Bethune-Hill v. Va. State Bd. of Elections

368 F. Supp. 3d 872

• Court: District Court, E.D. Virginia
• Decided: February 14, 2019
• Docket: Civil Action No. 3:14cv852
• Status: Published

Opinion

BARBARA MILANO KEENAN, Circuit Judge:

In June 2018, on remand from the Supreme Court of the United States, we held that eleven majority-minority Virginia House of Delegates districts were racial gerrymanders in violation of the Equal Protection Clause of the Fourteenth Amendment to the United States Constitution. See Bethune-Hill v. Va. State Bd. of Elections , 326 F.Supp.3d 128 (E.D. Va. 2018) ( Bethune II ). We ordered the Virginia General Assembly to adopt a new redistricting plan by October 30, 2018 to remedy the identified constitutional violations. Id. at 181. Although several plans were introduced in the Virginia House of Delegates, the legislature failed to enact any of the proposals submitted for its consideration. As a result, we appointed Dr. Bernard Grofman¹ as a special master to assist us in preparing a remedial plan. We also instructed the parties and any interested *874non-parties to file proposed plans for consideration by the special master and by this court.

The Virginia Department of Elections and its officials (collectively, the state defendants) have ceased defending the composition of the eleven unconstitutional districts. The House of Delegates and the Speaker of the House, who intervened in this action (the intervenors) and took primary responsibility at trial for defending the existing eleven challenged districts, see Bethune II , 326 F.Supp.3d at 139, filed an appeal to the Supreme Court. On November 13, 2018, the Supreme Court agreed to hear oral argument in the intervenors' appeal, postponing the question of jurisdiction until consideration of the case on the merits. See Va. House of Delegates v. Bethune-Hill , --- U.S. ----, 139 S.Ct. 481, 202 L.Ed.2d 374 (2018) (mem.). The Supreme Court also ordered that the parties file briefs on the question whether the intervenors have standing to appeal our decision. Id. The Supreme Court has not yet heard argument in that appeal.²

The parties and the interested non-parties now have submitted a total of seven proposed plans for our consideration in this remedial phase of the litigation. Dr. Grofman has filed his final report, in which he evaluated these plans and offered several alternative remedial plans. We also have received extensive briefing from the parties and the interested non-parties, and we have heard testimony from Dr. Grofman and oral argument from counsel.

Upon consideration of all the proposed remedial plans, as well as the special master's recommendations, we conclude that a map composed of four of the special master's regional proposals remedies the constitutional deficiencies identified in our prior opinion, complies with traditional districting criteria, defers to the priorities of the legislature, and does not undermine minorities' rights under the Voting Rights Act of 1965, 52 U.S.C. § 10101 et seq. (VRA). We therefore direct that the state defendants implement the Final Remedial Plan included in docket entry number 355 for use in the 2019 Virginia House of Delegates elections.

I.

The facts of this case are set forth in detail in our liability-phase opinion in Bethune II , 326 F.Supp.3d 128. Accordingly, we recount only briefly the procedural history of the case before the present remedial phase of this litigation.

The plaintiffs are Virginia registered voters living in twelve Virginia House of Delegates districts (the challenged districts). Id. at 136, 139. The plaintiffs alleged that during the 2011 redistricting cycle, their House districts were drawn primarily on the basis of race in violation of the Equal Protection Clause of the Fourteenth Amendment. Id. at 136-37. After a July 2015 bench trial, a divided panel of this court concluded that the legislature did not rely predominantly on race in drawing eleven of the twelve districts. See Bethune-Hill v. Va. State Bd. of Elections , 141 F.Supp.3d 505, 510-11 (E.D. Va. 2015) ( Bethune I ). With respect to the twelfth *875district, District 75, we concluded that the legislature had used race as its predominant criterion, but that this use of race was narrowly tailored to achieve a compelling state interest. Id. at 511. On appeal, the Supreme Court reversed with respect to the eleven districts, instructing us to apply on remand a "holistic analysis" regarding the issue of racial predominance. Bethune-Hill v. Va. State Bd. of Elections , --- U.S. ----, 137 S.Ct. 788, 800, 197 L.Ed.2d 85 (2017). The Supreme Court affirmed our conclusion that the composition of District 75 did not violate the Equal Protection Clause. Id. at 801-02.

After extensive briefing on remand, we held a second trial in October 2017, at which both the plaintiffs and the intervenors introduced substantial new evidence. Bethune II , 326 F.Supp.3d at 140. In June 2018, we issued an opinion holding that the legislature had subordinated traditional districting criteria to race in its construction of the eleven remaining districts. See id. at 173. In reaching this conclusion, we considered as an important, but not dispositive, factor that the legislature had applied a minimum 55% black voting age population (BVAP) requirement to all the remaining eleven challenged districts (the invalidated districts). Id. at 144-45, 174. As a result, the legislature had shifted substantial groups of voters in and out of those districts primarily on the basis of race, in derogation of traditional districting criteria. Id. at 146, 155-72, 174. We also held that the legislature failed to produce evidence to support its predominant use of race and, thus, that this use of race did not satisfy the required standard of strict scrutiny. Id. at 175-77. Judge Payne dissented because, in his view, the plaintiffs had not met their burden to prove the constitutional violations that they alleged. Id. at 181-227 (Payne, J., dissenting).

We ordered the General Assembly to adopt a new redistricting plan to remedy the identified constitutional violations. Id. at 181 (majority opinion). When we were informed that the General Assembly would not be enacting a remedial plan, we appointed Dr. Grofman as special master to assist us in the map-drawing process.³ Additionally, the plaintiffs submitted two proposed remedial plans; the intervenors submitted two proposals, both of which had been introduced in the House of Delegates but had not been enacted by the legislature; and the state defendants declined to submit a proposed plan. Interested non-party the Virginia State Conference of NAACP Branches (the NAACP) submitted one proposal, and student groups from the College of William & Mary, Marshall-Wythe School of Law, also interested non-parties, submitted two proposals.

Dr. Grofman reviewed these seven proposed plans, submitted his report, and later filed four addenda to that report. In his report, detailed further below, Dr. Grofman declined to recommend any of the plans offered by the parties and the interested non-parties. Grofman Rep. App. A.⁴

*876Instead, Dr. Grofman constructed alternative proposed remedial maps, which he presented in a regional "module" format. See infra p. 881-82.

