymandering may not be particularly praiseworthy, see infra, at 1038-1040, it may nonetheless provide the race-neutral explanation necessary for a State to avoid strict scrutiny of the district lines where gerrymandering is the “dominant and controlling” explanation for the odd district shapes.27
The District Court’s error had an apparently dispositive effect on its assessment of whether strict scrutiny should apply at all. Although aspects of our dispute with the plurality are “largely factual,” ante, at 971, n., they arise not out of our disagreement with the District Court’s credibility assessments, but out of that court’s erroneous conclusion that the State’s overwhelming reliance on this race-neutral factor was illegitimate and irrelevant to its evaluation of the factors involved in the shifting of this district’s lines. A fair evaluation of the record made in light of appropriate legal standards requires a conclusion very different from the District Court’s. By following the District Court down its misdirected path, the plurality itself goes astray.
*1030 Race as a Proxy
Faced with all this evidence that politics, not race, was the predominant factor shaping the district lines, the plurality ultimately makes little effort to contradict appellants’ assertions that incumbency protection was far more important in the placement of District 30’s lines than race. See ante, at 967-969. Instead, it adopts a fallback position based on an argument far removed from even the “analytically distinct” claim set forth in Shaw I, 509 U. S., at 652. In it, the plurality suggests that even if the predominant reason for the bizarre features of the majority-minority districts was incumbency protection, the State impermissibly used race as a proxy for determining the likely political affiliation of blocks of voters. See ante, at 968-971 (opinion of O’Connor, J.).
The effect of this process, in all likelihood, was relatively unimportant to the overall shape of the district. A comparison of the 1992 precinct results with a depiction of the proportion of black population in each census block reveals that Democratic-leaning precincts cover a far greater area than majority-black census blocks. Compare State’s Exh. 9A with State’s Exh. 45. One would expect the opposite effect if the single-minded goal of those drawing the districts was racial composition rather than political affiliation. At the very least, the maps suggest that the drawing of boundaries involves a demographic calculus far more complex than simple racial stereotyping.
Furthermore, to the extent that race served as a proxy at all, it did so merely as a means of “fine tuning” borders that were already in particular locations for primarily political reasons. This “fine tuning” through the use of race is, of course, little different from the kind of fine tuning that could have legitimately occurred around the edges of a compact majority-minority district.28 I perceive no reason why a *1031legitimate process — choosing minority voters for inclusion in a majority-minority district — should become suspect once nonracial considerations force district lines away from its core.
Finally, I note that in most contexts racial classifications are invidious because they are irrational. For example, it is irrational to assume that a person is not qualified to vote or to serve as a juror simply because she has brown hair or brown skin. It is neither irrational, nor invidious, however, to assume that a black resident of a particular community is a Democrat if reliable statistical evidence discloses that 97% of the blacks in that community vote in Democratic primary elections. See Brief for United States 44. For that reason, the fact that the architects of the Texas plan sometimes appear to have used racial data as a proxy for making political judgments seems to me to be no more “unjustified,” ante, at 969 (opinion of O’Connor, J.), and to have no more constitutional significance, than an assumption that wealthy suburbanites, whether black or white, are more likely to be Repub*1032licans than Communists.29 Requiring the State to ignore the association between race and party affiliation would be no more logical, and potentially as harmful, as it would be to prohibit the Public Health Service from targeting African-American communities in an effort to increase awareness regarding sickle-cell anemia.30
Despite all the efforts by the plurality and the District Court, then, the evidence demonstrates that race was not, in all likelihood, the “predominant” goal leading to the creation of District 30. The most reasonable interpretation of the record evidence instead demonstrates that political considerations were. In accord with the presumption against interference with a legislature’s consideration of complex and competing factors, see n. 9, supra, I would conclude that the configuration of District 30 does not require strict scrutiny.
*1033V
The Houston districts present a closer question on the application of strict scrutiny. There is evidence that many of the same race-neutral factors motivating the zigzags of District 30 were present at the creation (or recreation) of Districts 29 and 18. In contrast to District 30, however, there is also evidence that the interlocking shapes of the Houston districts were specifically, and almost exclusively, the result of an effort to create, out of largely integrated communities, both a majority-black and a majority-Hispanic district. For purposes of this opinion, then, I am willing to accept, arguendo, the plurality’s conclusion that the Houston districts should be examined with strict scrutiny.31 Even so, the plurality errs by concluding that these districts would fail that test.
The plurality begins with the perfectly obvious assumptions that a State has a compelling interest in complying with § 2 of the Voting Rights Act and that Texas had a strong basis for believing that it would have violated that Act in 1991 if it did not create three new majority-minority districts.32 The plurality goes on to conclude, however, that because the final shape of these districts is not coextensive with the community that would form the core of a § 2 violation, these districts would not be “narrowly tailored” to further that state interest. Ante, at 979. I respectfully disagree.
Neither evidence nor insinuation suggests that the State in the redistricting process considered race for any reason *1034other than as a means of accomplishing its compelling interest of creating majority-minority districts in accord with the Voting Rights Act. The goal was, by all accounts, achieved, for these districts would certainly avoid liability under § 2 of the Voting Rights Act.33 For reasons that continue to escape me, however, the plurality simply insists that the lack of compactness in the districts prevents them from being “narrowly tailored” solutions to the State’s interests.
