Covington v. North Carolina

MEMORANDUM OPINION

James A. Wynn, Jr., United States Circuit Judge

Circuit Judge James A. Wynn, Jr., wrote the opinion, in which District Judge Thomas D. Schroeder and District Judge Catherine C. Eagles joined:

More than two decades ago, the Supreme Court considered a legal challenge to election districts that assigned voters to districts primarily on the basis of race. Shaw v. Reno (Shaw I), 609 U.S. 630, 113 S.Ct. 2816, 126 L.Ed.2d 611 (1993). In holding that the plaintiffs had stated a claim under the Equal Protection Clause of the Fourteenth Amendment, the Court explained that racial gerrymandering “reinforces the perception that members of the same racial group ... think alike, share the same political interests, and will prefer the same candidates at the polls.” Id. at 647, 113 S.Ct. 2816. Race-based districting also sends the “pernicious” message to representatives that “their primary obligation is to represent only the members of [a single racial] group.” Id. at 648, 113 S.Ct. 2816. In light of these harms, the Supreme Court later invalidated the redistricting plan. Shaw v. Hunt (Shaw II), 517 U.S. 899, 902, 116 S.Ct. 1894, 136 L.Ed.2d 207 (1996).

Plaintiffs charge that in 2011 the North Carolina General Assembly created State House and Senate redistricting plans through the predominant and unjustified use of race. Defendants contend that race was not the primary factor used in the redistricfc-ing, and that even if it was, their use of race was reasonably necessary to serve a compelling state interest — namely, compliance with Section 2 and Section 5 of the Voting Rights Act (“VRA”).

After careful consideration of the evidence presented, we conclude that race was the predominant factor motivating the drawing of all challenged districts. Moreover, Defendants have not shown that their use of race to draw any of these districts was narrowly tailored to further a compelling state interest. In particular, Defendants have not shown that their use of race was reasonably necessary to remedy a violation of Section 2 of the VRA, since they have not demonstrated that any challenged district was drawn with a strong basis in evidence that the “majority votes sufficiently as a bloc to enable it ... usually to defeat the minority’s preferred candidate.” Thornburg v. Gingles, 478 U.S. 30, 61, 106 S.Ct. 2752, 92 L.Ed.2d 25 (1986). Similarly, Defendants have not provided a strong basis in evidence demonstrating that them use of race was reasonably necessary to comply with Section 5, i.e., to prevent “retrogression in respect to racial minorities’ ‘ability ... to elect their preferred candidates of choice.’ ” Alabama Legislative Black Caucus v. Alabama, — U.S. -, 135 S.Ct. 1257, 1263, 191 L.Ed.2d 314 (2015) (quoting 52 U.S.C. § 10304(b)).

Because Defendants have failed to demonstrate that their predominant use of race was reasonably necessary to further a compelling state interest, the twenty-eight challenged districts in North Carolina’s 2011 State House and Senate redistricting plans constitute racial gerrymanders in violation of the Equal Protection Clause of the United States Constitution. We therefore must order that new maps be drawn.1

This opinion proceeds as follows: Part I outlines the federal and state legal background relevant to redistricting in North Carolina, the 2011 redistricting process, and the litigation stemming from the 2011 redistricting. Part II analyzes statewide and district-specific evidence regarding the use of *125race in the 2011 redistricting, finding that race-based criteria predominated over race-neutral criteria in creating the challenged districts. Part III concludes that the race-based districting does not survive strict scrutiny because Defendants have failed to show a strong basis in evidence that their use of race was reasonably necessary to comply with the VRA. Part IV addresses the proper remedy.

I. Factual and Procedural Background A. Legal Context for Redistricting

Every ten years, the North Carolina General Assembly — comprised of the North Carolina House of Representatives and the North Carolina Senate — must conduct a statewide redistricting based on the latest decennial census. N.C. Const, art. II, §§ 3, 5. Redistricting legislation must comply with a complex array of federal and state legal requirements, all of which combine to make redistricting perhaps “the most difficult task a legislative body ever undertakes.” Page v. Va. State Bd. of Elections, No. 3:13CV678, 2015 WL 3604029, at *7 (E.D.Va. June 5, 2015) (quoting Smith v. Beasley, 946 F.Supp. 1174, 1207 (D.S.C.1996)), appeal dismissed sub nom. Wittman v. Personhuballah, — U.S. -, 136 S.Ct. 1732, 195 L.Ed.2d 37 (2016).

Federal election law requirements include the one person, one vote standard, see Baker v. Carr, 369 U.S. 186, 82 S.Ct. 691, 7 L.Ed.2d 663 (1962), and the provisions of the VRA. Section 2 of the VRA, as relevant to this ease, prohibits redistricting plans that result in vote dilution, which occurs when “based on the totality of circumstances, it is shown that ... members of a [protected group] have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice.” 52 U.S.C. § 10301(b).

Section 5 of the VRA applies only to covered jurisdictions and prohibits retrogression, i.e., the adoption of any electoral change affecting a covered jurisdiction “that has the purpose of or will have the effect of diminishing the ability of any [protected group] to elect their preferred candidates of choice.” Id. § 10304(b). Because the Supreme Court invalidated the criteria used to determine which jurisdictions are covered by Section 5’s requirements, Shelby Cty. v. Holder, — U.S.-, 133 S.Ct. 2612, 2631, 186 L.Ed.2d 651 (2013), no North Carolina jurisdictions remain subject to Section 5. However, during the 2011 redistricting and prior to Shelby County, forty North Carolina counties were subject to those requirements, meaning that any state laws affecting voting in those jurisdictions, including new election districts, had to be precleared as non-retrogressive by the Justice Department or a three-judge panel of the U.S. District Court for the District of Columbia. See 52 U.S.C. §§ 10303-10304; 28 C.F.R. pt. 51 app. (2016).

Finally, any state legislative redistricting plan must also comply with various state legal requirements. The North Carolina Constitution’s “Whole-County Provision” (the WCP) requires that “[n]o county shall be divided in the formation of a senate district,” N.C. Const, art. II, § 3(3), or “representative district,” id. § 5(3). Because this proscription is often impossible to implement without violating federal law, the Supreme Court of North Carolina has reinterpreted the WCP to require that redistricting planners group counties together in drawing districts, generally keeping such groups as small as possible and minimizing the number of traverses across county boundaries within groups. See Stephenson v. Bartlett (Stephenson I), 355 N.C. 354, 562 S.E.2d 377, 396-98 (2002); Stephenson v. Bartlett (Stephenson II), 357 N.C. 301, 582 S.E.2d 247, 250-51 (2003); see also infra section II.A1.b.

B. Factual Context for Redistricting

Between 1991 and 2010, the number of “majority-black districts” — i.e., districts with a black voting-age population (“BVAP”) above fifty-percent — in North Carolina’s state House and Senate districting plans gradually declined.2 In the House, for instance, between 1991 and 2010, the number *126of majority-black districts decreased from a high of thirteen in 1991 to a low of nine starting in 2002. Defs.’ Ex. 3001 at 19-24, 42-48, 83-87, 143-50, 227-34, 298-300, 315-17 (Historical House Maps 1991-2010). Similarly, in the Senate, the number of majority-black districts decreased from four in 1991 to zero beginning in 2003. Defs.’ Ex. 3000 at 18-20, 30-32, 46-47, 73-76, 116-19, 151 (Historical Senate Maps 1991-2010). The redistricting plan that was in place in the House at the time of the 2011 redistricting (the “Benchmark House Plan”) was enacted in 2009, and as drawn it had nine majority-black districts.3 Defs.’ Ex. 3001 at 315-17. The “Benchmark Senate Plan,” enacted in 2003, had zero majority-black districts. Defs.’ Ex. 3000 at 151.

Many African-American General Assembly candidates, however, had electoral success even when running in non-majority-black districts. African-American candidates certainly experienced losses in such districts, but their overall success was significant. For example, in the three election cycles preceding the 2011 redistricting, African-American candidates for the North Carolina House won thirty-nine general elections in districts without a majority BVAP (including eleven such elections in 2010 alone), and African-American candidates for the North Carolina Senate won twenty-four such elections (including seven such elections in 2010). Defs.’ Ex. 3020-14 at 2-5 (North Carolina House of Representatives Election Contest Winners); Defs.’ Ex. 3001 at 315-17; Defs.’ Ex. 3020-13 at 2-3 (North Carolina Senate General Election Winners); Defs.’ Ex. 3000 at 151.

C. The 2011 Redistricting Process

The redistricting process at issue here began and ended within the 2011 calendar year. On January 27, 2011, Senator Robert Rucho was appointed Chair of the Senate Redistricting Committee. Second Joint Stip. ¶ 1, ECF No. 80; Defs.’ Ex. 3013 at 2 (Aff. of Robert Rucho in Dickson v. Rucho). On February 15, 2011, Representative David Lewis was appointed Chair of the parallel House Redistricting Committee.4 Second Joint Stip. ¶ 1; Defs.’ Ex. 3037 at 2 (Aff. of David Lewis in Dickson v. Rucho). Together, Senator Ru-cho and Representative Lewis (the “Redistricting Chairs” or “Chairs”) were “effectively ... the manager[s] of the [redistricting] process,” Trial Tr. vol. Ill, 119:1-4 (Lewis), and they “worked very closely” with each other throughout that effort, Defs.’ Ex. 3013 at 2 (Rucho).

Although the Redistricting Chairs led the redistricting, they did not actually draw the maps. That work was done by Dr. Thomas Hofeller, whom the General Assembly’s private counsel engaged to design the 2011 redistricting plans. Second Joint Stip. ¶ 3. Dr. Hofeller was to be the “chief architect” of the plans. Trial Tr. vol. IV, 41:22-24 (Rucho); Second Joint Stip. ¶ 6; Joint Ex. 1051 at 71. In other words, the Chairs would rely on him to translate their policy directives into actual districts. Trial Tr. vol. IV, 17:21-18:3 (Rucho); see also id. at 41:11-16 (“Dr. Hofeller was given clear instructions as to what was required of him ... just as much as I would do if I were asking an architect to build my home.”).

In March 2011, soon after receiving the 2010 census data, Dr. Hofeller began his work. Second Joint Stip. ¶ 5. The Redistricting Chairs were the only ones who gave him instructions, see Trial Tr. vol. IV, 216:2-9 (Hofeller); Second Joint Stip. ¶ 7, and they only communicated with Dr. Hofeller orally, Trial Tr. vol. Ill, 199:5-8 (Lewis); Trial Tr. vol. IV, 216:2-9 (Hofeller).

It appears that no one besides the two Chairs and Dr. Hofeller had any substantive role in designing the 2011 districts. Dr. Ho-feller never attended a Redistricting Committee meeting or reviewed any Redistricting Committee meeting transcripts. Trial Tr. vol. V, 89:5-10 (Hofeller). The Redistricting Com*127mittees did not participate in defining redis-trieting criteria for Dr. Hofeller, nor were Dr. Hofeller’s draft maps presented to the Redistrieting Committees for their input pri- or to public release. Trial Tr. vol. Ill, 213:17-23 (Lewis); Defs.’ Ex. 3013-1 at 1-3 (Timeline of 2011 Redistrieting Process); Joint Ex. 1022 at 37 (June 15, 2011, Joint Redistrieting Committee Meeting) (Rep. Joe Hackney). Dr. Hofeller did not attend any of the public hearings on redistrieting, review any transcripts of those hearings, or confer with anyone other than Representative Lewis and Senator Rucho about the redistrieting. Trial Tr. vol. V, 88:23-89:4, 89:11-16 (Hofeller).

The Redistrieting Chairs instructed Dr. Hofeller to begin the line-drawing process by identifying geographically compact minority populations and then drawing majority-minority districts in those locations, where possible, so that African-American voters would have a roughly proportional opportunity statewide to elect their preferred candidates of choice. Trial Tr. vol. V, 96:7-97:15 (Hofeller). They termed these majority-minority districts “VRA districts,” which they considered to be districts with geographically compact, politically cohesive minority populatins, where there was some evidence of racially polarized voting. Trial Tr. vol. III, 222:23-223:24 (Lewis); see also Trial Tr. vol, IV, 49:1-17 (Rucho). The Chairs instructed Dr. Hofeller that each of these districts was to have at least 50%-plus-one BVAP. See infra section II.A1.5

On June 17, 2011, as the first step in making Dr. Hofeller’s plans public, the Chairs released a map for both the House and Senate (the “VRA maps”)6 that included only the purported “VRA districts” they claimed were necessary for compliance with the VRA Second Joint Stip. ¶ 8; Defs.’ Ex. 3013-1 at 2, 3. On July 12, the Chairs proposed full House and Senate redistrieting plans to the public. Second Joint Stip. ¶ 11; Defs.’ Ex. 3013-1 at 2, 3.

On July 20, a slightly modified version of that full Senate plan (“Rucho Senate 2”) was released to the public and, the following day, presented to the Senate Redistrieting Committee. Second Joint Stip. ¶ 12; Defs.’ Ex. 3013-1 at 3. On July 25 and July 27, the Senate and House, respectively, passed that modified plan. Second Joint Stip. ¶¶ 15-16; Defs.’ Ex. 3013-1 at 4-5.

A similarly streamlined sequence of proposal and passage unfolded in the House. After a full House plan was released to the public on July 12, a slightly modified version of that plan (“Lewis-Dollar-Dockham 2”) was released to the public on July 20 and presented to the House Redistrieting Committee the following day. Second Joint Stip. ¶ 17; Defs.’ Ex. 3013-1 at 3. The full House passed that plan on July 25. Id. After a few further modest revisions made by the Senate Redistricting Committee, the full Senate approved the House plan (later termed “Lewis-Dollar-Dockham 4”) on July 27. Second Joint Stip. ¶¶ 20-21; Defs.’ Ex. 3013-1 at 3.

Rucho Senate 2 (the “Enacted Senate Plan”) became law on July 27, 2011. Second Joint Stip. ¶ 16; Defs.’ Ex. 3013-1 at 4; see 2011 N.C. Sess. Law 402, Lewis-Dollar-Dock-ham 4 (the “Enacted House Plan”) became law on July 28, 2011.7 Second Joint Stip. ¶ 22; Defs.’ Ex. 3013-1 at 2; see 2011 N.C. Sess. Laws 404. Those plans were subsequently precleared by the Justice Department pursuant to Section 5 of the VRA

In short, within a month-and-a-half, Dr. Hofeller’s draft maps were released in near-final form to the public, presented to the Redistrieting Committees, and passed without significant modification by the General *128Assembly. And because those maps were the work of Dr. Hofeller, who was in turn directed only by the two Redistricting Chairs, it is clear that three individuals substantially carried out North Carolina’s 2011 statewide redistricting effort. See Trial Tr. vol. Ill, 213:24-214:2 (Lewis) (“[Q:] [W]ith only a few minor changes, those districts, your districts and Dr. Hofeller’s districts, became the law of North Carolina, didn’t they? [A:] Yes, sir.”).

D. Litigation Challenging the 2011 Enacted Plans

In November 2011, two sets of plaintiffs collectively challenged in North Carolina state court twenty-seven state House and Senate districts, as well as three Congressional districts, alleging that they were unconstitutional racial gerrymanders. See Mem. in Supp. of Mot. to Stay, Defer, or Abstain Ex. 1, at 7, 15, N.C. Superior Ct. Op. in Dickson v. Rucho, ECF No. 32. A three-judge panel was appointed, the two cases were consolidated, and a two-day bench trial was held in June 2013. Id. at 7-8. In July 2013, the court issued a decision upholding the challenged districts. Id. at 48-49.

The Supreme Court of North Carolina affirmed the trial court’s judgment. Dickson v. Rucho, 367 N.C. 542, 766 S.E.2d 238 (2014). The U.S. Supreme Court then granted cer-tiorari, vacated, and remanded the case for further consideration in light of Alabama, — U.S.-, 135 S.Ct. 1257, 191 L.Ed.2d 314. Dickson v. Rucho, — U.S. -, 135 S.Ct. 1843, 191 L.Ed.2d 719 (2015) (Mem). On December 18, 2015, the North Carolina Supreme Court reaffirmed the trial court’s judgment. Dickson v. Rucho, 368 N.C. 481, 781 S.E.2d 404, 410-11 (2015).

Meanwhile, in October 2013, two plaintiffs uninvolved in the Dickson litigation brought suit in federal district court, alleging that two Congressional districts drawn during the 2011 redistricting were racial gerrymanders. Harris v. McCrory, 159 F.Supp.3d 600, 609-10 (M.D.N.C. Feb. 5, 2016), prob. juris, noted, — U.S. -, 136 S.Ct. 2512, — L.Ed.2d-(2016). After the appointment of a three-judge panel, a three-day bench trial was held in October 2015. Id. That court found, on February 5 of this year, that both challenged Congressional districts were unconstitutional racial gerrymanders, and it ordered the General Assembly to draw remedial districts. Id. at 604-06, 627-28.

Plaintiffs, who are thirty-one U.S. citizens registered to vote in North Carolina,8 brought this action on May 19, 2015, against the State of North Carolina, the Redistricting Chairs, the North Carolina Board of Elections, and other state officials. Compl., ECF No. 1; First Am. Compl. ¶¶ 10-49, ECF No. 11.

Plaintiffs allege that North Carolina Senate Districts 4, 5, 14, 20, 21, 28, 32, 38, and 40 and North Carolina House of Representatives Districts 5, 7, 12, 21, 24, 29, 31, 32, 33, 38, 42, 43, 48, 57, 58, 60, 99, 102, and 107 (together the “challenged districts”) are racial gerrymanders in violation of the Equal Protection Clause. First Am. Compl. ¶ 1. Each of the challenged districts was included as a purported “VRA district” in the House and Senate VRA maps released on June 17, 2011. Plaintiffs seek a judgment declaring unconstitutional the challenged districts and a permanent injunction blocking their use.9 First Am. Compl. at 92.

Because the Plaintiffs’ action “challeng[es] the constitutionality of ... the apportionment of a[] statewide legislative body,” 28 *129U.S.C. § 2284(a), the Chief Judge of the U.S. Court of Appeals for the Fourth Circuit granted Plaintiffs’ request for a three-judge panel on August 18, 2015. Order, EOF No. 18. After an expedited discovery period, this court held a five-day bench trial from April 11 to April 15, 2016. The issues are now ready for our consideration.

II. Analysis of Evidence of Racial Predominance

At the outset of our analysis, it is important to emphasize that a finding that race was the predominant motive in drawing a district does not automatically render that district unconstitutional. Nor does it signify that the legislature acted in bad faith or with discriminatory intent in its redistricting. Indeed, redistricting legislatures -will almost always be aware of racial demographics, but “[t]hat sort of race consciousness does not lead inevitably to impermissible race discrimination.” Shaw I, 509 U.S. at 646, 113 S.Ct. 2816.

As it must in do when undertaking any official action, a state must draw electoral districts in accordance with equal protection principles. Miller v. Johnson, 515 U.S. 900, 905, 115 S.Ct. 2475, 132 L.Ed.2d 762 (1995). Consequently, if Plaintiffs show that race predominated over traditional race-neutral redistricting principles, we apply strict scrutiny, and Defendants have the burden of “show[ing] not only that [their] redistricting plan was in pursuit of a compelling state interest, but also that ‘[their] districting legislation is narrowly tailored to achieve [that] compelling interest.’ ” Shaw II, 517 U.S. at 908, 116 S.Ct. 1894 (quoting Miller, 515 U.S. at 920, 115 S.Ct. 2475).

In proving whether race predominated in a racial gerrymandering case, a plaintiffs burden is a “demanding one,” Easley v. Cromartie (Cromartie II), 532 U.S. 234, 241, 121 S.Ct. 1452, 149 L.Ed.2d 430 (2001) (quoting Miller, 515 U.S. at 928, 115 S.Ct. 2475 (O’Connor, J., concurring)), because “the underlying districting decision is one that ordinarily falls within a legislature’s sphere of competence,” id. at 242, 121 S.Ct. 1452. Specifically, a plaintiff must “show, either through circumstantial evidence of a district’s shape and demographics or more direct evidence going to legislative purpose, that race was the predominant factor motivating the legislature’s decision to place a significant number of voters within or without a particular district.” Alabama, 135 S.Ct. at 1267 (quoting Miller, 515 U.S. at 916, 115 S.Ct. 2475). In general, that requires proof that “the legislature subordinated traditional race-neutral districting principles, including ... compactness, contiguity, and respect for political subdivisions ... to racial considerations.” Miller, 515 U.S. at 916, 115 S.Ct. 2475. Put differently, the predominance of racial considerations is evident where “[r]ace was the criterion that, in the State’s view, could not be compromised,” such that traditional districting principles were applied “only after the race-based decision had been made.” Shaw II, 517 U.S. at 907, 116 S.Ct. 1894.

In evaluating whether racial considerations predominated in a districting decision, the Supreme Court has considered both direct and circumstantial evidence of legislative intent, including statements by legislators identifying race as a chief districting criterion, see Miller, 515 U.S. at 917-18, 115 S.Ct. 2475; indications that attaining a racial percentage within a given district was nonnegotiable, see Shaw II, 517 U.S. at 906-07, 116 S.Ct. 1894; bizarre or non-compact district shape, see Shaw I, 509 U.S. at 646-48, 113 S.Ct. 2816; and district lines that cut through traditional geographic boundaries or local election precincts, see Bush v. Vera, 517 U.S. 952, 974, 116 S.Ct. 1941, 135 L.Ed.2d 248 (1996) (plurality opinion). Moreover, in light of Alabama, we are mindful that a legislature’s “policy of prioritizing mechanical racial targets above all other districting criteria (save one-person, one-vote)” provides particularly strong evidence of racial predominance. 135 S.Ct. at 1267.

As elaborated below, we find evidence of all of the above here.

A. Analysis of Statewide Evidence of Racial Predominance

Although racial gerrymandering claims are properly brought district by district, evidence that applies statewide — espe*130cially direct evidence revealing legislative intent — will frequently also be material. See Alabama, 136 S.Ct. at 1266. Therefore, before proceeding to an analysis of the particular geographic and demographic features of individual districts, we begin by considering evidence relevant to all challenged districts.

1. The Chairs’ Redistrieting Criteria and Instructions to Dr. Hofeller

There is overwhelming and consistent evidence before us as to the intentions of the Enacted Plans’ authors. The primary criteria that guided North Carolina’s 2011 redistricting were articulated repeatedly and with little variation by the Chairs throughout the redistrieting process and have been affirmed under oath by the Chairs and Dr. Hofeller on numerous occasions since. These countless statements show without real dispute that there were three main instructions the Chairs gave to Dr. Hofeller about redistricting. All three instructions centered around the creation of what the Chairs called “VRA districts,” which, as noted above, see supra section I.C, the Chairs considered to be districts with geographically compact, politically cohesive minority populations where there was some evidence of racially polarized voting. The Chairs instructed Dr. Hofeller (1) to draw each of these VRA districts with at least 50%-plus-one BVAP; (2) to draw these districts first, before drawing the lines of other districts; and (3) to draw these districts everywhere there was a minority population large enough to do so and, if possible, in rough proportion to their population in the state.

a. VRA Districts at 50%-Plus-One BVAP

First, the Redistrieting Chairs instructed Dr. Hofeller to draw all purported VRA districts to reach a 50%-plus-one BVAP threshold. Trial Tr. vol. V, 97:9-11 (Hofeller) (“[Q:] [Y]ou were ... instructed to draw each VRA district at 50 percent plus one or greater; isn’t that correct? [A:] If possible, yes.”); Trial Tr. vol. Ill, 201:19-21 (Lewis) (“[Q:] You told [Dr. Hofeller] to draw the [VRA] districts at more than 50 percent [BVAP]; right, Representative Lewis? [A:] I believe that would be correct, yes, sir.”); Joint Ex. 1018 at 28 (July 21, 2011, House Redistricting Committee Meeting Transcript) (Lewis) (“[W]e felt that it was a prudent course of action to draw the districts that we were going to call and think of as VRA districts at 50 percent plus one.”); Joint Ex. 1013 at 10 (July 21, 2011, Senate Redistrieting Committee Meeting Transcript) (Rucho) (“[I]n constructing the VRA districts, the Chairs recommended, where possible, these districts be drawn at a level equal to 50 percent plus one BVAP.”); Joint Ex. 1015 at 94-95 (July 25, 2011, Senate Floor Session) (Rucho) (“In the absence of any ... legal or factual basis for not doing so, we have decided to draw all of our Voting Rights Act districts at a 50 percent-plus level.”).

The Chairs’ 50%-plus-one instruction was based on their interpretation of the plurality opinion in Bartlett v. Strickland, 556 U.S. 1, 129 S.Ct. 1231,173 L.Ed.2d 173 (2009), which they took to mean “that if you were going to draw districts to comply with the Voting Rights Act ... the districts had to exceed 50 percent in minority population.” Trial Tr. vol. Ill, 120:20-23 (Lewis).10

It is clear that the name of the Strickland case operated as a shorthand for the Chairs’ 50%-plus-one instruction whenever it was employed. In other words, throughout the redistrieting process, “complying with Strickland” meant drawing all purported VRA districts at 50%-plus-one BVAP. The Chairs made that clear in them very first public statement, which accompanied the June 17 release of the House and Senate VRA maps. See Joint Ex. 1005 at 2 (“Under the Strickland decisions, districts created to comply with section 2 of the Voting Rights Act, must be created with a [BVAP] at the level of at least 50% plus one. Thus, in constructing VRA majority black districts, the Chairs recommend that, where possible, these districts be drawn at a level equal to at least 50% plus one ‘BVAP.’ ”).

*131The 50%-plus-one requirement, always tied to Strickland, was then repeated in every formal public statement issued by the Chairs, and in nearly every explanatory comment made to other legislators throughout the redistricting process. See Joint Ex. 1006 at 7 (Joint Statement by Redistricting Chairs Pri- or to June 23, 2011, Public Hearing) (“[VRA] districts must comply with Strickland ... and [therefore] be drawn at a level that constitutes a true majority of black voting age population.”); Joint Ex. 1007 at 4-5 (July 12, 2011, Joint Statement by Redistricting Chairs) (“[Strickland] require[s] that [VRA] districts ... be drawn with a [BVAP] in excess of 50% plus one.... [I]n light of [Strickland], we see no principled legal reason not to draw all VRA districts at the 50% or above level when it is possible to do so.”); Joint Ex. 1015 at 89 (July 25, 2011, Senate Floor Session) (Rucho) (“Strickland ... said that any district drawn to comply with or avoid liability under Section 2 of the Voting Rights Act must be drawn at ... 50 percent or more of black voting age population.”); Joint Ex. 1020 at 52 (July 25, 2011, House Floor Session) (Lewis) (“[0]ur proposed plan complies with Section 2 of the Voting Rights Act under the decision by the United States Supreme Court in Strickland .... The state is now obligated to draw majority black districts with true majority black voting age population.”); see also Trial Tr. vol. Ill, 195:15-18 (Lewis) (“[W]e felt ... that the Strickland requirement of majority-minority districts ... meant that we should draw the VRA districts at over 50 percent ....”); id. at 195:25-196:2 (“It was my understanding of the Strickland decision that drawing the districts at 50 percent plus one was the threshold for creating a VRA district.”).