After receiving additional briefing, we held a hearing on January 10, 2019 to address the remedial plans submitted by the parties and the interested non-parties, and the special master's proposals. At the hearing, Dr. Grofman testified under oath in response to questioning by counsel for the plaintiffs, counsel for the intervenors, and the court. We also heard extensive argument from the parties and the NAACP. At the end of the hearing, we ordered Dr. Grofman to submit a final report incorporating the information from his original report and all four addenda. Upon review of the final report of the special master, we ordered Dr. Grofman to submit a Final Remedial Plan incorporating four specific modules offered in his report and encompassing all 100 House of Delegates districts. Dkt. No. 353. For the reasons set forth below, we now adopt the Final Remedial Plan filed by the special master on January 29, 2019. Dkt. No. 355.

II.

We begin by reviewing the criteria that we apply in our evaluation of the proposed remedial plans. We later discuss the extent to which the plans proposed by the special master and the other submitted plans are consistent with these redistricting goals.

A.

The foundational purpose of the 2011 redistricting in Virginia was to redistribute population among the 100 House of Delegates districts to achieve the constitutional requirement of equal population based on the results of the 2010 census. See Reynolds v. Sims , 377 U.S. 533, 568-69, 84 S.Ct. 1362, 12 L.Ed.2d 506 (1964) (discussing equal population requirement for state legislative districts); see also Va. Const. art. 2, § 6 (Virginia constitutional requirement that "[e]very electoral district ... give, as nearly as is practicable, representation in proportion to the population of the district"). As the Supreme Court has explained, "the requirement that districts have approximately equal populations is a background rule against which redistricting takes place." Ala. Legislative Black Caucus v. Alabama , --- U.S. ----, 135 S.Ct. 1257, 1271, 191 L.Ed.2d 314 (2015). Thus, our court-imposed plan also must satisfy the requirement of population equality.

During the 2011 redistricting process, the legislature determined that each House of Delegates district must have 80,000 residents, with a maximum population deviation of plus or minus one percent. Bethune II , 326 F.Supp.3d at 138. This population deviation allowance is narrower than that applied to the 2001 House of Delegates plan, see 1st Trial Tr. at 275, and falls well within the constitutional requirement for equal population in state legislative districts, see Evenwel v. Abbott , --- U.S. ----, 136 S.Ct. 1120, 1124, 194 L.Ed.2d 291 (2016) ("Where the maximum population deviation between the largest and smallest district is less than 10%, the Court has held, a state or local legislative map presumptively complies with the one-person, one-vote rule." (citing Brown v. Thomson , 462 U.S. 835, 842-43, 103 S.Ct. 2690, 77 L.Ed.2d 214 (1983) ).

During the remedial phase of this litigation, the parties agreed that the one percent figure also should govern the population distribution in our court-imposed plan. We therefore ordered the special master to construct all districts in his proposed *877remedial plans with 80,000 residents, with a maximum population deviation of plus or minus one percent. The special master's proposed plans, as well as those submitted by the parties and the interested non-parties, achieve this population-equality metric.

B.

In addition to the background principle of population equality, our chosen plan also must remedy the Equal Protection violations that we identified in the 2011 plan. As explained above, in Bethune II we concluded that the legislature had subordinated traditional districting criteria to racial considerations in the eleven invalidated districts, and had failed to show a "strong basis in evidence" for its race-based decisionmaking. See generally 326 F.Supp.3d at 138, 144-45, 155-74. To remedy these Equal Protection violations, we now draw a plan consistent with traditional districting criteria. See Personhuballah v. Alcorn , 155 F.Supp.3d 552, 561-62, 565 (E.D. Va. 2016) (explaining that the court-imposed remedial plan "remedies the [Equal Protection] violation ... by drawing districts based on neutral, traditional criteria"). In Virginia, such traditional criteria include "the constitutional requirements of compactness and contiguity, respect for political subdivisions, and consideration of communities of interest." Id. at 561 (citing Va. Const. art. 2, § 6 and Page v. Va. State Bd. of Elections , No. 3:13cv678, 2015 WL 3604029, at *10 (E.D. Va. June 5, 2015) ). The House of Delegates similarly identified contiguity, compactness, and communities of interest to include "governmental jurisdictions," as criteria governing the 2011 redistricting process.⁵ Pl. Ex. 16.

Although we must ensure that our plan remedies the Equal Protection violations, we are mindful that redistricting "is primarily a matter for legislative consideration and determination." White v. Weiser , 412 U.S. 783, 794, 93 S.Ct. 2348, 37 L.Ed.2d 335 (1973). We are "guided by the legislative policies underlying the existing plan, to the extent those policies do not lead to violations of the Constitution or the Voting Rights Act." Personhuballah , 155 F.Supp.3d at 563 (quoting Abrams v. Johnson , 521 U.S. 74, 79, 117 S.Ct. 1925, 138 L.Ed.2d 285 (1997) ). Thus, our role is a narrow one. Our modifications to the existing plan will be "limited to those necessary to cure any constitutional or statutory defect." Upham v. Seamon , 456 U.S. 37, 43, 102 S.Ct. 1518, 71 L.Ed.2d 725 (1982) ; see also Personhuballah , 155 F.Supp.3d at 563. Once "the racial gerrymanders at issue in this case [are] remedied," our role in Virginia's redistricting process is "at an end." North Carolina v. Covington , --- U.S. ----, 138 S.Ct. 2548, 2555, 201 L.Ed.2d 993 (2018) (per curiam).

In the present case, the task of balancing these considerations is especially complex. The eleven invalidated districts are located in four distinct groupings, and some, but not all, of these groups of districts *878are adjacent to one another. The invalidated districts themselves frequently span multiple municipalities, and many cities and counties have been split between invalidated districts and surrounding non-challenged districts. In choosing a remedial plan, we endeavor to minimize the number of districts affected by our revisions, recognizing that districts immediately adjacent to the invalidated districts may be subject to significant changes. See Abrams , 521 U.S. at 86, 117 S.Ct. 1925 (concluding that "substantial changes to the existing plan consistent with ... traditional districting principles" was warranted, given that a "large geographic area of the state" was impacted by the constitutional violation).

C.