The plurality uses two premises to reach its conclusion that compactness is required to meet the “narrow tailoring” requirement: (i) §2 would not have been violated unless a reasonably compact majority-minority district could have been created; and (ii) nothing in § 2 requires the creation of a noncompact district. I have no quarrel with either proposition, but each falls far short of mandating the conclusion that the plurality draws from it. While a State can be liable for a § 2 violation only if it could have drawn a compact district and failed to do so, it does not follow that creating such a district is the only way to avoid a § 2 violation. See generally Shaw II, ante, at 946-950 (Stevens, J., dissenting). The plurality admits that a State retains “a limited degree of leeway” in drawing a district to alleviate fears of § 2 liability, ante, at 977, but if there is no independent constitutional duty to create compact districts in the first place, and the plurality suggests none, there is no reason why noncompact districts should not be a permissible method of avoiding violations of law. The fact that they might be unacceptable judicial remedies does not speak to the question whether they *1035may be acceptable when adopted by a state legislature. Because these districts satisfy the State’s compelling interest and do so in a manner that uses racial considerations only in a way reasonably designed to ensure such a satisfaction, I conclude that the districts are narrowly tailored.
<1
I cannot profess to know how the Court s developing jurisprudence of racial gerrymandering will alter the political and racial landscape in this Nation — although it certainly will alter that landscape. As the Court’s law in this area has developed, it has become ever more apparent to me that the Court’s approach to these cases creates certain perverse incentives and (I presume) unanticipated effects that serve to highlight the essentially unknown territory into which it strides. Because I believe that the social and political risks created by the Court’s decisions are not required by the Constitution, my first choice would be to avoid the preceding analysis altogether, and leave these considerations to the political branches of our Government.
The first unintended outcome of the legal reasoning in Shaw II and this case is the very result that those decisions seek to avoid: the predominance of race in the districting process, over all other principles of importance. Given the Court’s unwillingness to recognize the role that race-neutral districting principles played in the creation of the bizarrely shaped districts in both this case and Shaw II, it now seems clear that the only way that a State can both create a majority-minority district and avoid a racial gerrymander is by drawing, “without much conscious thought,” ante, at 967 (opinion of O’Connor, J.), and within the “limited degree of leeway” granted by the Court, ante, at 977, the precise compact district that a court would impose in a successful §2 challenge. See post, at 1066-1067 (Souter, J., dissenting). After the Court’s decisions today, therefore, minority voters can make up a majority only in compact districts, whether *1036intentionally or accidentally drawn, while white voters can be placed into districts as bizarre as the State desires.
The great irony, of course, is that by requiring the State to place the majority-minority district in a particular place and with a particular shape, the district may stand out as a stark, placid island in a sea of oddly shaped majority-white neighbors. See Karlan, Still Hazy After All These Years: Voting Rights in the Post-Stew Era, 26 Cumberland L. Rev. 287, 309 (1995-1996). The inviolable sanctity of the §2-eligible districts will signal in a manner more blatant than the most egregious of these racial gerrymanders that “a minority community sits here: Interfere with it not.” The Court-imposed barriers limiting the shape of the district will interfere more directly with the ability of minority voters to participate in the political process than did the oddly shaped districts that the Court has struck down in recent cases. Unaffected by the new racial jurisprudence, majority-white communities will be able to participate in the districting process by requesting that they be placed into certain districts, divided between districts in an effort to maximize representation, or grouped with more distant communities that might nonetheless match their interests better than communities next door. By contrast, none of this political maneuvering will be permissible for majority-minority districts, thereby segregating and balkanizing them far more effectively than the districts at issue here, in which they were manipulated in the political process as easily as white voters. This result, it seems to me, involves “discrimination” in a far more concrete manner than did the odd shapes that so offended the Court’s sensibilities in Miller, Shaw II, and these cases.
In light of this Court’s recent work extolling the importance of state sovereignty in our federal scheme, cf. Seminole Tribe of Fla. v. Florida, ante, p. 44, I would have expected the Court’s sensibilities to steer a course rather more deferential to the States than the one that it charts with its *1037decisions today. As we have previously noted, “[electoral districting is a most difficult subject for legislatures, and so the States must have discretion to exercise the political judgment necessary to balance competing interests.” Miller, 515 U. S., at 915; see also post, at 1047-1048 (Souter, J., dissenting). The record in these cases evidences the “complex interplay of forces that enter a legislature’s redistricting calculus,” 515 U. S., at 915-916, and the Court’s failure to respect those forces demonstrates even less respect for the legislative process than I would have expected after the decision in Miller.
The results are not inconsequential. After Miller and today’s decisions, States may find it extremely difficult to avoid litigation flowing from decennial redistricting. On one hand, States will risk violating the Voting Rights Act if they fail to create majority-minority districts. If they create those districts, however, they may open themselves to liability under Shaw and its progeny. See Miller, 515 U. S., at 949 (Ginsburg, J., dissenting). Perhaps States will simply avoid the problem by abandoning voluntary compliance with §2 of the Voting Rights Act altogether. See Shaw I, 509 U. S., at 672 (White, J., dissenting); post, at 1063-1064 (Souter, J., dissenting).34 This result would not necessarily bring peace to redistricting, for there is no guarantee that districts created by court order to comply with §2 will be immune from attack under Shaw; in both Florida and Illinois, for instance, that very sort of schizophrenic second-guessing has already occurred. See King v. State Bd. of Elections, *1038No. 95-C-827, 1996 WL 130439 (ND Ill., Mar. 15, 1996); Johnson v. Mortham, 926 F. Supp. 1460 (ND Fla. 1996). Given the difficulty of reconciling these competing legal responsibilities, the political realities of redistricting, and the cost of ongoing litigation, some States may simply step out of the redistricting business altogether, citing either frustration or hopes of getting a federal court to resolve the issues definitively in a single proceeding. See, e. g., Johnson v. Miller, 922 F. Supp. 1556, 1559 (SD Ga. 1995) (after remand from Miller, Georgia Legislature abdicated its redistricting responsibilities to Federal District Court); post, at 1064 (Souter, J., dissenting) (noting the likely “vacuum of responsibility” at the state level).