It is clear, then, that the 50%-plus-one BVAP target was of paramount concern for the Chairs as they drew purported VRA districts, including the challenged districts,

b. VRA Districts First

Another main instruction the Chairs gave Dr. Hofeller regarding the purported VRA districts was to draw those districts first, before any other “non-VRA” districts were drawn or any other redistricting criteria (besides the 50%-plus-one requirement) were considered. Trial Tr. vol. Ill, 207:12-14 (Lewis) (“[Q:] And one instruction was to draw the VRA districts first? You told Dr. Hofeller to draw th[ose] districts first? [A:] Yes, sir.”); Joint Ex. 1005 at 1 (‘VRA districts [must] be created before other legislative districts.”). In fact, as described above, the VRA districts were not only drawn first, but also released first — nearly one month before the release of the full redistricting maps. Second Joint Stip. ¶¶ 8, 11; Defs.’ Ex. 3013-1 at 2,3.

The Chairs’ instruction to draw VRA districts first was grounded in a pair of opinions issued by the Supreme Court of North Carolina, See Stephenson I, 355 N.C. 354, 562 S.E.2d 377; Stephenson II, 357 N.C. 301, 582 S.E.2d 247. As mentioned above, see supra section I.A, both Stephenson I and Stephenson II sought to harmonize the state constitution’s WCP with federal election law, including the one person, one vote requirement and the VRA. Recognizing the supremacy of federal legal requirements, the Stephenson decisions set forth an enumerated, hierarchical list of steps to guide the enactment of “any constitutionally valid redistricting plan.” Stephenson II, 582 S.E.2d at 250.

As explained in Stephenson II, step 1 of any North Carolina redistricting process is that “legislative districts required by the VRA shall be formed prior to creation of non-VRA districts,” and that “to the maximum extent practicable, such VRA districts shall also comply with the legal requirements of the WCP.” Id. Compliance with one person, one vote is step 2. Id. Later steps require the formation of single-district, one-county groups (step 3); the formation of mul-ti-district, single-county groups (step 4); the formation of multi-county groupings, with a preference for fewer counties per group (steps 5 and 6); and the consideration of communities of interest (step 7). Id.11

*132As discussed below, it is clear that as the map-drawing process unfolded, the Chairs and Dr. Hofeller did attempt to comply with the WCP as defined by the Stephenson cases. See infra section II.A.2. However, it is equally clear that, in accordance with those eases, the Chairs and Dr. Hofeller made drawing VRA districts — as they understood them — then first priority. See Trial Tr. vol. IV, 7:17-26 (Rucho) (“Stephenson required the General Assembly, as we were drawing these maps, to take on the Voting Rights Act issue first because of federal dominance over the state law. The second part was that we also needed to harmonize what is the [WCP].” (emphasis added)); id. at 48:22-49:4 C‘[Q:] You say you complied with Stephenson; that was your goal? [A:] Yes, sir. We followed the Stephenson decision to its letter. [Q:] And under that analysis, what was the first thing that you believed you had to do? [A:] Well, the first step, as required, would have been for us to identify potential! ] VRA districts.” (emphasis added)); Trial Tr. vol. IV, 219:2-9' (Hofeller) (“The Stephenson case instructed, according to my understanding of it, that the first thing that had to happen was an analysis of the areas of concentrations of minority voters in the state to determine where VRA districts could be drawn and then to proceed to draw districts.” (emphasis added)); Joint Ex. 1024 at 22 (House 2011 Section 5 Submission) (“[The Stephenson cases hold] that districts ‘required by the Voting Rights Act’ must be created before any other districts.”); Joint Ex. 1023 at 20 (Senate 2011 Section 5 Submission) (same).

Further, because the Chairs and Dr. Ho-feller believed that Strickland required all VRA districts to be drawn at 50%-plus-one BVAP, they applied that purported requirement at Stephenson’s first step. In other words, complying with Stephenson to the Chairs meant drawing 50%-plus-one districts, and drawing them first. See Trial Tr. vol. IV, 46:20-47:1 (Rucho) (“[Q:] And is it your understanding that Strickland is encompassed in the way that you understand the Stephenson ease? [A:] Yes, sir. The Stephenson case ... included the decision on Strickland requiring the 50 percent plus one, as far as saying that if you are building a VRA district, then you can go ahead and follow that as part of the law.”); id. at 32:25-33:3 (“[Q:] [Y]ou applied the 50 percent plus one rule across the state, didn’t you? [A:] That was what was expected of us as we followed the Stephenson criteria.”).

The Chairs’ combined understanding of the Strickland and Stephenson cases thus operate ed to make the 50%-plus-one BVAP threshold Dr. Hofeller’s first consideration — both in time and priority — in drawing all VRA districts and therefore all challenged districts.12

c. Near-Maximization of VRA Districts

Finally, in addition to instructing Dr. Ho-feller to draw all purported VRA districts first and to draw them at 50%-plus-one BVAP, the Chairs instructed him to draw enough VRA districts “to provide North Carolina’s African American citizens with a substantially proportional and equal opportunity to elect their preferred candidates of choice.” Joint Ex. 1005 at 3. According to the Chairs, this would mean “the creation of 24 majority African American House districts and 10 majority African American Senate districts.” Id. Like the invocation of Strickland and Stephenson, this proportionality criterion was stressed in each of the Chairs’ public statements, and it was repeated throughout the redistricting process. Id.; Joint Ex. 1006 at 7 (explaining that any proposed maps must “provide black voters with a substantially proportional state-wide opportunity to elect candidates of them choice”); Joint Ex. 1007 at 5 (“Our proposed plan provides black voters in North Carolina with substantial or rough proportionality in the number of VRA districts in which they have an equal opportunity to elect their preferred candidates of choice.”); see also Joint Ex. 1018 at 12-14 (July 21, 2011, House Redistricting Committee Meeting); Joint Ex. 1021 at 21 (July 27, 2011, Senate Redistricting Committee Meeting).

*133As with their 50%-plus-one BVAP target, the Chairs sought to ground their proportionality goal in ease law. Citing Johnson v. De Grandy, 512 U.S. 997, 114 S.Ct. 2647, 129 L.Ed.2d 775 (1994), they asserted that achieving proportionality would “further[ ] the State’s obligation to comply with Section 2 of the Voting Rights Act,” Joint Ex. 1005 at 4, in that it would “give the State an important defense to any lawsuit that might be filed challenging the plans under Section 2,” Joint Ex. 1007 at 5. Representative Lewis stated his belief that proportionality would likely “insulate [the state] from lawsuits,” Trial Tr. vol. Ill, 196:7-11, and Senator Ru-cho indicated at least once his understanding that proportionality was “required,” Joint Ex. 1021 at 21.

This was not a proper interpretation of the law. De Grandy considered rough proportionality — i.e., whether “minority voters form[ed] effective voting majorities in a number of districts roughly proportional to the minority voters’ respective shares in the voting-age population” — as one “relevant fact in the totality of circumstances” bearing on a Section 2 vote dilution claim. 512 U.S. at 1000, 114 5.Ct. 2647. That same case also clarified that under no circumstances is proportionality to be considered a “safe harbor” from Section 2 litigation, id. at 1017-21, 114 S.Ct. 2647, and that proportionality should not be sought if it requires destroying “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,” id. at 1020,114 S.Ct. 2647. In other words, proportionality is not required, not a safe harbor, and not to be pursued at the cost of fracturing effective coalitional districts.13

Though grounded in legal misconceptions, the proportionality goal was nevertheless applied to effect a major increase in the number of majority-black districts across the state. The Benchmark House Plan as drawn had only nine majority-black districts, Defs.’ Ex. 3001 at 315-17, meaning that the Chairs’ proportionality goal would require creating fifteen new majority-black districts to reach their twenty-four-district target. The Benchmark Senate Plan as drawn had no majority-black districts, Defs.’ Ex. 3000 at 151, meaning that the Chairs’ proportionality goal would require creating ten new majority-black districts to reach their ten-district target.14 Overall, the Chairs sought to more than triple the number of majority-black General Assembly districts, from nine to thirty-four.

The Enacted Plans fell just short of the stated goal.15 In the House, while the VRA *134map released on June 17 had included twenty-four majority-black districts, Joint Ex. 1005 at 3, the Chairs “elected to delete a majority black district [they] had proposed for southeastern North Carolina based upon ... strong statements opposing such a district,” Joint Ex. 1007 at 2. The Chairs also noted that, due to “the lack of black population,” they could not draw “two majority BVAP [House] districts .., in Forsyth County.” Joint Ex. 1007 at 5.

As for the Senate, the Chairs explained that they had “been unable to identify a reasonably compact majority African American population to create a tenth majority African American [Senate] district.” Joint Ex. 1005 at 4; see also Joint Ex. 1006 at 2; Joint Ex. 1007 at 4 (“[0]ur tenth [majority-black] senate district, District 32, cannot be drawn within Forsyth County in excess of 50% plus one.”).

Either way, the end result of the proportionality goal was a striking increase in the number of majority-black General Assembly districts. The Enacted House Plan contains twenty-three majority-black districts, and the Enacted Senate Plan contains nine, meaning that the total number of majority-black General Assembly districts increased from nine to thirty-two. Joint Ex. 1023 at 10; Joint Ex. 1024 at 10.

The fact that the goal sought to increase, significantly, the number of majority-black General Assembly districts suggests that here the proportionality target functionally operated as a goal to maximize the number of majority-black districts. In response to a question regarding the proportionality instruction, Senator Rucho responded that his instruction was for Dr. Hofeller to draw VRA districts “wherever he could ... but with no requirement of maximization.” Trial Tr. vol. IV, 36:4-5. It is difficult to see, though, how the instruction to draw VRA districts “wherever” one can or “where possible,” Trial Tr. vol. V, 81:17-20 (Hofeller); Joint Ex. 1005 at 2-3, is meaningfully different from a goal to maximize such districts.

Even if the proportionality goal was not quite a maximization policy, there is no doubt that this statewide numerical target was based on race, and that it was of principal importance during the 2011 redistricting process. Because the proportionality goal was considered to be a component of VRA compliance, that priority — along with the 50%-plus-one requirement — was folded into the first Stephenson step, and thereby the first step of the redistricting. See Trial Tr. vol. IV, 36:6-10 (Rucho) (“[Q:] And the [Senate] plan that was enacted included ten districts that you believed met the State’s obligations under the Voting Rights Act; correct? [A:] In our understanding of what was expected of us following the Stephenson decision and the criteria, yes.”).

That meant that two numerical racial targets — the 50%-plus-one goal and the proportionality goal — took precedence in the redistricting process.

d. Three “Primary” Criteria

It is not just that the 50%-plus-one instruction, the proportionality goal, and compliance with Stephenson (including its requirement to draw VRA districts first) were the criteria most frequently or prominently cited by the Chairs. Those three considerations were also often identified by the Chairs themselves and by Dr. Hofeller as the only “primary” criteria.

For example, in his affidavit, Dr. Hofeller identified Strickland (i.e., the 50%-plus-one requirement), Stephenson (including the instruction to draw VRA districts first), and the proportionality goal as the three “primary criteria used to draw [the redistricting] plans.” Defs.’ Ex. 3026 at 4. And when the Chairs first proposed a complete plan, they announced in their accompanying public statement that their “primary goal [was] to propose maps that will survive any possible legal challenge,” and then identified Stephenson, Strickland (meaning the 50%-plus-one requirement), and compliance with the VRA (meaning proportionality, in addition to the *13550%-plus-one requirement) as the “legal requirement[s]” that would ward off such litigation. Joint Ex. 1007 at 1-2.

The Chairs also indicated these were the criteria that “could not be compromised.” Shaw II, 517 U.S. at 907,116 S.Ct. 1894. For instance, after announcing their proposed VRA districts, the Chairs stated that any alternative proposals should comply with three criteria: the Stephenson cases, Strickland, and the proportionality goal. Joint Ex. 1005 at 8. Six days later, the Chairs again stressed in a public statement that they would “entertain any specific suggestions” for alternative districts, but only if those suggested alternatives satisfied the proportionality and 50%-plus-one targets. Joint Ex. 1006 at 7. Even before proposing any plans, they asked the Legislative Black Caucus to “take into consideration the requirements of Strickland ... as well as the Stephenson line of cases” in suggesting possible districts. Defs.’ Ex. 3013-5 at 2 (Apr. 5, 2011, Email and Letter from Redistricting Chairs to Leaders of the Legislative Black Caucus).

By their own characterization, then, the 50%-plus-one target and the proportionality goal were two of the three “primary” criteria the Chairs and Dr. Hofeller employed. And the third — compliance with the Stephenson eases — assured that those two “mechanical racial targets,” which the Chairs took to be necessary for compliance with the VRA, would be “prioritiz[ed] ... above all other districting criteria (save one-person, one-vote).” Alabama, 135 S.Ct. at 1267.16

In other words, the overriding priority of the redistricting plan was to draw a predetermined race-based number of districts, each defined by race.

2. Dr. Hofeller’s Implementation of the Chairs’ Instructions

Given clear instructions, Dr. Hofeller closely followed them.

One of Dr. Hofeller’s first tasks, conducted in March 2011 soon after receipt of the 2010 census data, was the creation of a spreadsheet calculating the exact number of majority-black districts in the House and Senate that would achieve the Chairs’ proportionality goal. Pis,’ Ex. 2037 (Carolina Proportionality Chart); Second Joint Stip. ¶ 4; Trial Tr. vol. V, 89:17-91:16 (Hofeller).

Next, for both the House and Senate, Dr. Hofeller conducted “a demographic analysis ... to determine where in the State sufficiently populous, compact minority populations were present to form single-member African-American ... districts containing minority population percentages in excess of 50%.”17 Defs.’ Ex. 3030 at 5 (Second Expert Report of Thomas B. Hofeller, Ph.D.). Based on this analysis, he drew VRA “exemplar districts,” which were “racially defined” in that they embodied nothing more than “concentrations of minority voters” capable of constituting a district that could satisfy the 50%-plus-one BVAP threshold. Trial Tr. vol, IV, 228:5-12 (Hofeller); Trial Tr. vol. V, 104:4-105:1 (Hofeller). Dr. Hofeller drew the VRA exemplar districts without reference to any communities of interest or geographic subdivisions, such as county lines and precinct lines. Trial Tr. vol. V, 104:21-105:6 (Ho-feller).18

*136After drawing exemplar 50%-plus-one BVAP districts across the state, Dr. Hofeller then drew a separate “optimum [county] grouping map” in accordance with the criteria outlined by the Stephenson cases. Trial Tr. vol. IV, 240:8-11 (Hofeller). Because the boundary lines of the optimum county grouping map often crossed the lines of the VRA exemplar districts, Dr. Hofeller then engaged in what he termed “an iterative harmonization project” in order to create county groups that could accommodate 50%-plus-one VRA districts, and vice versa. Id. at 240:11-241:1; see also id. at 237:10-14; Trial Tr. vol. V, 28:9-12 (Hofeller) (“[T]he optimal county groups ..., because of the Voting Rights Act provision in Stephenson, had to be modified in order to create the districts that we felt needed to be created.”).

Ultimately, the optimum county groups were substantially modified, such that in many of those areas of the state where purported VRA districts were drawn, the optimum county groups were not enacted. See Defs.’ Ex. 3030 at 100-01 (Maps 9 and 10) (optimum groups excluded from Enacted Plans in white); Trial Tr. vol. IV, 236:10-25 (Hofeller); Trial Tr. vol. V, 28:7-17 (Hofeller).

In contrast, “[a]ll of the 2011 enacted VRA districts for the General Assembly are substantially based on the compact minority populations found in the corresponding exemplar districts.” Defs.’ Ex. 3029 at 9 (First Decl. of Thomas B. Hofeller, Ph.D.). Indeed, the vast majority of the African-American population included in Dr. Hofeller’s VRA exemplar districts was eventually included in enacted VRA districts. Trial Tr. vol. IV, 238.T8-21 (Hofeller) (“[T]he plans created and enacted have to contain a very high percentage of those minority areas from one or more of those [exemplar] districts in the district which is enacted.”). On average, 90.25% of the total BVAP in the House VRA exemplar districts and 83.64% of the total BVAP in the Senate VRA exemplar districts were incorporated into an enacted VRA district. Defs.’ Ex. 3029 at 28.

Most significantly, although the boundaries of some VRA exemplar districts did shift as Dr. Hofeller sought to “harmonize” them with the county groupings, the Chairs’ 50%-plus-one BVAP target was not compromised.19 Indeed, Dr. Hofeller did whatever it took to meet that racial threshold, even where doing so required major sacrifices in terms of respect for other traditional district-ing principles. Trial Tr. vol. V, 20:12-19 (Ho-feller) (“[A]s you tried to lift the black voting-age population in the districts up above 50 percent, it became increasingly difficult to include territory in those districts which had the requisite number of African-American adults in them percentage-wise. So as you were reaching out to do that, it became more and more difficult, and that, in turn, governed the shapes of those districts.”); id. at 32:17-21 (“[A]s you attempted] to raise the minority percentage in some of these districts ... it became increasingly difficult to find areas that had high percentages of African-Americans to raise that district up.”); see also id. at 105:7-14; Trial Tr. vol. IV, 231:2-4 (Hofeller).

As Dr. Hofeller strived to keep VRA districts at 50%-plus-one BVAP throughout the “harmonization” process, the boundaries of those districts generally became less compact. Although some lines in multi-county groupings did follow county boundaries, that often had the effect of making any remaining, non-county lines more irregular so that the 50%-plus-one BVAP threshold could be attained. See Trial Tr. vol. IV, 237:10-14 (Hofeller) (explaining that the exemplar version of House District 12 was more compact prior to harmonization with the optimum *137county groupings); id. at 238:5-8 (explaining that “the compact version” of House District 48, i.e., the exemplar district, crossed into multiple county groups).

In short, Dr. Hofeller drew race-defined exemplar districts across the state in order to implement the 50%-plus-one BVAP and proportionality goals for the purported VRA districts. Those exemplar districts, while modified somewhat in their boundaries to accommodate the Stephenson criteria, were nevertheless substantially enacted as drawn to achieve the uncompromising 50%-plus-one target. Trial Tr. vol. IV, 231:2-4 (Hofeller) (“[Tjhose [exemplar] areas quickly morph[ed] into actual districts, which would be the proposed districts in the state_”).

3. The Subordination of Race-Neutral Districting Criteria

As might now be clear, because race-based goals were primary in the 2011 redistricting process, other “traditional race-neutral dis-tricting principles, including ... compactness, contiguity, and respect for political subdivisions or communities defined by actual shared interests,” Miller, 515 U.S. at 916,115 S.Ct. 2475, were secondary, tertiary, or even neglected entirely in the Chairs’ instructions to Dr. Hofeller, and in his implementation of those instructions.

a. Political Subdivisions and Communities of Interest

The Supreme Court has indicated that one “ti’aditional districting principien” whose disregard may indicate racial predominance includes “respect for political subdivisions.” Shaw I, 509 U.S. at 647, 113 S.Ct. 2816. For example, the division of counties, municipalities, and precincts may be evidence of racial predominance. Miller, 515 U.S. at 908, 918,115 S.Ct. 2475. Additionally, if the legislature has split “communities of interest” and instead grouped areas with “fractured political, social, and economic interests,” connected solely by race, that too may indicate that race was the predominant factor in redistricting, Id. at 919, 115 S.Ct. 2475.

The Chairs did not give Dr. Hofeller any instructions to keep towns or cities whole, to preserve communities of interest, or to avoid splitting precincts. Trial Tr. vol. Ill, 202:1-203:22 (Lewis). Consequently, aside from seeking to create county groupings that were compliant with Stephenson (and even then only after satisfying the 50%-plus-one goal), Dr. Hofeller paid little attention to political subdivisions or communities of interest as he drew his lines, and he divided precincts as necessary in order to satisfy the 50%-plus-one target. Trial Tr. vol. V, 104:21-105:6 (Hofeller); Trial Tr. vol. IV, 43:16-20 (Rucho) (“[Q:] You and Dr. Hofeller divided precincts as necessary to get to your 50 percent goal; correct? ... [A:] Yes ... it was essential for us to be able to do whatever — to use whatever tools were necessary for Dr. Hofeller to harmonize the criteria.”); id. at 44:6-8 (“[Q:] Following the law required you to divide precincts; is that your testimony? ... [A:] Following the law — yes, it did”).

As a result, it is not surprising that the Enacted House and Senate Plans split a high number of precincts. Of the 2,692 precincts in North Carolina, the Enacted House Plan splits 395 precincts, where the Benchmark House Plan split only 285. Pis.’ Ex. 2091 at 3 (First Aff. of Theodore S. Arrington, Ph.D. in N.C. State Conference of Branches of the NAACP v. North Carolina, No. 11-CV-01640 (N.C.Super.Ct.)); Defs.’ Ex. 3017-7 at 2 (Aff. of Dan Frey in Dickson v. Rucho, Ex. 7). The contrast is even starker in the Senate. While the Enacted Senate Plan splits 257 precincts, the Benchmark Senate Plan split only 79. Pls.’ Ex. 2091 at 3; Defs.’ Ex. 3017-7 at 2.

Aside from the large overall number of precinct splits, there is statistically significant evidence that precincts were generally divided for the purpose of separating voters according to race. As an initial matter, it should be noted that racial data — but not, for example, political data — is available below the precinct level, and is reported for every census block. Pis.’ Ex. 2091 at 3-4; Trial Tr. vol. I, 113:19-114:14, 161:9-11 (Arrington). Not surprisingly, then, when precincts were split in both the House and Senate Enacted Plans, the portions that were more heavily African-American in population were systematically assigned to predominantly black districts, and the predominantly white portions *138to white districts. Pls.’ Ex. 2091 at 7-10, 14, 18-19; Trial Tr. vol. 1,117:10-121:16 (Arring-ton). Precincts were “almost never” split between two white districts. Trial Tr. vol. I, 121:23 (Arrington). And precinct splitting occurred most often in the most racially diverse areas of the state, i.e., those areas with both substantial white and substantial black populations. Id. at 122:20-124:11; Pis.’ Ex. 2092 at 10-11, 22, 24-26 (Second Aff. of Theodore S. Arrington, Ph.D. in N.C. State Conference of Branches of the NAACP v. North Carolina, No. 11-CV-01640 (N.C.Super.Ct.)).

Generally, it appears that little to no attention was paid to political subdivisions, communities of interest, or precinct boundaries when drawing the challenged districts’ lines. All such criteria were “subordinated ... to racial considerations.” Miller, 516 U.S. at 916, 115 S.Ct. 2475.

b. Compactness

The Supreme Court has also identified “compactness” as among those “traditional districting principles” whose disregard may indicate the predominance of race in redistricting. Shaw I, 509 U.S. at 647, 113 S.Ct. 2816. The regularity and compactness of a district “may be persuasive circumstantial evidence that race for its own sake, and not other districting principles, was the legislature’s dominant and controlling rationale in drawing its district lines.” Miller, 515 U.S. at 913,115 S.Ct. 2475; see also League of United Latin Am, Citizens (LULAC) v. Perry, 548 U.S. 399, 433, 126 S.Ct. 2594 (2006) (“In the equal protection context, compactness focuses on the contours of district lines to determine whether race was the predominant factor in drawing those lines.”); Shaw II, 517 U.S. at 905-06, 116 S.Ct. 1894 (considering a district’s bizarre shape and non-compactness to affirm a finding of racial predominance).

As with other traditional redistricting principles, Dr. Hofeller was not instructed to make compactness a “primary” or even “secondary” redistricting criterion. Trial Tr. vol. V, 97:16-18 (Hofeller). In Dr. Hofeller’s words, “Compactness would not ... ‘trump’ the need to create 50 percent districts.” Id. at 97:14-15. In fact, although the map-drawing software program Dr. Hofeller used could calculate eight compactness measures, Dr. Hofeller never ran those numbers at any time prior to the Plans’ enactment.20 Id. at 99:8-21.

The results show. The Enacted House Plan is less compact than the Benchmark Plan according to all eight of the measures calculated by Dr. Hofeller’s map-drawing program, Maptitude. Pis.’ Ex. 2094 at 10 (Second Aff. of Anthony E. Fairfax in Dickson v. Rucho). The Enacted Senate Plan is less compact than the Benchmark Plan on seven of those eight measures. Id. at 9.

Of course, statewide compactness scores cannot establish whether race predominated in drawing a particular district, and numerical compactness scores have limited value in determining racial predominance. See generally section II.B. But the fact that these particular compactness scores were available to Dr. Hofeller as he drew district lines and yet were apparently given little consideration does suggest that compactness was subordinated to the Chairs’ racial goals throughout the redistricting.

c. The WCP

As an initial matter, the WCP cannot be relied upon as having determined the shape of district lines drawn within a single county. See Defs.’ Ex. 3028 at 11 (Third Aff. of Thomas B. Hofeller, Ph.D. in Dickson v. Rucho) (“The 16 minority districts drawn within single counties did not require reconfigurations of the county groups.”).

However, as we have seen, in seeking to comply with the WCP as articulated in the Stephenson cases, the Chairs and Dr. Hofel-ler did pay substantial attention to county groupings. Those groupings, in turn, certainly influenced the overall shape and location of some of the challenged districts. But there is *139no doubt that WCP compliance was nevertheless subordinated to the 50%-plus-one and proportionality targets.

First, in applying the Stephenson criteria, the Chairs directed that all purported VRA districts be drawn first, before any other districts. See section ILAl.b. Because the Chairs’ 50%-plus-one and proportionality goals applied to VRA districts, those targets necessarily predominated over the remaining Stephenson criteria, including the requirement to minimize the size of county groupings. In addition, Dr. Hofeller testified that such subordinate Stephenson rules as the county traversal rule — under which the crossing of county lines is minimized within a county group — could only be broken if doing so facilitated the creation of districts required by the VRA, which under the Chairs’ definition meant districts greater than 50% BVAP. See Trial Tr. vol. V, 11:8-23.

Second, as discussed above, where there was conflict between the optimum county groupings and the VRA exemplar districts, Dr. Hofeller generally resolved that conflict in favor of the latter. The optimum county groupings were often excluded from the Enacted Plans in challenged areas of the state, and the vast majority of the BVAP encompassed in VRA exemplar districts was included in enacted VRA districts. See supra section II.A.2.

Finally, and perhaps most importantly, even where county groupings or county lines played some role in the eventual shape of the enacted district, what was never compromised was the 50%-plus-one BVAP target.

