Full opinion text
MEMORANDUM OPINION GRIFFITH, Circuit Judge: Table of Contents I. Background.............................................................138 II. Principles of Section 5 Analysis...........................................139 A.Retrogression.......................................................139 1. Texas’s Burden of Proof..........................................140 2. Election Analysis Methodologies..................................141 a. Types of Elections ...........................................141 b. Election Analysis Sample Sets................................143 3. Statewide Retrogression Analysis.................................144 4. Coalition and Crossover Districts .................................147 a. Section 5 Analysis............................................147 b. Standard of Proof............................................149 B.Discriminatory Intent...............................................151 III. Congressional Plan......................................................152 A.Retrogression in the Congressional Plan..............................152 1. Congressional District 27.........................................153 2. Congressional District 23.........................................154 3. Retrogression with New Congressional Seats.......................156 B.Discriminatory Intent in the Congressional Plan.......................159 IV. State Senate Plan .......................................................162 A.Retrogression in the Senate Plan.....................................162 B.Discriminatory Intent in the Senate Plan .............................163 V. State House Plan........................................................166 A.Retrogression in the State House Plan................................166 1. Alleged Retrogressive Districts....................................167 a. State House District 33.......................................167 b. State House District 35.......................................167 c. State House District 41.......................................169 d. State House District 117......................................170 e. State House District 149......................................172 f. State House Districts 26,106, and 144..........................175 2. Alleged New Ability Districts.....................................175 B.Discriminatory Intent in the State House Plan 177 VI. Conclusion....................... 178 The latest Census reports that since 2000 the population of Texas grew by over four million. This dramatic increase required the Texas legislature to create new voting districts for the four seats added to the State’s congressional delegation, U.S. Const, art. I, § 2, cl. 3; id. amend. XIV, § 2, and draw new boundaries for the state and congressional voting districts to comply with the mandate of one-person, one-vote, see Georgia v. Ashcroft, 539 U.S. 461, 488 n. 2, 123 S.Ct. 2498, 156 L.Ed.2d 428 (2003). Because Texas is a covered jurisdiction under section 5 of the Voting Rights Act of 1965(VRA), 42 U.S.C. § 1973, the Attorney General of the United States or a three-judge panel of this Court must approve, or “preclear,” any redistricting plan before it can take effect. Id. § 1973e(a). Texas chose not to seek administrative preclearance and instead seeks from this Court a declaratory judgment that its redistricting plans will neither have “the purpose nor will have the effect of denying or abridging the right to vote on account of race or color, or [language minority group].” Id. The United States opposes preclearance of the redistricting plans for Texas’s congressional delegation and the State House of Representatives, but has no quarrel with the plan for the Texas Senate. Seven Intervenors raise a variety of challenges that collectively encompass all three plans. We conclude that Texas has failed to show that any of the redistricting plans merits preclearance. I. Background On July 19, 2011, Texas filed a complaint in this Court seeking a declaratory judgment that its newly enacted redistricting plans for the U.S. House of Representatives (Plan C185 or Congressional Plan), the Texas House of Representatives (Plan H283 or House Plan), and the Texas Senate (Plan S148 or Senate Plan) comply with section 5 of the VRA. This Court has been properly convened as a three-judge court, 28 U.S.C. § 2284; 42 U.S.C. § 1973c(a), and we took jurisdiction under 42 U.S.C. § 1973c and 28 U.S.C. §§ 1346(a)(2), 2201. After the United States and several Intervenors filed answers, Texas moved for summary judgment for all three plans on September 14, 2011. We heard argument on the motion on November 2, 2011, and issued an order denying summary judgment on November 8, 2011, 2011 WL 5402888. Our memorandum opinion followed on December 22, 2011. The same three redistricting plans have been challenged under section 2 of the VRA before a three-judge district court in the Western District of Texas. The State’s population growth and the addition of four seats to its congressional delegation make it impossible for Texas to conduct elections using the district boundaries last approved under section 5. Our denial of Texas’s motion for summary judgment required the district court in the section 2 litigation to draw interim maps for the State’s fast-approaching primaries and the ensuing general election. After the Supreme Court invalidated those maps, see Perry v. Perez, — U.S. -, 132 S.Ct. 934, 181 L.Ed.2d 900 (2012), the court issued a second set, which have not been challenged. See Feb. 28, 2012 Order, Perez v. Perry, No. 11-cv-360 (W.D.Tex. filed May 9, 2011), ECF No. 681 (Congressional Plan interim map); Feb. 28, 2012 Order, Perez, No. 11-cv-360, ECF No. 682 (House Plan interim map); Feb. 28, 2012 Order, Davis v. Perry, No. 5:11-ev-00788 (W.D.Tex. filed May 9, 2011), ECF No. 141 (Senate Plan interim map). Meanwhile, after expedited discovery, this Court sat for trial January 17-26, 2012, with closing arguments on January 31, 2012. The voluminous trial record includes evidence taken in open court, party exhibits, expert reports, post-trial briefing, and designated portions of the transcript from the section 2 trial in Texas. After reviewing this record and carefully considering the arguments of all parties, we now deny Texas preclearance and enter judgment for the defendants. In the discussion that follows, we do not recount the extensive background of the Voting Rights Act or of this case. Much of that is contained in our opinion at summary judgment. In addition, we do not repeat many of the factual findings set out in the appendix to this opinion. Using the framework for applying section 5 described in our summary judgment opinion, we first address a series of legal issues that remain outstanding after trial about what section 5 requires for preclearance. Then, we examine the Congressional, Senate, and House Plans in turn. II. Principles of Section 5 Analysis A. Retrogression Texas must show that its redistricting plans have neither the effect nor the purpose of abridging minority voting rights. 