Finally, we also seek to ensure that in remedying the identified Equal Protection violations, we do not select a plan under which black voters' rights are diminished when compared with the unconstitutional 2011 plan. We thus consider compliance with Section 2 of the VRA as an "equitable factor" in our redistricting process, and will "implement a plan that complies with federal policy disfavoring discrimination against minority voters."⁶ See Personhuballah , 155 F.Supp.3d at 564 ; see also Abrams , 521 U.S. at 90, 117 S.Ct. 1925 (noting that "[o]n its face, § 2 does not apply to a court-ordered remedial redistricting plan," but assuming without deciding that "courts should comply with [Section 2] when exercising their equitable powers to redistrict"). Section 2 prohibits "vote dilution," which occurs if election processes "are not equally open to participation by members of a [racial minority]" such that members of that minority 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) ; see also Personhuballah , 155 F.Supp.3d at 564-65 (describing factors to consider in a Section 2 analysis). Accordingly, in evaluating the proposed remedial plans, we will assess black voters' ability to elect their preferred candidates in the redrawn districts.

With these principles in mind, we turn to identify the proposed plans that remedy the identified constitutional deficiencies while balancing these priorities that occasionally conflict with one another.

III.

We begin with general observations about the challenges inherent in this remedial process. As discussed above, the number of invalidated districts, and their proximity to and interconnectedness with one another, renders our task especially complex. A single change in one invalidated district will, at a minimum, impact an immediately adjacent district and could impact numerous other districts, both invalidated and non-challenged. We thus agree with Dr. Grofman's commonsense observation *879that, in practice, crafting a plan consistent with traditional districting criteria requires accepting certain tradeoffs among priorities. Grofman Rep. at 30. For example, the creation of a maximally compact district likely would require changing more boundaries than strictly necessary to remedy a particular Equal Protection violation.

We also observe that an inevitable shift of black voters will result from remedying the specific Equal Protection violations we identified in the 2011 plan. In Bethune II , we found that the legislature had sorted black voters into the invalidated districts predominantly on the basis of their race, thereby creating in the invalidated districts BVAP levels much higher than necessary to comply with Section 5 of the VRA.⁷ See 326 F.Supp.3d at 177-80. By reversing this violation and redrawing the districts according to neutral districting criteria, many black voters formerly subjected to race-based inclusion in the invalidated districts will be assigned to surrounding non-challenged districts. Accordingly, the BVAP in the invalidated districts will decrease under our remedial plan, and the BVAP of adjacent non-challenged districts will increase. See Grofman Rep. at 36. In our view, this effect is not evidence of race-based decisionmaking, but rather is a foreseeable and necessary result of a remedial plan that does not subordinate traditional districting factors to race. See Grofman Rep. at 158-59.

We turn to consider Dr. Grofman's methodology. Consistent with his task of drawing remedial districts according to traditional districting criteria, Dr. Grofman identified the following nine criteria governing his construction of the proposed remedial maps:

(1) population equality;

(2) avoiding dilution in the voting strength of minorities and avoiding retrogression in minority groups' opportunity to elect preferred candidates, in compliance with Section 2 of the VRA and the Equal Protection Clause;

(3) avoiding using race as a predominant consideration;

(4) contiguity;

(5) avoiding splits of political subdivisions such as cities and counties;

(6) compactness;

(7) avoiding changes to the 2011 plan not required to remedy the identified constitutional violations, by limiting changes to the invalidated districts and immediately adjacent districts and by minimizing the number of non-challenged districts so affected;

(8) partisan neutrality;

(9) avoiding incumbency pairings, to the extent feasible.

Grofman Rep. at 25-28.

Among these criteria, Dr. Grofman was "especially attentive to issues of contiguity, compactness, and avoiding splitting of existing political subunit boundaries," because those political units represent identifiable communities of interest. Grofman Rep. at 32 & nn. 20-22, 53-54, 165 ("[T]he most appropriate way to remedy the constitutional violation ... is to replace the present unconstitutional districts with contiguous equipopulous districts with fewer city or county splits than are found in the 2011 plan and with at least as high average level of compactness."). He also viewed partisan neutrality as a "necessity" in any court-ordered plan. Grofman Rep. at 32-33. And finally, none of Dr. Grofman's proposed *880remedial maps paired any incumbent delegates who held office in 2017. Grofman Rep. at 175.

Dr. Grofman also sought to confine the impact of his proposals in recognition of our limited remedial role. First, the only non-challenged districts that Dr. Grofman changed are located adjacent to invalidated districts. Grofman Rep. at 60-61. Second, he made changes to adjacent, non-challenged districts only where such districts contained a portion of a city or county that also was included in one of the invalidated districts. Grofman Rep. at 61-62. Dr. Grofman concluded that these two constraints resulted in the creation of plans that adequately remedied the constitutional violations in the invalidated districts while avoiding unnecessary changes to non-challenged districts.⁸ Grofman Rep. at 62-63. After a thorough evaluation of Dr. Grofman's qualifications, report, and testimony, we find that Dr. Grofman was a credible witness and that he used an appropriate methodology.

We reject the intervenors' assertion that Dr. Grofman's methodology used race improperly as the predominant criterion by applying a 55% BVAP "ceiling" to the invalidated districts. Throughout his report and testimony, Dr. Grofman emphasized under oath that he never sought to achieve a predetermined BVAP level in any of the proposed districts. Grofman Rep. at 51, 158-59, 166; Jan. 10, 2019 Tr. at 30-32, 66, 115. He averred that only after redrawing the invalidated districts according to traditional districting criteria did he seek to ensure that the new districts had not inadvertently resulted in minority vote dilution.⁹ Grofman Rep. at 51-52, 54, 158-59, 166; Jan. 10, 2019 Tr. at 66. And, as Dr. Grofman explained, the fact that the BVAP in the invalidated districts fell below 55% in his remedial maps was a foreseeable consequence of applying traditional districting criteria to "the geography and demography" of Virginia. See Jan. 10, 2019 Tr. at 30, 32. We therefore credit Dr. Grofman's explanations regarding the manner in which he considered race in constructing his proposals.

We also reject the intervenors' contention that Dr. Grofman intentionally sought to reduce voter support for certain Republican House leaders in their redrawn districts. Nothing in the record suggests that Dr. Grofman acted with animus toward any incumbents, or toward any party. We credit Dr. Grofman's contention that he constructed his proposals without regard to partisan outcome in the non-challenged districts, and that he treated all incumbents *881equally.¹⁰ Grofman Rep. at 65-66, 176-77.

