Rodney Pierce v. North Carolina State Board of Elections

97 F.4th 194

Opinion

PUBLISHED

UNITED STATES COURT OF APPEALS FOR THE FOURTH CIRCUIT

No. 24-1095

RODNEY D. PIERCE; MOSES MATTHEWS, Plaintiffs – Appellants,

v.

THE NORTH CAROLINA STATE BOARD OF ELECTIONS; ALAN HIRSCH, in his official capacity as Chair of the North Carolina State Board of Elections; JEFF CARMON, III, in his official capacity as Secretary of the North Carolina State Board of Elections; STACY "FOUR" EGGERS, IV, in his official capacity as a member of the North Carolina State Board of Elections; KEVIN LEWIS, in his official capacity as a member of the North Carolina State Board of Elections; SIOBHAN MILLEN, in her official capacity as a member of the North Carolina State Board of Elections; PHILLIP E. BERGER, in his official capacity as President Pro Tem of the North Carolina Senate; TIMOTHY K. MOORE, in his official capacity as Speaker of the North Carolina House of Representatives,

Defendants – Appellees.

------------------------------ ROY A. COOPER, III; JOSHUA STEIN, Amici Supporting Appellants.

Appeal from the United States District Court for the Eastern District of North Carolina, at Greenville. James C. Dever III, District Judge. (4:23-cv-00193-D-RN)

Argued: February 15, 2024 Decided: March 28, 2024

2

Before WILKINSON, GREGORY, and RUSHING, Circuit Judges.

Affirmed by published opinion. Judge Rushing wrote the majority opinion, in which Judge Wilkinson joined. Judge Gregory wrote a dissenting opinion.

ARGUED: Elisabeth S. Theodore, ARNOLD & PORTER KAYE SCHOLER LLP, Washington, D.C., for Appellants. Phillip John Strach, NELSON MULLINS RILEY & SCARBOROUGH, LLP, Raleigh, North Carolina, for Appellees. ON BRIEF: Edwin M. Speas, Jr., POYNER SPRUILL LLP, Raleigh, North Carolina; R. Stanton Jones, Samuel I. Ferenc, ARNOLD & PORTER KAYE SCHOLER LLP, Washington, D.C., for Appellants. Thomas A. Farr, Alyssa M. Riggins, Cassie A. Holt, Alexandra M. Bradley, NELSON MULLINS RILEY & SCARBOROUGH LLP, Raleigh, North Carolina; Richard B. Raile, Katherine L. McKnight, Trevor M. Stanley, Benjamin D. Janacek, Washington, D.C., Patrick T. Lewis, Cleveland, Ohio, Rachel Palmer Hooper, Tyler G. Doyle, BAKER & HOSTETLER LLP, Houston, Texas, for Appellees. Joshua H. Stein, Attorney General, Ryan Y. Park, Solicitor General, James W. Doggett, Deputy Solicitor General, Lindsay Vance Smith, Deputy Solicitor General, South A. Moore, Deputy General Counsel, Mary Elizabeth D. Reed, Solicitor General Fellow, NORTH CAROLINA DEPARTMENT OF JUSTICE, Raleigh, North Carolina, for Amici Curiae.

3

RUSHING, Circuit Judge:

North Carolinians are currently going to the polls to vote in primary elections for their state senators, among many other public offices. 1 Each voter casts a ballot for a candidate to represent their respective district among the State’s 50 Senate districts, as recently reconfigured by the General Assembly in October 2023. In November, Plaintiffs—two North Carolina voters—sued the State Board of Elections and its members, along with the President pro tempore of the North Carolina Senate and the Speaker of the North Carolina House of Representatives, alleging that the boundaries of Senate Districts 1 and 2 in eastern North Carolina violate Section 2 of the Voting Rights Act of 1965 (VRA). In addition to permanent injunctive relief, Plaintiffs also sought a preliminary injunction barring use of Senate Districts 1 and 2 and ordering use of new districts drawn by Plaintiffs in the 2024 elections.