Regardless of the route taken by the States, the Court has guaranteed that federal courts will have a hand — and perhaps the only hand — in the “abrasive task of drawing district lines.” Wells v. Rockefeller, 394 U. S. 542, 553 (1969) (White, J., dissenting). Given the uniquely political nature of the redistricting process, I fear the impact this new role will have on the public’s perception of the impartiality of the Federal Judiciary. I can only reiterate the Court’s cautionary admonition, issued over two decades ago, that “[i]n fashioning a reapportionment plan or in choosing among plans, a district court should not pre-empt the legislative task nor 'intrude upon state policy any more than necessary.’” White v. Weiser, 412 U. S., at 795 (citing Whitcomb v. Chavis, 403 U. S. 124, 160 (1971)).
I do not wish to leave the impression that decisions of the Court from Shaw I to the present are focusing on entirely nonexistent problems. I merely believe that the Court has entirely misapprehended the nature of the harm that flows from this sort of gerrymandering. Rather than attach blameworthiness to a decision by the majority to share political power with the victims of past discriminatory practices, the Court’s real concern should be with the more significant harms that flow from legislative decisions that “serve no *1039purpose other than to favor one segment — whether racial, ethnic, religious, economic, or political — that may occupy a position of strength at a particular point in time, or to disadvantage a politically weak segment of the community.” Karcher v. Daggett, 462 U. S. 725, 748 (1988) (Stevens, J., concurring). These cases are as good an illustration of such self-serving behavior on the part of legislators as any — but not with respect to racial gerrymandering. The real problem is the politically motivated gerrymandering that occurred in Texas. Many of the oddest twists and turns of the Texas districts would never have been created if the legislature had not been so intent on protecting party and incumbents. See also Shaw II, ante, at 937-938 (Stevens, J., dissenting) (noting the same influences behind the bizarre shape of North Carolina’s District 12).
By minimizing the critical role that political motives played in the creation of these districts, I fear that the Court may inadvertently encourage this more objectionable use of power in the redistricting process.35 Legislatures and elected representatives have a responsibility to behave in a way that incorporates the “elements of legitimacy and neutrality that must always characterize the performance of the sovereign’s duty to govern impartially.” Cleburne, 473 U. S., at 452. That responsibility is not discharged when legislatures permit and even encourage incumbents to use their positions as public servants to protect themselves and their parties rather than the interests of their constituents. See Karcher v. Daggett, 462 U. S., at 748, 754 (Stevens, J., concurring). If any lines in Texas are worth straightening, *1040it is those that were twisted to exclude, not those altered to include.36
VII
The history of race relations in Texas and throughout the South demonstrates overt evidence of discriminatory voting practices lasting through the 1970’s. Brischetto, Richards, Davidson, & Grofman, Texas, in Quiet Revolution in the South: The Impact of the Voting Rights Act, 1965-1990, pp. 233-248 (C. Davidson & B. Grofman eds. 1994). Even in recent years, Texans have elected only two black candidates to statewide office; majority-white Texas districts have never elected a minority to either the State Senate or the United States Congress. Brief for Appellants in No. 94-806, p. 53. One recent study suggests that majority-white districts throughout the South remain suspiciously unlikely *1041to elect black representatives. Davidson & Grofman, The Effect of Municipal Election Structure on Black Representation in Eight Southern States, in Quiet Revolution in the South, at 344. And nationwide, fewer than 15 of the hundreds of legislators that have passed through Congress since 1950 have been black legislators elected from majority-white districts.37 In 1994, for example, 36 of the Nation’s 39 black Representatives were elected from majority-minority districts, while only 3 were elected from majority-white districts.38 See post, at 1050-1051 (Souter, J., dissenting).
Perhaps the state of race relations in Texas and, for that matter, the Nation, is more optimistic than might be expected in light of these facts. If so, it may be that the plurality’s exercise in redistricting will be successful. Perhaps minority candidates, forced to run in majority-white districts, will be able to overcome the long history of stereotyping and discrimination that has heretofore led the vast majority of majority-white districts to reject minority candidacies. Perhaps not. I am certain only that bodies of elected federal and state officials are in a far better position than anyone on this Court to assess whether the Nation’s long history of discrimination has been overcome, and that nothing in the Constitution requires this unnecessary intrusion into the ability of States to negotiate solutions to political differences while providing long-excluded groups the opportunity to participate effectively in the democratic process. I respectfully dissent.
[Appendixes to opinion of Stevens, J., follow this page.]
*1042APPENDIX A TO OPINION OF STEVENS, J.
[[Image here]]
TEXAS CONGRESSIONAL DISTRICT 3
*1043APPENDIX B TO OPINION OF STEVENS, J.
[[Image here]]
TEXAS CONGRESSIONAL DISTRICT 6
*1044APPENDIX C TO OPINION OF STEVENS, J.
[[Image here]]
TEXAS CONGRESSIONAL DISTRICT 25
*1045[[Image here]]
The District Court recognized, but erroneously ignored, the overwhelming weight of evidence demonstrating that political considerations dominated the shaping of Texas’ congressional districts. See Vera v. Richards, 861 F. Supp. 1304, 1331, 1334-1336 (SD Tex. 1994); infra, at 1027-1029.
Because I believe that political gerrymanders are more objectionable than the “racial gerrymanders” perceived by the Court in recent cases, see Karcher v. Daggett, 462 U. S. 725, 748 (1983) (Stevens, J., concurring); Davis v. Bandemer, 478 U. S. 109,161-162, 166 (1986) (Powell, J., concurring in part and dissenting in part), I am not entirely unsympathetic to the Court’s holding. I believe, however, that the evils of political gerrymandering should be confronted directly, rather than through the race-specific approach that the Court has taken in recent years. See also infra, at 1038-1040.