In view of our credibility findings and our approval of the special master's methodology, we adopt Dr. Grofman's recommendation that we reject the remedial maps proposed by the parties and the interested non-parties.¹¹ See generally Grofman Rep. App. A. Some of those maps made more extensive changes than are required to remedy the constitutional violations, and some failed to impose an adequate remedy for those violations. Grofman Rep. at 29-30, 108. Most notably, each of the proposed plans from the parties and the interested non-parties includes changes to 30 or more districts. Grofman Rep. at 120. Dr. Grofman explained that changing so many districts was "certainly" unnecessary to remedy the Equal Protection violations in the invalidated districts. Grofman Rep. at 120. In contrast, Dr. Grofman's proposals would change between 21 and 26 districts.¹² Grofman Rep. at 120. Given the narrow scope of our remedial mandate, we agree with Dr. Grofman's assessment that our court-ordered plan should avoid these excessive and unnecessary changes. Grofman Rep. at 120-21.

Dr. Grofman also criticized the proposed plans submitted by the parties and the interested non-parties as imposing "an excessive number of avoidable [political subdivision] splits," Grofman Rep. at 131, thereby unnecessarily dividing identifiable communities of interest. Because a plan that disregards political subdivision boundaries is inconsistent with traditional districting criteria, see Alabama , 135 S.Ct. at 1270, such a plan is inadequate to address the racial sorting violations we identified in the invalidated districts. For these reasons, we decline to adopt any of the plans proposed by the parties and the interested non-parties.

We proceed to consider the special master's various remedial maps. Dr. Grofman presented his proposals in a regional "module" format, in which he provided multiple versions of a remedial map for each geographical region in which the invalidated districts are located, namely, Petersburg, Richmond, North Hampton Roads (the peninsula), and South Hampton Roads (Norfolk). Grofman Rep. at 70. The creation of multiple modules allowed us to select the most appropriate module from each region to create a final remedial map that includes all 100 House of Delegates districts. See Grofman Rep. at 70-71.

We begin with the Petersburg region, in which we must remedy the racial gerrymander in District 63. Non-challenged Districts 62, 64, 66, and 75¹³ also are located *882in the Petersburg region. Grofman Rep. at 83. In concluding that the 2011 version of District 63 was drafted primarily on the basis of race, we relied on the "avowedly racial" decision to split Dinwiddie County with District 75 in order to distribute the black population between those two majority-minority districts, as well as on the significant reduction in compactness of District 63 caused by the 2011 redistricting. Bethune II , 326 F.Supp.3d at 159-60.

After considering the potential reconfigurations of District 63, we conclude that the "Petersburg 2" module offered by Dr. Grofman best remedies the constitutional violations we identified previously and complies with traditional districting criteria.¹⁴ First, and most notably, the Petersburg 2 module includes all of Dinwiddie County within District 63, directly remedying the explicitly race-based split of that county that we criticized in Bethune II . Grofman Rep. at 89. The reconfigured district also includes the entire city of Petersburg. Grofman Rep. at 89. In total, District 63 now comprises some or all of three political subdivisions, including a portion of Chesterfield County, instead of the five political subdivisions included in the 2011 version of District 63. See Grofman Rep. at 83, 89. All the non-challenged districts in the Petersburg region similarly include fewer segments of political subdivisions, or retain the same number, as in the 2011 map. Grofman Rep. at 83, 89. In total, the Petersburg 2 module allocates 20 segments of cities and counties to the five districts in the Petersburg region, while, in the 2011 map, the political subdivisions were divided into 28 segments in these districts. Grofman Rep. at 83, 89. Accordingly, the Petersburg 2 module advances the important goal of respecting political subdivision boundaries and associated communities of interest.¹⁵

The Petersburg 2 module also substantially increases the compactness scores of District 63, more than doubling the Reock score from .25 to .57, and nearly doubling the Polsby-Popper score.¹⁶ Grofman Rep. at 83, 89. All the non-challenged districts in the region either maintain nearly identical compactness scores, or improve in compactness compared with the 2011 plan. Grofman Rep. at 83, 89. Upon review of the map of the Petersburg 2 module, we conclude that the map was drawn in accordance with traditional districting criteria.

Based on the data in the record, we further conclude that black voters in Districts 63 and 75 will continue to have a reasonable opportunity to elect candidates of their choice. Under the 2011 map, racially polarized voting in these two districts was higher than in any of the other majority-minority districts. See Bethune II , 326 F.Supp.3d at 178. Therefore, we are mindful of the possibility that a reduced BVAP in these districts could impair the voting rights of black voters. Under the Petersburg 2 module, the BVAP in *883District 63 is reduced from 59.53% in the 2011 plan to 47.47% in the remedial plan. Grofman Rep. at 83, 89. The BVAP in neighboring District 75 decreases from 55.43% in the 2011 map to 52.45% in the remedial map. Grofman Rep. at 83, 89.

Notably, Dr. Maxwell Palmer, who was qualified by this court as an expert witness in political science, testified credibly at the second trial that despite some racially polarized voting, black voters in the 2011 version of District 63 could elect their preferred candidates even at a 45% BVAP.¹⁷ Bethune II , 326 F.Supp.3d at 145, 147, 178 ; Pl. Ex. 71 at 68; see also 2nd Trial Tr. at 427-30 (describing methodology). According to Dr. Palmer, black voters in District 75 could do so with a 50% BVAP. Pl. Ex. 71 at 68. Dr. Grofman agreed with these assessments. Grofman Rep. at 46.

With respect to the Petersburg 2 module specifically, Dr. Grofman examined the vote share of black Democratic candidates in two recent elections, namely, President Barack Obama's 2012 general election and Justin Fairfax's 2013 primary election for Virginia attorney general,¹⁸ to estimate the likely success of future minority candidates in the redrawn invalidated districts.¹⁹ Grofman Rep. at 42-43. More than 59% of voters in the reconfigured versions of Districts 63 and 75 voted for President Obama in 2012. Grofman Rep. at 89. In reconfigured District 63, Mr. Fairfax won 68.5% of the Democratic primary vote in 2013. Grofman Rep. at 89. He won 58.23% of that vote in reconfigured District 75. Grofman Rep. at 89. Put simply, these two black Democratic candidates were preferred by a significant majority of the voters who will reside in the versions of Districts 63 and 75 drawn in the Petersburg 2 module.²⁰

Accordingly, upon our review of the voting practices in the Petersburg region, we are satisfied that our adoption of the Petersburg 2 module will not result in dilution of the voting rights of black voters. The Petersburg 2 module also remedies the identified constitutional violation and complies with traditional districting criteria. We therefore adopt the configuration of the districts in the Petersburg 2 module into our remedial map.