42 U.S.C. § 1973c(a). We will take up the “purpose” prong below in section B. The goal of the “effect” prong is “to insure that no voting-procedure changes would be made that would lead to a retrogression in the position of racial minorities with respect to their effective exercise of the electoral franchise,” Beer v. United States, 425 U.S. 130, 141, 96 S.Ct. 1357, 47 L.Ed.2d 629 (1976), regardless of whether the change was intended to do so. “Effective exercise,” in turn, has long been understood to include not only the “ability of minority groups to participate in the political process,” but also the ability “to elect their choices to office.” Id. (quoting H.R.Rep. No. 94-196, at 60 (1975)). In the most recent reauthorization of the VRA, Congress further reinforced the meaning of the effect prong by stating that minority voters’ “ability to elect” their candidates of choice is the appropriate measure of whether a proposed change will be retrogressive. See 42 U.S.C. § 1973c(b) (stating that section 5 blocks voting changes that diminish minority citizens’ “ability ... to elect their preferred candidates of choice”), id. § 1973c(d) (explaining that the “purpose of subsection (b) ... is to protect the ability of [minority] citizens to elect their preferred candidates of choice”). As we explained in our summary judgment opinion, ensuring that a proposed plan will not undo the gains minority voters have achieved in electoral power requires a multi-factored, functional analysis. Texas v. United States, 831 F.Supp.2d 244, 262-64 (D.D.C.2011). A single-factor inquiry, such as the test Texas proposed relying on racial and ethnic population statistics alone, is inconsistent with precedent and too limited to provide an accurate picture of the on-the-ground realities of voting power. Id.; see also, e.g., Ashcroft, 539 U.S. at 480, 123 S.Ct. 2498 (“The ability of minority voters to elect a candidate of their choice is important but often complex in practice to determine.”). We do not repeat here the rationale for our conclusion, but instead address the additional arguments raised at trial about the appropriate standard to determine retrogression. 1. Texas’s Burden of Proof Texas bears the burden of proving by a preponderance of the evidence that its redistricting plans are not retrogressive. City of Pleasant Grove v. United States, 479 U.S. 462, 469, 107 S.Ct. 794, 93 L.Ed.2d 866 (1987). Texas does not deny that it bears this burden. Instead, relying on the Supreme Court’s observation that a state is entitled to select its “own method of complying with the Voting Rights Act,” Bartlett v. Strickland, 556 U.S. 1, 23, 129 S.Ct. 1231, 173 L.Ed.2d 173 (2009) (plurality opinion), Texas claims that “the flexibility to choose one theory of effective representation over the other,” Ashcroft, 539 U.S. at 482, 123 S.Ct. 2498, gives it significant latitude in how to prove its case. Tex. Post-Trial Br. 3. We agree that section 5 does not interfere with many of the policy judgments a state must make during redistricting, such as whether to retain an ability district — a district in which minority citizens have the ability to elect their preferred candidates — or create a new one elsewhere. Yet Texas takes this point too far, claiming that the prerogative to choose among methods of redistricting extends to the type of evidence we should use to measure retrogression. For example, Texas argues that we must defer to its decision to use the results of statewide elections to measure compliance with section 5. Id. at 5. We disagree. Ashcroft holds that states may choose between “theories] of effective representation,” 539 U.S. at 482, 123 S.Ct. 2498 (emphasis added), but gauging effectiveness is a legal judgment that we must make. Texas is entitled to advocate its preferred methods of measuring minority voting strength, and we address those arguments below, but we need not defer to a state’s legal theory on how best to measure minority voters’ ability to elect. That is a measure at the heart of the preclearance analysis that section 5 has left to the Attorney General or the judiciary. 2. Election Analysis Methodologies The parties have submitted reports and testimony from fourteen experts in fields such as redistricting, election analysis, voting rights law, and the history of voting discrimination in Texas. Although we do not find the analysis of any one expert sufficient to guide our retrogression inquiry, we rely most heavily on the reports and testimony of Dr. Lisa Handley, expert for the United States; Dr. Richard Engstrom, expert for the Texas Latino Redistricting Task Force (TLRTF); and Dr. Stephen Ansolabehere, expert for the Gonzales Intervenors. We find their methodologies sound and their conclusions helpful to our analysis of the State’s redistricting. To explain our use of these experts we address two areas of disagreement between the parties about the merits of the various approaches the experts use: which type of elections to examine and the appropriate sample sets to use. a. Types of Elections Endogenous analysis examines the results of elections held within a district to determine how often minority-preferred candidates succeed. See, e.g., Defs.’ Ex. 326, Dr. Lisa Handley, A Section 5 Voting Rights Analysis of the Proposed Texas State House Plan 3 [hereinafter Handley House Rep.]. Because endogenous analysis is based on actual election results within a single district, it is necessarily retrospective. It can only be used to determine whether a district in the existing, or benchmark, plan has an ability to elect. It cannot be used to assess whether a proposed district does as well, because a proposed district has not yet conducted any district-wide elections. Exogenous election analysis examines how minority-preferred candidates fared in a particular district in statewide or national elections. See, e.g., Pl.’s Ex. 175, Direct Written Test, of Dr. John Alford 5-6 [hereinafter Alford Rep.]. Take the 2008 presidential election as an example. In a state where minority voters almost always prefer Democratic candidates, exogenous election analysis suggests that minority voters lack an ability to elect in a benchmark district carried by John McCain over Barack Obama. Because exogenous analysis considers results from elections that occur across all districts in a state, such analysis allows comparison between benchmark and proposed districts. Precinct-level data from statewide or national elections can show if the minority-preferred candidate won the benchmark district, and by assembling, or “reconstituting,” the precinct-level returns into a district’s proposed new shape, exogenous election analysis can indicate whether the minority-preferred candidate would have won in the proposed district as well. Texas urges us to consider exogenous election analysis alone, see Tex. Post-Trial Br. 4-5, but we conclude that endogenous results are often more probative of ability to elect. As Dr. Engstrom explained, exogenous elections are “not a good basis for predicting the specific number of elections in many new districts that will result in Hispanic preferred candidates winning,” partly because there are significant contextual differences between exogenous and endogenous elections. Defs.’ Ex. 747, Rebuttal Report of Dr. Richard Engstrom 6 [hereinafter Engstrom Reb. Rep.]. Likewise, Dr. Handley concluded that “the most essential piece of information” when determining benchmark ability districts “is whether minority voters have been successful at electing their preferred candidates to the legislative office at issue in the district.” Defs.’ Ex. 794, Rebuttal Report of Dr. Lisa Handley to Supplement Expert Report of Dr. John Alford 3 [hereinafter Handley Reb. Rep.]