In elections since 1980, the State has elected a Democrat in only two of four gubernatorial races, and in only two of six races for the United States Senate. America Votes 21: A Handbook of Contemporary American Election Statistics 417 (R. Scammon & A. MeGillivray eds. 1995). Furthermore, in 1994, Republican candidates received a total of 550,000 more votes than Democratic candidates in Texas’ 30 races for the United States House of Representatives, id., at 4, while in 1992, Democratic House candidates outpolled Republicans by only 147,000 votes (despite winning 27 of 30 districts). America Votes 20: A Handbook of Contemporary American Election Statistics 474 (R. Scammon & A. MeGillivray eds. 1993).
Then-State Senator from Dallas, Eddie Bernice Johnson, who was chair of the Senate Subcommittee on Congressional Districts, maneuvered to construct District 30 in a manner that would ensure her election. 861 *1006F. Supp., at 1313; Politics in America 1994: The 103rd Congress 1536 (1993) (“This is the District Eddie Bernice Johnson drew”). Vice chair of the same committee, Frank Tejeda, also “attempted to draw a district [District 28] that would facilitate his potential candidacy.” 861 F. Supp., at 1326. And State Senator Gene Green and State Representative Roman Martinez, both Houston-area officials with designs on Congress, competed in an effort to design District 29 in a way that would guarantee their own election. Id., at 1324, n. 27. (Martinez later dropped out of the congressional race to run for State Senate.) Because the role that these legislators played in the redistricting process was largely identical to that played by sitting incumbents, my references to the role of “incumbents” in the redistricting process generally refer to these individuals as well.
As did many other States, Texas kept track of the shapes of its post-1990 districts with a computer districting program loaded with 1990 census information and geographic information at scales ranging from statewide to that of a city block. See generally Shaw v. Hunt, 861 F. Supp. 408, 457 (EDNC 1994) (describing computer programs); 861 F Supp., at 1318-1319. The dramatic increase in bizarrely shaped districts after 1990 can be traced, at least in part, to the fact that computers allowed legislators to achieve their political goals geographically in a manner far more precise than heretofore possible. See Pildes & Niemi, Expressive Harms, “Bizarre Districts,” and Voting Rights: Evaluating Election-District Appearances After Shaw v. Reno, 92 Mich. L. Rev. 483, 574 (1993); Note, The Illegitimacy of the Incumbent Gerrymander, 74 Texas L. Rev. 913, 924 (1996).
The State added District 28 (a majority-Hispanie district in south Texas), District 29 (a majority-Hispanie district in Houston), and District 30 (a majority-black district in Dallas). In addition, the State reconfigured Houston’s District 18. That district had elected African-American Representatives to Congress since the early 1970’s and remained majority-minority in 1990, although a plurality of its population was by then Hispanic. To create District 29, the legislature altered the shape of District 18 to move parts of its Hispanic population into that neighboring district while retaining a majority-black population.
To the extent that the precise shape of these districts relied on race rather than other factors, that racial gerrymandering was somewhat less *1008effective than the political gerrymandering had been: District 29, created as a majority-Hispanic district, elected an Anglo, former State Senator Green, in 1992, and reelected him in 1994. America Votes 21, at 437. Given his substantial role in crafting the district to meet his electoral needs, see n. 4, supra, Green’s success suggests the power of incumbency over race.
I do not agree with the Court’s approach to these cases. Nonetheless, given that the Court seems settled in its conclusion that racial gerrymandering claims such as these may be pursued, I endorse this proposition.
Though expressly reserving the issue in Shaw I, we noted there that appellants wisely conceded that while “race-conscious redistricting is not always unconstitutional_This Court has never held that race-conscious state decisionmaking is impermissible in all circumstances.” 509 U. S., at 642 (emphasis in original). The threshold test for the application of strict scrutiny as set forth in Miller v. Johnson, 515 U. S. 900 (1995), implicitly accepts this as true, concluding that strict scrutiny applies not when race merely influences the districting process, but only when “the legislature subordinated traditional race-neutral districting principles ... to racial considerations.” Id., at 916 (emphasis added); see also id., at 928-929 (O’Connor, J., concurring) (test does not “throw into doubt the vast majority of the Nation’s 435 congressional districts ... even though race may well have been considered in the redistricting process”). Shaw II similarly recognizes that intent does not trigger strict scrutiny: Although the District Court concluded that the State “deliberately drew” the district in question to ensure that it included a majority of African-American citizens, see Shaw, 861 F. Supp., at 473; Shaw II, ante, at 905, the Court reviews the District Court’s findings regarding the demographics of the district to determine whether the strict scrutiny was appropriately applied. See ante, at 905-906; cf. ante, at 999 (Thomas, J., concurring in judgment) (where State intends to create majority-minority district, application of strict scrutiny not even a “close question”).
Justice Thomas takes a strong view on this matter, arguing that a majority-minority district should escape strict scrutiny only when it is created “in spite of,” not “because of,” the race of its population. Ante, at 1001. But because minorities are, by definition, minorities in the population, it will be rare indeed for a State to stumble across a district in which the minority population is both large enough and segregated enough to allow majority-minority districts to be created with at most a “mere awareness” that the placement of the lines will create such a district. See ibid. Indeed, I doubt that any such district exists in the entire Nation; the creation of even the most compact majority-minority district will gen*1010erally require a conscious decision to draw its lines “just so” to ensure that the group is not a minority in the district population. It appears, however, that even when a district is placed “just so” in order to include a traditional community in which race does correlate with community interests (consider, for example, New York District 15, which is centered on Harlem), Justice Thomas would review that district with the same presumption of invidiousness with which we viewed the district in Gomillion v. Lightfoot, 364 U. S. 339 (1960). Cf. Miller, 515 U. S., at 944 (Ginsburg, J., dissenting) (noting that “ethnicity itself can tie people together” in communities of interest). Because the creation of such a district threatens neither the harms of Go-million nor, I believe, any harms against which the Fourteenth Amendment was intended to protect, I cannot accept his conclusion.