In other words, the concern with WCP compliance “came into play only after the race-based deeision[s] had been made.” Shaw II, 517 U.S. at 907, 116 S.Ct. 1894. Consequently, that the WCP influenced the redistricting efforts “does not in any way refute ... that race was the ... predominant consideration.” Id.

d. Politics

Finally, there is no evidence in this record that political considerations played a primary role in the drawing of the challenged districts. Indeed, the evidence suggests the opposite. In one of their first public statements, the Chairs made a point of responding to the claim that “[t]he proposed VRA districts plan is solely an attempt to maintain Republicans’ political power,” which they characterized as an “erroneous statement ].” Joint Ex. 1006 at 1, 3. The Chairs’ response to the claim was that “[t]he State has an obligation to comply with the [VRA],” and that any increased competitiveness for Republicans merely “resulted] from [that] compliance.” Id. at 3-4. Senator Rucho said the same thing the next month during a Senate Redistricting Committee Meeting. Joint Ex. 1013 at 36 (“I mean, very simply, we’re following ... the letter of the law. And if it makes the rest of the districts more competitive, then very simply, you know, is that wrong that they’re more competitive? I don’t think so.”). Although there was an increase in the maps’ political favorability for North Carolina Republicans, see Defs.’ Ex. 3031 (Revised Aff. of Sean P. Trende in Dickson v. Rucho), these statements suggest that such an increase was attributable to VRA compliance.

In other words, according to the Chairs’ statements, politics was an afterthought. And aside from a few scattered references in the record to the “political” nature of redistricting, see Trial Tr. vol. Ill, 123:23-124:5 (Lewis), or the fact that “[p]olitics has traditionally played a role in redistricting,” Defs.’ Ex. 3069 at 15 (2011 Legislator’s Guide to North Carolina Legislative and Congressional Redistricting), there is nothing in the record in connection with the districts at issue here to suggest that statewide political considerations motivated the 2011 redistrieting process.21

*140In sum, there is copious statewide evidence that race-based criteria predominated — and that race-neutral criteria were subordinated — in the creation of the Chairs’ 50%-plus-one purported VRA districts.

B. Analysis of District-Specific Evidence of Racial Predominance

As a racial gerrymandering claim “applies district-by-district,” and not to the state “as an undifferentiated “whole,’ ” we must also consider district-specific evidence signifying that race predominated in drawing the challenged districts. Alabama, 135 S.Ct. at 1265. For all the challenged districts, the overwhelming statewide evidence provides decisive proof that race predominated. See id. (“Voters, of course, can present statewide evidence in order to prove racial gerrymandering in a particular district.”). But a look to the district-specific evidence in this case supports and confirms that conclusion, and provides concrete illustrative examples of how compactness, traditional political and geographic boundaries, communities of interest, and the WCP were compromised in order to meet the 50%-plus-one target and proportionality goal.

Before we delve into the district-specific evidence, certain key concepts and categories of evidence warrant further explanation.

First, in certain cases the Supreme Court has emphasized districts’ compactness, see Shaw II, 517 U.S. at 905-06, 116 S.Ct. 1894; Miller, 515 U.S. at 913, 115 S.Ct. 2475, but it is important to note that a district’s degree of compactness is usually not dispositive in a racial gerrymandering claim. “In some exceptional cases, a reapportionment plan may be so highly irregular that, on its face, it rationally cannot be understood as anything other than an effort to ‘segregate] ... voters’ on the basis of race.” Shaw I, 509 U.S. at 646-47, 113 S.Ct. 2816 (alteration in original) (quoting Gomillion v. Lightfoot, 364 U.S. 339, 341, 81 S.Ct. 125, 5 L.Ed.2d 110 (1960)). However, it is not the case that “a district must be bizarre on its face before there is a constitutional violation.” Miller, 515 U.S. at 912, 115 S.Ct. 2475. Compactness is simply one factor that can indicate whether race played a predominant role in drawing a district, and here our discussions of compactness merely serve to reinforce our conclusions regarding racial predominance.

There are two primary ways that courts evaluate compactness. One way is through quantitative measures of compactness, like the eight measures available in Dr. Hofeller’s map-drawing software and entered into evidence as part of Mr. Fairfax’s expert report. See Pis.’ Ex. 2094 at 2. “Substantial divergences from a mathematical standard of compactness may be symptoms of illegitimate gerrymandering.” Karcher v. Daggett, 462 U.S. 725, 755, 103 S.Ct. 2653, 77 L.Ed.2d 133 (1983) (Stevens, J., concurring). However, the Supreme Court has not established clear numerical standards defining when a district becomes non-compact. See Bethune-Hill v. Va. State Bd. of Elections, 141 F.Supp.3d 505, 535 (E.D.Va.2015) (discussing the challenges of utilizing compactness scores when “no one can agree what [compactness] is or, as a result, how to measure it”), problems. noted, — U.S. -, 136 S.Ct. 2406, 195 L.Ed.2d 779 (2016). And there is no clear consensus among scholars defining the exact score on a particular measure that divides compact from non-compact districts. Trial Tr. vol. I, 183:7-13 (Fairfax). Instead, compactness scores are most useful to show relative compactness, by comparing one district to alternative or benchmark versions of that district, or comparing scores to the statewide or nationwide average. See Vera, 517 U.S. at 960, 116 S.Ct. 1941 (citing a study measuring *141the relative compactness of districts nationwide).

The other way to measure compactness is “by an ‘eyeball’ approach,” Vera, 517 U.S. at 960,116 S.Ct. 1941, or what has been called the “interocular test,” Trial Tr. vol. I, 157:18-158:7 (Arrington). Although visually assessing districts necessarily involves some subjective judgment, the Supreme Court has repeatedly relied upon such assessments to determine if a district is “bizarre” or “irregular.” See, e.g., Vera, 517 U.S. at 965-66, 116 S.Ct. 1941; Shaw II, 517 U.S. at 905-06, 116 S.Ct. 1894; Shaw I, 509 U.S. at 646-47, 113 S.Ct. 2816.

Besides compactness, another districting concept in need of further explanation is “contiguity.” The Supreme Court has identified contiguity as one of the “traditional race-neutral districting principles,” Miller, 515 U.S. at 916, 115 S.Ct. 2475, and the North Carolina Constitution requires that each of the state’s “district[s] shall at all times consist of contiguous territory,” N.C. Const, art. II, §§ 3(2), 5(2). In Stephenson II, the Supreme Court of North Carolina affirmed a lower court’s finding that “a district whose parts are ‘held together’ by the mathematical concept of ‘point contiguity’ does not meet the ... criteria for contiguity.” 582 S.E.2d at 254. The U.S. Supreme Court has likewise noted point contiguity as a sign that traditional districting criteria were compromised. See Shaw I, 509 U.S. at 636, 113 S.Ct. 2816 (“At one point the district remains contiguous only because it intersects at a single point with two other districts before crossing over them.”). Additionally, districts may be contiguous, but only because they are connected by narrow “land bridges,” which connect the more populous parts of the district (or in the case of a racial gerrymandering claim, the parts with higher concentrations of minority voters) with a narrow, sparsely populated strip. Miller, 515 U.S. at 908, 917, 115 S.Ct. 2475. This, too, can be a sign of race predominating.

Finally, racial demographic data may help explain the location and idiosyncrasies of a district boundary. Id at 917, 115 S.Ct. 2475 (noting that even if a district is not “bizarre on its face,” the predominance of race may become clearer “when its shape is considered in conjunction with its racial and population densities”); see also Vera, 517 U.S. at 961-62, 116 S.Ct. 1941. Thus, in this case we may look to “racial density maps,” which are shaded to indicate the percentage of the population in each census block that identified as any-part black. Because Dr. Ho-feller testified that he used data of this nature while drawing the challenged districts, the racial density maps can provide useful insights into whether district boundaries reflect racial differences in the population. Trial Tr. vol. V, 100:14-101:18; see Pis.’ Ex. 2062 (providing a screenshot of racial density data projected onto a map in the map-drawing software used by Dr. Hofeller).22

Racial demographic data can also be useful because it may signify whether “race was the predominant factor motivating the legislature’s decision to place a significant number of voters within or without a particular district.” Miller, 515 U.S. at 916,115 S.Ct. 2475. Thus, the Supreme Court has considered the race of the individuals who were added to or subtracted from the benchmark district in order to form the enacted district, See Alabama, 135 S.Ct. at 1263, 1271. We also may consider testimony providing an explanation for the “contours” of the district. LULAC, 548 U.S. at 433, 126 S.Ct. 2594.

Keeping these principles in mind, we turn to our district-specific analysis. Based on the following evidence regarding each district, and our analysis of the statewide evidence above, we find that race was the predominant *142criterion in drawing all of the challenged districts.

1, Senate District 4

The enacted version of Senate District 4 is a majority-black district in northeastern North Carolina. The benchmark version of Senate District 4 had a BVAP of 49.14% under the 2000 census, which grew to a BVAP of 49.70% under the 2010 census. Defs.’ Ex. 3000 at 151, 158. Enacted Senate District 4 was drawn with a BVAP of 52.75%, thus achieving the Chairs’ goal of drawing each VRA district above 50% BVAP. Third Joint Stip. ¶ 3, EOF No. 90; Joint Ex. 1003 at 120 (Enacted Senate Plan map and statistical information). Senate District 4 was one of the VRA districts originally identified in the June 17 Senate VRA map and was enacted without substantial changes to the shape, location, or BVAP level of the proposed VRA district. Joint Ex. 1001 at 1, 67 (Senate VRA Map and statistical information); Joint Ex. 1003 at 1,120; Pis.’ Ex. 2072 at 1.

Enacted Senate District 4 encompasses the entirety of Vance, Warren, and Halifax Counties, then snakes through portions of Nash County and Wilson County, ending just south of the city of Wilson, North Carolina. Not only is enacted Senate District 4 less visually compact than the benchmark version, which was made entirely of whole counties, Pis.’ Ex. 2012 at 1, but it is also quantitatively less compact than the benchmark district on six of the eight measures analyzed by Mr. Fairfax, Pis.’ Ex, 2094 at 18, 20.

Enacted Senate District 4 divides two counties, five municipalities,23 and two precincts. Third Joint Stip. ¶ 5; Joint Ex. 1003 at 6-8,104. The demographic evidence suggests that these political and geographic units were divided on the basis of race. For instance, the portion of Nash County drawn into Senate District 4 has a BVAP of 51.03%, whereas the remainder of Nash County has a BVAP of only 25.78%. Defs.’ Answer to Am. Compl. (“Answer”) ¶ 75, EOF No. 14. Similarly, the portion of Wilson County assigned to Senate District 4 has a BVAP of 63.62%, whereas the remainder of the county has a BVAP of only 24.10%. Id. ¶ 77. Additionally, Senate District 4 captures roughly 52% of the city of Rocky Mount in eastern Nash County, but manages to grab 84.26% of the voting-age African-Americans who reside in that city. Third Joint Stip. ¶ 8; Joint Ex. 1003 at 104. In the two split precincts, 82.2% of the voting-age African-Americans were assigned to Senate District 4. Third Joint Stip. ¶ 6. Finally, the racial density map demonstrates that the boundaries in the split counties in enacted Senate District 4 seem to trace areas that have a high proportion of African-Americans. Id. ¶ 18.

When viewed in conjunction with the strong statewide evidence, the district-specific evidence confirms that race was the predominant motive in drawing Senate District 4.

2. Senate District 5

Enacted Senate District 5 is a majority-black district located in Wayne, Greene, Pitt, and Lenoir Counties. The benchmark version of Senate District 5 had a BVAP of 30.14% under the 2000 census and 30.99% under the 2010 census. Third Joint Stip. ¶ 20. Enacted Senate District 5 was drawn with a BVAP of 51.97%, thus achieving the Redistricting Chairs’ 50%-plus-one target. Id. ¶ 21; Joint Ex. 1003 at 120. The Chairs identified Senate District 5 as a proposed VRA district in the Senate VRA map, and the enacted district substantially comports with the proposed district’s shape, location, and BVAP. Joint Ex. 1001 at 1, 67; Joint Ex. 1003 at 1, 120; Pis.’ Ex. 2072 at 2. Further, Senator Rucho stated during debate on the Senate floor that “Senate District 5 has been drawn in such a way as to include a 50 percent-plus b[l]ack voting *143age population to rectify” a perceived VRA violation. Joint Ex. 1015 at 93.

Enacted Senate District 5 is centered on Greene County, which is kept whole. The rest of the district, however, is much more irregular: one portion expands east to capture a substantial but oddly shaped portion of Pitt County; a narrow, hook-shaped appendage reaches west from Greene County into Wayne County; and the final appendage stretches south from Greene County to capture portions of Lenoir County. The district is visually less compact than the benchmark district, Pis.’ Ex. 2012 at 2, and is less compact on eight of the eight measures presented by Mr. Fairfax, Pis.’ Ex. 2094 at 18, 20.

Enacted Senate District 5 substantially overlaps with the benchmark version of the district. It too was centered on Greene County and contained portions of Pitt and Wayne Counties, although it did not reach into Lenoir County. Defs.’ Ex. 3000 at 155. To increase the BVAP of this district by roughly 21% — without drastically relocating the district — the Chairs necessarily carved out white voters and added a large number of African-American voters. Thus, enacted Senate District 5 was drawn to include 38,250 fewer white persons and 38,181 more African-Americans than the benchmark version. Third Joint Stip. ¶ 22.

To achieve such a dramatic change in de-mographies, enacted Senate District 5 divides three counties, six municipalities, and forty precincts. Id. ¶ 23; Joint Ex. 1003 at 8-9, 104-05. These divisions appear to be motivated largely by race. For instance, the portion of Pitt County in Senate District 5 has a BVAP of 64.59%, while the remainder of Pitt County — assigned to majority-white Senate District 7 — has a BVAP of only 16.16%. Answer ¶ 85. Lenoir and Wayne Counties were similarly divided such that the portions of the counties in Senate District 5 contain a substantially higher BVAP than the portions excluded from the district. Id. ¶ 83, 87. Further, 70.6% of the voting-age African-Americans in the forty divided precincts were assigned to Senate District 5. Third Joint Stip. ¶ 24.

Finally, the racial density map supports the conclusion that race was the predominant motive in drawing the district’s lines. Third Joint Stip. ¶ 33. And Plaintiff Julian Pridgen, a resident of Lenoir County, testified that certain irregular portions of Senate District 5 capture areas that are “heavily concentrated with African-Americans” such as the cities of Kinston and La Grange. Trial Tr. vol. I, 211:1-8.

When viewed in conjunction with the strong statewide evidence, the district-specific evidence confirms that race was the predominant motive in drawing Senate District 5.

3. Senate District 14

Enacted Senate District 14 is a majority-black district located entirely in Wake County. The 2003 version of Senate District 14 had a BVAP of 41.01% under the 2000 census and 42.62% under the 2010 census. Third Joint Stip. ¶ 35. Enacted Senate District 14 was drawn with a BVAP of 51.28%, thus achieving the Redistricting Chairs’ 60%-plus-one goal. Id. ¶36; Joint Ex. 1003 at 120.

In Dr. Hofeller’s initial study of the state’s demographics, he determined that a majority-black district could be drawn in Wake County, and proceeded to do so. Defs.’ Ex. 3028 at 18. The Chairs identified Senate District 14 as a proposed VRA district in the Senate VRA map, and the enacted district substantially matches the proposed district’s shape, location, and BVAP. Joint Ex. 1001 at 1, 67; Joint Ex. 1003 at 1, 120; Pis.’ Ex. 2072 at 3.

Enacted Senate District 14 is located entirely within Wake County, but is part of a two-county grouping of Wake and Franklin Counties. The majority of the population of Senate District 14 comes from the southeastern portion of the city of Raleigh. Joint Ex. 1003 at 108. The benchmark version of Senate District 14 was also located entirely within Wake County, and included a portion of Raleigh. Defs.’ Ex. 3000 at 155. Senator Dan Blue, who represents Senate District 14, testified that the Benchmark Senate Plan divided Wake County roughly “into four quadrants.” Trial Tr. vol. I, 50:21-51:6. Benchmark Senate District 14 was more rectangular and traveled “along the county *144line” for a significant portion of the boundary. Id. at 51:6-7. In contrast, Senator Blue described enacted Senate District 14 as a “crab” with “things that look like claws” and “pincers” reaching out from the core of the district. Id. at 52:7-10. Our own independent assessment confirms that Senate District 14 is visually less compact and more irregular than its predecessor. Pis.’ Ex. 2012 at 3. Additionally, it is less compact on eight of the eight compactness measures evaluated by Mr. Fairfax. Pis.’ Ex. 2094 at 18, 20.

Benchmark Senate District 14 was overpopulated by 41,804 persons according to the 2010 census. Third Joint Stip. ¶ 37. The General Assembly thus had to remove a large number of individuals from the district to comply with one person, one vote requirements. However, compared to its benchmark, enacted Senate District 14 contains 2,145 fewer African-Americans and 38,040 fewer white persons. Id. Senator Blue testified that “effectively what they did is [they] took only the white voters out in order to get [the population] down.” Trial Tr. vol. I, 68:21-24.

Additionally, Senate District 14 divides three municipalities, seemingly on the basis of race. Joint Ex. 1003 at 108. For instance, the district includes 36.29% of the city of Raleigh, but this portion of Raleigh contains 65.44% of the city’s African-American voting-age population. Third Joint Stip. ¶ 40; Joint Ex. 1003 at 108. The enacted district also divided twenty-nine of the fifty-one precincts in the district. Third Joint Stip. ¶ 38, Joint Ex. 1003 at 18-19. Dr. Hofeller testified that he divided these precincts to achieve the “goal” of “creat[ing] a majority-minority district.” Pis.’ Notice of Filing of Designated Dep. Test. Ex. 1 (“Pis.’ Designated Deps.”), at 306, EOF No. 102. In addition, Senate District 14’s racial density map supports the conclusion that the strange appendages that cut through precincts assign areas with a greater proportion of African-Americans to Senate District 14, leaving areas with a small proportion of African-Americans in the neighboring districts.24 See Third Joint Stip. ¶ 51.

Thus it appears that traditional districting criteria were subordinated to race to draw this district. When viewed in conjunction with the direct evidence of the legislature’s intent to create a majority-black district in Wake County, we conclude that race predominated in drawing Senate District 14.

4. Senate District 20

Enacted Senate District 20 is a majority-black district made up of Granville County and part of Durham County. The 2003 version of Senate District 20 had a BVAP of 44.58% under the 2000 census, and 44.64% under the 2010 census. Third Joint Stip. ¶ 53. Enacted Senate District 20 was drawn with a BVAP of 51.04%, and therefore achieves the Chairs’ goal of drawing each VRA district above 50% BVAP. Id. ¶ 54; Joint Ex. 1003 at 120.

In Dr. Hofeller’s initial study of the state’s demographics, he determined that a majority-black district could be drawn in this area of the state. Defs.’ Ex. 3028 at 18. He then drew Senate District 20 as a proposed VRA district in the Senate VRA map, and the enacted district substantially accords with the proposed VRA district’s shape, location, and BVAP. Joint Ex. 1001 at 1, 67; Joint Ex. 1003 at 1,120; Pis.’ Ex. 2072 at 4.

Enacted Senate District 20 includes the entirety of Granville County and part of Durham County. Benchmark Senate District 20 was contained entirely within Durham County. Defs.’ Ex. 3000 at 155. While the Granville County portion of the enacted district follows county lines, the Durham portion is oddly shaped. Plaintiff Milo Pyne, a resident of Durham, described the shape as 'Very eccentric,” noting that it “goes off in little squiggles to capture particular census blocks.” Trial Tr. vol. II, 155:19, 22-24. Representative Larry Hall, who represents House District 29 in Durham County, testified that enacted Senate District 20 is a “cutout” of areas in the city of Durham with a high African-American population, connected by a “bridge” to the Granville County portion of the district. Id. at 198:22-25. We find that *145enacted Senate District 20 is less visually compact than its predecessor, Pis.’ Ex. 2012 at 4, and note that it is less compact on seven of the eight compactness measures assessed by Mr. Fairfax.25 Pis.’ Ex. 2094 at 18, 20.

Plaintiffs allege, and the evidence introduced at trial suggests, that the pairing of Durham and Granville Counties demonstrates a disregard for communities of interest. See Pis.’ Post-trial Revised Proposed Findings of Fact and Conclusions of Law (“Pis.’ Post-trial Findings”) 23, ECF No. 113. Representative Hall testified that many residents in Durham are part of the “university economy,” and described the county’s economy overall as being dominated by “light industry, eommerc[e] and .., [the] service industry, [the] university industry, and [the] medical [industry].” Trial Tr. vol. II, 186:2, 199:15-16. Granville County, on the other hand “is a primar[ily] agricultural area with some light industry.” Id. at 199:11-12. In Representative Hall’s opinion, there are almost no commonalities between the communities in Granville and those in Durham, aside from the fact that both contain significant African-American populations. Id. at 199:5-10.

Additionally, Senate District 20 divides the city of Durham, seemingly on the basis of race. The enacted district contains 53.29% of the city of Durham, but manages to grab 76.94% of Durham’s African-American voting-age population. Third Joint Stip. ¶ 58; Joint Ex. 1003 at 109.

The way that the city of Durham was divided is notable: the Enacted Plan splits thirty-five of the forty-nine precincts contained in the Durham County portion of Senate District 20. Joint Ex. 1003 at 23-24; Trial Tr. vol. II, 155:18-21 (Pyne). In other words, over 70% of the precincts in the Durham County portion of Senate District 20 are split. In contrast, benchmark Senate District 20 split only four precincts. Third Joint Stip. ¶ 56; Trial Tr. vol. II, 155:8-12 (Pyne). Moreover, in the enacted district’s split precincts, 63.8% of the African-American voting-age population is assigned to Senate District 20. Third Joint Stip. ¶ 57. Dispelling the possibility that such racial divisions occurred by chance, Dr. Hofeller testified that the precincts in District 20 were split in order to draw the district above 50% BVAP. Pis.’ Designated Deps. at 308-09.

According to an election administrator for the Durham County Board of Elections, the enacted district not only splits significantly more precincts, but also splits precincts “in a much more complicated manner” than any previous plan. Pis.’ Ex. 2101 at 1, 3. For instance, some of the splits run along “minor roads that only span one or two blocks.” Id. at 3. Within the split precincts, the boundaries of Senate District 20 divide neighborhoods, apartment complexes, and even individual homes. Pis.’ Ex. 2102 at 1-12 (Letter from Joseph Fedrowitz, Geographer and Absentee by Mail Coordinator for the Durham County Board of Elections). This evidence strongly suggests that Defendants were unwilling to compromise on their 50%-plus-one goal in this district, even at the expense of traditional considerations such as existing political boundaries and communities of interest.

Finally, Plaintiffs put forth evidence that the precise contours of the district were determined by race. Mr. Pyne testified that predominantly white neighborhoods, such as Forrest Hills and communities around South-point Mall, were notably excluded from Senate District 20. Trial Tr. vol. II, 163:11-165:7. On the other hand, neighborhoods with substantial African-American populations, such as West End, Old Farm, and the area surrounding North Carolina Central University, were captured by the bizarre district lines. Id, at 164:3-8, 165:16-25, 166:14-18. The racial density map supports Mr. Pyne’s testimony that race substantially explains the placement of the district’s boundaries. Third Joint Stip. ¶ 68.

When viewed in conjunction with the strong statewide evidence, the district-specific evidence confirms that race was the predominant motive in drawing Senate District 20.

*1465. Senate District 21

Senate District 21 is a majority-black district which gets more than half of its population from the city of Fayetteville, North Carolina. Joint Ex. 1003 at 109. In his initial review of the state’s demographics, Dr. Hotelier determined that he could create a majority-black Senate district “anchored in Cumberland County.” Defs.’ Ex. 3028 at 18. The benchmark version of Senate District 21 had a BVAP of 41.00% based on the 2000 census, which grew to 44.93% under the 2010 census. Third Joint Stip. ¶ 70. Enacted Senate District 21 has a BVAP of 51.53%, thus meeting the Chairs’ goal of drawing each VRA district above 50% BVAP. Id. ¶ 71; Joint Ex. 1003 at 120.

Benchmark Senate District 21 was a “squarely shaped” district located in the northwestern quadrant of Cumberland County. Trial Tr. vol. II, 99:8-13 (Covington); Defs.’ Ex. 3000 at 155. During the 2011 redistricting process, the Chairs released the Senate VRA map, which contained a proposed version of Senate District 21 located in the same northwestern quadrant of Cumberland County but with a more contorted and irregular shape. Joint Ex. 1001 at 1. After the release of the Senate VRA map, Senator Rucho significantly altered the proposed district by adding the entirety of Hoke County, Cumberland County’s neighbor to the west. Id.; Joint Ex. 1003 at 1; Pis,' Ex. 2072 at 5. The evidence indicates that Hoke County was added to this district because it was a Section 5 county, and therefore the Chairs determined that it should be in a 50%-plus-one district. Joint Ex. 1007 at 3; Joint Ex. 1013 at 41-42; Trial Tr. vol. IV, 45:13-21 (Rucho).

To accommodate the population of Hoke County, while maintaining a BVAP above 50%, the portion of the district in Cumberland County became even more bizarre in its shape, The portion of enacted Senate District 21 in Cumberland County contains multiple appendages, which are so thin and oddly shaped that it is hard to see exactly where the district begins and ends. See Pis.’ Ex. 2012 at 5. Some portions of the district are so narrow that the district is nearly non-contiguous.

Eric Mansfield, the former Senator from Senate District 21, described the enacted district as “squid”- or “crab”-shaped. Trial Tr. vol. II, 120:6-7. Roberta Waddle, a Cumberland County resident, stated that the district had “long reaching fingers that divide our community in a nonsensical manner.” Pis.’ Ex. 2105 at 2 (Aff. of Roberta Waddle in Dickson v. Rucho). Plaintiff Sandra Coving-ton, who also lives in Cumberland County, described the district as “fragmented” and “non-compact.” Trial Tr. vol. II, 99:16. She explained that the bizarre shape of the district has created a lot of confusion among voters, since individuals in the same neighborhoods are often assigned to different districts. Id at 100:18-25. Our own assessment of the district accords with these descriptions. Enacted Senate District 21 is not only less visually compact than the benchmark district, Pis.’ Ex. 2012 at 5, but it is also less compact on all eight compactness measures presented by Mr. Fairfax, Pis.’ Ex. 2094 at 18, 20.

Enacted Senate District 21 also appears to divide traditional political boundaries on the basis of race. First, the district divides three municipalities. Joint Ex. 1003 at 109. Most notably, enacted Senate District 21 divides the city of Fayetteville: it contains 55.16% of Fayetteville’s overall population, but 75.70% of the voting-age African-Americans in the city. Third Joint Stip. ¶ 75; Joint Ex. 1003 at 109. Enacted Senate District 21 also includes 45.20% of the town of Spring Lake, but 69.87% of that town’s African-American voting-age population. Third Joint Stip. ¶76; Joint Ex. 1003 at 109.

Additionally, the enacted district divides thirty-three of the forty-one precincts located in the Cumberland County portion of the district (roughly 80%). Third Joint Stip. ¶ 73; Joint Ex, 1003 at 24-25. Only one precinct was divided in the benchmark district. Third Joint Stip. ¶ 73; Trial Tr. vol. II, 121:4-8 (Mansfield). Within those split precincts, 60.3% of the African-American voting-age population was assigned to Senate District 21. Third Joint Stip. ¶ 74. According to Dr. Hofeller, these precincts had to be divided in order to achieve the 50%-plus-one BVAP *147goal for this district. Pis.’ Designated Deps. at 307.