. Candidates in endogenous elections live in a particular district and focus their campaigns on local voters. Candidates in statewide elections are likely to make an appeal with a less direct connection to voters in that district. Nationwide contests are even more attenuated. Local connections and direct campaigning, then, may allow a minority-preferred candidate to win an endogenous election in a district the minority-preferred candidate for statewide office could not carry. We agree with Dr. Engstrom and Dr. Handley. Given the numerous and difficult-to-quantify factors that go into determining ability to elect, the best evidence is whether and how often minority voters have actually elected their candidate of choice to the position at issue, not the indirect proxy offered by exogenous analysis. Texas argues that endogenous analysis is an “impracticable” tool because it is available only for benchmark plans and does not provide the “common unit of measurement” available with exogenous results. Tex. PosWTrial Br. 4. As we have stated, we agree that endogenous elections are not well suited to prospective analysis, but when predicting the impact of redistricting changes on minority voters’ ability to elect, more information is better than less. We should not discount the powerful evidence of minority voting power that endogenous elections provide in favor of a single tool that may be a less accurate gauge. When endogenous and exogenous analyses yield different results, we will give special attention to other relevant characteristics of the voting district. Texas argues that endogenous analysis may overvalue minority voting power and undervalue the advantage of incumbency in districts where the minority-preferred candidate has been repeatedly reelected. See id. at 5. We disagree with the premise that an incumbent’s advantage does “not bear on the ability-to-elect inquiry.” Id. The advantage incumbents enjoy during reelection campaigns is a factor that minority voters, like any other voters, often use to help elect their preferred candidate. Ability to elect is not less real simply because subsequent elections are easier to win than the first. Texas raises the more specific objection that endogenous results may be misleading in a district in which ability status is closely contested if a long-term incumbent plans to retire. Id. Yet as our analysis below bears out, our finding that endogenous elections are particularly probative evidence does not mean that a high endogenous score automatically implies ability status. Careful consideration of all factors matters, especially in close cases. We thus see no reason to exclude all endogenous election data from our analysis, nor to weigh exogenous data more heavily. Both types of data provide information about whether minority voters are or will be able to participate in the political process. b. Election Analysis Sample Sets The experts also vary widely in which elections they used for their sample sets. All use a similar methodology for their exogenous analysis. Starting with the boundaries in the benchmark plan, they count the number of times the minority-preferred candidate carried the district. Reconfiguring the districts by regrouping precincts as called for in the enacted plan, their analyses then look to see how many times the minority-preferred candidate would have carried that district. Outcomes are determined by inputs, of course, and whether the analysis shows an ability to elect turns on variations in the sample set such as the number of elections chosen, the length of time they span, whether the sample is weighted toward more recent contests, and the offices at stake. For example, Texas’s expert, Dr. Alford, relies on reconstituted election results from a'set of ten statewide elections weighted toward more recent years provided by the Texas Office of the Attorney General (the OAG 10). See Alford Rep. 9 tbl.2. Texas argues that we should give greatest weight to these exogenous results because they used a larger data set and relied more heavily on recent elections than did any other expert in the case. Tex. Post-Trial Br. 5. Although we agree that a larger data set generally improves accuracy, we are not persuaded that the OAG 10 is the best indicator of minority voting strength. A preference for recent elections may in fact distort the results. Dr. Handley, the expert for the United States, cautions against giving more weight to some years than others. To do so, she warns, would allow atypical election years to skew the picture of long-term minority voting power. See Handley Reb. Rep. 4 n. 6. This caution is especially appropriate here because three of the OAG 10 elections are from the 2010 election cycle. As the evidence in this case shows, 2010 was an unusual year with low Democratic turnout in which Republicans won several seats that had long been held by Democrats. See, e.g., Defs.’ Ex. 776, Seliger Dep. 15:1— 7, Sept. 1, 2011, Perez, No. ll-cv-360 [hereinafter Seliger San Antonio Dep.] It is too soon to tell if 2010 was an aberration or marked the start of a lasting change in Texas politics. Our concerns with the OAG 10 extend to the other sample sets used by the parties’ experts. Dr. Engstrom’s exogenous election sample also places greater weight on recent years, considering elections from only 2006-2010. See Defs.’ Ex. 726, Supplemental Expert Report of Dr. Richard Engstrom 2 [hereinafter Engstrom Suppl. Rep.]; Defs.’ Ex. 799, Dr. Richard Engstrom Analysis: Retrogression in State’s Adopted House Plan [hereinafter Engstrom Chart]. And all the experts in this case use relatively small sample sets. Dr. Handley, for example, uses only five elections from 2002-2010, and Dr. Engstrom uses just seven general elections. Handley House Rep. 3-4; Engstrom Suppl. Rep. 2; Engstrom Chart. Where there are so many elections from which to choose — the record contains analysis using races ranging from governor to railroad commissioner — it is hard to assess the merits of any one expert’s data when the sample sets are small and often do not overlap. In short, we are uncomfortable relying exclusively on the exogenous analysis of any single expert. Our solution is to consider the exogenous results from all three of these sources — the OAG 10, Dr. Handley, and Dr. Engstrom — as well as the analysis for the Congressional Plan conducted by Dr. Stephen Ansolabehere, the expert for the Gonzales Intervenors, finding all to be probative but none dispositive. 