Because the Court’s approach to eases of this kind seeks to identify the “predominant” motive of the legislature, it is worth pointing out, as we have on so many prior occasions, that it is often “difficult or impossible for any court to determine the ‘sole’ or ‘dominant’ motivation behind the choices of a group of legislators.” Palmer v. Thompson, 403 U. S. 217, 225 (1971). As in every other legislative body, each of the members of Texas’ Legislature has his or her own agenda and interests — particularly in the “complicated process” of redistricting, in which every decision “inevitably has sharp political impact.” White v. Weiser, 412 U. S. 783, 795-796 (1973). In these circumstances, “[rjarely can it be said that a legislature ... operating under a broad mandate made a decision motivated solely by a single concern, or even that a particular purpose was the ‘dominant’ or ‘primary’ one. In fact, it is because legislators ... are properly concerned with balancing numerous competing considerations that courts refrain from reviewing the merits of their decisions, absent a showing of arbitrariness or irrationality.” Arlington Heights v. Metropolitan Housing Development Corp., 429 U. S. 252, 265 (1977) (footnote omitted); see also Edwards v. Aguillard, 482 U. S. 578, 636-639 (1987) (Scalia, J., dissenting); Shaw II, ante, at 940 (Stevens, J., dissenting).
Not only is this a case in which a legislature is operating under a “broad mandate,” but other factors weigh in favor of deference as well. First, the inherently political process of redistrieting is as much at the core of state sovereignty as any other. Second, the “motive” with which we are concerned is not per se impermissible. (For that reason, this litigation is very different from Price Waterhouse v. Hopkins, 490 U. S. 228 (1989), and Arlington Heights v. Metropolitan Housing Development Corp., 429 U. S. 252 (1977), in which the plaintiffs alleged that the defendant’s action was motivated by an intent to harm individuals because of their status as members of a particular group. Where there is “proof that a discriminatory purpose has been a motivating factor in the decision,” the “judicial deference” due to the legislative process is no longer justified. Id., at 265-*1013266.) Finally, those that are injured by the allegedly discriminatory districts can alleviate their injury through the democratic process: Those in the district could elect a representative who is not a part of their racial group, while the population at large could elect a legislature that refused to rely on racial considerations in the drawing of districts. In such circumstances we should take particular care in questioning the legislature’s motives and, if in doubt, presume that the legislature has acted appropriately. See post, at 1058-1062 (Soutek, J., dissenting).
We require state legislatures to ensure that populations, from district to district, are “as mathematically equal as reasonably possible,” with de minimis exceptions permissible only in “unavoidable” instances. White v. Weiser, 412 U. S., at 790; see also Karcher, 462 U. S., at 734-735. Population variances are not permissible even “ ‘if they necessarily result from a State’s attempt to avoid fragmenting political subdivisions by drawing congressional district lines along existing . .. political subdivision boundaries.’ ” White, 412 U. S., at 791 (citing Kirkpatrick v. Preisler, 394 U. S. 526, 533-534 (1969)). The legislature, therefore, understandably felt compelled to achieve mathematical equality regardless of other concerns. Rather surprisingly, they were able to do so: Every one of Texas’ 30 congressional districts contains precisely 566,217 persons. Of course, this precision could not have been accomplished without breaking apart counties, cities, neighborhoods, and even pre-existing voting precincts.
This phenomenon is not unique to Dallas County: Throughout the State, “incumbent residences repeatedly fall just along district lines.” 861 F. Supp., at 1318 (giving examples); see State’s Exhs. 10A and 10B (showing incumbent residences). District 6, for instance, changed from a rural district stretching far to the southeast of Dallas to a more suburban district wrapping around Fort Worth. As it did so, however, the district pivoted around the home of incumbent Representative Joe Barton, whose residence sits at the extreme southeastern end of a district stretching in a 100-mile-long loop around Fort Worth. See Appendix D, infra.
The plurality suggests that these communities were shed from District 30 in a “suspect use of race as a proxy to further neighboring incumbents’ interests.” Ante, at 979; see also ante, at 971-972, n. I had thought, however, that the Court’s concern in these cases was the “resemblance to political apartheid” involved in the creation of majority-minority districts. Shaw I, 509 U. S. 630, 647 (1993). I do not see how the decision to include minority communities in a neighboring majority-white district bears any resemblance to such “apartheid” or, for that matter, how it has any rele*1016vance to the validity of the creation of a district from which those minority communities have been excluded. See also infra, at 1030-1032.
See, e. g., 3 Tr. 187 (testimony of Christopher Sharman: “[A]ny time you took part of a district away on one end, you would usually squeeze or push the district out on another end; and in this case, most of the time the district would get pushed to the north”).
The author of the District Court opinion was herself aware of these political realities. See id., at 194 (Jones, J., noting that Johnson did not want anything to do with the Park Cities because she “[djidn’t want competition from Ross Perot”). In light of this recognition, it is difficult to understand why the District Court described District 30’s efforts to avoid that community as a contributing factor to the allegedly race-based bizarreness of the district borders. See 861 F. Supp., at 1337; ante, at 967.
Because political boundaries are more closely packed in urban than in rural areas, drawing lines based on such boundaries will almost always require tighter twists and turns in urban districts than in rural districts. Significantly, the three districts struck down by the District Court are the only three districts in the entire State with population densities of over 2,000 persons per square mile. See U. S. Dept, of Commerce, Bureau of the Census, Population and Housing Characteristics for Congressional Districts of the 103d Congress: Texas 40-44 (Feb. 1993). If enough empty land were added to these districts that they matched the sparse densities of rural districts (such as District 28, which was upheld by the District Court), their turns would not appear so sharp, and the open space, without its demographic implications, could smooth the deepest of the districts’ notches.