Finally, the racial density map supports the conclusion that race was the predominant motive in drawing the district’s lines. Third Joint Stip. ¶ 86. Testimony from those familiar with the Fayetteville area supports this conclusion. Former Senator Mansfield said that the lines appear to be “capturing black neighborhoods,” Trial Tr. vol. II, 122:6-9. Plus, both Sandra Covington and Reva McNair — residents and voters in Cumberland County — identified specific predominantly African-American neighborhoods that were encompassed by the oddly shaped lines of the enacted district. Id. at 100:2-7 (Covington); Pis.’ Ex. 2108 at 4 (Second Aff. of Reva McNair in Dickson v. Rucho).

In conclusion, both the statewide and the district-specific evidence confirms that race was the predominant motive in drawing Senate District 21,

6. Senate District 28

Enacted Senate District 28 is a majority-black district in Guilford County. The benchmark version of Senate District 28, enacted in 2003, had a BVAP of 44.18% under the 2000 census and 47.20% under the 2010 census. Third Joint Stip. ¶88, As enacted in 2011, Senate District 28 has a BVAP of 66.49%. Third Joint Stip. ¶ 89; Joint Ex. 1003 at 120.

In Dr. Hofeller’s initial study of the state’s demographics, he determined that a majority-black district could be drawn in Guilford County, and proceeded to do so. Defs.’ Ex. 3028 at 18. The Redistricting Chairs identified Senate District 28 as a proposed VRA district in the Senate VRA map, and the district was enacted without substantial changes to the proposed district’s shape, location, or BVAP. Joint Ex. 1001 at 1, 67; Joint Ex. 1003 at 1, 120; Pis.’ Ex. 2072 at 6. Dr. Hofeller testified that Senate District 28 was drawn to achieve the 50%-plus-one goal. Pis.’ Designated Deps. at 309.

The city of Greensboro forms the primary population center in Senate District 28. Joint Ex. 1003 at 111. Although the portion of the district in Greensboro is not particularly strange in its shape, an arm of the district protrudes west, then hooks south, to capture part of the city of High Point. See Trial Tr. vol. I, 197:8-11 (Yvonne Johnson) (describing the enacted district as more “far reaching” and “fragmented” than its benchmark). The enacted district is visually less compact than the benchmark district, Pis.’ Ex. 2012 at 6, and is less compact on five of the eight compactness measures reported by Mr. Fair-fax, Pis.’ Ex. 2094 at 18, 20.

Both the enacted district and the benchmark district were located entirely within Guilford County. The benchmark district was underpopulated by 13,673 people. Third Joint Stip. ¶ 90. As drawn in 2011, the enacted district includes 12,508 fewer white persons and 30,773 more African-Americans. Id.

To achieve this dramatic demographic change, enacted Senate District 28 splits two municipalities, seemingly on racial lines. Joint Ex. 1003 at 111. The enacted district contains 57.69% of the population of the city of Greensboro, but manages to capture 82.45% of the African-American voting-age population in that city. Third Joint Stip. ¶ 93; Joint Ex. 1003 at 111. Enacted Senate District 28 also includes only 35.26% of the population of High Point, but over 60% of High Point’s African-American voting-age population. Third Joint Stip. ¶ 94; Joint Ex. 1003 at 111.

Enacted Senate District 28 also splits fifteen precincts, more than twice as many as the benchmark district. Third Joint Stip. ¶ 91. In those fifteen split precincts, 70.4% of the voting-age African-Americans were assigned to Senate District 28. Id. ¶ 92. Dr. Hofeller testified that the precincts were divided in this district in order to achieve the goal of drawing it above 50% BVAP. Pis.’ Designated Deps. at 309.

Finally, Yvonne Johnson, a long-time Greensboro resident who has served as the mayor of Greensboro and currently serves as a Greensboro City Council member, testified that the portion of the district that reaches into High Point grabs “an African-American community” in that area. Trial Tr. vol. I, 191:15-192:10, 197:8-11. The racial density map supports this conclusion: the boundary lines of Senate District 28 outline areas with *148a high proportion of African-Americans. Third Joint Stip. ¶ 104.

Based on this district-specific evidence, in conjunction with the statewide evidence of legislative intent, we conclude that race predominated in drawing Senate District 28.

7. Senate District 32

Senate District 32, as enacted, has a BVAP of 42.53% and is located entirely in Forsyth County. Third Joint Stip. ¶ 107. Enacted Senate District 32 is unique in this case, because it is the only challenged district that was not drawn above 50% BVAP. Thus, enacting this district did not further the Chairs’ statewide goal to increase the number of districts with a 50%-plus-one BVAP. Our analysis of Senate District 32 therefore relies primarily upon district-specific evidence to determine whether race predominated in drawing this district. We find that it did.

The benchmark version of Senate District 32 was located in roughly the same location as the enacted district — also entirely within Forsyth County. Pis.’ Ex. 2012 at 7. Benchmark Senate District 32 had a BVAP of 41.42% under the 2000 census and 42.52% under the 2010 census. Third Joint Stip. ¶ 106.

When the 2011 redistrieting process began, the Chairs concluded that it was not possible to draw a majority-black district in Forsyth County. Joint Ex. 1005 at 6 (June 17, 2011, Joint Statement by Redistrieting Chairs) (“Chairman Rucho believes that it is not possible to create a majority black Senate district in Forsyth.”); Joint Ex. 1006 at 2 (Joint Statement by Redistrieting Chairs Pri- or to June 23, 2011, Public Hearing) (“Senate District 32 is not a majority black district because of the absence of sufficient black population in Forsyth County.”).

The Senate VRA map, released on June 17, included a version of Senate District 32 with a BVAP of 39.32%. Joint Ex. 1001 at 67; Joint Ex. 1005 at 6. Proposed Senate District 32, as it appeared on that map, was fairly regular in shape. It followed the county boundary on its south side, and only split one precinct. Pis.’ Ex. 2072 at 7; Pis.’ Designated Deps. at 310. Senator Rucho explained to the Senate Redistrieting Committee that the “Forsyth County population doesn’t meet the level of a complete Voting Rights Act district, and, therefore, what we tried to do is develop what would be a coalition district with the black voting age population and the Hispanic voting age population.” Joint Ex. 1013 at 17.

On June 23, a group called the Alliance for Fair Redistrieting and Minority Voting Rights (AFRAM) submitted information and proposed district maps to the General Assembly. Defs.’ Ex. 3013-11 at 2. On its proposed Senate map, AFRAM recommended that Senate District 32 be drawn with a 41.95% BVAP. Defs.’ Ex. 3000 at 169.26 After receiving this information, the Chairs announced that the BVAP in Senate District 32 would be increased to 42.53%, which would “exceed[ ] the percentage suggested for that district by [AFRAM],” Joint Ex. 1007 at 3, and exceed the BVAP of the benchmark district.

Both Senator Rucho and Dr. Hofeller testified at trial that the BVAP of Senate District 32 was purposefully increased to exceed the percentage recommended in the AFRAM map. Trial Tr. vol. IV, 38:9-12 (Rucho) (“We tried to go along with the AFRAM plan.”); Trial Tr. vol. V, 32:1-7 (Hofeller) (“[W]e felt that in order to avoid a complaint about that district, we should raise that district’s percentage slightly above the percentage of that district on the AFRAM map.”); see also Defs.’ Ex. 3028 at 20 (Hofeller) (“This minority percentage was intended to meet the minority percentage contained in the [AF-RAM] map for this district, which it did.”). Senator Rucho also opined that increasing the BVAP above that of the AFRAM map and the benchmark district would improve the state’s chances of obtaining Section 5 preclearance from the Justice Department. Trial Tr. vol. IV, 29:19-30:2.

To reach a BVAP of 42.53% for Senate District 32, Dr. Hofeller had to markedly *149reduce its compactness, carving out neighborhoods in the center of the district and adding jagged protrusions on the outer edge. Pis.’ Ex. 2072 at 7. In drawing the enacted district, he changed almost every mile of proposed Senate District 32’s boundaries. Id. While the proposed district split one precinct, the enacted district split forty-three. Third Joint Stip. ¶ 109; Joint Ex. 1003 at 38-40. The result was a district that was less visually compact than the versions of Senate District 32 in both the Benchmark Plan and the Senate VRA map.27

At trial, Senator Rucho testified that “the change in the shape” which occurred between the release of the Senate VRA map and the enactment of the district was “a result of the increase in the black voting-age population in the district,” which was done to “go along with the AFRAM plan.” Trial Tr. vol. IV, 38:9-12.

Senator Rucho’s testimony that the change in shape was explainable only by race is supported by the demographic evidence. While Senate District 32 — located entirely within Forsyth County — has a BVAP of 42.53%, the remainder of Forsyth County has a BVAP of only 7.19%. Answer ¶¶ 126-127. Senate District 32 also splits two municipalities, seemingly on the basis of race. Third Joint Stip. ¶¶ 110-11; Joint Ex. 1003 at 112. For instance, enacted Senate District 32 includes 73.62% of the city of Winston-Salem, but 94.27% of Winston-Salem’s African-American voting-age population. Third Joint Stip. ¶ 111; Joint Ex. 1003 at 112. Additionally, roughly 80% of the voting-age African-Americans who lived in the precincts split by this district were assigned to Senate District 32. Third Joint Stip. ¶ 110.

One split precinct in Senate District 32 is particularly notable, because Senator Rucho split the precinct in order to carve a white incumbent out of the district. Linda Garrou, a white Democrat, had represented Senate District 32 since 1999. Third Joint Stip. ¶ 112; Defs.’ Ex. 3016 at 10. She resided in Precinct 908, which was excluded from the proposed version of Senate District 32 on the Senate VRA map. Joint Ex. 1001 at 13; Joint Ex. 1050 at 164. Precinct 908 was then partially added back into the enacted district as a split precinct — although Linda Garrou’s residence, notably, remained excluded. Joint Ex. 1001 at 13; Joint Ex. 1003 at 40; Joint Ex. 1050 at 164-65.

When they released the Senate VRA map, the Chairs publicly announced Senator Ru-cho’s recommendation “that the current white incumbent for the Forsyth Senate district not be included in the proposed Senate District 32.” Joint Ex. 1005 at 6. Senator Rucho, during both the 2011 redistricting and the trial in this case, made clear that Ms. Garrou was drawn out of her district because of her race. See Joint Ex. 1015 at 91 (“We have also removed the white incumbent from the district who has previously defeated African-American primary challenges, and we think that this will provide the minority community within the district with a better opportunity to elect a candidate of their choice.”); Trial Tr. vol. IV, 55:13-19 (Rucho) (agreeing that the incumbent was drawn out of Senate District 32 “because the candidate was white and had defeated a b[l]ack candidate”). Thus, the division of Precinct 908 appears to have been predominantly motivated by race — the race of Linda Garrou.28

The statements of Senator Rucho and Dr. Hofeller establish that race was the key factor that explains the contours of Senate District 32. Although the Redistricting Chairs were willing to draw Senate District 32 below their 50%-plus-one BVAP target, they substituted a different racial target: the BVAP of the AFRAM map’s proposed district. To ac*150complish their goal of reaching a BVAP around 42%, they compromised compactness and respect for political boundaries. In sum, the district-specific evidence clearly demonstrates that race predominated in drawing Senate District 32.

8. Senate Districts 38 and 40

Enacted Senate Districts 38 and 40 are both majority-black districts located entirely within Mecklenburg County. Under the Benchmark Plan, Mecklenburg County was paired with Union County to form a two-county group. There were four Senate districts located entirely within Mecklenburg County, and one comprised of a portion of Mecklenburg and the entirety of Union County. Defs.’ Ex. 3000 at 165. None of the Mecklenburg County districts in the Benchmark Plan had a BVAP above 50%. Id. at 151, 158. Benchmark Senate District 38 was the only one with a BVAP above 40%: it was drawn in 2003 at 47.69% BVAP based on the 2000 census, and decreased to 46.97% under the 2010 census. Third Joint Stip. ¶ 121. Benchmark Senate District 40 had a BVAP of 31.11% under the 2000 census and 35.43% under the 2010 census. Id. ¶ 138.

By the 2011 redistricting, population growth in Mecklenburg County allowed for the county to be a single-county group made up of five complete Senate districts. In Dr. Hofeller’s “initial study of the State’s demographics,” he “quickly determined” that it was possible to draw two majority-black districts in Mecklenburg County. Defs.’ Ex. 3028 at 18 (Hofeller). He proceeded to do so, and the Chairs subsequently released Senate Districts 38 and 40 as proposed VRA districts on the Senate VRA map. Joint Ex. 1001 at 1. Senate Districts 38 and 40 were enacted without substantial changes to the shape, location, or BVAP of the proposed VRA districts. Id. at 1, 67; Joint Ex. 1003 at 1, 120; Pis.’ Ex. 2072 at 8, 9. As enacted in 2011, Senate District 38 had a BVAP of 52.51% and Senate District 40 had a BVAP of 51.84%, thus meeting the 50%-plus-one target. Third Joint Stip. ¶¶ 122,139.

Senate Districts 38 and 40 are not as sprawling or bizarre in shape as many of the other challenged districts. Both are located in the densely populated urban area of Charlotte, North Carolina. Senate District 38 is the only challenged Senate district that was more compact than its benchmark version on five of the eight compactness measures calculated by Mr. Fairfax. Pis.’ Ex. 2094 at 18, 20. Senate District 40, on the other hand, was less compact on five of the eight compactness measures. Id. Of the two, Senate District 40 is more unusually shaped. It curves around the northern and western portions of Charlotte, almost encircling a portion of downtown Charlotte that was assigned to Senate District 37, a majority-white district. See Joint Ex. 1003 at 1,120.

The existence of a relatively compact African-American population in the Charlotte area suggests that it may have been possible to draw a majority-black district in this area without race predominating. But the fact that a district is somewhat compact, on its own, does not foreclose the possibility that race was the predominant factor in the creation of that district. See Shaw II, 517 U.S. at 907, 116 S.Ct. 1894 (explaining that a state’s attention to certain legitimate, traditional interests in drawing a district “does not in any way refute the fact that race was the legislature’s predominant consideration”); Miller, 515 U.S. at 912, 115 S.Ct. 2475 (explaining that the Court’s consideration of compactness “was not meant to suggest that a district must be bizarre on its face before there is a constitutional violation”). Here, Plaintiffs put forth extensive evidence that race did predominate in drawing these districts. Not only were Senate Districts 38 and 40 part of the statewide plan to increase the BVAP in numerous districts to hit a 50%-plus-one target, but the district-specific evidence suggests that race was the driving criterion explaining the “contours” of the district boundaries. See LULAC, 548 U.S. at 433, 126 S.Ct. 2594.

First, the impact of Senate Districts 38 and 40 on the compactness of the remaining districts in Mecklenburg County is notable. Because the Chairs drew Senate Districts 38 and 40 first, and determined that they had to be majority-black districts, the other Meck-lenburg County districts had to be drawn around them. As a result, majority-white Senate District 41 (which is not challenged in *151this case) had to contain the northernmost portion of Mecklenburg County, then follow a long, thin strip of land along the entire eastern border of Mecklenburg County to connect it to the southeastern corner of the county. Joint Ex. 1003 at 1, 120. This land bridge is made primarily of precincts that are split between Senate District 41 and Senate Districts 38 and 40.29 Pls.’ Ex. 2012 at 8-9. In fact, Senate District 41 is nearly non-contiguous: at one point the northern portion of the narrow land bridge is connected to the southern portion solely by a freeway interchange, where no individuals live. Trial Tr. vol. II, 62:6-19 (Daniel Clodfelter). The evidence thus suggests that the compactness and contiguity of Senate District 41 were compromised in order to create two majority-black districts in Senate Districts 38 and 40.

The demographic evidence also indicates that the city of Charlotte was divided along racial lines. Senate Districts 38 and 40 contain 23.36% and 24.64% of the population of the city of Charlotte, respectively. Joint Ex. 1003 at 114. Because Charlotte is such a large city — with a population over 731,000 according to the 2010 census — it necessarily had to be divided into multiple districts. Id.; see Defs.’ Ex. 3069 at 22. However, the way it is divided suggests that race played a significant role: 72.78% of the city’s African-American voting-age population is assigned to either Senate District 38 or 40, with the remainder divided among Senate Districts 37, 39, and 41. Third Joint Stip. ¶ 126.

Finally, viewing the district boundaries in conjunction with the racial make-up of the city of Charlotte confirms that the precise contours of these districts are explainable by race. See Third Joint Stip. ¶ 161 (racial density map for the Charlotte area). The former mayor of Charlotte, former state Senator from Senate District 40, and long-term Charlotte resident Daniel Clodfelter testified extensively about the precise neighborhoods and communities depicted by the racial density map of Senate Districts 38 and 40. Trial Tr. vol. II, 42:14-44:17; 56:22-67:11. Clodfel-ter identified particular neighborhoods and communities with large African-American populations that were assigned to Senate Districts 38 and 40, even when that required splitting precincts or transgressing traditional boundary lines. Id. at 66:22-67:11. He also pointed to predominantly white neighborhoods that appear to have been intentionally carved out of the majority-black districts. Id. at 67:7-58:9, 64:24-65:19. Altogether, the evidence suggests that Dr. Hofeller carefully drew Senate Districts 38 and 40 to reach a BVAP above 50%, without regard for political, natural, or community boundaries, Id. at 63:24-64:1.

When viewed together with the statewide evidence, the district-specific evidence confirms that race predominated in drawing Senate Districts 38 and 40.

9. House District 5

Enacted House District 5 is a majority-black district in northeastern North Carolina. The benchmark version of House District 5 had a BVAP of 49.02% under the 2000 census and 48.87% under the 2010 census. Third Joint Stip. ¶ 153. Enacted House District 5 has a BVAP of 54.17%, thus achieving the Redistricting Chairs’ goal of increasing the BVAP above 50%. Id. ¶ 154; Joint Ex. 1004 at 147. The Chairs identified House District 5 as an intended VRA district in them House VRA map.30 Joint Ex. 1002 at 1. House District 5 was thereafter enacted without substantial changes to its shape, location, or BVAP. Joint Ex. 1002 at 1, 99; Joint Ex. 1004 at 1,147; Pis.’ Ex. 2072 at 10.

Both the benchmark and the enacted version of House District 5 contain three whole counties: Gates, Hertford, and Bertie. However, the benchmark contained a fourth whole county, Perquimans, which was re*152moved from the district in the enacted plan and replaced with a portion of Pasquotank County. Joint Ex. 1004 at 1; Pis.’ Ex. 2022 at 1; Defs.’ Ex. 3001 at 326. Enacted House District 5 is less compact than the benchmark on four of the eight compactness measurements calculated by Mr. Fairfax. Pis.’ Ex. 2094 at 45, 58. Although enacted House District 5 is not significantly less compact than its predecessor, other factors indicate that race predominated in drawing the district. See Shaw II, 517 U.S. at 907, 116 S.Ct. 1894; Miller, 515 U.S. at 912-13, 115 S.Ct. 2475.

House District 5 splits Pasquotank County, seemingly on the basis of race. Within Pas-quotank County, House District 5 splits six of the eleven precincts that are included in the district. Third Joint Stip. ¶ 156; Joint Ex. 1004 at 5. In the split precincts, 74.5% of the voting-age African-Americans were assigned to House District 5. Third Joint Stip. ¶ 157. The enacted district also splits Elizabeth City, the “economic and cultural center” of Pasquotank County and the surrounding area. Joint Ex. 1004 at 124; Trial Tr. vol. II, 88:2-5 (Claude Harris). Enacted House District 5 includes 86.57% of Elizabeth City, and 94.74% of the city’s black voting-age population. Third Joint Stip. ¶ 158; Joint Ex. 1004 at 124. In addition, the racial density map supports the conclusion that the district divides Pasquotank County along racial lines. Third Joint Stip. ¶ 168.

Plaintiff Claude Harris testified at trial that, as a resident of Pasquotank County, he does not consider himself “economically or culturally” tied to the other counties in House District 5. Trial Tr. vol. II, 86:15-22. Mi\ Harris explained that Pasquotank County is considered one of the five “Finger Counties” in the northeastern corner of the state, and it used to be in a district with two other Finger Counties: Camden and Curri-tuck, Id. at 86:8-88:2; Defs.’ Ex. 3001 at 326. Under the Enacted Plan, the portion of Pas-quotank County in House District 5 is no longer part of a district that includes any of the other Finger Counties. Thus, there is some circumstantial evidence that House District 6 divided communities of interest in order to meet the 50%-plus-one goal.

Finally, there is strong direct evidence that Representative Lewis and Dr. Hofeller drew House District 5 on the basis of race. Representative Lewis testified that House District 5 was “one of the districts that [he] drew to reach the [Chairs’] proportionality goal.” Trial Tr. vol. Ill, 214:13-19. Representative Lewis also testified that he divided both Elizabeth City and “Pasquotank County so that [he] could get to 50 percent [BVAP] for House District 5.” Id. at 215:1-6.

House District 5 also provides an example of how race predominated over attempts to comply with the WCP. Dr. Hofeller explained that the Stephenson cases would have required the eight-county group containing House District 5 (as well as House District 1) to “be split” unless there was a VRA district in that location. Defs.’ Ex. 3028 at 14. Therefore, this district illustrates that when the 50%-plus-one goal and the WCP were in conflict, the WCP gave way to the racial target.

Thus, in accordance with their statewide goals, Dr. Hofeller and Representative Lewis drew House District 5 in a way that would increase its BVAP to 50%-plus-one. We conclude that race predominated in drawing this district.

10. House District 7

Enacted House District 7 is a majority-black district located in Franklin and Nash Counties. House District 7 has a BVAP of 50.67%. Third Joint Stip. ¶ 169; Joint Ex. 1004 at 147.

For House District 7, comparisons to the Benchmark Plan are particularly limited in value. Benchmark House District 7, which had a BVAP of 56.03% under the 2000 census and 60.77% under the 2010 census, Defs.’ Ex. 3001 at 315, 332, included the majority of Halifax County and a small portion of Nash County, id. at 326. That small portion in Nash County, however, is the only part of benchmark House District 7 that remains in the enacted version.31 Pis.’ Ex. 2022 at 2. *153Otherwise, the district now contains a much larger portion of Nash County and a significant section of Franklin County, Joint Ex. 1004 at 1, areas that used to be in House Districts 25 and 49 under the Benchmark Plan, Defs.’ Ex. 3001 at 326. Therefore, it is more appropriate to compare House District 7 to benchmark House Districts 25 and 49, which as drawn had BVAPs of 25.87% and 28.49%, respectively. Defs.’ Ex. 3001 at 315.

The relocation of House District 7 appears to be the result of the Chairs’ 50%-plus-one and proportionality goals, since it allowed them to draw two majority-black districts in this area of the state — House Districts 7 and 27 — where there was previously only one. House District 7 was identified in the House VRA map as a proposed VRA district,32 and was enacted without substantial changes to the shape, location, or BVAP of the proposed district. Joint Ex. 1002 at 1, 99; Joint Ex. 1004 at 1,147.

The shape of House District 7 strongly suggests that race predominated in drawing this district. While the district’s northern border tracks the northern boundaries of Franklin and Nash Counties, the remainder of the distinct takes the form of several strangely shaped protrusions, extending south into the two counties. Joint Ex. 1004 at 1. State Senator Angela Bryant, the former Representative for House District 7, testified that the district “was almost impossible to describe to constituents” because its “tentacles” cut through the counties in such strange ways. Tidal Tr. vol. II, 14:22-25. Senator Bryant expressed her concerns before the General Assembly during the redistricting process, noting the “odd and irregular shape” of the district and its “lack of compactness.” Joint Ex. 1020 at 150. Our own visual assessment of the district confirms that it is non-compact and bizarrely shaped. Enacted House District 7 is also less compact than the benchmark district on eight of the eight compactness measures reported by Mr. Fairfax. Pis.’ Ex. 2094 at 45, 58.

Additionally, there is evidence that House District 7 divides political and geographic boundaries in order to gather a sufficient number of African-American voters to reach the 50%-plus-one goal. First, Nash and Franklin Counties are divided between House District 7 and House District 25. House District 25 has a BVAP of only 16.05%, compared to House District 7’s BVAP of 50.67%. Joint Ex. 1004 at 147. Further, House District 7 divides seven municipalities. Id. at 124. Most notably, 61.78% of the population of Rocky Mount was assigned to House District 7, but the lines were drawn such that the district managed to capture almost all of the city’s voting-age African-American population in Nash County: 96.16%. Third Joint Stip. ¶ 174; Joint Ex. 1004 at 124. Likewise, House District 7 includes only 38.93% of the population of the city of Dortehes, but 70.65% of that city’s black voting-age population. Third Joint Stip. ¶ 173; Joint Ex. 1004 at 124. Finally, House District 7 includes 48.18% of the city of Spring Hope, but 76.63% of the voting-age African-Americans in that city. Third Joint Stip. ¶ 175; Joint Ex. 1004 at 124.

The enacted district also divides twenty-two of the thirty-two precincts comprising the district. Third Joint Stip. ¶ 170; Joint Ex. 1004 at 6; Trial Tr. vol. II, 14:14-18 (Bryant). In other words, nearly 69% of the precincts included in the district were split. Of the voting-age African-Americans residing in the twenty-two split precincts, 83.1% were assigned to House District 7, and the remainder were assigned to the majority-white House District 25. Third Joint Stip. ¶ 171; Joint Ex. 1004 at 1, 147. The racial density map also indicates that race explains the placement of the unusual boundaries for this district. Third Joint Stip. ¶ 185.

Senator Bryant testified that the benchmark district was connected through Interstate 95 and U.S. Route 301, but enacted House District 7 does not even have a major road connecting the portions in Franklin County to the portions in Nash County. Trial Tr. vol. II, 15:8-10; see Karcher, 462 U.S. at 757 n. 20,103 S.Ct. 2653 (Stevens, J., concurring) (noting that even oddly shaped districts *154may be compact from a “sociopolitical” standpoint if, for instance, the district follows a “major transport corridor”). In fact, she and her volunteers attempted to drive the district “to figure out where the boundaries are” and found that it was “virtually impossible.” Trial Tr. vol. II, 15:13-16, 18:17-20 (Bryant). According to Senator Bryant, a driver on Highway 64, which is a major corridor through that portion of the state, would cross in and out of House District 7 roughly “five times.” Id. at 18:3-7. Unsurprisingly, the bizarre boundaries of enacted House District 7 have caused significant voter confusion. Id. at 17:17-23.

The circumstantial evidence strongly suggests that race was the predominant motivation for drawing the enacted district. But even if the circumstantial evidence were less clear, the direct evidence of legislative intent removes any doubt. At trial, Representative Lewis testified that race explains the line dividing House Districts 7 and 25. Trial Tr. vol. Ill, 217:25-218:4. Plus, Dr. Hofeller noted that House District 7 would have violated the WCP if it were not a VRA district because it included a “double traverse of the boundary between Franklin and Nash Counties.” Defs.’ Ex. 3028 at 14. Once again, this illustrates that the WCP, along with other traditional districting criteria, was subordinated to race in drawing this district. Id.; see also Defs.’ Proposed Findings of Fact and Conclusions of Law at 60 (“Defs.’ Post-trial Findings”), ECF No. 118.