We turn to consider the Richmond region, which includes invalidated Districts 69, 70, 71, and 74, as well as non-challenged Districts 27, 68, 72, and 73. Grofman Rep. at 76. In Bethune II , we identified numerous instances of race-based line drawing involving all four invalidated districts. In several cases, for example, high *884BVAP VTDs²¹ were transferred into an invalidated district to ensure that the recipient district achieved a BVAP of over 55%. See, e.g. , Bethune II , 326 F.Supp.3d at 157, 161 (high BVAP VTDs moved from Districts 70 and 74 into District 71). VTDs also were split with the purpose of allocating high-BVAP areas to the invalidated districts, and heavily white areas to the non-challenged districts. See, e.g. , id. at 161-62. The legislature's many race-based maneuvers caused "ripple effects throughout" the region. Id. at 155, 157.

We begin by highlighting two factors complicating our task in the greater Richmond area. First, given the dense population and the number of districts in the region, even a small change to an invalidated district may affect multiple nearby non-challenged districts. We therefore recognize the need to ensure that given changes actually are required to remedy the violations in the invalidated districts. Second, the population numbers would allow the drawing of between two and three districts wholly within Richmond's city boundaries. Grofman Rep. at 67. The 2011 map, however, split Richmond into five districts, with four 2011 incumbents living in the part of the city north of the James River. Grofman Rep. at 67, 75-76. As a result, the reallocation of parts of Richmond according to traditional districting criteria, while retaining incumbents in their districts, was challenging. We are mindful that we cannot "perfectly" achieve all desirable goals, but must balance competing considerations in pursuit of an adequate remedy. See Grofman Rep. at 30.

Upon review of the submissions of the parties, the interested non-parties, and the special master, we conclude that the "Richmond 1A" module offered by Dr. Grofman remedies the identified Equal Protection violations, avoids vote dilution, and complies with traditional districting principles in accordance with our remedial mandate. One feature of the Richmond 1A module is especially compelling in our evaluation. In this module, six of the eight districts in the region, including three of the four invalidated districts, are contained wholly within single political subdivisions. Grofman Rep. at 78. Specifically, reconfigured Districts 69 and 71 are located entirely within the city of Richmond, Districts 72, 73, and 74 are located entirely within Henrico County, and District 27 is located wholly within Chesterfield County. Grofman Rep. at 78. In contrast, in the 2011 map, only three of the eight districts, and none of the invalidated districts, were encompassed within a single political subdivision.²² Grofman Rep. at 76. In total, political subdivisions are split into 13 segments in the Richmond 1A module, compared with 16 segments in the 2011 map.²³ Grofman Rep. at 76, 78. This *885increased consistency between political subdivision and district boundaries substantially improves the map's compliance with traditional districting criteria by focusing on identifiable communities of interest.²⁴

The NAACP and the intervenors assert that because the Richmond 1A module retains certain lines in the invalidated districts we have identified as racially motivated, the Richmond 1A module fails to remedy the constitutional violations in the Richmond region.²⁵ We reject that argument. We have found that the special master did not act with race as a predominant motive in constructing the remedial modules. And, as plaintiffs' counsel explained, the goal of the remedial process is not to "correct" every suspicious boundary. Jan. 10, 2019 Tr. at 85-86. Our identification of race-based lines during the liability phase was necessary to evaluate the legislature's motive in drawing the 2011 plan. Indeed, as the Supreme Court has explained, a legislature may act with a predominantly racial motive and still draw lines consistent with traditional districting criteria, requiring a nuanced inquiry into legislative purpose. See Bethune-Hill , 137 S.Ct. at 797-98. In contrast, at the remedy phase, our task is to construct districts based on traditional districting criteria in order to ensure the voting rights of Virginia voters.

Additionally, as noted above, our mandate is circumscribed by our remedial role. We will not redraw a non-challenged district to improve its compliance with traditional districting criteria when such a change is divorced from the constitutional deficiencies we identified in the invalidated districts.²⁶ In our view, the Richmond 1A module appropriately avoids unnecessary changes to non-challenged districts by slightly altering the boundaries of non-challenged District 72 and by otherwise leaving the non-challenged districts unchanged. Grofman Rep. at 78. Given our remedial task, and the substantially improved preservation of political subdivision boundaries offered by the Richmond 1A module, we conclude that this module best remedies the constitutional violations we found in the Richmond-area invalidated districts.

We further hold that implementation of the Richmond 1A module will not affect black voters' ability to elect their preferred candidates. In the 2011 map, the BVAP percentages in the four invalidated Richmond districts ranged from 55.19% in District 69 to 57.24% in District 74. Grofman Rep. at 76. In the Richmond 1A module, these BVAP levels decrease slightly, ranging from 52.29% in District 70 to 54.38% in District 69. Grofman Rep. at 78.

*886Despite these reduced BVAP levels in the Richmond 1A module, black voters will retain their ability to elect preferred candidates. Dr. Palmer estimated that with a 45% BVAP in the 2011 version of the Richmond-area invalidated districts, candidates preferred by black voters would win between 65.4% and an overwhelming 81% of the vote. Pl. Ex. 71 at 68. Consistent with this prediction, President Obama won between 72.38% and 86.72% of the vote in the reconfigured invalidated districts. Grofman Rep. at 78. Mr. Fairfax won majorities in all four redrawn invalidated districts as well. Grofman Rep. at 78. We easily conclude that the reduction in BVAP levels in the Richmond-area invalidated districts will not dilute the voting strength of black voters.

For these reasons, we conclude that the Richmond 1A module is drawn according to traditional districting criteria and remedies the Equal Protection violations we identified in Districts 69, 70, 71, and 74. Black voters will retain their ability to elect their candidates of choice in these redrawn districts. We therefore adopt the Richmond 1A module as part of our remedial map.

D.

We proceed to consider the North Hampton Roads area, commonly known as the peninsula. See Bethune II , 326 F.Supp.3d at 162. Invalidated Districts 92 and 95, as well as non-challenged Districts 91, 93, and 94, are located in the peninsula region. Grofman Rep. at 92, 96. In Bethune II , we strongly criticized the long, narrow appendage added to District 95 during the 2011 redistricting, "which on its face disregarded traditional districting criteria" and caused the "separation of predominantly black neighborhoods from predominantly white neighborhoods with striking precision." Bethune II , 326 F.Supp.3d at 162. Race also was the predominant factor in drawing District 92, which "was controlled by the race-based decisions in District 95." Id. at 165.