As Democratic communities were identified, they had to be connected with the core of the district. Although Texas has no state statutory or constitutional requirement to that effect, state legislators agreed that each of the 30 districts should be entirely contiguous, permitting any candidate, map in hand, to visit every residence in her district without leaving it.
Incumbents influenced the shape of districts in other ways. Both District 30 and District 29, for instance, detoured to include portions of the state legislative districts that were being represented by the state legislators who hoped to run for Congress. See, e. g., State’s Exh. 31 (showing that portion of Tarrant County included in District 30 had been part of Johnson’s State Senate district). In some cases, legislators drew districts to avoid the residences of potential primary challengers. See 3 Tr. 192-193; 4 id., at 46. Incumbents also sought to include communities that they expected (or knew) to contain particularly active supporters; this interest in “active” voters often trumped any desire to ensure a particular racial makeup. See 3 id., at 190; 4 id., at 40-41; 861 F. Supp., at 1320.
While two extremely noncompaet majority-Anglo districts in Texas (Districts 3 and 25) might be able to blame part (though by no means all) of their contortions on their contiguity with the majority-minority districts, District 6 has little excuse. Although it shares a border with District 30 for a short distance, that stretch is one of the straightest in either of the districts, running almost entirely along the county line through the Dallas-Fort Worth International Airport. See Appendix D, infra.
As for the obligatory florid description: District 6 has far less of an identifiable core than any of the majority-minority districts struck down by the District Court. To the extent that it “begins” anywhere, it is probably near the home of incumbent Rep. Barton in Ennis, located almost 40 miles southwest of downtown Dallas. From there, the district winds across predominantly rural sections of Ellis County, finally crossing into Tarrant County, the home of Fort Worth. It skips across two arms of Joe Pool Lake, noses its way into Dallas County, and then travels through predominantly Republican suburbs of Fort Worth. Nearing the central city, the borders dart into the downtown area, then retreat to curl around the city’s northern edge, picking up the airport and growing suburbs north of town. Worn from its travels into the far northwestern corner of the county (almost 70 miles, as the crow flies, from Ennis), the district lines plunge south into Eagle Mountain Lake, traveling along the waterline for miles, with occasional detours to collect voters that have built homes along its shores. Refreshed, the district rediscovers its roots in rural Parker *1020County, then flows back toward Fort Worth from the southwest for another bite at Republican voters near the heart of that city. As it does so, the district narrows in places to not much more than a football field in width. Finally, it heads back into the rural regions of its fifth county— Johnson — where it finally exhausts itself only 50 miles from its origin, but hundreds of “miles apart in distance and worlds apart in culture.” Miller, 515 U. S., at 908 (describing a similar combined rural/urban district).
Seeking specific examples, the plurality makes much hay over a portion of Collin County located just over the county line north of Dallas. See ante, at 965, 971. There, District 30 excludes a portion of a precinct that voted Democratic in 1990, and maps “exactly onto the only area in the southern half of th[e] county with a [minority] percentage population in excess of 50%.” Ante, at 965.
The map to which the plurality refers, however, groups the minority percentage by precinct, and since precincts are defined by the district boundaries, it is no surprise that the district maps “exactly” onto the precinct. See App. 153. (One might similarly argue that “District 30 maps exactly onto the only area in all of north Texas that is 50% black,” but such a statement reveals little about the underlying demographics of spe-*1021cifie sections of the district.) The more telling maps are the census block maps, which demonstrate that the Collin County section of District 30 contains many more census blocks of less than 25% minority population than it does blocks that are more than 50% minority. See State’s Exhs. 45 and 46 (Exh. 45 is reproduced, in part, as Appendix D, infra). Even if those majority-white blocks have relatively small populations, they were nonetheless included, suggesting that the creation of the district was not as single-mindedly focused on race as the Court and the District Court assume.
Even more significant is the fact that the new precinct leaned overwhelmingly Democratic in the 1992 election, while the portion of the precinct that was not included in District 30 voted overwhelmingly Republican. See State’s Exh. 9B (Collin County). While the excluded portion of the 1990 precinct may have been dropped, in part, to help comply with the State’s goals under the Voting Rights Act, it also involved a successful effort to maximize Democratic votes while avoiding Republican votes.
See 861 F. Supp., at 1312 (black population in Dallas County is 362, 130); Bureau of Census, Population and Housing Unit Counts 185 (Oct. 1993) (total population of Dallas County is 1,852,810).
Several responses to the plurality’s specific examples are worth making, however. In Collin County, the plurality relies on the fact that the “combined African-American and Hispanic” population in the Collin County extremity of the northern appendage to District 30 is in excess of 50%. Ante, at 971. But District 30 was created with an eye to a majority-6Zacfc population, rather than a majority-minority population, so the more relevant facts are that (i) African-Americans make up only 19.8% of the Collin County appendage, App. 331, (ii) those African-Americans consist of only two-tenths of 1% of the entire population in the district, ibid., and (iii) this appendage contains more majority-white census blocks than it does majority-minority census blocks, see State’s Exh. 45.
The plurality also points out that a small portion of one of the tentacles — the one that extends west into Tarrant County — contains an African-American majority. Ante, at 965. It would be implausible to claim, however, that race was the “predominant” reason that this community was included in District 30. First, the community had been part of Senator Johnson’s state legislative district, see n. 17, supra; second, it also includes majority-white census blocks; and third, the total population in that portion of the district is less than 2,000 people. App. 331. Finally, and more important, the population of the entire western tentacle (at the tip of which is the Tarrant County community) is only 29% black, see State’s Exh. 33 — less than half the proportion of minorities in the core of the district.