After conducting a hearing and considering all the parties’ evidence, the District Court for the Eastern District of North Carolina denied the requested preliminary injunction. The district court concluded that Plaintiffs have not shown the extraordinary circumstances necessary to justify disrupting the status quo before trial; that Plaintiffs have not proven they are likely to succeed on the merits of their VRA claim; and that equitable factors, including proximity to the 2024 elections, counsel against preliminary injunctive

1

In the time since this opinion was written, the primary election has ended, the county boards of elections have conducted their post-election canvass, and the state and county boards of elections have certified the final results in all contests. Voting in runoff primaries, if any, will begin in April.

4

relief. Plaintiffs appealed, and we granted their motion to expedite our review. We now affirm the judgment of the district court and remand for continued proceedings.

I.

Plaintiffs challenge the electoral map the North Carolina General Assembly enacted in Senate Bill 758 (SB 758) in 2023. That map wasn’t drawn in a vacuum, so we begin with some legal and historical context.

A.

Under the Equal Protection Clause of the Fourteenth Amendment to the United States Constitution, “districting maps that sort voters on the basis of race ‘are by their very nature odious.’” Wis. Legislature v. Wis. Elections Comm’n, 142 S. Ct. 1245, 1248 (2022) (per curiam) (quoting Shaw v. Reno, 509 U.S. 630, 643 (1993)). States cannot enact such maps “‘unless they are narrowly tailored to achieving a compelling state interest.’” Id. (quoting Miller v. Johnson, 515 U.S. 900, 904 (1995)). At the same time, compliance with the VRA “often insists that districts be created precisely because of race.” Abbott v. Perez, 138 S. Ct. 2305, 2314 (2018). “In an effort to harmonize these conflicting demands,” the Supreme Court has assumed that complying with the VRA, and Section 2 in particular, is a compelling interest. Id. at 2315; Wis. Legislature, 142 S. Ct. at 1248.

As relevant here, a State violates Section 2 of the VRA “if, based on the totality of circumstances, it is shown that the political processes leading to nomination or election in the State or political subdivision are not equally open to participation by members of a [racial minority 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

5

choice.” 52 U.S.C. § 10301(b). The Supreme Court has construed Section 2 “to prohibit the distribution of minority voters into districts in a way that dilutes their voting power.” Wis. Legislature, 142 S. Ct. at 1248. In Thornburg v. Gingles, 478 U.S. 30 (1986), the Supreme Court articulated a framework for demonstrating this sort of violation. “First, three ‘preconditions’ must be shown: (1) The minority group must be sufficiently large and compact to constitute a majority in a reasonably configured district, (2) the minority group must be politically cohesive, and (3) a majority group must vote sufficiently as a bloc to enable it to usually defeat the minority group’s preferred candidate.” Wis. Legislature, 142 S. Ct. at 1248 (quoting Gingles, 478 U.S. at 50–51); see also Allen v. Milligan, 143 S. Ct. 1487, 1503 (2023). Then, “a plaintiff who demonstrates the three preconditions must also show, under the ‘totality of circumstances,’ that the political process is not ‘equally open’ to minority voters.” Milligan, 143 S. Ct. at 1503 (quoting Gingles, 478 U.S. at 45–46).

Before a State may engage in race-based districting, it must have “‘a strong basis in evidence’ for concluding that [Section 2] required its action,” i.e., “that it would transgress the [VRA] if it did not draw race-based district lines.” Cooper v. Harris, 137 S. Ct. 1455, 1464 (2017) (quoting Ala. Legislative Black Caucus v. Alabama, 575 U.S. 254, 278 (2015)). Put differently, the Equal Protection Clause “does not allow a State to adopt a racial gerrymander that the State does not, at the time of imposition, ‘judg[e] necessary under a proper interpretation of the VRA.’” Wis. Legislature, 142 S. Ct. at 1250 (quoting Cooper, 137 S. Ct. at 1472).