At the outset, we observe that as a result of the patently race-based appendage in District 95, the 2011 version of that district shared borders with four other districts on the peninsula. See Grofman Rep. at 91; DI Ex. 94 at 14. Excising the appendage from District 95 will have a significant impact on the surrounding districts, a result that is required to remedy the racial gerrymander. See Grofman Rep. at 95. Upon review of the submitted proposals, we adopt the "Peninsula 2" module offered by Dr. Grofman as the appropriate remedy for the unconstitutional racial gerrymanders in Districts 92 and 95.²⁷

In the Peninsula 2 module, Districts 92 and 95 both are contained entirely within single municipalities, the city of Hampton and the city of Newport News, respectively. Grofman Rep. at 96. In the 2011 map, District 92 was split between Hampton and Newport News, and Hampton was divided into three separate districts (Districts 91, 92, and 93). Grofman Rep. at 92. District 95 improves significantly in compactness in our remedial plan, with the district's Polsby-Popper score more than doubling. Grofman Rep. at 92, 96. District 92 retains nearly identical compactness scores as in the 2011 map. Grofman Rep. at 92, 96.

With respect to surrounding non-challenged districts, District 94 previously was divided into four municipalities, but is now contained wholly within the city of Newport News. Grofman Rep. at 92, 96. Although Districts 91 and 93 now include more portions of municipalities than in the *8872011 map, the Peninsula 2 module reduces overall the number of political subdivision splits in the region from 11 in the 2011 plan to 10 in the remedial module. Grofman Rep. at 92, 96. Similarly, the mean compactness score of all the districts in the region improves according to both the Reock and Polsby-Popper measures.²⁸ Grofman Rep. at 92, 96. Balancing all the relevant factors, we conclude that the Peninsula 2 module complies with traditional districting criteria and remedies the constitutional violations in Districts 92 and 95.

Our consideration of potential vote dilution in the invalidated districts leads us to conclude that black voters will retain their ability to elect their preferred candidates in Districts 92 and 95 as drawn in the Peninsula 2 module. In the 2011 map, both districts had high BVAP levels of around 60%, and President Obama and Mr. Fairfax won nearly 80% of the vote in those districts. Grofman Rep. at 92. The Peninsula 2 module would reduce the BVAP in District 92 to 53.87%, and to 47.36% in District 95. Grofman Rep. at 96.

Although these are significant decreases in BVAP levels, Dr. Palmer found that candidates preferred by black voters would prevail with more than 60% of the vote with a BVAP of only 45% in the 2011 versions of Districts 92 and 95. Pl. Ex. 71 at 68. Consistent with this prediction, President Obama won 75.75% of the vote in District 92 as drawn in the Peninsula 2 module, and 72.42% of the vote in reconfigured District 95. Grofman Rep. at 96. Similarly, Mr. Fairfax won 77.9% of the vote in redrawn District 92, and 71.03% of the vote in redrawn District 95. Grofman Rep. at 96. Given these substantial margins of victory by two recent black Democratic candidates, the implementation of the Peninsula 2 module will not diminish black voters' ability to elect their preferred candidates.

For these reasons, we conclude that the Peninsula 2 module is an appropriate remedy for the identified constitutional violations in the peninsula region. We therefore adopt the Peninsula 2 module as part of our remedial plan.

E.

Finally, we turn to consider Dr. Grofman's remedial proposals for the South Hampton Roads region, encompassing the cities of Norfolk, Chesapeake, Portsmouth, and surrounding areas. Bethune II , 326 F.Supp.3d at 165. South Hampton Roads includes invalidated Districts 77, 80, 89, and 90, as well as eight non-challenged districts, namely, Districts 21, 76, 78, 79, 81, 83, 84, and 85. Grofman Rep. at 99. As we explained in Bethune II , aspects of the invalidated districts in this region plainly manifested a predominant reliance on race, involving "complicated population-shifting maneuvers to sweep concentrations of black residents into one of the challenged districts," and efforts "to respond to the ripple effects of such population shifts throughout the region." Bethune II , 326 F.Supp.3d at 166. As in the Richmond area, a single change to the boundaries of an invalidated district can impact districts throughout South Hampton Roads.

*888Upon consideration of the proposed remedial plans, we adopt the "Norfolk 1A" module as the configuration that best remedies the identified constitutional violations in the region, is consistent with traditional districting criteria, and avoids diluting the voting strength of black voters. The benefits of the Norfolk 1A module are apparent. Notably, under this remedial plan, the four invalidated districts in the region now are contained wholly within single municipalities: District 77 within the city of Chesapeake; District 80 within the city of Portsmouth; and Districts 89 and 90 within the city of Norfolk. Grofman Rep. at 101.

By contrast, District 89 was the only invalidated district in the region contained within a single city under the 2011 map. Grofman Rep. at 99. District 80 was split in the 2011 map among the four distinct cities of Chesapeake, Norfolk, Portsmouth, and Suffolk, thereby "render[ing] the shape of the district bizarre on its face, resembling a sideways 'S.' " Bethune II , 326 F.Supp.3d at 167 ; see Grofman Rep. at 99. District 77, comprising portions of two cities in the 2011 map, connected its center in Chesapeake to a smaller Suffolk population by creating "an exceptionally narrow corridor to connect pockets of black residents in two cities, without including an avenue for constituents or delegates to travel along that corridor." Bethune II , 326 F.Supp.3d at 171 ; Grofman Rep. at 99. Under the Norfolk 1A module, District 77 no longer includes this patently race-based extension into Suffolk. Grofman Rep. at 100-01. Overall, the Norfolk 1A module reduces the number of municipality splits from 22 in the 2011 map to 18 in the remedial map. Grofman Rep. at 99, 101.

The Norfolk 1A module also improves the compactness of the districts in the region overall, increasing the average Polsby-Popper score from .24 to .35, and slightly improving the overall Reock score. Grofman Rep. at 99, 101. Invalidated Districts 77 and 80 increase substantially in both measures of compactness. By correcting the bizarre shape of District 80, that district's Reock score more than doubled and its Polsby-Popper score more than tripled. Grofman Rep. at 99, 101. Similarly, the Norfolk 1A module nearly triples the Reock score for District 77, and increases that district's Polsby-Popper score from just .15 to a highly compact .52. Grofman Rep. at 99, 101. Districts 89 and 90 also show marked improvement in their Polsby-Popper scores in this remedial module. Grofman Rep. at 99, 101. Given these significant improvements in compliance with traditional districting criteria, the Norfolk 1A module clearly remedies the violations we identified in South Hampton Roads.