In conclusion, the direct and circumstantial district-specific evidence, in addition to the statewide evidence, confirms that race predominated in drawing House District 7.

11. House District 12

Enacted House District 12 is a majority-black district located in Greene, Lenoir, and Craven Counties. The benchmark version of House District 12 had a BVAP of 47.51% under the 2000 census and 46.45% under the 2010 census. Third Joint Stip. ¶ 187. Enacted House District 12 has a BVAP of 50.60%, thus achieving the Redistricting Chairs’ goal of increasing the BVAP above 50%. Id. ¶ 188; Joint Ex. 1004 at 147. The Chairs identified House District 12 as an intended VRA district on the House VRA map, and the district was enacted without substantial changes to the shape, location, or BVAP of the proposed VRA district. Joint Ex. 1002 at 1, 99; Joint Ex. 1004 at 1; Pis.’ Ex. 2072 at 11.

House District 12 is long and skinny, and visually stands out as one of the least compact districts in the state. The enacted district has a Reock score of 0.12.33 Pis.’ Ex. 2094 at 58. This is the lowest, and thus worst, Reock score of any district in the Enacted Plan, or in the Benchmark Plan. Pis.’ Ex. 2093 at 33-38. This score is also significantly lower than the average Reock scoi’e for the Enacted House Plan, which is 0.38. Id. at 35. Enacted House District 12 is less compact than the benchmark on six of eight measures presented by Mr. Fairfax, and ties the benchmark on a seventh measure. Pis.’ Ex. 2094 at 45, 58.

In Craven County, enacted House District 12 is almost non-contiguous at certain points. The district is connected by the narrowest of land bridges along Craven County’s western border. Pis.’ Ex. 2022 at 3; see Miller, 515 U.S. at 908, 920, 115 S.Ct. 2475 (holding that race predominated in drawing a district *155which contained narrow “land bridges” through unpopulated areas to connect African-American populations).

House District 12 divides three counties and four municipalities, seemingly on the basis of race. Joint Ex. 1004 at 1, 125. For example, House District 12 contains 42.49% of the population of the city of New Bern, but 72.70% of New Bern’s African-American voting-age population. Third Joint Stip. ¶ 193; Joint Ex. 1004 at 125. It also includes 81.99% of the city of Kinston, but 92.72% of the voting-age African-American population in that city. Third Joint Stip. ¶ 192; Joint Ex. 1004 at 125. House District 12 contains thirteen whole precincts, and thirty-four split precincts. Third Joint Stip. ¶ 190; Joint Ex. 1004 at 8-9. In other words, over 72% of the precincts in the district are split. Of those living in the split precincts, 65.99% of the voting-age African-Americans are assigned to House District 12. Third Joint Stip. ¶ 191.

The racial density map similarly suggests that some of the more bizarre portions of the district were drawn to capture areas with a high proportion of African-Americans. Id. ¶ 203. Plaintiff Julian Pridgen, a resident of Lenoir County, testified that particular neighborhoods and communities with high concentrations of African-American voters were drawn into House District 12. Trial TV. vol, I, 204:25-205:1, 212:12-23. For instance, a narrow protrusion in the northeastern corner of Lenoir County reaches out to grab the town of Griffon, which Mr. Pridgen stated is predominantly African-American. Id. at 212:18-213:5.

Finally, Defendants’ own statements support our conclusion that race predominated in drawing House District 12. In the state’s Section 5 preclearance filings, the state explained that the changes it made to the benchmark version of House District 12, such as adding additional areas in Greene and Lenoir Counties, “restored” House District 12 to “majority black status.” Joint Ex. 1024 at 15. Additionally, Defendants assert that the reason enacted House District 12 is less compact than some of the alternative proposed versions of the district is because “[n]one of the alternate versions of [House District] 12 were created with a majority black [voting-age population].” Defs.’ Post-trial Findings at 131. These statements reinforce our conclusion that it is race, and not any of the traditional redistricting criteria, that explains why House District 12 is one of the least compact districts in the state.

Viewed in conjunction with the statewide evidence, the district-specific evidence confirms that race predominated in drawing House District 12.

12. House District 21

House District 21 is a majority-black district containing portions of Sampson, Duplin, and Wayne Counties. The benchmark version of House District 21 had a BVAP of 48.35% under the 2000 census and 46.25% under the 2010 census. Third Joint Stip. ¶ 205. Enacted House District 21 has a BVAP of 51.90%, thus meeting the Redistricting Chairs’ 50%-plus-one target. Id. ¶ 206; Joint Ex. 1004 at 147.

In Dr. Hofeller’s initial study of the state’s demographics, he determined that a majority-black district could be drawn in this area of the state, and proceeded to do so. Defs.’ Ex. 3028 at 8. The Chairs identified House District 21 as a proposed VRA district on the House VRA map. Joint Ex. 1002 at 1. The proposed version of the district contained a portion in Pender County, which was later removed and replaced with a portion in Dup-lin County prior to the enactment of the final House map. Id.; Joint Ex. 1004 at 1. The only explanation put forth by the Redistricting Chairs for the addition of Duplin County is that it allowed them to “reach[ ] the threshold of getting above the 50 percent mark.” Joint Ex, 1018 at 31 (July 21, 2011, House Redistricting Committee Meeting Transcript) (Lewis); see also Joint Ex. 1024 at 15-16 (House 2011 Section 5 Submission) (“[Significant portions of Duplin County have been added to this district to restore its majority black status and to add population.”).

House District 21 contains the entire eastern edge of Sampson County, a substantial portion of western Duplin County, and a narrow appendage that reaches north into Wayne County, capturing parts of the city of Goldsboro. Joint Ex. 1004 at 1; see Trial Tr. vol. II, 152:6-15 (Albert Kirby). Albert Kirby, *156a County Commissioner in Sampson County, testified that the enacted district “looks like an animal eating something.” Id. at 142:1-4, 143:23-24. Based on our own assessment, the enacted district is visually less compact than the benchmark district. See Pis.’ Ex. 2022 at 4. It is also less compact on six of the eight compactness measures presented by Mr. Fairfax. Pis.’ Ex. 2094 at 45, 58. The Reock score for House District 21 is 0.19, which is the fourth-lowest score in the state and significantly worse than the Enacted House Plan’s district average of 0.38. Pis.’ Ex. 2093 at 35; Pis.’ Ex. 2094 at 43-44, 58.

Enacted House District 21 splits all three of the counties it crosses through. The part of Duplin County in House District 21 has a BYAP of 45.75%, whereas the part of Duplin County excluded from the district has a BYAP of only 15.13%. Answer ¶ 170. In Sampson County, the portion in House District 21 has a BVAP of 53.71%, while the remainder of the county has a BVAP of 21.28%. Answer ¶ 173. Similarly, the BVAP of the part of Wayne County in House District 21 is 54.08%, while the portion of the county in adjacent House District 4 is 16.91%. Id. 176.

House District 21 also divides seven municipalities and twenty-five precincts. Third Joint Stip. ¶ 208; Joint Ex. 1004 at 127. The demographic data suggests that these divisions were motivated by race. For example, 81.40% of the population of the city of Golds-boro is included in House District 21, but that portion of Goldsboro includes 92.10% of the city’s African-American voting-age population. Third Joint Stip. ¶ 211; Joint Ex. 1004 at 127. House District 21 also includes 46.15% of the city of Clinton, but manages to grab 72.67% of the voting-age African-American population in that city. Third Joint Stip. ¶ 210; Joint Ex. 1004 at 127. And in the twenty-five split precincts, 60.6% of the African-American voting-age population is assigned to House District 21. Third Joint Stip. ¶ 209.

Finally, the racial density map also indicates that areas with a high proportion of African-American voting-age population are enveloped by the protrusions and contours of House District 21. Third Joint Stip. ¶ 222. Consistent with the racial density map, Albert Kirby testified that even “small neighborhoods” in his hometown of Clinton were “carved” into House District 21 if they were predominantly black, but assigned to neighboring House District 22 if they were predominantly white. Trial Tr. vol. II, 145:14— 146:17.

In conclusion, when viewed in conjunction with the statewide evidence, the district-specific evidence confirms that race was the predominant motive for drawing House District 21.

13. House District 24

Enacted House District 24 is a majority-black district in Wilson and Pitt Counties. The benchmark version of House District 24 had a BVAP of 54.76% under the 2000 census and 56.07% under the 2010 census. Third Joint Stip. ¶ 224. Enacted House District 24 has a BVAP of 57.33%, thus achieving the Redistricting Chairs’ goal of drawing the intended VRA districts above 50% BVAP. Id. ¶ 225; Joint Ex. 1004 at 147.

Enacted House District 24 overlaps substantially with two of the proposed VRA districts identified on the House VRA map, both of which had a BVAP above 50%.34 Joint Ex. 1002 at 1, 99; Pis.’ Ex. 2072 at 13. One of these proposed VRA districts included Martin County, roughly half of Edgecombe County, and a portion of Wilson County. Joint Ex. 1002 at 1. The other proposed district was primarily in Pitt County, but reached into a small portion of Beaufort County. Id. After the release of the House VRA map, the locations of these two intended VRA districts were adjusted, but both districts retained a BVAP above 50%. Joint Ex. 1004 at 1, 147. One was House District 24, and the other was House District 23 — a majority-black district containing the entirety of Martin and Edgecombe Counties, and therefore forming a single-district, two-county group. Id. Thus, despite the changes to the map, the legisla*157ture still enacted two majority-black districts in this area of the state.

Enacted House District 24 is less visually compact than the benchmark district. Pis.’ Ex. 2022 at 5. The district connects portions of two cities: Wilson on the western edge of the district and Greenville on the eastern edge. Joint Ex. 1004 at 128. Between these two cities, House District 24 narrows to a small bridge of land connecting the part of the district in Wilson County with the part in Pitt County. Pis.’ Ex. 2022 at 5. Additionally, the enacted district is less compact on all eight of the compactness measures presented by Mr. Fairfax. Pis.’ Ex. 2094 at 45, 58.

House District 24 divides twelve precincts and two municipalities. Third Joint Stip. ¶ 227; Joint Ex. 1004 at 128. For instance, the district contains 51.94% of the city of Wilson, but manages to include 74.43% of Wilson’s black voting-age population. Third Joint Stip. ¶ 230; Joint Ex. 1004 at 128. The enacted district also contains 41.12% of the city of Greenville, and 58.28% of the city’s black voting-age population. Third Joint Stip. ¶ 229; Joint Ex. 1004 at 128.

Further, Dr. Hofeller explained that if there was not a VRA district located in this county grouping, the WCP would have required Wilson County to stand alone as its own single-county district. Defs.’ Ex. 3028 at 13 (‘Wilson County has the correct population to form a single house district entirely within its borders.”); see Joint Ex. 1052 at 3 (indicating that under the 2010 census, Wilson County deviated from the ideal House district population by roughly two percent). However, Dr. Hofeller identified the problem with this plan: Wilson County had a BVAP of only 38.19%. Defs.’ Ex. 3028 at 13. Thus, creation of such a single-county district would have required deviating from the goal of drawing 50%-plus-one districts. Id. Consistent with the statewide approach, when the 50%-plus-one target conflicted with the WCP, the WCP principles gave way. Dr. Hofeller therefore drew House District 24 in a two-county grouping and created a double traverse of the border between Pitt and Wilson Counties. Id. As Dr. Hofeller recognized, both changes were “departures from the strict Stephenson county-grouping criteria.” Id. Thus, it was the racial target, and not the WCP, that dictated the contours and location of House District 24.

In conclusion, when viewed in light of the strong statewide evidence, the district-specific evidence confirms that race predominated in drawing House District 24.

14. House Districts 29 and 31

Enacted House Districts 29 and 31 are two majority-black districts in Durham County. Under the Benchmark Plan, Durham County was paired with Person County in a two-county group containing four districts, all of which had a BVAP below 50%. Defs.’ Ex. 3001 at 315-16, 326, 332-33. Benchmark House District 29 had a BVAP of 44.71% under the 2000 census and 39.99% under the 2010 census. Third Joint Stip. ¶ 242. Benchmark House District 31 had a BVAP of 44.71% under the 2000 census and 47.23% under the 2010 census. Id. ¶260. In the Enacted Plan, which paired Durham with Orange County, both House Districts 29 and 31 are majority-black districts located entirely within Durham County, with BVAPs of 51.34% and 51.81%, respectively. Id. ¶¶ 243, 261; Joint Ex. 1004 at 1, 147. Thus, the Chairs achieved their goal of drawing both of these districts at 50%-plus-one BVAP.

In Dr. Hofeller’s initial study of the state’s demographics, he determined that two majority-black district could be drawn in Durham County, and proceeded to do so. Defs.’ Ex. 3028 at 9. House Districts 29 and 31 were identified as proposed VRA districts on the House VRA map, and were enacted without substantial changes to the location, shape, or BVAP of the proposed districts. Joint Ex. 1002 at 1, 99; Joint Ex. 1004 at 1, 147; Pis.’ Ex. 2072 at 14,15.35

Enacted House Districts 29 and 31 both became visually less compact than their corresponding benchmark districts. While the benchmark districts followed county lines for significant stretches, the enacted districts’ boundaries do not appear to follow county, or *158even precinct, lines. See Pis.’ Ex. 2022 at 6-7. However, both are relatively small districts in the urban area surrounding the city of Durham, and thus they are more compact than many of the other challenged districts. House District 29 was less compact on four of the eight compactness measures presented by Mr. Fairfax, and House District 31 was less compact on five of the eight compactness measures. Pis.’ Ex. 2094 at 45, 58. As noted previously, even if a district is not egregiously non-compact, race can still be the predominant factor in its creation. See Shaw II, 517 U.S. at 907, 116 S.Ct. 1894 (noting that evidence suggesting some traditional districting criteria were “addressed” does not foreclose the conclusion that race predominated); Miller, 515 U.S. at 912-13 (holding that race can still be the predominant motive in a compact district).

The evidence suggests that Durham County and the city of Durham were divided along racial lines to form these two majority-black districts. The other districts containing portions of the city of Durham — House Districts 30 and 50 — have BVAPs of 18.43% and 13.25%, respectively. Joint Ex, 1004 at 108, 147. House Districts 29 and 31 contain 35.02% and 30.58% of the population of the city of Durham, respectively, but together capture 82.81% of Durham’s black voting-age population. Third Joint Stip. ¶ 248; Joint Ex. 1004 at 108.

The benchmark versions of House Districts 29 and 31 each split only one precinct, Third Joint Stip. ¶¶ 245, 263. In contrast, the Enacted House Plan splits fourteen of the twenty-eight precincts that make up House District 29, and thirteen of the twenty-one precincts that make up House District 31. Id.; Joint Ex. 1004 at 18-19. In total, the Enacted Plan splits twenty-one precincts in Durham County. Third Joint Stip. ¶ 247. Of the voting-age African-Americans residing in those twenty-one split precincts, 75% were assigned to either House District 29 or 31. Id.

At tidal, Representative Larry Hall, who represents House District 29, testified that many of these precincts are split along racial lines. Trial Tr. vol. II, 177:19-179:4, 189:14-193:13. For instance, Precinct 53-1 in the southwest comer of House District 29 is divided so that the “predominantly white” population in the southern part of the precinct is cut out of the district, and the “predominantly black” section in the northern part of the precinct is kept in the district. Id. at 190:3-19. This dividing line becomes more irregular at one point, jutting out to grab one particular “densely populated African-American community.” Id. at 189:25-191:12; see also id. at 188:11-189:24 (discussing how race explains the split in Precinct 35); id. at 191:13-192:16 (same for Precinct 6); id. at 192:17-193:13 (same for Precinct 33); id. at 195:13-23 (same for Precinct 30-1 in House District 31). The racial density maps confirm this testimony, showing that district lines were drawn so as to keep areas with a high proportion of African-Americans in House Districts 29 and 31, and assign areas without a high proportion of African-Americans to the majority-white districts. Third Joint Stip. ¶¶ 258, 275.

When viewed in conjunction with the statewide evidence, the district-specific evidence supports a finding that race predominated in drawing House Districts 29 and 31.

15. House District 32

Enacted House District 32 is a majority-black district along the border between North Carolina and Virginia. The benchmark version of House District 32 had a BVAP of 36.22% under the 2000 census and 35.88% under the 2010 census. Third Joint Stip. ¶ 277. As enacted in 2011, House District 32 has a BVAP of 50.45%, thus achieving the Chairs’ 50%-plus-one goal. Id. ¶ 278; Joint Ex. 1004 at 147. The Chairs identified House District 32 as an intended VRA district on the House VRA map, and the district was enacted without substantial changes to the shape, location, or BVAP of the proposed VRA district.36 Joint Ex. 1002 at 1, 99; Joint Ex. 1004 at 1, 147; Pis.’ Ex. 2072 at 16.

*159House District 32 contains the entirety of Warren and Vance Counties, but splits Gran-ville County. Joint Ex. 1004 at 1. The part of the district in Granville County includes a rectangle-shaped portion along the northern boundary of the county, and a more unusual, oddly shaped portion that extends south toward the center of Granville County, encompassing parts of the city of Oxford. Benchmark House District 32, on the other hand, included all of Granville County, and roughly half of Vance County. Defs.’ Ex. 3001 at 326. Enacted House District 32 is visually less compact than the benchmark, particularly in Granville County, Pis.’ Ex. 2022 at 8, and is less compact than the benchmark on six of the eight compactness measures calculated by Mr. Fairfax, Pis.’ Ex. 2094 at 45, 58.

House District 32 splits ñve of the nine precincts in Granville County that are included in that district. Joint Ex. 1004 at 19. In contrast, the benchmark district did not divide any precincts. Third Joint Stip. ¶ 280. In the five split precincts, 82% of the voting-age African-Americans were assigned to House District 32. Id. ¶281. The enacted district also divides the city of Oxford, seemingly on the basis of race. House District 32 contains 77.49% of the city of Oxford, but manages to capture 92.92% of the African-American voting-age population in that city. Id. ¶ 282; Joint Ex. 1004 at 129. This demonstrates that Defendants split political subdivisions in order to accomplish their 50%-plus-one goal.

Finally, the racial density map suggests that the district boundaiy in Granville County, particularly near the city of Oxford, was drawn to capture areas with a high proportion of African-American voters. Third Joint Stip. ¶ 287.

We acknowledge that the part of House District 32 in Warren and Vance Counties does not contain these same irregularities. But even if other criteria — such as keeping counties and precincts whole — may explain some of the district’s boundaries, race still may predominate when a state has “mixed motive[s]” in drawing a district. Vera, 517 U.S. at 959, 116 S.Ct. 1941; see also Shaw II, 517 U.S. at 907, 116 S.Ct. 1894 (“That the legislature addressed [some race-neutral] interests does not in any way refute the fact that race was the legislature’s predominant consideration.”). The district-specific evidence indicates that traditional districting criteria — such as compactness, respect for traditional and political boundaries, and maintaining communities of interest — were compromised in the Granville County portion of this district. When viewed in conjunction with the strong statewide evidence, the fact that the Chairs increased the BVAP of House District 32 by roughly fifteen percent in order to meet their 50%-plus-one goal leads us to conclude that race was the predominant factor in drawing this district.

16. House Districts 33 and 38

Enacted House Districts 33 and 38 are two majority-black districts located entirely within Wake County. Under the Benchmark Plan, Wake County included only one district with a BVAP above 40%. Third Joint Stip. ¶ 447. Based on the 2010 census, Wake County had sufficient population to contain eleven complete districts. Answer ¶ 206. In Dr. Hotelier’s initial study of the state’s demographics, he determined that it was possible to draw two of these Wake County districts above 50% BVAP, and he did just that: enacted House District 33 has a BVAP of 51.42% and enacted House District 38 has a BVAP of 51.37%. Third Joint Stip. ¶¶288, 303; Joint Ex. 1004 at 147. The Chairs identified House Districts 33 and 38 as proposed VRA districts on the House VRA map and enacted those districts without substantial modification to their location, shape, or BVAP. Joint Ex. 1002 at 1, 99; Joint Ex. 1004 at 1, 147; Pis.’ Ex. 2072 at 17-18.37

Benchmark House District 33, which was in roughly the same location as enacted House District 38, was initially drawn with a BVAP of 49.97% based on the 2000 census, but increased to a BVAP of 51.74% by the 2010 census. Defs.’ Ex. 3001 at 315, 326, 332. Benchmark House District 38, which now forms part of enacted House District 33, had a BVAP of only 31.63% under the 2000 cen*160sus and 27.96% under the 2010 census. Id. The remainder of enacted House District 33 overlaps with several other benchmark districts, none of which had a BVAP above 35%. Pis.’ Ex. 2022 at 9-10; Defs.’ Ex. 3001 at 315, 326,332.

Visually, House Districts 33 and 38 became marginally less compact than the corresponding benchmark districts. Pis.’ Ex. 2022 at 9-10. Both are located in the urban area surrounding Raleigh, North Carolina, and thus did not have to be as sprawling as more rural districts to meet Defendants’ population requirements. House District 33 is less compact on six of the eight compactness measures presented by Mr. Fairfax, and House District 38 is less compact on five. Pis.’ Ex. 2094 at 45, 58. The relative compactness of these districts does not, however, preclude a finding that race predominated. See Miller, 515 U.S. at 913, 115 S.Ct. 2475 (“[Pjarties may rely on evidence other than bizarreness to establish race-based districting.”).

The evidence suggests that the Enacted Plan divides Wake County on the basis of race. First, other than House Districts 33 and 38, no Wake County House districts have a BVAP higher than 27%. Answer ¶ 207; Joint Ex. 1004 at 147. Additionally, House Districts 33 and 38 each split more than half of their precincts. Third Joint Stip. ¶¶289, 305; Joint Ex, 1004 at 20, 22. Both of the districts capture roughly 65% of the African-American voting-age population in the precincts that they split. Third Joint Stip. ¶¶ 290,306.

The two districts also appear to divide municipalities and communities of interest on the basis of race. House Districts 33 and 38, combined, contain 36.28% of the city of Raleigh, but manage to include 66.81% of Raleigh’s voting-age African-American population. Third Joint Stip. ¶ 307; Joint Ex. 1004 at 116. In addition to Raleigh, these two districts divide the towns of Knightdale and Garner as well as several neighborhoods. Joint Ex. 1004 at 129-30. For instance, the historic Raleigh neighborhoods of Oakwood and Mordecai in Precinct 14 appear to have been divided on the basis of race. See Joint Ex. 1004 at 20, 22 (showing that the portion of Precinct 14 in House District 34 has a BVAP of less than 19%, while the portions in House Districts 33 and 38 have BVAPs of 41.29% and 33.98%, respectively); Pis,’ Ex. 2104 at 2-3 (Aff. of Hugh Stohler in Dickson v. Rucho). The General Assembly also received information that eastern Wake County, which is split by the Enacted Plan, forms a community of interest. Joint Ex. 1018 at 33; Joint Ex. 1020 at 132-35.

Finally, the racial density map indicates that the district boundaries were drawn, in many places, to capture areas with a high proportion of voting-age African-Americans. Third Joint Stip. ¶¶ 300, 312.

When viewed in light of the strong statewide evidence, it is clear that Defendants drew the district boundaries in Wake County with the primary goal of creating two majority-black districts. The district-specific evidence supports our finding that race predominated in drawing House Districts 33 and 38.

17. House Districts 42 and 43

Enacted House Districts 42 and 43 are both majority-black districts located entirely within Cumberland County. Under the Benchmark Plan when drawn, Cumberland County did not contain any majority-black districts. Defs.’ Ex. 3001 at 308, 315.

Benchmark House District 42 had a BVAP of 45.11% under the 2000 census and 47.94% under the 2010 census. Third Joint Stip. ¶314. Enacted House District 42 has a BVAP of 52.56%, thus meeting the Chairs’ stated goal of drawing VRA districts above 50% BVAP. Id. ¶ 315. Benchmark House District 42 was underpopulated by 11,017 persons, according to the 2010 census. To remedy this, enacted House District 42 — which was located in roughly the same location as its predecessor — contains 137 fewer white persons and 9,681 more African-Americans than the benchmark. Id. ¶ 316; Pis.’ Ex. 2022 at 11.

House District 43 follows a somewhat different pattern, because although the benchmark was drawn with a BVAP of 48.69% based on the 2000 census, its BVAP had increased to 54.69% under the 2010 census. Third Joint Stip. ¶ 331. Therefore, the benchmark district already contained a sufficient *161black population to satisfy the Chairs’ 50%-plus-one goal. The challenge for Defendants was to keep the district’s BVAP above 50% even as they simultaneously increased the BVAP in neighboring House District 42 (all while adding population as necessary to meet the one person, one vote standard).38 Defendants accomplished that goal, drawing House District 43 with a BVAP of 51.45% while still creating a second majority-black district in Cumberland County. Id. ¶ 332; Joint Ex. 1004 at 147.

In Dr. Hotelier's initial study of the state’s demographics, he determined that two majority-black district could be drawn in this area of the state, and proceeded to do so. Defs.’ Ex. 3028 at 9. House Districts 42 and 43 were both identified as proposed VRA districts on the House VRA map, and were enacted without substantial modification to the location, shape, or BVAP of the proposed districts. Joint Ex. 1002 at 1, 99; Joint Ex. 1004 at 1,147; Pis.’ Ex. 2072 at 19-20.

While both districts contain a few odd appendages, House Districts 42 and 43 are relatively compact compared to many of the other challenged districts. House District 42 is less compact when compared to the benchmark on five of the eight compactness measures presented by Mr. Fairfax. Pis.’ Ex. 2094 at 45, 58. House District 43 is more compact than the benchmark on five of the eight measures. Id. But as stated previously, there is no requirement “that a district must be bizarre on its face before there is a constitutional violation.” Miller, 515 U.S. at 912, 115 S.Ct. 2475.

The evidence suggests that race was the predominant factor determining the contours of the enacted district. The most notable change in the shape of House District 42 is the removal of a large portion of the benchmark district in its northeastern corner. Pis.’ Ex. 2022 at 11. This area contains Fort Bragg, an Army base. Trial Tr. vol. II, 104:1-24 (Covington). The state explained in its Section 5 preclearance filing that House District 42 was established as a majority-black district, and “[tjhis was accomplished” by removing Fort Bragg, which would otherwise “dilute the minority voting strength of African-American voters in Cumberland County.” Joint Ex. 1024 at 16. In other words, removing Fort Bragg from the district allowed the General Assembly to enact House District 42 as a majority-black district.