Indeed, if the “appendages” to District 30 reaching into neighboring counties were cut off, the proportion of African-Americans in the resulting district would actually increase. See App. 331. As presently constituted, District 30 includes 566, 217 people, of which 283,225 (or 50.02%) are African-American. If the Tarrant County and Collin County portions of the district were removed, the resulting district would have 557,218 people, of which 280,620 (or 50.36%) would be African-American. While the resulting district would not include the “zero deviation” necessary under Reynolds v. Sims, 377 U. S. 533 (1964), and its progeny, see n. 10, supra, the missing population could easily be acquired in majority-black census blocks adjacent to District 30’s southern and eastern edge, thereby increasing the proportion of black population still further. Because the alleged racial goals of the district could be achieved more effectively by making the district more compact, I simply do not comprehend how the plurality can conclude that the effort to create a majority-minority district “predominated” over other, race-neutral goals.
Testimony by individuals is relevant, but hardly dispositive evidence of collective motivations. See n. 9, supra. It may be true that the most important concern motivating Senator Johnson, the Chairman of the Senate Districting Committee, was her desire to create the first Congressional District in the history of the State in which African-Americans were in the majority. Johnson never testified, however, that racial considerations were the sole concern motivating the changes to the shapes of the districts. See, e. g., App. 454-456 (certain areas that were minority communities were assigned to Anglo incumbents because of incumbent power), id., at 459 (“[J]ust as 30 went looking for friendly territory regardless of color, [the incumbents] went looking for friendly territory as well regardless of color”). Since this testimony was not only irrelevant to the §2 proceedings but arguably harmful to her claim there that racial considerations had been taken into account, these admissions are particularly telling.
To the extent that testimony of individual legislators is relevant, the following statements from the floor of the Texas House confirm that many legislators viewed these districts as political, not racial, gerrymanders:
“This plan was drawn to protect incumbents....
“[I]n order to protect an incumbent Dallas congressman and an incumbent Houston congressman, county lines were not respected, urban boundaries were not respected, precinct boundaries were not respected.” Id., at 374-375 (statement of Rep. Ogden).
“With the adoption of this plan, you will have 8 Republican Congressmen out of 30. That’s de facto regression and provides for less Republican representation in Washington, D. C.
“Communities throughout the State are surgically split in what appears to be illogical, irrational and erratic pattern[s]. But if you look at election result data throughout the State, you’ll find that these lines are very logi*1025cal and very rational. The lines have been drawn, dissecting communities very creatively in order to pack Republicans and maximize Democratic representation.” Id., at 376 (statement of Rep. Gusendorf).
See also id., at 377-380 (statement of Rep. Gusendorf illustrating the gerrymandering process by reference to District 6, not a majority-minority district).
These gerrymanders “d[o] not have to happen. It has nothing to do with fairness. It has nothing to do with minority representation because if we were really concerned about minority representation, we would have drawn this map in such a way that the minorities were considered and not simply to elect Democrats.” Id., at 384 (statement of Rep. Hill).
It is ironic and slightly unfair for the plurality and District Court to use the State’s §5 submission and Congresswoman Johnson’s testimony in a §2 challenge to the congressional district as evidence against them in these cases. See, e. g., 861 F. Supp., at 1319-1321, 1338-1339; ante, at 969-970. Both of those proceedings required the State to assure the Attorney General and a federal court, respectively, that the State had adequately considered the interests of minority voters in the 1991 redistricting process. Under such circumstances, it is not at all surprising that the relevant declarant would limit his or her comments to the role that race played in the redistricting process, for other considerations were largely irrelevant (the District Court’s opinion to the contrary notwithstanding, see 861 F. Supp., at 1339).
As Justice Ginsbueg noted in her dissent in Miller, “ethnicity itself can tie people together” in communities of interest. 515 U. S., at 944; see also Rogers v. Lodge, 458 U. S. 613, 651 (1982) (Stevens, J., dissenting) *1027(“Whenever identifiable groups in our society are disadvantaged, they will share common political interests and tend to vote as a ‘bloc’ ”). Furthermore, it may be that the very fact of racial bloc voting, a prerequisite to §2 liability, see Thornburg v. Gingles, 478 U. S. 30, 51 (1986) (and, under the Court’s recent jurisprudence, to the voluntary formation of a majority-minority district), demonstrates the presence of a minority community. While communities based on race may merit a more skeptical review to ensure that a bond, rather than mere stereotyping, ties the community, see 861 F. Supp., at 1338, recognition of such a community in an electoral district certainly could, in certain circumstances, serve as a legitimate race-neutral explanation for particularly odd district shapes. By suggesting the contrary, I believe that the District Court erred. See ibid,.; post, at 1060-1061 (Souter, J., dissenting).
The District Court’s legal analysis was probably flawed in part because its decision was issued before this Court announced its opinion in Miller.
While it may be that the political gerrymandering in this case is “different in degree” from that previously recognized, 861 F. Supp., at 1334,1 do not believe that the reference in Shaw I and Miller to “traditional” districting principles, see Shaw I, 509 U. S., at 642; Miller, 515 U. S., at 916, was intended to prohibit a State from changing the process or policies underlying the complex negotiating process that is modern redistricting.