3. Statewide Retrogression Analysis As the Supreme Court has made clear, our analysis of minority voting power “must encompass the entire statewide plan as a whole.” Ashcroft, 539 U.S. at 479, 123 S.Ct. 2498. Section 5 is not concerned with the location of particular ability districts, but rather with whether the enacted plan, in its entirety, preserves minority voters’ ability to elect. In other words, section 5 allows a state to dismantle an ability district as long as it offsets that loss by drawing a new ability district elsewhere. But Texas asks us to expand this principle to a point that is inconsistent with section 5. Texas’s expert submitted two reports to the Court, one at summary judgment and another at trial. His first report counted any district in which the number of registered Hispanic voters exceeded 50% or the Black Voting Age Population (BVAP) exceeded 40% as an ability district, without giving attention to actual election performance. See Texas, 831 F.Supp.2d at 263 n. 23. After we rejected this single-factor test, Dr. Alford changed tack in his trial report, which uses what he calls a “statewide functional analysis.” See Alford Rep. 7. Rather than determine if particular districts are ability districts, Dr. Alford’s latest approach examines changes in the degree of minority voting power across the entire plan. Using the benchmark ability districts the United States listed, Dr. Alford counted every instance in which a minority-preferred candidate carried the district in an exogenous election. He then counted how many times the minority-preferred candidate would have carried the district in the enacted plan. If the total number of “wins” in the enacted plan meets or exceeds the number in the benchmark, Dr. Alford concludes that the plan is not retrogressive. See id. at 7-12. Dr. Alford contrasts his statewide approach to what he calls the “binary” approach of every other expert in the case. Those experts examine each district individually, using exogenous results as one factor when determining if a district is an ability district. See id. at 12-13. They then compare the number of ability districts in the benchmark map with the number in the proposed plan. Dr. Alford’s method counts election victories across all districts and does not label a district as “ability” or not. Texas argues this approach is superior to the “blunt technique” of the binary method because it “captures the degree of minority voting strength across all relevant districts.” Tex. Post-Trial Br. 6. Perhaps, but this approach is a variation on the type of retrogression analysis that Congress rejected when it amended the VRA in 2006. In Georgia v. Ashcroft, 539 U.S. 461, 123 S.Ct. 2498, 156 L.Ed.2d 428 (2003), the Supreme Court concluded that courts “should not focus solely on the comparative ability of a minority group to elect a candidate of its choice,” but instead should consider the “totality of the circumstances” regarding minority participation in the electoral process. Id. at 479-80,123 S.Ct. 2498. Specifically, the Court concluded that states could draw maps containing a combination of two different types of districts to satisfy section 5: traditional majority-minority districts, and “influence districts,” which are not ability districts, but rather those in which minority voters play a “substantial, if not decisive, role in the electoral process.” Id. at 480-83,123 S.Ct. 2498. Congress rejected this holding in 2006 when it reauthorized section 5, making it clear that retrogression is not concerned with the degree of influence minority voters exert, but with their ability to elect their preferred candidates. See 42 U.S.C. § 1973e(b) (stating that voting changes must not diminish minority citizens’ “ability ... to elect their preferred candidates of choice”), id. § 1973c(d) (defining subsection (b)’s purpose as protecting “the ability of [minority] citizens to elect their preferred candidates of choice”). The House Report explained that the 2006 amendments were a response to Georgia v. Ashcroft, which allowed “the minority community’s own choice of preferred candidates to be trumped by political deals struck by State legislators purporting to give ‘influence’ to the minority community while removing that community’s ability to elect candidates.” H.R.Rep. No. 109-478, at 69 (2006), 2006 U.S.C.C.A.N. 618, 670. Congress decided that “[permitting these trade-offs is inconsistent with the original and current purpose of Section 5.” Id.; see also id. at 68-72; S.Rep. No. 109-295, at 18-20 (2006) (stating that the amendments “clarify that [section 5] protects the ability of minority voters ‘to elect their preferred candidates of choice,’ ” id. at 19). Congress does not view “ability to elect” in degrees; states may not add up districts in which minority voters have “partial” ability to elect to satisfy section 5. Instead, Congress views ability status as an on-off switch: minority voters either have an ability to elect in a district or they do not. Endorsing Dr. Alford’s analysis would be a return to the approach Congress rejected in 2006. Consider, for example, a benchmark map with three districts. In two of the districts, minority voters elect their preferred candidates in six out of ten elections in a sample set, but in the third, they fail to win a single election. In all three districts in the enacted plan, minority-preferred candidates win in four out of the ten elections. A traditional binary approach would likely conclude that the benchmark map has two ability districts (where minority voters can elect their candidate of choice more often than not), and the enacted plan has no ability districts, just three influence districts. Such a plan would be clearly retrogressive under the current version of section 5. Yet Dr. Alford’s approach would show no retrogression because the total number of minority electoral victories remains the same (6 + 6 + 0 = 12 in the benchmark; 4 + 4 + 4 = 12 in the enacted). Texas argues that Dr. Alford’s approach yields better policy results, but such determinations belong to Congress, not the courts. In any event, the “benefits” Texas touts are illusory. Texas argues that the binary approach “ignores gradations in minority abilities to elect and gives States no credit for improving electoral performance in districts that stay above or below the ability-to-elect cutoff.” Tex. Post-Trial Br. 6. In other words, Texas seeks credit for strengthening an already-performing district from, say, six out of ten victories to ten out of ten. Yet giving credit in a scenario like this would allow Texas to use those four “additional” victories to offset a four-election decrease elsewhere. Such an approach would create a legal tool to dismantle ability districts as long as the state increases the effectiveness of others. In short, it would give states credit for packing minority voters into districts. A starker example would be a plan in which six benchmark districts perform for minority voters nine out of ten times, but perform ten out of ten times in the enacted plan. Statewide functional analysis would allow a state to use this six-election “increase” in minority effectiveness to weaken or even destroy ability districts in other parts of the state. We also find it significant that Dr. Alford can point to no other advocates of his approach within the well-populated field of voting rights and redistricting. Statewide functional analysis is not only foreclosed by the 2006 amendments, but it lies outside accepted academic norms among redistricting experts. See, e.g., Engstrom Reb. Rep. 2-6 (critiquing Dr. Alford’s approach and noting he was “not aware of any analysis, prior to this one by Dr. Alford, by any expert that completely ignores the results of endogenous elections in a benchmark plan in a retrogression analysis,” id. at 2); Handley Reb. Rep. 2-6 (critiquing Dr. Alford’s approach). Moreover, statewide functional analysis would be much more difficult to administer than the already fact-intensive binary approach because courts would need to make even more precise findings than whether a district is or is not an ability district. Courts would need to determine, for example, the difference between districts with effectiveness levels of 60% and 70%. Dr. Alford claims he can make these fine distinctions based on a district’s electoral performance in the limited set of elections that he chose. Yet as the multitude of experts in this case demonstrates, there is no agreed-upon method to choose how many elections are necessary to demonstrate voting strength, much