The evidence further suggests that enacted House Districts 42 and 43 divide precincts and communities on the basis of race. House District 42 splits fifteen of its nineteen precincts. Third Joint Stip. ¶ 317; Joint Ex. 1004 at 23-24. House District 43 splits fifteen of the twenty-one precincts contained in that district. Third Joint Stip. ¶ 334; Joint Ex. 1004 at 24. Thus, House Districts 42 and 43 split 78.95% and 71.43% of their precincts, respectively. Roughly 67% of the voting-age African-Americans who reside in one of the split precincts in Cumberland County were assigned to either House District 42 or 43. Third Joint Stip. ¶ 335.

Since Fayetteville is a large city, it was necessarily split among multiple House districts. However, the evidence establishes that it was divided on the basis of race. House Districts 42 and 43, combined, contain 63.52% of the city of Fayetteville, but manage to grab 80.37% of Fayetteville’s voting-age African-American population. Id. ¶ 319; Joint Ex. 1004 at 109. Finally, the racial density maps indicate that many of the strange protrusions in the district reach out to capture areas with a high proportion of voting-age African-Americans. Third Joint Stip. ¶¶ 329, 345.

When viewed in light of the strong statewide evidence, it is clear that Defendants drew the district boundaries in Cumberland County with the primary goal of creating two majority-black districts. The district-specific evidence supports our finding that race predominated in drawing House Districts 42 and 43.

18. House District 48

Enacted House District 48 is a majority-black district along North Carolina’s southern border. The benchmark version of House District 48 had a BVAP of 45.46% under the *1622000 census and 45.56% under the 2010 census. Third Joint Stip. ¶ 347. Enacted House District 48 was drawn with a BVAP of 51.27%, thus achieving the Redistricting Chairs’ goal of drawing each VRA district above 50% BVAP. Id. ¶ 348; Joint Ex. 1004 at 147.

In Dr. Hofeller’s initial study of the state’s demographics, he determined that a majority-black district could be drawn in this area of the state, and proceeded to do so. Defs.’ Ex. 3028 at 8. The Chairs identified House District 48 as a proposed VRA district on the House VRA map, and the district was enacted without substantial changes to the shape, location, or BVAP of the proposed district. Joint Ex. 1002 at 1, 99; Joint Ex. 1004 at 1, 147.

Enacted House District 48 is located in roughly the same location as the benchmark district, although it incorporates Richmond County, which was previously in House District 66. Defs,’ Ex. 3001 at 326. Benchmark House District 48 was underpopulated by 13,018 persons according to the 2010 census. To remedy this, and increase the BVAP, the enacted district contains 12,908 more African-Americans and 6,751 more white people. Third Joint Stip. ¶ 349.

Visually, House District 48 is one of the most bizarre and sprawling districts in the Enacted House Plan. Joint Ex. 1004 at 1. The district is located in portions of Hoke, Robeson, Scotland, and Richmond Counties. The base of the district is a narrow strip of land along the border between North Carolina and South Carolina. See Trial Tr. vol. I, 78:18-23 (Blue) (describing portions of the base of the district as a “land bridge” between African-American populations). Three different arms reach north from this base: one that snakes through Robeson County, one that travels primarily through the eastern side of Scotland County before crossing into Hoke County and eventually veering into a small portion of Robeson County, and one that meanders erratically through Richmond County. House District 48 is significantly less visually compact than the benchmark district,39

Additionally, enacted House District 48 divides four counties and five municipalities. Joint Ex. 1004 at 1, 131. The evidence suggests that these divisions occurred largely along racial lines. Senator Dan Blue, who was bom and reared in Robeson County, explained that “[tjhere are ten [to] twelve towns in Robeson County, each having its own black population” and House District 48 “reaches most of them and sticks them into the district.” Trial Tr. vol. I, 42:25, 79:4-6. For instance, House District 48 includes 60.06% of the city of Ellerbe, but captures 96.24% of that city’s black voting-age population. Third Joint Stip. ¶ 352; Joint Ex. 1004 at 131. The district also includes 44.99% of the city of Hamlet, but manages to include 78.88% of the voting-age African-Americans in that city. Third Joint Stip. ¶ 354; Joint Ex. 1004 at 131. Similar patterns hold true for the cities of Laurinburg and Rockingham. Third Joint Stip. ¶¶355, 356; Joint Ex. 1004 at 131.

Further, House District 48 divides thirty-one precincts. Third Joint Stip. ¶ 350. The evidence suggests that the numerous precinct splits were motivated by race. Of the voting-age African-Americans residing in the district’s split precincts, 77.9% are assigned to House District 48. Id. ¶ 351. This information is consistent with what the racial density suggests: many of the unusual borders of the district capture areas with high proportions of voting-age African-Americans. Id. ¶ 366.

Despite this evidence, Defendants assert that the boundaries for House District 48 had to be “stretched to a greater length” because the district had to include an above-average population. Defs.’ Post-trial Findings at 138, The population in this district did have to be higher than average because it was placed in a twenty-county grouping that needed to include more populous districts in order to *163offset underpopulated districts elsewhere in the state. Trial Tr. vol. IV, 242:21-243:19 (Hofeller); Pis.’ Ex. 2085. However, there were many ways to draw a regularly shaped district in this region that would have contained the requisite population; those alternatives simply would have required compromising on the 50%-plus-one target. Additionally, to the extent that the one person, one vote standard influenced the drawing of lines, we must follow the Supreme Court’s instruction and view that as a background consideration that cannot negate the predominance of race in redistricting. See supra note 16.

Finally, Representative Lewis testified at trial that “[t]he shape of the lines in this particular VRA district” reflected an attempt to group together an African-American population. Trial Tr. vol. Ill, 221:13-18. In fact, he noted that the twenty-county group actually gave them more freedom to draw the district based on race, stating that “because the county grouping was so big, there was less of a restriction for how this particular ... minority population could be grouped and drawn.” Id. at 221:19-21.

In conclusion, the district-specific evidence provides abundant support for the finding that House District 48 was drawn to achieve the Chairs’ statewide goal of drawing VRA districts at 50%-plus-one. We find that race predominated in drawing House District 48.

19. House Districts 57, 58, and 60

Enacted House Districts 57, 58, and 60 are majority-black districts located entirely within Guilford County. Under the Benchmark Plan, Guilford County had two majority-black districts. Defs.’ Ex. 3001 at 316, 326, 333. During his initial review of the state’s demographics under the 2010 census, Dr. Hofeller determined that it was possible to increase the number of majority-black districts in Guilford County to three, in accordance with the proportionality goal and 50%-plus-one target. Defs.’ Ex. 3028 at 9. House Districts 57, 58, and 60, as enacted, have BVAPs of 50.69%, 51.11%, and 51.36%, respectively. Third Joint Stip. ¶¶ 369, 377, 389; Joint Ex. 1004 at 148. The Chairs identified these three districts as proposed VRA districts on the House VRA map, and they were enacted without any substantial modification to the proposed districts’ location, shape, or BVAP. Joint Ex. 1002 at 1, 99; Joint Ex. 1004 at 1, 148; Pis.’ Ex. 2072 at 21-23.40

It is challenging to compare these districts to their counterparts in the Benchmark Plan because the districts were moved and reshaped significantly in order to fit three majority-black districts in Guilford County. Joint Ex. 1004 at 1; Pis.’ Ex. 2022 at 14-16; Defs.’ Ex, 3001 at 326. For instance, House District 57 moved from the area west of downtown Greensboro to the northeastern portion of the city, with almost no discema-ble overlap. Pis.’ Ex. 2022 at 14. Thus, it is most useful to look at Guilford County as a whole.

In the Benchmark Plan, House District 58 had a BVAP of 53.35% under the 2000 census and 53.43% under the 2010 census. Defs.’ Ex, 3001 at 316, 333. House District 60 had a BVAP of 50.59% under the 2000 census and 54.36% under the 2010 census. Id. Of the remaining districts in Guilford County, the one with the highest BVAP was benchmark House District 59, which had a BVAP of 23.52% under the 2000 census, which grew to 30.15% under the 2010 census. Id. Benchmark House District 57 had a BVAP of 21.38% under the 2000 census and 29.93% under the 2010 census. Id. Therefore, increasing the BVAP in a third Guilford County district to 50%-plus-one required the Chairs to reconfigure all districts within the county, and actually reduce the BVAPs in House Districts 58 and 60 so that additional African-American communities could be added to enacted House District 57.

House Districts 57, 58, and 60 all became less visually compact when compared to the benchmark versions of those districts. However, they are all located in or near the city of Greensboro, an urban area, and therefore are not as bizarrely shaped as many of the other challenged districts. See Shaw II, 517 U.S. at 907, 116 S.Ct. 1894 (noting that evi*164dence suggesting some traditional districting criteria were “addressed” does not foreclose the conclusion that race predominated); Miller, 515 U.S. at 912-13, 115 S.Ct. 2475 (explaining that race can still be the predominant motive in a compact district). Of the three, House District 60 is the most bizarre on its face. House District 60 contains two substantial portions: one that captures the southwestern portion of Greensboro and one that captures part of the population of High Point. Joint Ex. 1004 at 133; Pis.’ Ex. 2022 at 16. These two poles are connected by a narrow land bridge made entirely of split precincts. Pis.’ Ex. 2022 at 16. That land bridge is made up almost completely of census blocks with less than 20% BVAP, which probably explains its narrowness. See Third Joint Stip. ¶¶ 376, 399 (racial density maps). On the eight compactness measures presented by Mi'. Fairfax, House Districts 57, 58, and 60 are less compact than the benchmark districts on eight, six, and four41 of the measures, respectively. Pis.’ Ex. 2094 at 24, 43.

Because of the size of the city of Greensboro, it necessarily had to be divided among multiple house districts, Joint Ex. 1004 at 110. However, the evidence suggests that the division of Greensboro was not race-neutral. House Districts 57, 58, and 60 together contain 70.67% of the city of Greensboro, but manage to capture 88.39% of Greensboro’s African-American voting-age population. Third Joint Stip. ¶ 373; Joint Ex. 1004 at 110. The remainder of the voting-age African-Americans in Greensboro are split among three other districts. Third Joint Stip. ¶ 373.

Additionally, House Districts 57, 58, and 60 divide fifteen, fifteen, and sixteen precincts, respectively. Id. ¶¶ 371, 378; Joint Ex. 1004 at 30-33. Of the total number of split precincts in Guilford County, 77.7% of the voting-age African-Americans who live in those split precincts were assigned to House Districts 57, 58, and 60, as opposed to the other three districts. Third Joint Stip. ¶372. This data suggests that precincts were split on the basis of race.

The racial density maps also indicate that the boundaries of these districts were drawn on the basis of race. Third Joint Stip. ¶¶ 376, 399. For instance, the map of House District 60 clearly shows that the two poles of the district encompass areas with a high proportion of voting-age African Americans, in both High Point and Greensboro. Id. Additionally, Yvonne Johnson, a long-time elected official in the city of Greensboro, testified that the portion of House District 57 that stretches like a tail into the eastern part of the county encompasses the “predominantly black community” of Sedalia. Trial Tr. vol. I, 200:14-201:2. This assertion is supported by the racial density map. Third Joint Stip. ¶ 376.

Overall, when viewed in conjunction with the strong statewide evidence of legislative intent, the district-specific evidence confirms that race predominated in drawing House Districts 57, 58, and 60.

20. House Districts 99, 102, and 107

Enacted House Districts 99, 102, and 107 are all majority-black districts located entirely within Mecklenburg County. Under the Benchmark Plan, Mecklenburg County contained only three districts with more than 40% BVAP, two of which were drawn above 50% BVAP. Third Joint Stip. ¶453; Defs.’ Ex. 3001 at 316-17, 326. Of the twelve total districts drawn entirely within Mecklenburg County, the Enacted Plan contains five districts that exceed 50% BVAP. Third Joint Stip. ¶454; Joint Ex. 1004 at 1, 148-49. Three of those five districts — House Districts 99, 102, and 107 — have been challenged in this case. Those three districts were drawn with BVAPs of 54.65%, 53.53%, and 52.52%, respectively. Joint Ex. 1004 at 148-49. The Chairs identified these three districts as proposed VRA districts on the House VRA map, and they were enacted without substantial modification to the proposed districts’ location, shape, or BVAP. Joint Ex. 1002 at 1, 99; Joint Ex. 1004 at 1, 148-49; Pls.’ Ex. 2072 at 24-26.42

Like in Guilford County, fitting additional majority-black districts in Mecklenburg *165County required the districts within the county to be significantly moved and reshaped. Therefore, the benchmark districts do not substantially correspond with the enacted districts. Pis.’ Ex. 2022 at 17-19. Additionally, since the districts are located in the urban area surrounding the city of Charlotte, they are more visually compact than many of the other challenged districts. The least visually compact of these three is House District 107, which is shaped like an arch, and stretches from downtown Charlotte on one side to the border of Cabarrus County on the other. Id. at 19.

Of the eight compactness measures presented by Mr. Fairfax, House Districts 99, 102, and 107 are less compact than their corresponding benchmark districts on three, eight, and four of the measures, respectively. Pis.’ Ex. 2094 at 24-25, 43-44. As noted previously, however, race can still predominate even in a relatively compact district. See Shaw II, 517 U.S. at 907, 116 S.Ct. 1894 (noting that evidence suggesting some traditional districting criteria were “addressed” does not foreclose the conclusion that race predominated); Miller, 515 U.S. at 912-13, 115 S.Ct. 2475 (explaining that race can still be the predominant motive in a compact district).

There are numerous indications that these districts were used to divide Mecklenburg County on the basis of race. The non-majority-black districts in Mecklenburg County have drastically lower BVAPs than the so-called VRA districts. See Answer ¶244. House District 88, for example, which shares its western boundary line with challenged House District 102, has a BVAP of only 7.94%. Joint Ex. 1004 at 1, 148. The boundary line between House Districts 102 and 88 includes split precincts, particularly in the northeastern portion of House District 102, which juts out irregularly from the rest of the district. Pis.’ Ex. 2022 at 18. On the other side of House District 102 sits House District 92, which was enacted with a BVAP of 18.18%. Joint Ex. 1004 at 1,148. Here again, the boundary — this time between House Districts 102 and 92 — has several split precincts, and includes an irregular appendage in the southwestern corner of House District 102, Pis.’ Ex. 2022 at 18. Similarly, the most irregular portion of the boundary for House District 99 is the only portion that separates it from a majority-white district: House District 103, which has a BVAP of 13.07%. Joint Ex. 1004 at 1, 148. That boundary also contains a number of precinct splits. Pis.’ Ex. 2022 at 17. House Districts 99, 102, and 107 divide seven, thirteen, and nine precincts, respectively. Third Joint Stip. ¶¶ 404, 419, 431.

Just as it racially divides Mecklenburg County, the Enacted House Plan also divides the city of Charlotte on the basis of race. The majority-black districts in Mecklenburg County, combined, contain 50.52% of Charlotte’s population, but manage to include 76.93% of the voting-age African-Americans who reside in that city. Id. ¶ 406; Joint Ex. 1004 at 106. The remainder of the African-American population in Charlotte is split among seven other districts. Third Joint Stip. ¶ 406.

Finally, the racial density maps indicate that the boundaries which separate House Districts 99 and 102 from neighboring predominantly white districts appear to trace areas with high proportions of voting-age African-Americans.43 Id ¶ 414.

When viewed in conjunction with the statewide, direct evidence of racial predominance, the district-specific evidence confirms that Defendants drew five majority-black VRA districts in Mecklenburg County with the predominant goal of meeting their racial targets. And at least for House Districts 99,102, and 107, the evidence suggests that traditional criteria were subordinated, where necessary, to meet this racial target. Therefore, we find that race predominated in drawing House Districts 99,102, and 107.

In sum, we find that race was the predominant factor motivating the drawing of all challenged districts.

*166III. Voting Rights Act Compliance as a Possible Defense

“Racial classifications are antithetical to the Fourteenth Amendment, whose ‘central purpose’ was ‘to eliminate racial discrimination emanating from official sources in the States.’ ” Shaw II, 517 U.S. at 907, 116 S.Ct. 1894 (quoting McLaughlin v. Florida, 379 U.S. 184, 192, 85 S.Ct. 283, 13 L.Ed.2d 222 (1964)). Consequently, when race is the predominant factor motivating the creation of a district, that district “cannot be upheld unless it satisfies strict scrutiny, our most rigorous and exacting standard of constitutional review.” Miller, 515 U.S. at 920, 115 S.Ct. 2475. Thus, the burden now shifts to Defendants to demonstrate that the challenged districts are narrowly tailored to serve a compelling state interest. Shaw II, 517 U.S. at 908, 116 S.Ct. 1894; Miller, 515 U.S. at 920, 115 S.Ct. 2475.

Defendants assert two compelling state interests for their race-based districting: compliance with Sections 2 and 5 of the VRA. The Supreme Court has yet to decide whether compliance with the VRA is a compelling state interest. Instead, the Court has assumed as much for the purpose of its analysis. See, e.g., Vera, 517 U.S. at 977,116 S.Ct. 1941 (“[W]e assume without deciding that compliance with the results test [of Section 2 of the VRA] ... can be a compelling state interest.”); Shaw II, 517 U.S. at 911, 116 S.Ct. 1894 (assuming but not deciding that VRA compliance can be a compelling interest). Thus, for the purpose of resolving this suit, we too assume, arguendo, that compliance with either Section 2 or Section 5 of the VRA can be a compelling state interest.44

However, even assuming such compelling interests, attempts at VRA compliance “cannot justify race-based districting where the challenged district was not reasonably necessary under a constitutional reading and application” of federal law. Miller, 515 U.S. at 921, 115 S.Ct. 2475. Thus, to satisfy strict scrutiny, Defendants must show that they had a “strong basis in evidence” or “good reasons to believe” that each of the challenged districts, as drawn, were required to comply with the VRA. Alabama, 135 S.Ct. at 1274. Narrow tailoring also requires that each district be drawn in a manner that actually remedies the potential VRA violation- Shaw II, 517 U.S. at 916, 116 S.Ct. 1894; cf. Vera, 517 U.S. at 979, 116 S.Ct. 1941 (“[A] district drawn in order to satisfy [the VRA] must not subordinate traditional districting principles to race substantially more than is ‘reasonably necessar/ to avoid [VRA] liability.”). We address Defendants’ defenses with respect to Sections 2 and 5 of the VRA in turn.

A. Section 2 as a Compelling Interest

Defendants principally argue that their predominant use of race was justified to avoid violating the VRA through vote dilution. As stated above, see supra section I.A, Section 2(a) of the VRA prohibits the imposition of any electoral practice or procedure that “results in a denial or abridgement of the right of any citizen ... to vote on account of race or color.” 52 U.S.C. § 10301(a). As relevant here, a Section 2 violation occurs if, “based on the totality of circumstances, it is shown that the political processes leading to nomination or election in the State or political subdivision are not equally open to participation by members of a [protected group] ... in that its members have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice,” Id. § 10301(b).

In Gingles, the Supreme Court established that a minority group alleging a Section 2 vote dilution claim must prove three threshold preconditions: first, “that [the minority group] is sufficiently large and geographically compact to constitute a majority in a single-member district”; second, “that it is politically cohesive”; and third, “that the white majority votes sufficiently as a bloc to *167enable it ... usually to defeat the minority’s preferred candidate.” 478 U.S. at 50-51, 106 S.Ct. 2752; see also Growe v. Emison, 507 U.S. 25, 40-41, 113 S.Ct. 1075, 122 L.Ed.2d 388 (1993) (affirming the applicability of the Gingles preconditions in the context of Section 2 challenges to single-member districts). “[0]nly when a party has established the Gingles requirements does a court proceed to analyze whether a violation has occurred based on the totality of the circumstances.” Strickland, 556 U.S. at 11-12, 129 S.Ct. 1231.

Because Defendants assert that each of the challenged districts was narrowly tailored to prevent a violation of Section 2, Defendants must establish that they had a “strong basis in evidence” for believing that the three Gin-gles factors were present in each of the districts at the time they were drawn. Vera, 517 U.S. at 978, 116 S.Ct. 1941 (“The State must have a ‘strong basis in evidence’ for finding that the threshold [Gingles] conditions for § 2 liability are present.”); Shaw II, 517 U.S. at 908 n. 4, 116 S.Ct. 1894 (“[T]he legislature must have ... a strong basis in evidence ... before it implements the [relevant racial] classification.” (emphasis added)); Smith v. Beasley, 946 F.Supp. 1174, 1210 (D.S.C.1996); cf. Wygant v. Jackson Bd. of Educ., 476 U.S. 267, 277, 106 S.Ct. 1842, 90 L.Ed.2d 260 (1986) (plurality opinion) (noting that, under strict scrutiny, a state must have “convincing evidence that remedial action is warranted” before implementing an affirmative action measure). Absent a strong basis in evidence for the three factors, Defendants would have had no reason to anticipate a potential Section 2 violation and therefore no reason to believe the race-based districting was necessary to comply with Section 2. See Harris, 159 F.Supp.3d at 623-24 (“A failure to establish any one of the Gingles factors is fatal to the defendants’ claim.”).

1. Defendants Never Analyzed Gingles’ Third Factor

The evidence in this case demonstrates that Defendants erred in drawing each of the challenged districts by failing to evaluate whether there was a strong basis in evidence for the third Gingles factor in any potential VRA district. We assume, without deciding, that Defendants had a strong basis in evidence for the first two Gingles factors regarding each challenged district. However, Defendants have failed to demonstrate that, for any challenged district, they had a strong basis in evidence for the third Gingles factor — racial bloc voting that, absent some remedy, would enable the majority usually to defeat the minority group’s candidate of choice. Gingles, 478 U.S. at 51, 106 S.Ct. 2752. This failure is fatal to their Section 2 defense.

“[R]acial bloc voting is a key element of a vote dilution claim.” Id. at 55, 106 S.Ct. 2752. However, not all racial bloc voting rises to a level that is cognizable within the meaning of Gingles’ third factor. “Racial bloc voting” or “racially polarized voting”45 refers to the circumstance in which “different races ... vote in blocs for different candidates.” Id. at 62, 106 S.Ct. 2752; see also id. at 58, 106 S.Ct. 2752 (characterizing “evidence that black and white voters generally prefer different candidates” as evidence of racially polarized voting). However, the third Gingles precondition requires racial bloc voting that is “legally significant” — that is, majority bloc voting at such a level that it enables the majority group “usually to defeat the minority’s preferred candidates.” Id. at 56, 106 S.Ct. 2752. To be sure, evidence of “especially severe” racially polarized voting, in which there are few majority-group “crossover” votes for the minority group’s preferred candidate, can help support finding the existence of Gingles’ third factor. See, e.g., LULAC, 548 U.S. at 427, 126 S.Ct. 2594 (explaining that Section 2 plaintiffs had demonstrated the third Gingles factor, in part, because the evidence indicated that racially polarized voting was “especially severe,” with 92% of Latinos voting against a candidate and 88% of non-Latinos voting for him). But a general finding regarding the existence of any racially polarized voting, no matter the level, is not enough.

Moreover, because “[minority] voters’ ability to elect representatives of *168their choice ... will vary from district to district according to a number of factors,” Gingles, 478 U.S. at 56, 106 S.Ct. 2752, “[generalized assumptions about the ‘prevalence of racial bloc voting’ do not qualify as a ‘strong basis in evidence,’ ” Harris, 159 F.Supp.3d at 623-24 (quoting Vera, 517 U.S. at 994, 116 S.Ct. 1941 (O’Connor, J., concurring)). The key inquiry under Gingles’ third factor, then, is whether racial bloc voting is operating at such a level that it would actually “ ‘minimize or cancel’ ... [minority] voters’ ability to elect representatives of their choice,” if no remedial district were drawn. Gingles, 478 U.S. at 56,106 S.Ct. 2752 (internal citation omitted).

Therefore, to have a strong basis in evidence for the third Gingles precondition, a legislature must give consideration to the actual and potential effect of bloc voting on electoral outcomes. See Lewis v. Alamance Cty., 99 F.3d 600, 608 (4th Cir.1996) (“[Assessing whether ‘the white majority votes sufficiently as a bloc to enable it ... usually to defeat the minority’s preferred candidate,’ [requires] at a minimum [considering] a representative cross-section of elections.”).

The evidence in this case demonstrates this is exactly what Defendants did not do. As explained below, Defendants never asked whether there was a strong basis in evidence that the “majority [was voting] sufficiently as a bloc to enable it ... usually to defeat the minority’s preferred candidate” before they drew the challenged districts. Gingles, 478 U.S. at 51,106 S.Ct. 2752.

It was revealed at trial that the Redistricting Chairs misconstrued what the third Gin-gles factor requires. Representative Lewis testified that he understood the third Gingles factor to be present so long as there is “some evidence of racially polarized voting.” Trial Tr. vol. Ill, 223:21-23 (“[T]he third cri-teri[on] in drawing the VRA [districts] is that you have to have some evidence of racially polarized voting ....”). Similarly, according to Senator Rucho, the third Gingles precondition would be satisfied if “racially polarized voting exist[s].” Trial Tr. vol. IV, 50:7-9 (“[T]he third [Gingles factor] is the ... requirement that there was evidence that racially polarized voting existed.”). Thus, during the 2011 redistricting, evidence of any racially polarized voting, regardless of its extent, was considered sufficient to show Gingles’ third precondition — racially polarized voting rising to the level that the majority group “vote[s] sufficiently as a bloc usually to defeat the minority’s preferred candidates.” Gingles, 478 U.S. at 56, 106 S.Ct. 2752.

In fact, the Redistricting Chairs testified that they never made any determination whether majority bloc voting existed at such a level that the candidate of choice of African-American voters would usually be defeated without a VRA remedy. Trial Tr. vol. IV, 52:8-18 (Rucho) (“[Q:] Did you do any examination ... as you were drawing up VRA districts as to whether black candidates of choice would be outvoted by whites unless you upped the BVAP? [A:] ... I don’t believe we did any of that_); see also id. at 33:9— 13 (confirming that VRA districts were drawn “regardless of voting patterns” in a given geographic area as long as there was evidence of racially polarized voting); Trial Tr. vol. Ill, 228:16-229:4 (Lewis) (explaining that VRA districts were drawn where “there was a demonstrated history of racially polarized voting”). Thus, Defendants never asked the right question to determine whether there was a strong basis in evidence for the third Gingles factor.

It is also clear that Dr. Hofeller never conducted an inquiry to determine whether racially polarized voting sufficient to enable the majority usually to defeat the candidate of choice of African-American voters was present in the challenged districts. Dr. Hofel-ler did not conduct any district effectiveness analysis46 prior to drawing the districts, Trial Tr. vol. V, 82:7-11 (Hofeller), nor did he perform a racial polarization analysis, id. at 82:1-5; Trial Tr. vol. IV, 213:3-5 (Hofeller). According to Dr. Hofeller, he drew the race-*169based districts without regard to whether African-American candidates of choice were actually being elected or defeated. Trial Tr. vol. V, 80:23-81:3, 88:5-12; see also Pis.’ Designated Deps. at 282-83 (Hofeller) (explaining that, when drawing the districts, he “was not making a judgment on what was required by Section 2”).