The plurality expresses particular concern over the use of computer programs, particularly the availability of block-by-block racial data, and *1031argues that the State’s effort to “compilfe] detailed racial data,” ante, at 967, is evidence of the controlling role of race in the computer-dominated process of redistricting. See ante, at 961-962; 861 F. Supp., at 1318-1319. It is worth noting, however, that the State made no particular “effort” to gather these data; it was included, along with similarly detailed information about sex, age, and income levels, in the data set provided by the Census Bureau and imported wholesale into the State’s redistricting computers. Cf. Shaw, 861 F. Supp., at 457. Furthermore, even if the computer was used to fine tune the district lines to ensure that minority communities were included in District 30 (rather than individualized requests from candidates and their staffers on the basis of block-level data, see supra, at 1017-1018), such a technique amounts to little more than the use of a particularly efficient and accurate means of ensuring that the intended nature of the district was not undermined as incumbency protection forced it out of a compact district. I do not suggest that the end can always justify the means, but if those means are no more invidious than the end itself, I do not understand why their use should affect the analysis. I would not condemn state legislation merely because it was based on accurate information.
“A prediction based on a racial characteristic is not necessarily more reliable than a prediction based on some other group characteristic. Nor, since a legislator’s ultimate purpose in making the prediction is political in character, is it necessarily more invidious or benign than a prediction based on other group characteristics. In the line-drawing process, racial, religious, ethnic, and economic gerrymanders are all species of political gerrymanders.” Mobile v. Bolden, 446 U. S. 55, 88 (1980) (Stevens, J., concurring in judgment) (footnote omitted).
To the extent that a political prediction based on race is incorrect, the voters have an entirely obvious way to ensure that such irrationality is not relied upon in the future: Vote for a different party. A legislator relying on racial demographics to ensure his or her election will learn a swift lesson if the presumptions upon which that reliance was based are incorrect.
1 find it particularly ironic that the Court considers the use of race verboten in this benign context, while the Court just recently, on the basis of evidence that, inter alia, “[m]ore than 90% of the persons sentenced in 1994 for crack cocaine trafficking were black,” dismissed out of hand the Ninth Circuit’s assumption that “people of all races commit all types of crimes.” United States v. Armstrong, ante, at 469. The Ninth Circuit’s conclusion, it seems to me, is a model of the sort of race-neutral decision-making that this Court insists should be a part of constitutional decision-making processes.
Although I conclude that no reasonable interpretation of the record would require the application of strict scrutiny to District 30,1 believe for the reasons that follow that it, too, would survive strict scrutiny if it were to be subject to that level of review.
fWhile I believe that the evidence supporting the State’s conclusions in this regard is stronger than that suggested by the plurality or Justice Kennedy in his concurring opinion, I will simply assume, argmndo, as the plurality does, that the State had a reasonable fear of liability under §2. See also supra, at 1007.
Even if the Court in Shaw II is correct in asserting that Noi th Carolina's District 12 would not have allowed the State to avoid liability under §2, see ante, at 916-918, no such plausible argument could be made in these cases. The core of District 30, for instance, contains more than half of all the African-American population in the district, and coincides precisely with the heart of the compact community that.the State reasonably believes would give rise to a § 2 violation were it not placed in a majority-minority district. The same facts are true with respect to the Houston districts.
The difficulty of balancing between these competing legal requirements will only be exacerbated by the ability of litigants (and courts) to use evidence proffered in defense by the State or its actors in one context as evidence against the State in another. See n. 24, supra. While there is nothing wrong with using prior inconsistent statements (to the extent that they really are inconsistent), States will be all the more unwilling to enter into the process at all given the certainty that they will be subject to suits in which evidence offered in one as defense will be fodder for the plaintiffs in another.
The contrary is also possible, of course. Perhaps the burgeoning role of federal courts in this process, along with their relative isolation from the political pressures that motivate legislatures to bend district lines, will mean that there will actually be fewer politically gerrymandered districts. Regardless of whether political gerrymanders are more or less prevalent after our decisions today, my point is the same: The Court has its hierarchy of values upside down.
My view that a State may act unconstitutionally by gerrymandering to minimize the influence of a group on the political process is consistent with the belief that there is no constitutional error in the drawing of district lines based on benign racial considerations. As Justice Powell noted in his opinion in Davis v. Bandemer, 478 U. S., at 165, there is a sharp distinction between “gerrymandering in the ‘loose’ sense” {i. e., the drawing of district lines to advance general political and social goals), and “gerrymandering that amounts to unconstitutional discrimination” (i, e., the drawing of district lines for the sole purpose of “ ‘occupying] a position of strength at a particular time, or to disadvantage a politically weak segment of the community,’ ” id., at 164 (citing Kareher, 462 U. S., at 748 (Stevens, J., concurring)). See also 478 U. S., at 125, n. 9 (“[A] preference for nonpartisan as opposed to partisan gerrymanders . . . merely recognizes that nonpartisan gerrymanders in fact are aimed at guaranteeing rather than infringing fair group representation”). While I believe that allegations of discriminatory intent and impact, if proved, should give rise to a constitutional violation, Shaw, Miller, and these cases all involve allegations of both impact and intent that are far more diffuse than the allegations to which we have traditionally directed our most rigorous review. See Shaw II, ante, at 921-923 (Stevens, J., dissenting); cf. Gomillion v. Lightfoot, 364 U. S. 339 (1960). Limiting the constitutional ban on gerrymandering to those claims alleging that a specific group (as opposed to every group) has been harmed would be far more consistent with prior precedent than the Court’s still-developing jurisprudence of racial gerrymandering.
Compare 51 Congressional Quarterly 10 (1993) (list of African-Americans who have served in Congress through the end of 1992) and Supplement to 52 Congressional Quarterly 10 (Nov. 12, 1994) (listing minorities in the 104th Congress) with biyearly publications of The Almanac of American Politics (published 1975-present).
D. Bositis, Joint Center for Political and Economic Studies, Affrican-Americans & the 1994 Midterms 22 (rev. May 1995). Fifteen black candidates ran for office in majority-white districts. Ibid.