less which elections and over how long a period of time. We lack confidence that we, or any court, would be able to make findings at the level of precision Dr. Alford’s approach requires. Finally, we reject Texas’s argument that refusing to accept statewide functional analysis would increase the “substantial federalism costs” of preclearance by further limiting state flexibility, at the risk of rendering section 5 unconstitutional. See Tex. Post-Trial Br. 7 (quoting Reno v. Bossier Parish Sch. Bd. (Bossier II), 528 U.S. 320, 120 S.Ct. 866, 145 L.Ed.2d 845 (2000)) (internal quotation marks omitted). The constitutional avoidance canon is no aid to Texas because we are not faced with two competing yet permissible interpretations of section 5. See United States v. X-Citement Video, Inc., 513 U.S. 64, 69, 115 S.Ct. 464, 130 L.Ed.2d 372 (1994) (describing the interpretative presumption “that a statute is to be construed where fairly possible so as to avoid substantial constitutional questions” (emphasis added)). As we have just discussed, retrogression analysis under section 5 as amended limits our analysis to ability to elect and does not permit us to weigh degrees of effectiveness. We cannot adopt an interpretation at odds with the statutory text to avoid possible constitutional concerns. 4. Coalition and Crossover Districts a. Section 5 Analysis In a crossover district, a minority group “is large enough to elect the candidate of its choice with help from voters who are members of the majority and who cross over to support the minority’s preferred candidate.” Bartlett, 556 U.S. at 13, 129 S.Ct. 1231. In a coalition district, two or more minority groups work together to elect their preferred candidate. Id. We held at summary judgment that because existing “coalition and crossover districts provide minority groups the ability to elect a preferred candidate, they must be recognized as ability districts in a Section 5 analysis of a benchmark plan.” Texas, 831 F.Supp.2d at 267-68. Texas asks us to reconsider our ruling in light of Bartlett v. Strickland, 556 U.S. 1, 129 S.Ct. 1231, 173 L.Ed.2d 173 (2009) (plurality opinion), and the Supreme Court’s recent decision in Perry v. Perez, — U.S. -, 132 S.Ct. 934, 181 L.Ed.2d 900 (2012). Having considered the parties’ arguments, we reaffirm our conclusion that coalition and crossover districts are protected under section 5. The Supreme Court has never directly addressed whether section 5 protects coalition or crossover districts. A close reading of Georgia v. Ashcroft, however, suggests that it does. The Court described districts with “coalitions of voters who together will help to achieve the electoral aspirations of the minority group,” 539 U.S. at 481, 123 S.Ct. 2498, concluding that such districts count as “effective representation” for purposes of section 5, just like “safe majority-minority districts.” Id. at 480-82, 123 S.Ct. 2498 (“Section 5 gives States the flexibility to choose one theory of effective representation over the other.” Id. at 482, 123 S.Ct. 2498.). The Court’s statements in Georgia v. Ashcroft are reinforced by the House Report accompanying the 2006 amendments, which spoke of coalition districts as a type of ability district: “Voting changes that leave a minority group less able to elect a preferred candidate of choice, either directly or when coalesced with other voters, cannot be precleared under Section 5.” H.R.Rep. No. 109-478, at 71, 2006 U.S.C.C.A.N. 618, 672 (emphasis added). In addition, the Court’s jurisprudence under section 2, a distinct yet related provision of the VRA mandating equal opportunity for minority voters to participate in the electoral process, supports protecting coalition and crossover districts under section 5. The Court has long acknowledged the existence of coalition and crossover districts, recognizing at times that they can provide the means for minority voters to elect their candidates of choice. See Johnson v. De Grandy, 512 U.S. 997, 1020, 114 S.Ct. 2647, 129 L.Ed.2d 775 (1994) (describing “communities in which minority citizens are able to form coalitions with voters from other racial and ethnic groups, having no need to be a majority within a single district in order to elect candidates of their choice ” (emphasis added)); Voinovich v. Quitter, 507 U.S. 146, 154, 113 S.Ct. 1149, 122 L.Ed.2d 500 (1993) (describing a district in which a minority group was not large enough to elect its preferred candidate operating alone but could do so if it “attract[ed] sufficient cross-over votes from white voters”). In fact, the Court has suggested that such districts will become more common over time, replacing majority-minority districts as waning racial polarization makes it easier for minority voters to elect their preferred candidates even when they do not make up the majority of a district’s voters. See De Grandy, 512 U.S. at 1019-20, 114 S.Ct. 2647. In other words, “ability” may look different now than it did when the VRA was first enacted. Our responsibility to protect the rights secured by section 5 calls that we be sensitive to these new, but real, forms of minority voting power. Texas argues that the Court’s decision in Bartlett precludes recognizing coalition and crossover districts under section 5. See Tex. Post-Trial Br. 8. But the Bartlett Court only concluded that section 2 does not compel states to draw new crossover districts under section 2, not that states can disregard the existence of established crossover and coalition districts in a section 5 inquiry. See Texas, 831 F.Supp.2d at 267-68. Significantly, Bartlett noted that it did not reach the question of whether states could choose to draw crossover districts “as a matter of legislative choice or discretion,” and cited Georgia v. Ashcroft to show that drawing such districts may be the most effective way to strengthen minority voting power. Bartlett, 556 U.S. at 23, 129 S.Ct. 1231. Far from revealing skepticism or hostility toward coalition districts, this language suggests that such districts can increase minority voters’ electoral ability, even while holding that states are not required to draw districts maximizing this potential. Nor do the Bartlett Court’s concerns under section 2 speak to our task under section 5. Part of the Court’s analysis rested on the difficulties of predicting whether a potential coalition would provide minorities with an opportunity to elect. Id. at 17, 129 S.Ct. 1231. Section 5, by contrast, asks whether an existing coalition has achieved an ability to elect. Section 5 does not call on us to guess the future, but to determine whether there is past evidence of a demonstrated ability to elect. And while section 2 does not demand granting “special protection to a minority group’s right to form political coalitions” or “impose on those who draw election districts a duty to give minority voters the most potential, or the best potential, to elect a candidate by attracting crossover voters,” id. at 15, 129 S.Ct. 1231, section 5 mandates that we ensure that “the gains thus far achieved in minority political participation [are] not destroyed,” Beer, 425 U.S. at 141, 96 S.Ct. 1357 (quoting S.Rep. No. 94-295, at 19 (1975), 1975 U.S.C.C.A.N. 774, 785). To be sure, forcing a state to create crossover districts would reach beyond section 2’s equality mandate, but nothing in Bartlett suggests that courts can turn a blind eye towards a district in which minority voters have already turned