We turn to evaluate whether black voters will experience vote dilution as a result of our implementation of the Norfolk 1A module. The BVAP levels in reconfigured Districts 77 and 90 are the lowest in any of the invalidated districts in our remedial map. See Grofman Rep. at 78, 89, 96, 101. The BVAP in District 77 decreases from 58.78% in the 2011 map to 40.23% in the Norfolk 1A module, the lowest BVAP in any invalidated district. Grofman Rep. at 99, 101. The BVAP in District 90 decreases from 56.59% in the 2011 map to 41.93% in the Norfolk 1A module. Grofman Rep. at 99, 101.

The plaintiffs, black voters who live in the affected districts, affirm explicitly that these relatively low BVAP figures do not risk diluting their ability to elect their preferred candidates. See Dkt. No. 328 at 7; Jan. 10, 2019 Tr. at 152. After examining the results of elections involving President Obama and Mr. Fairfax in the reconfigured versions of these districts, we agree with the plaintiffs' assessment. Dr. Grofman determined that in reconfigured District 77, both candidates won with substantial *889majorities, President Obama with 63.53% of the vote, and Mr. Fairfax with 73.03% of the vote. Grofman Rep. at 101. And, according to Dr. Grofman, in reconfigured District 90, President Obama won with nearly 70% of the vote and Mr. Fairfax earned nearly 60% of the vote. Grofman Rep. at 101. Additionally, Dr. Palmer estimated that, with a 45% BVAP level in the 2011 version of the districts, candidates preferred by black voters would receive 63.5% of the vote in District 77, and 66.2% of the vote in District 90. Pl. Ex. 71 at 68. Accordingly, despite the significant decrease in BVAP in some of the reconfigured districts in South Hampton Roads, black voters will retain their ability to elect candidates of their choice under the Norfolk 1A module.²⁹

In sum, we find that the Norfolk 1A module remedies the constitutional violations we identified in Districts 77, 80, 89, and 90, substantially improves compliance with traditional districting criteria, and does not dilute the voting strength of black voters. We therefore adopt the districts as drawn in the Norfolk 1A module into our remedial plan.

IV.

For these reasons and upon review of the proposed remedial plans, we adopt the Final Remedial Plan submitted by the special master [Dkt. No. 355], comprising modules Petersburg 2, Richmond 1A, Peninsula 2, and Norfolk 1A. The state defendants are directed to implement the Final Remedial Plan for use in the 2019 Virginia House of Delegates elections.

IT IS SO ORDERED.

ROBERT E. PAYNE, Senior District Judge, dissenting:

For the reasons set forth in Bethune-Hill v. Va. State Bd. of Elections , 326 F.Supp.3d 128, 181-227 (E.D. Va. 2018) (Payne, J., dissenting), I concluded that the plaintiffs had not carried their burden to prove a violation of the plaintiffs' constitutional rights. Therefore, I see no need for any change to the redistricting plan enacted by the General Assembly.

1 Dr. Grofman is the Distinguished Professor of Political Science and Jack W. Peltason Endowed Chair of Democracy Studies at the University of California, Irvine and a specialist on redistricting remedial plans. Grofman Rep. at 1.

2 On August 30, 2018, we denied the intervenors' motion for stay pending appeal to the Supreme Court, concluding that the intervenors had not satisfied their burden of showing a likelihood of success on the merits, and finding that a delay in implementing a remedial plan likely would result in the 2019 Virginia House of Delegates elections being conducted with unconstitutional districts. We later denied without prejudice the intervenors' renewed motion for stay. On January 8, 2019, the Supreme Court denied the intervenors' motion to stay filed in that court. Oral argument in the Supreme Court is scheduled for March 18, 2019.

3 The state defendants and the plaintiffs recommended Dr. Grofman for appointment as special master based on his experience serving as special master in Personhuballah v. Alcorn , 155 F.Supp.3d 552 (E.D. Va. 2016), in which the district court had adopted one of his remedial plans. The intervenors objected to the appointment of Dr. Grofman in the present case. They argued that Dr. Grofman had not provided adequate protection regarding certain incumbency issues in his proposed map selected by the district court in Personhuballah . After reviewing Dr. Grofman's record, we concluded that he was an appropriate choice to assist us in constructing the remedial map in this case.

4 Citations in this opinion to "Grofman Rep." refer to the Corrected Second Report of the Special Master submitted on February 5, 2019. See Dkt. No. 360. That report corrected certain typographical errors contained in the final report Dr. Grofman previously had submitted on January 17, 2019. See Dkt. No. 351.

5 The House Committee on Privileges and Elections adopted a resolution setting forth the following criteria governing the 2011 redistricting: (1) population equality; (2) compliance with the VRA; (3) contiguity and compactness; (4) single-member districts; and (5) maintaining communities of interest, considering "economic factors, social factors, cultural factors, geographic features, governmental jurisdictions and service delivery areas, political beliefs, voting trends, and incumbency considerations." The resolution specified that "[l]ocal government jurisdiction and precinct lines may reflect communities of interest to be balanced, but they are entitled to no greater weight as a matter of state policy than other identifiable communities of interest." Pl. Ex. 16; see also Bethune II , 326 F.Supp.3d at 144.

6 Section 5 of the VRA established a "non-retrogression" requirement, under which Virginia was not permitted to implement a redistricting plan that "diminishe[d] the number of districts in which minority groups c[ould] 'elect their preferred candidates of choice.' " Bethune-Hill , 137 S.Ct. at 795 (citations omitted). The parties agree that, following the Supreme Court's decision in Shelby County v. Holder , 570 U.S. 529, 133 S.Ct. 2612, 186 L.Ed.2d 651 (2013), Virginia no longer must comply with Section 5. Jan. 10, 2019 Tr. at 137-38, 142-43, 151. Nevertheless, for the reasons discussed below with respect to Section 2, we conclude that black voters will retain their ability to elect their preferred candidates under the plan we adopt.

7 At trial, the intervenors asserted compliance with Section 5, and not Section 2, of the VRA as the compelling state interest justifying the predominant use of race. Bethune II , 326 F.Supp.3d at 143.

8 Contrary to the NAACP's contention, we do not apply an independent "minimum change" requirement to our remedial plan. Rather, we make changes as necessary to effectuate an adequate remedy. By confining our changes to boundaries of invalidated and adjacent districts, we respect our limited role in imposing a court-ordered redistricting plan, namely, to ensure only that the identified constitutional violations are fully remedied within the bounds of federal law, and nothing more.