In sum, the evidence demonstrates that no analysis was conducted during the 2011 redistricting to determine whether there was a strong basis in evidence to believe the “majority [was voting] sufficiently as a bloc to enable it ... usually to defeat the minority’s preferred candidate.” Gingles, 478 U.S. at 51, 106 S.Ct. 2752. Thus, Defendants could not have determined any of the challenged districts to be reasonably necessary to cure a potential Section 2 violation. This fundamental oversight is fatal to Defendants’ Section 2 defense.

2. Evidence Proffered by Defendants

The foregoing discussion of the errors Defendants made in drawing the challenged districts demonstrates those districts were not drawn with a strong basis in evidence for the third Gingles factor. The evidence that Defendants put forth in support of them Section 2 defense is likewise insufficient. For the reasons stated below, we find that Defendants have failed to proffer evidence demonstrating they had a strong basis in evidence to fear Section 2 liability. Indeed, a review of Defendants’ arguments regarding this proffered evidence confirms their erroneous interpretation of the third Gingles precondition.

a. Racial Polarization Studies

Defendants have proffered two racial polarization reports which were considered during the 2011 redistricting process and which, they contend, provided a strong basis in evidence justifying their race-based choices. See, e.g., Defs.’ Post-trial Findings at 224-25.

First, Defendants rely on a report prepared by Dr. Ray Block and submitted to the General Assembly at a May 9, 2011, public hearing. See Defs.’ Ex. 3013-8 (Report by Dr. Ray Block, Jr., “Racially Polarized Voting in 2006, 2008, and 2010 in North Carolina State Legislative Contests”); Defs.’ Ex. 3013-6 at 9-10; Defs.’ Ex. 3013-7 at 3. The Block report examines data from a sampling of North Carolina Congressional and state legislative contests from 2006 to 2010 involving an African-American candidate. Using estimates derived from ecological inference methods,47 Dr. Block estimates “the proportion of African-American and non-Black voters in each electoral contest who preferred the candidate of color.” Defs.’ Ex. 3013-8 at 3. Based on these estimates, Dr. Block concludes that the voting patterns from the elections studied “suggest the presence of racially polarized voting” in the state. Id. at 4.

Second, Defendants proffer a June 14, 2011, report by Dr. Thomas Brunell. See Defs.’ Ex. 3033 (“Report on Racially Polarized Voting in North Carolina” by Thomas L. Brunell, Ph.D.). During the 2011 redistricting, the General Assembly retained Dr. Bru-nell to examine whether or not racially polarized voting exists in fifty-one North Carolina counties.48 Id. at 3. In his report, Dr. Brunell analyzes North Carolina precinct-level voting data principally derived from three elections. Subjecting this data to various scientific methods, including county-by-county bivariate regression analysis, Trial Tr. vol. IV, 134:13-19 (Brunell), Dr. Brunell concludes that there is “statistically significant racially polarized voting in 50 of the 51 counties” studied, Defs.’ Ex. 3033 at 3.

Contrary to Defendants’ contentions, the Block and Brunell reports do not establish a strong basis in evidence for Gingles’ third factor in any potential district.

First, while both reports conclude that there is evidence of racially polarized voting in North Carolina, neither report “speak[s]— *170one way or the other — to the effects of racially polarized voting,” i.e., to how racial polarization is affecting election outcomes in any geographic area. Rodriguez v. Pataki, 308 F.Supp.2d 346, 438 (S.D.N.Y.), aff'd, 543 U.S. 997, 125 S.Ct. 627, 160 L.Ed.2d 454 (2004); see id. (discussing racial bloc voting study without data on election outcomes as “incomplete and insufficient to address specific aspects of the third Gingles factor”). Indeed, the Block report does not mention specific election results at all. Defs.’ Ex. 3013-8. And the Brunell report does not indicate the prevailing candidate in the three principal elections examined. Defs.’ Ex. 3033; see also Trial Tr. vol. IV, 148:7-11; 149:21-24 (Brunell) (confirming that, for the purposes of his 2011 report, “it doesn’t matter ... if the candidate of choice of black voters wins [an] election or not.”).

Defendants’ reliance on the Brunell and Block reports’ generalized conclusions regarding racially polarized voting demonstrates them misunderstanding of Gingles’ third factor. As discussed above, the third Gingles inquiry is concerned only with “legally significant racially polarized voting,” which occurs when the “majority [group] votes sufficiently as a bloc to enable it ... usually to defeat the minority’s preferred candidate.” Gingles, 478 U.S. at 51, 55-56,106 S.Ct. 2752. On the other hand, the general term “racially polarized voting” is defined much more broadly and simply refers to when different racial groups “vote in blocs for different candidates.” Gingles, 478 U.S. at 62, 106 S.Ct. 2752. Dr. Brunell testified that his report’s reference to “statistically significant” racially polarized voting means only that the evidence of racially polarized voting cannot be attributed to “chance alone.” Trial Tr. vol. IV, 136:18-137:3; see Defs.’ Ex. 3033 at 4.

This crucial difference between legally significant and statistically significant racially polarized voting becomes clear when one considers the expansive set of circumstances the latter term can describe. It characterizes elections where 90% of a minority group’s voters but only 10% of the majority group’s voters prefer a certain candidate. Yet, the label applies equally well where there is only a “minimal degree of polarization,” such as when 51% of a minority group’s voters prefer a candidate and 49% of the majority group’s voters prefer that same candidate (so long as that result is statistically significant). Trial Tr. vol. Ill, 17:5-10 (Lichtman); see also Pis.’ Designated Deps. at 61-62 (Brunell).

For example, Dr. Brunell concluded that there was statistically significant racially polarized voting in every county he examined except one.49 Defs.’ Ex. 3033 at 3. Yet, the county-specific non-African-American crossover voting estimates for the three principal elections he analyzed run the gamut.50 See, e.g., id. at 7 (indicating non-African-American crossover voting ranging from 4.7% (Greene County) to 59.2% (Durham County) in the 2008 Democratic Presidential Primary Election); id. at 12-13 (reporting non-African-American crossover voting ranging from 20.1% (Cleveland County) to 61.7% (Robeson County) in the 2004 State Auditor Election).

Even more strikingly, the estimates in the Block report indicate that, in thirty-three out of the fifty-three elections Dr. Block studied, African-American and non-African-American voters preferred the same candidate. Defs.’ Ex. 3013-8 at 6-8 (Tables 1-3). That is, in thirty-three of the elections, a majority of non-African-American voters preferred the African-American voters’ candidate of choice. Id, Such data cannot be construed as conclusive evidence of the third Gingles factor— *171bloc voting causing African-American voters’ candidate of choice to usually be defeated.

Finally, Defendants make no argument that the differing levels of non-African-American crossover voting in the elections encompassed in the polarization studies informed their decision to draw the challenged districts,51 nor have they made any effort to link such evidence to other district-specific data reflecting that the minority group’s candidates of choice were usually defeated by majority bloc voting. Had Defendants done so, perhaps this would be a different case.

Here, the evidence demonstrates that the Defendants — consistent with the misinterpretation that any racially polarized voting can constitute a strong basis in evidence for Gin-gles’ third factor — applied the 50%-plus-one rule across the state without regard to the differing levels of crossover voting reported in either racial polarization study.52 See, e.g„ Trial Tr. vol. IV, 33:9-13 (Rucho) (“[Q:] And you applied the 50 percent rule regardless of voting patterns in the county; correct? [A:] The fact that there was racially polarized voting that was clearly outlined ... we felt that was what our responsibility was.”); Defs.’ Ex. 3028 at 23 (Hofeller) (explaining reliance on the “assumption” and “universally accepted conclusion” that racial bloc voting existed in the state of North Carolina, a conclusion which was “confirmed by Dr. Bru-nell’s subsequent study.” (emphasis added)).

Accordingly, we conclude that the two racial polarization studies on which Defendants rely fail to demonstrate a strong basis in evidence justifying the challenged districts as drawn.

b. Election Outcomes

The second type of evidence Defendants have put forth as justification for the race-based challenged districts is evidence of African-American candidates losing elections.

Defendants’ post-trial briefing identifies some specific elections under the Benchmark Plans in which African-American candidates were defeated. See, e.g., Defs.’ Post-trial Findings at 21-23, 122, 152, 159, 167. Yet, Defendants make no coherent arguments explaining if and how these losses influenced their drawing of particular districts or demonstrating how these occasional election losses establish that, absent a Section 2 remedy, the majority can usually defeat the minority group’s candidate of choice in the specific districts.

In fact, there are only two challenged districts — Senate Districts 5 and 21 — where we can piece together from the evidence an argument that losses by African-American candidates provided even partial motivation for increasing the BVAP to make a particular district majority-black. See Joint Ex. 1005 at 3 (June 17, 2011, Joint Statement by Redistricting Chairs) (suggesting that a loss by an African-American candidate in Senate District 5 motivated the creation of a majority-black district); Joint Ex. 1015 at 93 (July 25, 2011, Senate Floor Session) (same); Joint Ex. 1023 at 15-16 (Senate 2011 Section 5 Submission) (discussing changes to Senate District 21 in the Enacted Plan in the context of past election results); Trial Tr. vol. IV, 27:11-29:9 (Rucho) (same).

But Defendants have provided no explanation for how these losses informed a district-specific assessment regarding the third Gin-gles factor. For example, with respect to Senate District 5, Defendants have not explained why a single loss by Senator Don Davis provided a strong basis in evidence to believe that racially polarized voting in that district was so severe that the African-American voters’ candidate of choice would usually be defeated unless Senate District 5 was drawn to be majority-black. See Defs.’ Ex. *1723020 at 10 (indicating that African-American Democrat Don Davis won Senate District 5’s general election in 2008, but was defeated in 2010).

Additionally, Defendants’ citation to election losses by African-American candidates, without further explanation, reveals a reliance on a faulty premise: that the African-American voters’ candidate of choice will always be African-American.53 However, operating on the basis of such an assumption is improper, as it is clear that a non-minority candidate can be the candidate of choice of a minority group. Lewis, 99 F.3d at 607 (“[T]he minority-preferred candidate may be either a minority or a non-minority.”); Rodriguez, 308 F.Supp.2d at 441 (“The question is not whether Latinos can elect a preferred candidate who is Hispanic, but a preferred candidate period.”); see LULAC, 548 U.S. at 444-5, 126 S.Ct. 2594 (indicating that testimony that an “Anglo” candidate was the favored candidate of the African-American community and served that community’s interests supported a finding that the “Anglo” candidate was African-American voters’ candidate of choice, but holding that the district court did not err in crediting contradictory testimony to reach the opposite conclusion).

Moreover, even if we accept, for the purpose of argument, Defendants’ assumption that an African-American candidate will always be the African-American voters’ candidate of choice, Defendants nonetheless seem to have ignored the elections showing African-American candidates’ success when they drew the challenged districts. For example, African-American candidates were elected in 2004, 2006, 2008 and 2010 (that is, every election held under the Benchmark Plans) in benchmark Senate Districts 4, 14, 20, 28, 38 and 40 and benchmark House Districts 5, 12, 21, 29, 31, 42 and 48. Defs.’ Ex. 3020 at 10-11, 13-14, 21-25.54 Each of the aforementioned districts had a BVAP below 50% according to 2000 and 2010 census data. Defs.’ Ex. 3000 at 151, 158; Defs.’ Ex. 3001 at 315, 332. Yet, for each of these districts, Defendants used race to increase the BVAP percentage to 50%-plus-one in the Enacted Plans.55 See supra section II.B. It would therefore be difficult to accept any assertions on Defendants’ part that there was a strong basis in evidence to believe that these challenged districts were necessary, as drawn, to remedy racial bloc voting that would “usually be able to defeat” African-American voters’ candidate of choice. Gingles, 478 U.S. at 49,106 S.Ct. 2752.

*173During the redistricting process, the Redistricting Chairs’ fellow legislators repeatedly asked the Chairs why such BVAP increases would be necessary to comply with Section 2. For example, at a House Redistricting Committee Meeting, Representative Marvin Lucas, an African-American legislator representing House District 42, Defs.’ Ex. 3020-14 at 3, asked why there was a “need to increase the [BVAP] percentages” in districts like his own “where ... history tends to show that there has been no problem” electing African-American candidates, Joint Ex. 1019 at 7. In response, Representative Lewis simply stated: “The VRA districts that appear in [the proposed plan] were all drawn at ... a 50 percent plus one[ ] level to foreclose the possibility of any Section 2 lawsuit.” Id. at 8. Similarly, when confronted with a question about why a BVAP percentage increase was necessary in a particular district even though African-American candidates were already winning, Senator Rucho replied: “We’re just following the law as devised by the Strickland decision for a minority-majority [sic] district_” Joint Ex. 1013 at 40-41.

Suffice it to say that Defendants knew they were increasing the BVAP in districts where African-American candidates, who were purportedly also the African-American voters’ candidates of choice, were already consistently winning under the Benchmark Plans. We can only conclude that such information was irrelevant to them when it came to determining the existence of Gingles’ third precondition and applying their 50%-plus-one rule.

In their post-trial briefing, Defendants have also cited those instances where an African-American candidate elected under the benchmark version of a challenged district was uncontested, an incumbent, or a more successful fundraiser than his or her opponent. What Defendants have not done is explain how any of this information is relevant to justifying their race-based districting. Defendants’ citations to fundraising numbers and incumbency status do nothing to demonstrate that the challenged districts were drawn with a strong basis in evidence that the white majority voted as a bloc to usually defeat the candidate of choice of African-American voters.

Lacking such explanations, we take Defendants at their word: the 50%-plus-one rule was applied to create majority-black districts, including the challenged districts, “whenever] it [was] possible to do so,” Joint Ex. 1007 at 5, without any district-specific determination that racially polarized voting was significant enough to enable the majority to usually defeat the candidate of choice of African-American voters, see Trial Tr. vol. IV, 52:8-13 (Rucho).

c. Defendants’ Other Evidence

The remaining evidence Defendants proffer in support of their Section 2 defense for the challenged districts warrants little attention.

Defendants point to lay testimony given during the 2011 redistricting, including testimony from voting rights advocates, to the effect that racially polarized voting existed in the state and that majority-black districts were still needed, see Defs.’ Ex. 3015A, Defs.’ Ex. 3013-7; the history and locations of prior “VRA districts” enacted by the General Assembly; alternative redistricting plans submitted during the 2011 redistricting; a law review article discussing the state’s history of VRA litigation, see Defs.’ Ex. 3013-9; and a letter from the University of North Carolina (UNC) School of Government discussing redistricting, Defs.’ Ex. 3014-11.

While we would not dispute that some of this information is relevant and should be considered during legislative districting, none of the evidence Defendants have cited — without additional proof and district-specific analysis — can constitute a strong basis in evidence demonstrating that any of the challenged districts were reasonably necessary as drawn to avoid a Section 2 violation. For instance, the lay testimony Defendants cite is overwhelmingly general, and any evidence regarding Gingles’ third factor in any particular district is sparse to non-existent.

Moreover, Section 2 does not force the states to perpetuate race-based districts simply because they may have been necessary in the past, or because advocates lobby for them. See Strickland, 556 U.S. at 23-24, *174129 S.Ct. 1231 (“Our holding also should not be interpreted to entrench majority-minority districts by statutory command, for that, too, could pose constitutional concerns.”). In any event, the Enacted Plans include a substantially higher number of majority-black districts as compared to the Benchmark Plans or any alternative plan that was proposed. Compare Defs.’ Ex. 3000 at 191, and Defs.’ Ex. 3001 at 500-02, with Defs.’ Ex, 3000 at 151,169, 201, 212, and Defs.’ Ex. 3001 at 315-17, 356-58, 428-30, 452-54. See also supra section I.B.

Additionally, the law review article that Defendants proffer, while helpful context, is “no substitute for proof [or a strong basis in evidence] that [legally significant] bloc voting [is] oecur[ring]” in a particular district. Growe, 507 U.S. at 42, 113 S.Ct. 1075. The article discusses the history of voting rights litigation in North Carolina, but it makes no findings as to whether the third Gingles precondition is present in any particular areas of North Carolina today. See Anita S. Earls et. al., Voting Rights in North Carolina: 1982-2006. 17 S. Cal. Rev. L. & Soc. Just. 577 (2008). In fact, the article does not discuss Gingles’ third precondition at all. Id.

Likewise, the letter from the UNO School of Government does not provide Defendants with a strong basis in evidence to fear Section 2 liability. The letter itself makes that clear: “recent analysis of voting patterns and the other Section 2 elements would be necessary to assert with any confidence that a Section 2 violation might be found in a particular part of the state today.” Defs.’ Ex. 3014-11 at 6.

In summary, the testimony in this litigation demonstrates that, when drawing the challenged districts, Defendants made no district-specific assessment regarding the third Gingles factor (as properly understood). Moreover, Defendants have failed to establish during this litigation that the challenged districts were justified by a strong basis in evidence. We therefore conclude that the challenged districts were not narrowly tailored to comply with Section 2.

B. Section 5 as a Compelling Interest

Finally, we turn to whether the challenged districts were narrowly tailored to comply with Section 5 of the VRA. Section 5 “prohibits a covered jurisdiction from adopting any change that ‘has the purpose of or will have the effect of diminishing the ability of [the minority group] to elect their preferred candidates of choice.’ ” Alabama, 135 S.Ct. at 1272 (alteration in original) (quoting 52 U.S.C. § 10304(b)). In other words, this section of the VRA prohibits any redistricting “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).

As an initial matter, we note that eleven of the challenged districts do not include any county, in whole or in part, that was covered by Section 5 in 2011, and therefore those districts could not have been drawn to remedy a Section 5 violation.56

With regard to the challenged districts that were covered by Section 5, we conclude that Defendants have not put forth a strong basis in evidence that any of those districts were narrowly tailored to avoid retrogression.

To satisfy strict scrutiny, “the remedy [must be] narrowly tailored to the asserted end.” Shaw II, 517 U.S. at 915, 116 S.Ct. 1894; see also id. at 916,116 S.Ct. 1894 (“[T]he legislative action must, at a minimum, remedy the anticipated violation or achieve compliance to be narrowly tailored.”). In this context, a district survives strict scrutiny if Defendants put forth a strong basis in evidence that their race-based redistricting decision was reasonably necessary to comply with Section 5, i.e., to prevent “retrogression in respect to racial minorities’ ‘ability ... to elect their preferred candidates of choice.’ ” Alabama, 135 S.Ct. at 1263 (quoting 52 U.S.C. § 10304(b)). “A reapportionment plan [is] not ... narrowly tailored to the goal of avoiding retrogression if the State went be*175yond what was reasonably necessary to avoid retrogression.” Shaw I, 509 U.S. at 655, 113 S.Ct. 2816. In other words, Section 5 does not “give covered jurisdictions carte blanche to engage in racial gerrymandering in the name of nonretrogression,” Id.

The Supreme Court also has made clear that Section 5 “does not require a covered jurisdiction to maintain a particular numerical minority percentage.” Alabama, 135 S.Ct. at 1272. “Rather, [Section] 5 is satisfied if minority voters retain the ability to elect their preferred candidates.” Id. at 1273. Therefore, states should not rely “upon a mechanically numerical view as to what counts as forbidden retrogression.” Id. As the Supreme Court stated in Alabama:

[T]he legislature asked the wrong question with respect to narrow tailoring. They asked: “How can we maintain present minority percentages in majority-minority districts?” But given § 5’s language, its purpose, the Justice Department Guidelines, and the relevant precedent, they should have asked: “To what extent must we preserve existing minority percentages in order to maintain the minority’s present ability to elect the candidate of its choice?”

Id. at 1274.

Here, Defendants surely failed to ask the right question. Instead, they drew every “VRA district” at 50%-plus-one BVAP or higher, regardless of whether the benchmark BVAP was 21% or 55%, and regardless of whether a BVAP of 50%-plus-one was reasonably necessary “to maintain the minority’s present ability to elect the candidate of its choice.” Id.; see Trial Tr. vol. IV, 33:4-13 (Rucho) (“[Q:] [Y]ou applied the 50 percent rule in Section 2 counties and Section 5 counties alike, didn’t you? [A:] ... When we had Voting Rights Act districts where we could and achieved them, yes, sir, that was what was done. [Q:] And you applied the 50 percent rule regardless of voting patterns in the county; correct? [A:] The fact that there was racially polarized voting that was clearly outlined ... we felt that was what our responsibility was.”).

Alabama makes clear that such a “mechanically numerical view” is not narrowly tailored to avoid retrogression. 135 S.Ct. at 1273; see also Harris, 159 F.Supp.3d at 627-28 (holding that North Carolina “legislators had no basis — let alone a strong basis — to believe that an inflexible racial floor of 50 percent plus one person was necessary” to comply with Section 5 in the challenged district); Page, 2015 WL 3604029, at *18 (“The legislature’s use of a BVAP threshold, as opposed to a more sophisticated analysis of racial voting patterns, suggests that voting patterns in the [challenged district] were not considered individually.” (internal quotations omitted)).57 Further, this racial target was applied to increase some districts’ BVAPs by over twenty percent, despite the fact that such a dramatic increase could never be required to prevent “retrogression,” See Vera, 517 U.S. at 983, 116 S.Ct. 1941 (“The problem with the State’s [Section 5] argument is that it seeks to justify not maintenance, but substantial augmentation, of the African-American population percentage in [the challenged district].”); Shaw I, 509 U.S. at 655, 113 S.Ct. 2816.

Although “we do not insist that a legislature guess precisely what percentage reduction a court or the Justice Department might eventually find to be retrogressive,” the legislature must have a “strong basis in evidence” to support its use of racial classifications in redistricting. Alabama, 135 S.Ct. at 1273-74. For instance, the Alabama Court noted with approval the Justice Department’s Section 5 redistricting guidelines, which stated that a determination of the BVAP required to avoid retrogression under Section 5 should be based on a “functional analysis of the electoral behavior within the particular jurisdiction or election district.” Id. at 1272 (quoting Guidance Concerning Redistricting Under Section 5 of the Voting Rights Act, 76 Fed. Reg. 7470, 7471 (Feb. 9, 2011)). Here, Defendants have put forth no evidence that they performed any analysis to determine the appropriate BVAP for the *176challenged districts. In fact, they acknowledge that they used the 50%-plus-one target for each of the districts covered by Section 5. Defs.’ Post-trial Findings at 228. Therefore, Defendants have failed to show a “strong basis in evidence” that the BVAP for each challenged district was reasonably necessary to avoid retrogression.

In determining whether redistricting plans were narrowly tailored to comply with Section 5, the Supreme Court has also repeatedly rejected “maximization” policies that require the state to draw majority-minority districts wherever possible, or to achieve a pre-determined number of districts. Miller, 515 U.S. at 924r-25, 115 S.Ct. 2475; Shaw II, 517 U.S. at 913,116 S.Ct. 1894.

In Shaw II, for example, the state legislature tried to follow the Justice Department’s instruction that when “you have twenty-two percent black people in this State, you must have as close to twenty-two percent black Congressmen, or black Congressional Districts in this State.” 617 U.S. at 913,116 S.Ct. 1894. That essentially describes the policy of “rough proportionality” Defendants applied throughout this redistricting. See supra section II.A.1.C. But as the Supreme Court has stated, such a proportionality target, pursued without any district-specific analysis, “is not pi’operly grounded in § 5,” Shaw II, 517 U.S. at 913, 116 S.Ct. 1894, and so cannot justify the districts drawn in this ease.

In sum, Defendants have not shown a strong basis in evidence for their conclusion that the race-based redistricting in this case — in particular them use of the mechanical 50%-plus-one target and their proportionality goal — was reasonably necessary to avoid a Section 5 violation.

Defendants nevertheless assert that “[a] challenged district furthers a compelling interest if it was reasonably necessary to obtain preelearance of [a redistricting] plan” or “to avoid preclearance objections.” Defs.’ Post-trial Br. at 7, EOF No. 116 (internal quotations omitted). However, the Supreme Court has made quite clear that avoiding preclearance objections cannot be a compelling interest justifying the use of racial classifications. In Miller, even when the creation of a challenged district was explicitly “required in order to obtain preclearance,” the Supreme Court “d[id] not accept the contention that the State has a compelling interest in complying with whatever preclearance mandates the Justice Department issues.” 515 U.S. at 921-22, 115 S.Ct. 2475. Instead, the Miller Court struck down a challenged district because the use of race was not “reasonably necessary under a constitutional reading and application” of the VRA. Id. at 921,115 S.Ct. 2475.

Obtaining preelearance was not a compelling interest in Miller, and it is even less of one here. Defendants in this case point to no evidence that the Justice Department “required” the state to draw the challenged districts, as was the case in Miller, or that the Justice Department issued “demands” for additional majority-black districts, as it did in Shaw II, 517 U.S. at 911, 116 S.Ct. 1894. Rather, Defendants wish for us to insulate them from constitutional review for any race-based classification they unilaterally determined might expedite preclearance.58 That we cannot do.

***

In conclusion, we hold that Defendants have not carried their burden to show that each of the challenged districts was supported by a strong basis in evidence and narrowly tailored to comply with either Section 2 or Section 5. Therefore, all districts challenged in this case violate the Equal Protection Clause and are unconstitutional.

IV. Remedy

Having found that twenty-eight districts in the Enacted Plans are racial gerrymanders in violation of the Equal Protection Clause, we must now address the proper remedy. Plaintiffs have asked for an immediate injunction blocking the use of the unconstitutional districts in any future elections. We agree that these unconstitutional, challenged districts have already caused Plaintiffs sub*177stantial stigmatic and representational injuries, and that Plaintiffs are entitled to vote under constitutional districting plans as soon as possible.

However, we are also cognizant that the timing for the implementation of injunctive relief is particularly delicate in this case. The next general elections for the North Carolina House and Senate are scheduled to take place in November 2016, less than three months from now. In addition, the 2016 House and Senate primary elections were already held under the challenged plans on March 15,2016.

The Supreme Court has stated that:

[U]nder certain circumstances, such as where an impending election is imminent and a State’s election machinery is already in progress, equitable considerations might justify a court in withholding the granting of immediately effective relief in a legislative apportionment case, even though the existing apportionment scheme was found invalid.

Reynolds v. Sims, 377 U.S. 533, 585, 84 S.Ct. 1362, 12 L.Ed.2d 506 (1964).

Moreover, “because it is the domain of the States, and not the federal courts, to conduct apportionment,” Voinovich v. Quilter, 507 U.S. 146, 156, 113 S.Ct. 1149, 122 L.Ed.2d 500 (1993), we must provide the North Carolina General Assembly with a “reasonable opportunity” to draw remedial districts in the first instance, Wise v. Lipscomb, 437 U.S. 535, 540, 98 S.Ct. 2493, 57 L.Ed.2d 411 (1978); see also id. at 539, 98 S.Ct. 2493 (“[Rjedistricting ... is a legislative task which the federal courts should make every effort not to pre-empt.”). In addition, remedying such a large number of unconstitutional districts will likely require changes to districts this decision has not directly rendered invalid.

Based on the schedules put forth by the parties in their post-trial briefing, we regrettably conclude that due to the mechanics of state and federal election requirements, there is insufficient time, at this late date, for: the General Assembly to draw and enact remedial districts; this Court to review the remedial plan; the state to hold candidate filing and primaries for the remedial districts; absentee ballots to be generated as required by statute; and for general elections to still take place as scheduled in November 2016.