electoral opportunity into ability to elect. And nothing in Perez extends the reasoning in Bartlett to section 5. Perez held only that the district court had no basis to draw a new coalition district under section 2, without addressing the separate question before us: whether preexisting coalition or crossover districts merit protection under section 5. See Perez, 132 S.Ct. at 944. Thus, although section 2 does not require states to draw new crossover districts, we read section 5’s ban on retrogression to extend protection to districts in which minority voters have demonstrated an ability to elect their preferred candidates via either assembling a coalition or attracting sufficient crossover votes, or both. b. Standard of Proof As we stated in our summary judgment opinion, proving the existence of coalition and crossover districts “require[s] more exacting evidence than would be needed to prove the existence of a majority-minority district.” Texas, 831 F.Supp.2d at 268. The discussion that follows explains the test we have applied. At the outset, the minority group or groups must vote cohesively in coalition and crossover districts, just as they must in protected majority-minority districts. See Growe v. Emison, 507 U.S. 25, 41, 113 S.Ct. 1075, 122 L.Ed.2d 388 (1993) (noting that proving political cohesion across an “agglomerated political bloc” — ie., a coalition- — “is all the more essential” than the need to prove cohesion within a single minority group). If minority groups split their vote between opposing candidates in the general election, there is by definition no candidate of choice, and the district is not protected under section 5. While the first inquiry considers whether minority voters have a candidate of choice, the next inquiry is grounded in a different part of section 5: do minority voters have the “ability to elect” their preferred candidate? See 42 U.S.C. § 1973c. In other words, are the groups large enough, motivated enough, or influential enough to elect their candidate of choice— and have they in fact done so? This question is in many respects similar to that for majority-minority ability districts. There is no single, clearly defined metric to determine when a minority group has an ability to elect, so we use a multi-factored approach to determine when a coalition or crossover district achieves that ability. See Growe, 507 U.S. at 41, 113 S.Ct. 1075 (pointing to anecdotal evidence, statistical evidence of minority political cohesion, and racial bloc voting as some of the factors relevant to prove the existence of a coalition district under section 2); Texas, 831 F.Supp.2d at 268 (“[Tjhere must be discrete data, by way of election returns, to confirm the existence of a voting coalition’s electoral power.”). A coalition district is protected under section 5 if there is sufficient evidence to find that minorities voté cohesively and have the ability to elect their preferred candidates. The same two inquiries apply to a crossover district, but the ability-to-elect analysis is more complicated. Although election returns are necessary to show that minority voters in a crossover district have a track record of success— results are the coin of the realm — it is not enough that they simply go along with the electoral decisions of some of the district’s Anglo voters. We must also be satisfied that it is the minority voters themselves who have the ability to elect their preferred candidate. The test to establish this ability must be rigorous enough to avoid the scenario Texas describes: that section 5 will be interpreted to protect any district that elects a Democrat, no matter how small its minority population. In other words, that minority voters are needed to win an election does not in itself prove that they have an ability to elect. As an extreme example of this concern, consider a district with a 90% Anglo and 10% minority population. If the Anglo vote splits evenly between Democrats and Republicans and minorities vote overwhelmingly Democratic, then the Democratic candidate will win with approximately 55% of the vote, and the minority vote will properly be viewed as essential to victory every time. Yet this would not be a district in which the minority group has an ability to elect; the Anglos do. Such a district would merely be a Democratic district that happens to contain a minority group. If we were to hold otherwise, then every district that consistently elects a Democrat with the minority vote providing the margin of victory, no matter how small, would qualify for protection under section 5. This would stretch the scope of section 5 too far. A protected crossover district is not created each time Anglos and minorities vote together to elect a candidate. With respect to both coalition and crossover districts, we require “more exacting evidence” to prove that minority voters have an ability to elect than we do for majority-minority ability districts. Texas, 831 F.Supp.2d at 268. Doing so ensures that we stay within the boundaries of section 5 and protect only those districts in which minority voters have demonstrated their effectiveness. Yet where that standard is met — where minority voters themselves “pull, haul, and trade” to elect their preferred candidates, De Grandy, 512 U.S. at 1020,114 S.Ct. 2647 — then the district is one in which minority voters have an ability to elect, and section 5’s safeguards apply. B. Discriminatory Intent In Reno v. Bossier Parish School Board (Bossier II), 528 U.S. 320, 120 S.Ct. 866, 145 L.Ed.2d 845 (2000), the Supreme Court considered whether section 5 barred a plan that “would have no retrogressive effect” but “nonetheless ... was enacted for a discriminatory ‘purpose.’ ” Id. at 325, 120 S.Ct. 866. The Court held that it did not, concluding that the purpose prong extended only to intent to retrogress, not to all intentional discrimination. Thus, section 5, the Court wrote, would catch only an “incompetent retrogressor,” but offered no recourse against a mapdrawer who intended to discriminate against minority voters using methods that did not create retrogression. Id. at 332, 120 S.Ct. 866. In direct response, the 2006 amendments to section 5 clarified that the term “purpose” must be read more broadly and includes “any discriminatory purpose.” 42 U.S.C. § 1973c(c); see also H.R.Rep. No. 109-478, at 93, 2006 U.S.C.C.A.N. 618, 678 (stating that Congress “rejects the Supreme Court’s holding in Reno v. Bossier Parish”). As a result, we may not preclear any redistricting plan enacted with discriminatory intent. Texas argues that it should not be required to prove that it lacked any discriminatory purpose. Saddling a state with that burden, so the argument goes, adds too much to the serious federalism costs already imposed by preclearance and could “exceed Congress’ enforcement authority under the Fifteenth Amendment and violate the Tenth Amendment.” Tex. Post-Trial Br. 17-18. The only way to avoid this problem, Texas claims, is to shift the burden of proof for discriminatory intent from Texas onto the United States and the Intervenors. Id. at 18. We acknowledge the substantial federalism costs of section 5, see Nw. Austin Mun. Util. Dist. No. One v. Holder, 557 U.S. 193, 203-04, 129 S.Ct. 2504, 174 L.Ed.2d 140 (2009) (stating that the preclearance remedy implicates serious federalism concerns), and recognize the difficulty of proving a negative. Yet it is settled law that Texas bears the burden of proving lack of discriminatory intent. See, e.g., Pleasant Grove, 479 U.S. at 469, 107 S.Ct. 794 (“The burden of proving absence of discriminatory purpose and effect is on [the covered jurisdiction].”); City of Rome v. United States, 446 U.S. 156, 183 n. 18, 100 S.Ct. 1548, 64 L.Ed.2d 119 (1980) (“Under § 5, the city bears the burden of proving lack of discriminatory purpose and effect.”); Beer, 425 U.S. at 140-41, 96 S.Ct. 1357; Georgia v. United States, 411 U.S. 526, 538, 93 5.Ct. 1702, 36 L.Ed.2d 472 (1973); South Carolina v. Katzenbach, 383 U.S. 301, 335, 86 S.Ct. 803, 15 L.Ed.2d 769 (1966). Texas has pointed to no evidence that Congress intended to modify this established understanding. Moreover, Texas’s burden is not insurmountable. There is no question, as the Supreme Court has previously stated, that “assessing a jurisdiction’s motivation in enacting voting changes is a complex task requiring a ‘sensitive inquiry into such circumstantial and direct evidence as may be available.’ ” Reno v. Bossier Parish Sch. Bd. (Bossier I), 520 U.S. 471, 488, 117 S.Ct. 1491, 137 L.Ed.2d 730 (1997) (quoting Vill. of Arlington Heights v. Metro. Hous. Dev. Corp., 429 U.S. 252, 266, 97 S.Ct. 555, 50 L.Ed.2d 450 (1977)). And as Texas rightly argues, this task is all the more difficult because disparate impact alone is insufficient to establish discriminatory purpose, see Bush v. Vera, 517 U.S. 952, 968, 116 S.Ct. 1941, 135 L.Ed.2d 248 (1996) (plurality opinion) (“If district lines merely correlate with race because they are drawn on the basis of political affiliation, which correlates with race, there is no racial classification to justify____”). But we have clear direction how to conduct this “complex task” from Village of Arlington Heights v. Metropolitan Housing Development Corp., 429 U.S. 252, 97 S.Ct. 555, 50 L.Ed.2d 450 (1977). See Bossier I, 520 U.S. at 488, 117 S.Ct. 1491 (“In conducting [a section 5 purpose] inquiry, courts should look to ... Arlington Heights for guidance.”); see also H.R.Rep. No. 109-478, at 68, 2006 U.S.C.C.A.N. 618, 669 (“[T]he factors set out in [Arlington Heights ] provide an adequate framework for determining whether voting changes submitted for preclearance were motivated by a discriminatory purpose.... ”). There, the Court set forth a framework for analyzing “whether invidious discriminatory purpose was a motivating factor” in a government body’s decisionmaking. Arlington Heights, 429 U.S. at 266, 97 S.Ct. 555; see also Bossier I, 520 U.S. at 488-89, 117 S.Ct. 1491 (collecting cases in which courts have applied Arlington Heights in the section 5 context). We follow this well-worn path and base our inquiry upon the five Arlington Heights factors: (1) discriminatory impact, (2) historical background, (3) sequence of events leading up to the decision, (4) procedural or substantive deviations from the normal decisionmaking process, and (5) contemporaneous viewpoints expressed by the decisionmakers. Arlington Heights, 429 U.S. at 266-68, 97 S.Ct. 555. Texas can carry its burden by showing that these factors — the longstanding yardstick for determining discriminatory intent — do not, taken together, show discriminatory purpose. III. Congressional Plan We now turn to the merits of the three plans before us, considering in turn whether Texas has carried its burden to prove the absence of discriminatory purpose and effect in the Congressional, Senate, and House Plans. A. Retrogression in the Congressional Plan There are thirty-six districts in the enacted Congressional Plan. Certain Intervenors argue that the enacted plan has one fewer ability district than the benchmark because three ability districts — Congressional Districts (CDs) 28, 25, and 27 — are lost and only two ability districts — CDs 34 and 35 — are added. There is no dispute that these two new districts are Hispanic ability districts. Texas agrees that CD 27 is a lost ability district, but disputes that benchmark CDs 23 and 25 are ability districts. Under Texas’s theory, the Congressional Plan results in a net increase of one Hispanic ability district. The United States and certain Intervenors argue that the enacted Congressional Plan retrogresses by failing to draw an additional Hispanic ability district. They assert that CDs 23 and 27, but not CD 25, were Hispanic ability districts in the benchmark whose loss in the enacted plan is offset by the gain of CDs 34 and 35. Nevertheless, in light of the growth in the State’s Hispanic population, they argue that failing to draw one of the four new congressional districts as a Hispanic ability district increases the degree of Hispanic disenfranchisement from the benchmark level and thus violates section 5. In addition to these arguments about Hispanic ability districts, some of the Intervenors argue that the Congressional Plan is retrogressive with respect to Black voters as well. All parties agree that CDs 9, 18, and 30 are ability districts for Black voters in both the benchmark and enacted congressional maps. Some of the Intervenors allege that the enacted plan “packed” these districts with Black voters from neighboring jurisdictions that were not performing for minority voters. But because section 5’s effect prong does not prohibit reductions in minority voting power in nonability districts, we find no retrogression in Black ability districts in the Congressional Plan. We do, however, conclude that the enacted Congressional Plan is retrogressive and thus cannot be precleared under section 5. Although we differ among ourselves whether benchmark CD 25 was an ability district, this disagreement does not affect our overall conclusion. At the outset, we discuss the two disputed ability districts upon which we agree, then explain the majority’s conclusion that Texas was required to draw a new ability district under section 5. We set out our separate views on CD 25 at the end of the opinion. 1. Congressional District 27 Benchmark CD 27 includes the cities of Corpus Christi and Brownsville in southeastern Texas. With a Hispanic Citizen Voting Age Population (HCVAP) of 63.8%, Pl.’s Ex. 11, at 9, and, until 2010, a twenty-seven year history of representation by a Hispanic Democrat, benchmark CD 27 is a clear Hispanic ability district. Although an Anglo Republican won the seat with a 775 vote margin in 2010, Pl.’s Ex. 32, at 13, no party argues that this anomalous result is reason to doubt the district’s status as an ability district. Indeed, Texas’s own expert conceded that the district had “performed” from the time of its creation for close to thirty years until the 2010 election, Defs.’ Ex. 581, Trial Tr. 1870:16-1871:4, Sept. 14, 2011, Perez, No. ll-cv-360, and Kel Seliger, chairman of the Texas Senate Select Committee on Redistricting, testified that benchmark CD 27 is clearly protected by the VRA and that he felt the legislature needed to draw another district to compensate for its loss, Seliger San Antonio Dep. 25:22-26:13; see also Trial Tr. 17:19-18:11, Jan. 24, 2012 AM. The enacted plan pivots CD 27 roughly 180 degrees such that the old northern boundary of the district is now the new southern boundary, with new CD 34 filling in much of CD 27’s old geography. The result is that enacted CD 27 is a majority-Anglo district: HCVAP drops to only 41.1%. PL’s Ex. 12, at 9. All parties agree that these significant geographic and demographic shifts mean that CD 27 will no longer perform for minority voters. We agree. 