9 The intervenors asserted at oral argument that, because this court did not credit the testimony of Delegate Steven Christopher Jones and John Morgan regarding their use of race in drafting the 2011 map, we should not accept Dr. Grofman's testimony that he did not apply a 55% BVAP ceiling. Jan. 10, 2019 Tr. at 118-21. Because this argument lacks a logical basis, we cannot respond to the argument directly. However, we have evaluated all witnesses' credibility in context, including the witnesses' demeanor and the quality of their recollection, as well as the plausibility of their testimony in light of all the other evidence. In our view, the fact that Dr. Grofman's proposed districts have BVAP levels below 55% is not evidence of racial motive, but rather is a foreseeable consequence of applying traditional districting criteria.

10 We note that Dr. Grofman considered "electability," i.e. , the strength of partisan preference, in the invalidated districts when evaluating whether minority voters in those districts would retain an opportunity to elect their preferred candidates under his proposed plans. Grofman Rep. at 65-66. This attention to race was required only to ensure that any court-ordered plan does not inadvertently dilute the voting strength of protected minority groups.

11 Dr. Grofman rejected outright the plans proposed by the William & Mary students due to several errors in census block assignments. Grofman Rep. at 23 n.15. For the same reason, we do not consider those plans.

12 The version of Dr. Grofman's plan we have selected includes changes to 25 districts.

13 As noted above, we found in Bethune I , and the Supreme Court affirmed, that the legislature predominantly used race to construct District 75, another majority-minority district located to the immediate south of District 63, but that this use of race was justified. 141 F.Supp.3d at 555-59 ; Bethune-Hill , 137 S.Ct. at 800-02.

14 The plaintiffs, NAACP, and Princeton Gerrymandering Project agree that the Petersburg 2 module is the most appropriate module of the three that Dr. Grofman submitted for that region.

15 Although the Petersburg 2 module changes the boundaries of five districts, more than the number of boundaries altered in Dr. Grofman's other proffered modules, we conclude that these changes are necessary to remedy the constitutional violation. Grofman Rep. at 81. Dr. Grofman's modules "Petersburg 1A" and "Petersburg 1B" maintain, either entirely or in large part, the race-based split of Dinwiddie County. Grofman Rep. at 81. As this split was among the most egregious examples of racial sorting in the 2011 map, we decline to implement a remedial plan that does not correct that problem.

16 See Bethune I , 141 F.Supp.3d at 535, for a detailed description of the various compactness measures.

17 Dr. Palmer did not identify a specific minimum BVAP percentage required for black voters to retain their ability to elect their preferred candidates in these or any other districts. 2nd Trial Tr. at 429-30.

18 Dr. Grofman explained that "the Fairfax vote share in a statewide contest recompiled within a (new) district boundary is a conservative estimate of potential minority support in a Democratic primary," because "a minority candidate residing locally with some degree of name recognition within the much smaller confines of a legislative district could be expected to win more votes (likely, considerably more votes) in the Democratic primary in the district [than] what was obtained by Mr. Fairfax." Grofman Rep. at 44.

19 President Obama and Mr. Fairfax were the candidates preferred by black voters in these elections. Grofman Rep. at 42-44; see also Pl. Ex. 71 at 26 ¶ 136 (noting that in all twelve challenged districts, black voters voted for Democratic candidates at an average rate of 95%).

20 The NAACP agrees that none of Dr. Grofman's proposed modules for any of the invalidated districts results in dilution of black voters' voting rights under Section 2 of the VRA. See Jan. 10, 2019 Tr. at 103.

21 See Bethune II , 326 F.Supp.3d at 137 n.4, for a description of VTDs, or "voting tabulation districts."

22 The 2011 version of non-challenged District 27 was located wholly within Chesterfield County, and non-challenged Districts 72 and 73 were located wholly within Henrico County. Grofman Rep. at 76.

23 In the Richmond 1A module, the city of Richmond is split into four segments. Grofman Rep. at 78. Although Dr. Grofman averred that he could have constructed a map in which Richmond was split into three segments, he declined to do so in order to avoid pairing any of the four 2017 incumbents residing within the city limits. Grofman Rep. at 67-68, 151.

In contrast, the intervenors' proposed plan, HB 7002, splits Richmond into five segments, a result that is not supported by the population of Richmond. Grofman Rep. at 126. HB 7002 also splits Chesterfield County into seven segments, compared with three segments in the Richmond 1A module. Grofman Rep. at 78, 126. These unnecessary city and county splits in HB 7002 lead us to favor the Richmond 1A module over the intervenors' proposal.

24 In contrast to our selected Petersburg and peninsula modules, see Grofman Rep. at 83, 89, 92, 96, the Richmond 1A module does not improve the overall compactness of districts in the region. Grofman Rep. at 76, 78. Although compactness is a traditional districting criterion and a desirable goal, other important considerations in drawing the Richmond-area districts precluded a remedy that offered more compact districts.

25 We observe that the intervenors assert both that the special master made unnecessarily extensive changes to the 2011 map, but also failed to make enough changes to remedy the identified constitutional violations. These contradictory positions taken by the intervenors underscore the delicate "balancing act" inherent in drawing remedial legislative districts.

26 For this reason, we decline to adopt the "Richmond 1B" module prepared by Dr. Grofman. The Richmond 1B module proposes substantial changes to non-challenged Districts 72 and 73 that Dr. Grofman concedes "are not required to effectuate a constitutional remedy," but instead improve the compactness of District 72. Grofman Rep. at 8, 73, 141 n.59.

27 Between Dr. Grofman's two proposed modules for the peninsula, the plaintiffs, NAACP, and Princeton Gerrymandering Project prefer the Peninsula 2 module.

28 Non-challenged District 91 is not contiguous by land in the Peninsula 2 module. See Dkt. No. 355 Ex. 3. However, Dr. Grofman testified that the district is contiguous by water based on the census block geography underlying the area of water contiguity. See Jan. 10, 2019 Tr. at 38-40; see also Grofman Rep. at 32 n.20, 162. Although water contiguity is not an ideal result, we are satisfied that, under the circumstances, District 91 is sufficiently contiguous in our remedial map to comply with traditional districting criteria. See Wilkins v. West , 264 Va. 447, 571 S.E.2d 100, 109-10 (2002) (holding that water contiguity can satisfy the Virginia constitutional requirement of contiguity).

29 The BVAP levels in invalidated Districts 80 and 89 also decrease in the Norfolk 1A module, but remain above 50%. Grofman Rep. at 101. For the reasons explained above with respect to Districts 77 and 90, we conclude that the redrawn versions of Districts 80 and 89 will not impair black voters' ability to elect their preferred candidates.