When “[n]ecessity” so requires, the Supreme Court has “authorized District Courts to order or to permit elections to be held pursuant to apportionment plans that do not in all respects measure up to ... constitutional requirements.” Upham v. Seamon, 456 U.S. 37, 44, 102 S.Ct. 1518, 71 L.Ed.2d 725 (1982). After careful consideration, and with much reluctance, we conclude that necessity demands such a result today. We decline to order injunctive relief to require the state of North Carolina to postpone its 2016 general elections, as we believe such a remedy would cause significant and undue disruption to North Carolina’s election process and create considerable confusion, inconvenience, and uncertainty among voters, candidates, and election officials. Instead, like other courts confronted with similarly difficult circumstances, we will allow the November 2016 elections to proceed as scheduled under the challenged plans, despite their unconstitutionality. See, e.g., Page v. Va. State Bd. of Elections, 58 F.Supp.3d 533, 554 (E.D.Va.2014), vacated on other grounds sub norm, Cantor v. Personhuballah, — U.S.-, 135 S.Ct. 1699, 191 L.Ed.2d 671 (2015); Vera v. Richards, 861 F.Supp. 1304, 1351 (S.D.Tex.1994), aff'd sub nom., Bush v. Vera, 517 U.S. 952, 116 S.Ct. 1941, 135 L.Ed.2d 248 (1996).

Nonetheless, Plaintiffs, and thousands of other North Carolina citizens, have suffered severe constitutional harms stemming from Defendants’ creation of twenty-eight districts racially gerrymandered in violation of the Equal Protection Clause. These citizens are entitled to swift injunctive relief.

Therefore, we hereby order the North Carolina General Assembly to draw remedial districts in their next legislative session to correct the constitutional deficiencies in the Enacted Plans. By separate order, we will direct the parties to file supplemental briefs on an appropriate deadline for such action by the legislature, on whether additional or other relief would be appropriate before the *178regularly scheduled elections in 2018, and, if so, the nature and schedule of that relief.

V. Conclusion

Section 2 of the VRA continues to play an important role in redistricting, and legislatures must undertake a district-specific analysis to identify and cure potential Section 2 violations, Our decision today should in no way be read to imply that majority-black districts are no longer needed in the state of North Carolina, Nor do we suggest that majority-black districts could not be drawn— lawfully and constitutionally — in some of the same locations as the districts challenged in this case. Rather, our holding today is attributable primarily to the explicit and undisputed methods that the General Assembly employed in the construction of these districts, and to the inadequacy of the district-specific evidence and arguments put forth by Defendants in this case.

For instance, if during redistricting the General Assembly had followed traditional districting criteria and, in doing so, drawn districts that incidentally contained majority-black populations, race would not have predominated in drawing those districts. See Shaw I, 509 U.S. at 646, 113 S.Ct. 2816 (“[Rjace consciousness does not lead inevitably to impermissible race discrimination.... [W]hen members of a racial group live together in one community, a reapportionment plan that concentrates members of the group in one district and excludes them from others may reflect wholly legitimate purposes.”). In this case, for example, Plaintiffs did not even challenge House Districts 23 and 27, which are reasonably compact majority-black districts that follow county lines. See Defs.’ Ex. 3001 at 494, 500.59

Similarly, if the General Assembly had demonstrated a strong basis in district-specific evidence that the “majority votes sufficiently as a bloc to enable it ... usually to defeat the minority’s preferred candidate,” Gingles, 478 U.S. at 51, 106 S.Ct. 2752, and that the other requirements for a Section 2 violation were present in a particular area, the State could have drawn an appropriately tailored remedial district. Evidence of a potential Section 2 violation may exist in some parts of the state, and if such evidence is properly examined and demonstrated, it certainly could justify future majority-minority districts.

But the General Assembly did not do any of this in carrying out its mechanical approach to districting. Further, although the legislature repeatedly identified certain cases as key in its purported attempt to draw maps that would “survive any possible legal challenge,” Joint Ex. 1007 at 1, it misinterpreted parts of these cases, see, e.g., Strickland, 556 U.S. at 24, 129 S.Ct. 1231 (“In areas with substantial crossover voting it is unlikely that the plaintiffs would be able to establish the third Gingles precondition — bloc voting by majority voters. In those areas majority-minority districts would not be required in the first place.”). The legislature also failed to heed the cases that would have helped to prevent this very litigation, See, e.g., Shaw I, 509 U.S. at 648, 113 S.Ct. 2816 (“[A] racial gerrymander may exacerbate the very patterns of racial bloc voting that majority-minority districting is sometimes said to counteract.”); Miller, 515 U.S. at 927-28,115 S.Ct. 2475 (“It takes a shortsighted and unauthorized view of the Voting Rights Act to invoke that statute, which has played a decisive role in redressing some of our worst forms of discrimination, to demand the very racial stereotyping the Fourteenth Amendment forbids”).

This state’s citizens have the right to vote in districts that accord with the Constitution. We therefore order that new maps be drawn that comply with the Constitution and the Voting Rights Act.

SO ORDERED, this the 11th day of August, 2016.

. In reaching this conclusion, we make no finding that the General Assembly acted in bad faith or with discriminatory intent in drawing the challenged districts, which were precleared by the Justice Department pursuant to Section 5 of the VRA. Nor do we consider whether the challenged districts involved any impermissible "packing” of minority voters, as Plaintiffs acknowledge that they bring no such claim. Finally, we do not reach the issue of whether majority-minority districts could be drawn in any of the areas covered by the current districts under a proper application of the law. See infra Part V.

. As used in this opinion, and in accordance with the statistics utilized by Defendants throughout the redistricting process, "BVAP” refers to the "total black” portion of the voting-age population, i.e., the portion that is "any-part black.”

. The benchmark districts corresponding to the House districts challenged in this case were not changed between the 2003 and 2009 House redistricting plans. Thus, for the purposes of this opinion, the term "Benchmark House Plan” accurately refers to both the 2003 and 2009 House redistricting plans.

. The House also appointed Representatives Nelson Dollar and Jerry Dockham as redistricting chairs, but Representative Lewis was the "senior chair and the manager of the process” in the House. Trial Tr. vol. Ill, 119:1-4 (Lewis).

. The Chairs — and not any court or regulator— designated these "VRA districts” as such. The Chairs' use of the term "VRA district” reflected their stated goal of complying with the VRA. Throughout this opinion, our use of the Chairs’ term "VRA district” in no way indicates this court's view, one way or the other, regarding whether the VRA required such districts.

. Throughout this opinion, "Senate VRA map” refers to "Rucho Senate VRA Districts.” Joint Ex. 1001. "House VRA map” refers to "Lewis House VRA- — Corrected,” Joint Ex. 1002, which was released June 21 and reflects an immaterial change to the June 17 map, Defs.' Ex. 3013-1 at 3; infra.

.Together, we refer to the Enacted House Plan and the Enacted Senate Plan as the "Enacted Plans.”

. Plaintiffs collectively reside in each of the challenged districts. Second Joint Stip. ¶¶ 35-65.

. We reject Defendants' contention that Plaintiffs’ claim is barred by res judicata as a result of the Supreme Court of North Carolina’s decision in Dickson, 368 N.C. 481, 781 S.E.2d 404. Under North Carolina law, which governs here, see Marrese v. Am. Acad, of Orthopaedic Surgeons, 470 U.S. 373, 375, 105 S.Ct. 1327, 84 L.Ed.2d 274 (1985), the doctrine of res judicata applies only where the parties in a later case are the same as or in privity with the parties in a prior case, Williams v. Peabody, 217 N.C.App. 1, 719 S.E.2d 88, 94 (2011), Recognizing none of the Plaintiffs in this action was either a plaintiff in the Dickson litigation or in privity with one, Defendants argue that the exception to privity recognized in Thompson v. Lassiter, 246 N.C. 34, 97 S.E.2d 492 (1957), applies. However, Defendants have not produced sufficient evidence to prove the elements of the Lassiter exception. See Williams, 719 S.E.2d at 94-95.

. In light of our conclusion that Defendants failed to demonstrate a strong basis in evidence for any potential Section 2 violation, see infra section III.A, we need not decide here whether this interpretation of Strickland was proper.

. The final two requirements are that multi-member districts be avoided unless "necessary to advance a compelling governmental interest” and, generally, that any departures from "the legal requirements set forth herein" occur "only to the extent necessary to comply with federal law.” Stephenson II, 582 S.E,2d at 250-51.

. We express no view as to whether the Stephenson cases require that VRA districts be drawn first both in priority and in time. As the record shows, and as is sufficient for our analysis, the Chairs interpreted those cases to require that VRA districts be drawn before all other districts.

. The Chairs also came close to conflating rough proportionality, a permissible redistricting consideration, with "proportional representation,” which Section 2 expressly excludes from the scope of its protections. See 52 U.S.C. § 10301(b). While proportionality "links the number of majority-minority voting districts to minority members' share of the relevant population,” proportional representation relates the number of elected representatives who are members of a particular minority group to that minority's population. De Grandy, 512 U.S. at 1014 n. 11, 114 S.Ct. 2647 (emphasis added). The Chairs suggested a concern with proportional representation, for instance, when they cited the number of African-American candidates who had actually been elected to the General Assembly in justifying their proportionality goal. See Joint Ex. 1005 at 3. Similarly, later in the redistricting process, the Chairs spoke of providing a certain number of "seats” for African-American representatives. Joint Ex. 1018 at 12 (Lewis); Joint Ex. 1021 at 21 (Rucho).

. We recognize that comparisons to benchmark plans and districts may be of limited value because such plans are based on outdated census information, and so could not have been reenacted without modification to comply with one person, one vote requirements. However, comparisons to benchmarks may still yield useful insights, particularly where there are marked differences between past and present plans. See Alabama, 135 S.Ct. at 1271 (examining "individuals that the new redistricting laws added to the population of” a benchmark district as relevant to the predominance inquiry).

.Even though the Chairs fell one majority-black district short in each chamber of the targets they set forth in their June 17 statement, they nevertheless insisted that they had succeeded in achieving their proportionality goal. Joint Ex. 1007 at 2 ("[The twenty-three majority-black districts,] combined with two over 40% BVAP districts, continue to provide black voters with a substantially proportional and equal opportunity to elect candidates of their choice.”); Joint Ex. 1005 at 6 ("If adopted by the General Assembly, proposed [Senate] District 32 will provide African American citizens with a more equal, and *134tenth opportunity, to elect a candidate of choice."); Joint Ex. 1007 at 3-5 (explaining that Senate District 32, drawn with a BVAP of 42.53%, counted as one of the ten Senate VRA districts). This is a conclusion difficult to square with the Chairs’ repeated assertions that only 50%-plus-one BVAP districts could satisfy the VRA.

. It should be noted that the Chairs also referenced and sought to comply with the one person, one vote requirement, which was incorporated by the Stephenson cases. See, e.g., Joint Ex. 1005 at 5; Joint Ex. 1007 at 1-2. That does not affect our predominance analysis, however. See Alabama, 135 S.Ct. at 1270 ("[A]n equal population goal is not one factor among others to be weighed against the use of race to determine whether race ‘predominates.’ Rather, it is part of the redistricting background, taken as a given, when determining whether race, or other factors, predominate ..,,”).

. That "demographic analysis” also accounted for one area of the state with a high concentration of Native American voters, which was in-eluded as House District 47 in the proposed House VRA map and Enacted House Plan. Defs.' Ex. 3030 at 5; Joint Ex. 1005 at 6; Joint Ex. 1004 at 1 (Enacted House Plan Map).

.In this opinion, we use the term "precincts" to refer to "voter tabulation districts” (VTDs). Counties in North Carolina draw precinct lines based on the latest census. Joint Ex. 1012 at 19. The General Assembly created VTDs on January 1, 2008, defined by the precinct lines as they existed on that date. N.C. Gen. Stat. § 163— 132.IB. For the most part, precincts and VTDs in North Carolina remain the same, although since January 1, 2008, some counties have divided certain VTDs into multiple precincts. Only twenty-three VTDs that were further divided into pre*136cincts were split by the Enacted House Plan, and only sixteen VTDs that were further divided into precincts were split by the Enacted Senate Plan. Pls. Ex. 2092 at 3 (Second Aff. of Theodore Arrington, Ph.D. in N.C. State Conference of Branches of the NAACP v. North Carolina, No. 11-CV-01640 (N.C.Super.Ct.)). Significantly, the Enacted Plans did not split any of these further-divided VTDs along the newly formed precinct lines. Id. In other words, for our purposes, "VTD” and "precinct” are essentially synonymous.

. There was one exception among the challenged districts: Senate District 32 was enacted with a BVAP of 42.53%. Third Joint Stip. ¶ 107, ECF No. 90. We therefore consider it separately and independently in our predominance analysis. See infra section II.B.7.

. Plaintiffs' expert Anthony Fairfax testified that the eight measures available in the map-drawing software program are known as Reock, Ehren-burg, Polsby-Popper, Schwartzberg, Perimeter, Length-Width, Population Circle, and Population Polygon. Trial Tr. vol. I, 168:24-173:12. Using data from the General Assembly and the same software utilized by Dr. Hofeller, Mr. Fairfax generated scores on all eight measures for the Enacted House and Senate Plans, as well as the Benchmark House and Senate Plans. Pis.’ Ex. 2094 at 2-3.

. That easily distinguishes this case from the Cromartie cases, where there was substantial direct evidence supporting the State’s "legitimate political explanation for its districting decision." Cromartie II, 532 U.S. at 242, 121 S.Ct. 1452; Hunt v. Cromartie (Cromartie I), 526 U.S. 541, 549, 119 S.Ct. 1545, 143 L.Ed.2d 731 (1999) (detailing evidence, including affidavit testimony from the two legislators responsible for the relevant redistricting plan to the effect that their aims were "to protect incumbents, to adhere to traditional districting criteria, and to preserve the existing partisan balance in the State's congressional delegation”). The Court in Cromartie II also stressed that the direct evidence of racial predominance was weak. See 532 U.S. at 254, *140121 S.Ct. 1452 (finding evidence “less persuasive than the kinds of direct evidence ... found significant in other redistricting cases,” including concessions by the state that its goal was the creation of majority-minority districts). Given these considerable distinctions between the cases, we see no basis for requiring Plaintiffs to present us with alternative plans showing that the "legislature could have achieved its legitimate political objectives in alternative ways.” Id. at 258, 121 S.Ct. 1452; see Harris, 159 F.Supp.3d at 622-23; Page, 2015 WL 3604029, at *7 n. 12. Indeed, Defendants have not identified with any specificity which "legitimate political objectives" any alternative plans ought to have "achieved."

. We acknowledge that the racial density maps are limited in their value because they do not indicate the total number of African-American or white voters in a particular census block. See Trial Tr. vol. V, 59:9-60:17 (Hofeller). Particularly in rural locations, the census blocks may be shaded very dark (indicating a high proportion of BVAP) when only a few individuals live in that block, or may be completely white when there are no inhabitants in the census block at all. Id, Despite this shortcoming, these maps provide useful information regarding the racial make-up of each district. See Vera, 517 U.S. at 961-62, 116 S.Ct. 1941 (finding that the state used “racial data at the block-by-block level" to "make more intricate refinements on the basis of race than on the basis of other demographic information”).

. The record contains statistical information on each enacted district, including a list of the municipalities in each district and the percentage of the municipalities’ population found in that district, See Joint Ex. 1003 at 86-117 (for the Enacted Senate Plan); Joint Ex, 1004 at 104-41 (for the Enacted House Plan). Municipalities are listed in these exhibits even when a district splits a city by capturing a de minimis amount of the population. See, e.g., Joint Ex. 1003 at 104 (noting that Senate District 4 contains 0.55% of the city of Red Oak). We adopt a definition more friendly to Defendants and consider a municipality "split” when the population is divided between two districts, and each district contains at least 10% of the voters in that city.

. Senator Blue testified, based on his own knowledge of Wake County, that the "oddly shaped” appendages in the district can be explained by "the effort ... to bring black voters into District 14." Trial Tr. vol. I, 52:7-13.

. The only measure on which Senate District 20 became marginally more compact was the Eh-renburg score, and the increase was minimal. Benchmark Senate District 20 had an Ehrenburg score of 0.32, while the enacted version had a score of 0.35. Pis.' Ex. 2094 at 18, 20.

. The map was submitted on behalf of AFRAM by a representative from the Southern Coalition for Social Justice. Defs.’ Ex. 3013-11 at 2. Thus, it is labeled in the record as the Southern Coalition for Social Justice map. Defs.' Ex. 3000 at 166.

. Enacted Senate District 32 is also less compact than the benchmark district on seven of the eight compactness measures presented by Mr. Fairfax. Pis.’ Ex. 2094 at 18, 20.

. It is not clear whether the addition or removal of a single person on the basis of race could be sufficient to establish that race predominated in drawing a district. See Miller, 515 U.S. at 916, 115 S.Ct. 2475 (holding that a plaintiff must show that race motivated "the legislature’s decision to place a significant number of voters within or without a particular district” in order to establish racial predominance (emphasis added)). We need not reach this issue, however, in light of the substantial direct evidence that a significant number of voters, in addition to Gar-rou, were excluded from Senate District 32 on the basis of race.

. Senate Districts 38 and 40 split eight and sixteen precincts, respectively. Third Joint Stip. ¶¶ 124, 141.

. Enacted House District 5 was labeled as district 2 on the House VRA map. The first proposed VRA map, released on June 17, 2011, had a different configuration for this district but accidentally excluded the residence of incumbent Representative Annie Mobley from the proposed district. According to Representative Lewis, they corrected the map and redrew this district so that Representative Mobley, an African-American, would be in the majority-black House District 5, and in doing so they changed the county groupings. Trial Tr. vol. Ill, 182:16-183:1.

. Benchmark House District 7 has for the most part been incorporated into enacted House District 27, a majority-black district that has not been challenged in this case. Joint Ex. 1004 at 1, 147; Defs.' Ex. 3001 at 326.

. The district that was enacted as House District 7 was labeled as district 23 on the House VRA map. Joint Ex. 1002 at 1.

. The Reock score is one of the most "widely used” compactness measures. Pls.' Ex. 2093 at 3-4; see also Karcher, 462 U.S. at 756, 756 n. 19, 103 S.Ct. 2653 (Stevens, J., concurring) (citing the Reock measure, among others, as a way to mathematically calculate compactness). While there is no particular score that divides compact from non-compact districts, Mr. Fairfax testified that some scholars believe scores under 0.19 or 0,30 indicate that the district "should be reevaluated.” Trial Tr. vol. I, 181:18-182:8. Defendants suggest we rely on Cromartie v. Hunt, 133 F.Supp.2d 407, 415-16, 421 (E.D.N.C.2000), rev'd 532 U.S. 234, 121 S.Ct. 1452, 149 L.Ed.2d 430 (2001), which said that 0.15 constitutes a "low” Reock score, and which held that race predominated in drawing a district with a Reock score of 0.31. See, e.g., Defs.’ Post-trial Findings at 49, 69, 73-74. The Supreme Court reversed the Cromartie district court decision, finding that politics and not race was the predominant factor in drawing the district, but it did not cite or mention Reock scores in its decision. Cromartie XX, 532 U.S, 234, 121 S.Ct. 1452, 149 L.Ed.2d 430. Based on the variety of numbers suggested by the parties, we cannot discern any clear indicator of what is a compact "enough" Reock score. But we note that House District 12’s Reock score of 0.12 is below every suggested threshold.

. These two districts were labeled districts 8 and 9 on the House VRA map. Joint Ex. 1002 at l.

. On the House VRA map, the district that became House District 29 was labeled as district 31, and the district that became House District 31 was labeled as district 30. Joint Ex. 1002 at 1.

. On the House VRA map, the district that corresponds to enacted House District 32 was labeled district 27. Joint Ex. 1002 at 1.

. On the House VRA map, what became enacted House District 33 is labeled district 38, and the area that became enacted House District 38 is labeled district 33. Joint Ex. 1002 at 1.

. House District 43 was underpopulated by 28,-637 persons according to the 2010 census. Third Joint Stip. ¶ 333.

. House District 48 provides a useful example of how the numerical estimates of compactness do not always give an accurate portrayal of the district’s shape. House District 48 is less compact than its counterpart in the Benchmark Plan on four of the eight measures presented by Mr. Fairfax, and ties the benchmark on a fifth measure. Pis.' Ex. 2094 at 24, 43. While statistically it seems that the districts may be comparable, the appearances of the districts stand in shatp contrast.

. On the House VRA map, House Districts 57 and 58 are labeled districts 64 and 63. Joint Ex. 1002 at 1.

. House District 60 tied the benchmark on a fifth measure. Pis.' Ex. 2094 at 24, 43.

. On the House VRA map, House Districts 99, 102, and 107 are labeled districts 82, 87, and 86, respectively. Joint Ex. 1002 at 1.

. The racial density map does not tell as clear of a story for House District 107, which mainly borders other majority-African-American districts. However, the totality of the evidence still supports a finding that race predominated in drawing that district.

. Although North Carolina's counties are no longer covered by Section 5 following the Supreme Court's 2013 decision in Shelby County, — U.S.-, 133 S.Ct. 2612, 186 L.Ed.2d 651, that does not foreclose the possibility that compliance with Section 5 was a compelling interest at the time of the 2011 redistricting. See Alabama, 135 S.Ct. at 1274; Page, 2015 WL 3604029, at *16.

. We use the terms “racially polarized voting” and "racial bloc voting” interchangeably throughout this opinion. See Gingles, 478 U.S. at 52 n. 18, 106 S.Ct. 2752.

. A "district effectiveness analysis” is a district-specific evaluation used to determine the minority voting-age population level at which a district "become[s] effective in providing [a] realistic opportunity for ... voters [of that minority group] to elect candidates of their choice.” Trial Tr. vol. Ill, 14:1-12 (Lichtman).

. Ecological regression analysis is a standard technique used to infer voting behavior among distinct population groups. Gingles, 478 U.S. at 52-53, 53 n. 20, 106 S.Ct. 2752.

. The North Carolina counties Dr. Brunell examined were the "40 counties [that were] covered by Section 5 of the VRA, along with 11 other counties: Columhus, Duplin, Durham, For-syth, Jones, Mecklenburg, Richmond, Sampson, Tyrrell, Wake, and Warren." Defs.’ Ex. 3033 at 3.

. The report indicates that Camden County had too few precincts to make a county-specific determination regarding the existence of statistically significant racially polarized voting. Defs.' Ex. 3033 at 3; see also Trial Tr. vol. IV, 134:22-24 (Brunell),

. At trial, Dr, Brunell testified that he made an "arithmetic error” when calculating the figures in the “Black Voter %" column in each of the tables of his 2011 report. Trial Tr. vol. IV, 137:20-140:4, However, the error did not implicate the non-African-American crossover voting estimates, and Dr. Brunell testified that correcting the error did not change his conclusion that there is statistically significant racially polarized voting in fifty North Carolina counties. Id. at 139:16-25. In addition, neither party alleges that Defendants were aware of the error at the time the Enacted Plans were adopted.

. For example, in post-trial briefing, Defendants simply cite the Brunell study for the proposition that each of the challenged districts encompassed counties that "were analyzed by Dr. Bru-nell and confirmed as continuing to experience statistically significant racially polarized voting.” See, e.g„ Defs.' Post-trial Findings at 71, 79, 83, 88, 93.

. Similarly, Representative Lewis was asked by a fellow legislator during the 2011 redistricting, "[D)id the Brunell ... study provide any specific recommendations to develop ... 50 percent districts statewide? In other words, was that contained in the study?” Joint Ex. 1019 at 35. Representative Lewis responded, "I don’t know, but it is irrelevant, as that is our understanding of the law_"Id. (emphasis added).

.Indeed, Defendants seem to have operated under this assumption when they created enacted Senate District 32. In drawing the district, Defendants purposefully drew Linda Garrou out of its boundaries, citing the fact that she was a "white incumbent” who had "defeated African American candidates in Democratic Primaries in 2004 and 2010." Joint Ex. 1005 at 6; see also Trial Tr. vol. IV, 55:13-19 (Rucho); supra section II.B.7. However, Garrou defeated African-American candidates to win the 2004 and 2010 democratic primaries by large margins, capturing over 80% of the vote in each election. Joint Ex. 1048 at 25; Defs.' Ex. 3020 at 13. And, according to Defendants themselves, "African Americans [comprised] 68.71% of [the] registered Democrats” in Senate District 32 under the benchmark plan. Defs.’ Post-trial Findings at 94. Not surprisingly, Senator Rucho confirmed at trial that he did not make a determination as to whether Linda Garrou was the African-American voters' candidate of choice when he decided to draw her out of the district. Trial Tr. vol. IV, 55:13-19. Without a strong basis in evidence that Linda Garrou herself was not the African-American voters’ candidate of choice, Defendants could not have concluded that drawing Senate District 32 based on race was "reasonably necessary” to remedy legally significant racially polarized voting.

. This data was readily available to the Chairs at the time of the 2011 redistricting. The General Assembly's legislative staff submitted to the Redistricting Chairs and Committees charts detailing the results of all House and Senate elections between 2006 and 2010 involving minority candidates. See Joint Ex. 1048; Joint Ex. 1049; Trial Tr. vol. Ill, 129:17-130:2 (Lewis); Trial Tr. vol. IV, 13:15-19 (Rucho); Trial Tr. vol. IV, 101:23-102:15, 107:18-108:17 (Churchill). For each election reported, the charts indicate the name, race, and party affiliation of the competing candidates, identify which candidate won, and specify the percentage of the vote captured. See Joint Ex. 1048; Joint Ex. 1049. The charts also indicate the racial makeup of each district using 2000 census data. Ici

. This list of electoral successes in certain benchmark districts does not include additional, similar successes in benchmark districts that do not correspond geographically with challenged districts. The list only includes successes in benchmark districts that overlap substantially with the geographic locations of their enacted counterparts.

. These districts are: Senate Districts 14, 32, 38, and 40, and House Districts 29, 31, 33, 38, 99, 102, and 107. Joint Ex. 1012 at 10; Joint Ex. 1003; Joint Ex. 1004.

. The one exception to this rule is Senate District 32, which Defendants drew with a BVAP of 42.53%. However, Senate District 32 is located entirely within Forsyth County, which was not covered by Section 5.

. Senator Rucho explained at trial that "what we tried to do is put together a map that would absolutely pass preclearance approval because, without that, we could not continue on our effort to be prepared for the 2012 election cycle.” Trial Tr.vol. IV, 21:16-19.

. We note that preserving traditional district boundaries is a race-neutral districting criterion that may be lawfully considered in drawing districts. See Alabama, 135 S.Ct. at 1271; Vera, 517 U.S. at 977, 116 S.Ct. 1941. Therefore, maintaining the rough location and boundaries of existing majority-minority districts, if the evidence does not otherwise suggest that race was the predominant motive, might not trigger strict scrutiny review.