2. Congressional District 23 West Texas’s CD 23 has a complicated history under the VRA. In 2006, the Supreme Court held that CD 23, as then constituted, violated section 2. See LULAC v. Perry, 548 U.S. 399, 425-42, 126 S.Ct. 2594, 165 L.Ed.2d 609 (2006). In response, the U.S. District Court for the Eastern District of Texas redrew its boundaries in 2006 to be an “opportunity district,” or one in which Hispanic voters would have an opportunity to elect their preferred candidates, as required by section 2. See Defs.’ Ex. 575, Trial Tr. 300:13-18, Sept. 7, 2011, Perez, No. ll-cv-360. We now find that the Hispanic voters in CD 23 turned that opportunity into a demonstrated ability to elect, but that the 2010 redistricting took that ability away. Benchmark CD 23 has an HCVAP of 58.4%. Pl.’s Ex. 11, at 9. During the most recent redistricting, the mapdrawers in the Texas legislature acknowledged that CD 23 was a protected district under the VRA. See, e.g., Seliger San Antonio Dep. 13:19-15:11, 30:6-15, 31:6-16 (testimony of Chairman Seliger describing his belief during the redistricting process that CD 23 was a protected Hispanic district); Defs.’ Ex. 978 (email from congressional mapdrawer Doug Davis to National Republican Congressional Committee staffer noting VRA concerns when drawing CD 23). CD 23 elected the minority-preferred candidate in two out of the three endogenous elections since its boundaries were redrawn in 2006. Defs.’ Ex. 327, Dr. Lisa Handley, A Section 5 Voting Rights Analysis of the Proposed Texas Congressional Plan 5 [hereinafter Handley Cong. Rep.]. The one narrow loss was in 2010, a year that Chairman Seliger described as “a bit of an aberration because of things like the Tea Party influence,” further noting that he “didn’t know if [that election] was reliable.” Seliger San Antonio Dep. 15:5-7; see also Trial Tr. 11:15-21, Jan. 24, 2012 AM. Texas counters that none of the experts found that benchmark CD 23 clearly performs as an ability district and points to the weak showing of minority voters in exogenous elections: only three out of ten victories in the OAG 10 and two out of five victories in Dr. Handley’s election set. Alford Rep. 23 tbl.4b; Handley Cong. Rep. 5. But these numbers do not tell the full story. Every expert save Dr. Alford concluded that benchmark CD 23 is an ability district despite marginal exogenous performance. Dr. Handley concluded that endogenous results are more probative than exogenous for this district, see Handley Cong. Rep. 5-6, and, as we have already discussed, we agree that this assessment is generally accurate. Dr. Ansolabehere’s analysis shows that minority-preferred candidates won the district “more often than not.” Defs.’ Ex. 724, Expert Witness Report of Dr. Stephen Ansolabehere 36-37 [hereinafter Ansolabehere Rep.]. And the TLRTF argues that a larger election sample set is necessary to make an informed judgment. When four additional racially contested contests are added to the OAG 10, the district’s exogenous success rises to seven out of fourteen. See Trial Tr. 111:14-113:4, Jan. 18, 2012 AM; Defs.’ Exs. 390, 647. These election results, combined with the endogenous elections discussed above, the fact that CD 23 was drawn to be an opportunity district, and the contemporary views of redistricting officials, are enough for us to find that benchmark CD 23 lived up to its potential as drawn in 2006 and became an ability district. But enacted CD 23 is not. Even though the district’s demographics remain relatively unchanged — HCVAP actually increased 0.1% from the benchmark to the enacted plan, Pl.’s Ex. 12, at 9 — this fact is inconclusive. Instead, we must look to other factors, including exogenous elections, testimony, and other evidence about changes made in the district. Enacted CD 23’s exogenous election results are significantly worse than those in benchmark CD 23. In the OAG 10, the number of victories decreases from three of ten to one. In Dr. Handley’s sample the number decreases from two of five to none. Alford Rep. 23 tbl.4b; Handley Cong. Rep. 7; see also Ansolabehere Rep. 37 (concluding that the enacted plan “lowers the electoral performance of minority-preferred candidates in the District to the point that it is likely no longer a minority opportunity seat”). Minority voter turnout in enacted CD 23 declines. While Hispanic voters accounted for an average of 39% of total votes cast in benchmark CD 23 over the past decade, they made up only 36.5% in enacted CD 23. Defs.’ Ex. 365, at 5-12; see also, e.g., Defs.’ Ex. 575, Trial Tr. 450:19-454:11, Sept. 7, 2011, Perez, No. ll-ev-360 (testimony of Dr. Henry Flores, noting that Hispanic voter turnout was higher in areas moved out of the district than in areas that were moved in; turnout in some excluded areas was consistently over 30%, while turnout in areas that replaced them was only 25-30%). The changes were enough to “nudge” a district that was an ability district, but barely so, to a nonperforming district. See Ansolabehere Rep. 37 (noting that “in a competitive district such as this one,” seemingly small changes “made a huge difference”). Even Texas’s expert testified that CD 23 “is probably less likely to perform than it was, and so I certainly wouldn’t count and don’t [and] haven’t counted the 23rd as an effective minority district in the newly adopted plan.” Defs.’ Ex. 581, Trial Tr. 1839:2-7, Sept. 14, 2011, Perez, No. 11-cv-360. Thus, CD 23 is an ability district in the benchmark, but would be no longer in the enacted plan. Texas claims that the enacted district has remained functionally identical to the benchmark, but these claims are undermined by the mapdrawers’ own admissions that they tried to make the district more Republican — and consequently, less dependable for minority-preferred candidates — without changing the district’s Hispanic population levels. The mapdrawers consciously replaced many of the district’s active Hispanic voters with low-turnout Hispanic voters in an effort to strengthen the voting power of CD 23’s Anglo citizens. In other words, they sought to reduce Hispanic voters’ ability to elect without making it look like anything in CD 23 had changed. See, e.g., Defs.’ Ex. 304 (email from Eric Opiela, counsel to Texas House Speaker Joe Straus, to mapdrawer Gerardo Interiano in November 2010 urging Interiano to find a metric to “help pull the district’s Total Hispanic Population] and Hispanic CVAPs up to majority status, but leave the Spanish Surname [Registered Voter] and [turnout numbers] the lowest,” which would be “especially valuable in shoring up [CD 23 incumbent] Canseco”); id. (email from Interiano responding that he would “gladly help with this”); Defs.’ Ex. 739, at 40 (email indicating that Opiela provided sample maps to Interiano as late as June 11, 2011, that would “improve CD 23’s [H]ispanic performance while maintaining it as a Republican district”). We also received an abundance of evidence that Texas, in fact, followed this course by using various techniques to maintain the semblance of Hispanic voting power in the district while decreasing its effectiveness. See, e.g., Defs.’ Ex. 436 (evidence showing that over 600,000 persons were moved into and out of the district to redress overpopulation of only 149,000); Defs.’ Ex. 903, at 1 (email noting that a draft map of CD 23 was “over 59% HCVAP, but still at 1/10 [exogenous election performance],” and commenting that there must be an HCVAP level high enough that low election results would not raise tro