A state legislature attempting to produce a districting plan that comports with both the Equal Protection Clause (which “restricts consideration of race”) and the VRA (which

6

“demands consideration of race”) is thus “vulnerable to ‘competing hazards of liability.’” Perez, 138 S. Ct. at 2315 (quoting Bush v. Vera, 517 U.S. 952, 977 (1996) (plurality opinion)). Over the past three decades, the North Carolina General Assembly has attempted to navigate those hazards with mixed success. A brief survey of its efforts provides helpful context for understanding the current case.

B.

Though we could go back further, 2 we begin in 2003 when the North Carolina General Assembly adopted a redistricting plan that divided Pender County in southeastern North Carolina between two state House districts. See Pender County v. Bartlett, No. 04CVS06966, 2006 WL 4077037 (N.C. Super. Ct. Jan. 9, 2006). Pender County sued, arguing the 2003 redistricting plan violated the North Carolina Constitution’s Whole County Provision, which prohibits counties from being divided “in the formation of a representative district.” N.C. Const. art. II, § 5(3); see also id. § 3(3) (same, Senate districts). The State defended the map as an effort to comply with Section 2 of the VRA by creating a crossover district—i.e., a district in which the minority population is not a

2

In the 1990 redistricting cycle, North Carolina created two majority-minority congressional districts in response to the Attorney General’s preclearance demand under Section 5 of the VRA. The ensuing lawsuit resulted in the Supreme Court’s first case recognizing a racial gerrymandering claim. In Shaw v. Reno, 509 U.S. 630 (1993), the Court held that challengers to North Carolina’s two majority-minority districts stated a claim under the Equal Protection Clause. The Supreme Court subsequently invalidated one of those congressional districts because it was not narrowly tailored to the State’s asserted interest in complying with Section 2 of the VRA. Shaw v. Hunt, 517 U.S. 899, 918 (1996). After North Carolina redrew that congressional district, a district court held it unconstitutional and the Supreme Court reversed, concluding that the district court clearly erred in finding that race, rather than politics, predominantly explained the district’s boundaries. Easley v. Cromartie, 532 U.S. 234, 243–244 (2001).

7

majority but is large enough to elect the candidate of its choice with help from voters who are members of the majority and who cross over to support the minority’s preferred candidate. See Pender County v. Bartlett, 649 S.E.2d 364, 367 (N.C. 2007) (“Past election results in North Carolina demonstrate[d] that a legislative voting district with . . . an African-American voting age population of at least 38.37 percent, create[d] an opportunity to elect African-American candidates,” so “the General Assembly fashioned House District 18 with . . . an African-American voting age population of 39.36 percent” to create an “effective black voting district.” (internal quotation marks omitted)).

The Supreme Court rejected the State’s VRA defense, explaining that “Section 2 does not impose on those who draw election districts a duty to give minority voters the most potential, or the best potential, to elect a candidate by attracting crossover voters.” Bartlett v. Strickland, 556 U.S. 1, 15 (2009) (plurality opinion). Instead, the Court held that the first Gingles precondition requires “the minority population in the potential election district [to be] greater than 50 percent.” Id. at 20. Because Section 2 did not require crossover districts, it could not justify a violation of state law, namely the Whole County Provision. Id. at 14; see also id. at 21 (“If § 2 were interpreted to require crossover districts throughout the Nation, it would unnecessarily infuse race into virtually every redistricting, raising serious constitutional questions.” (internal quotation marks omitted)).

The General Assembly then adopted a policy of creating majority-minority districts.

In the 2011 redistricting plans, the legislature created twenty-three majority-black state House districts and nine majority-black state Senate districts. Covington v. North Carolina, 316 F.R.D. 117, 134 (M.D.N.C. 2016) (three-judge district court), aff’d, 137 S. Ct. 2211

8

(2017) (mem.). A group of voters challenged twenty-eight of those districts, which included northeastern counties relevant to this case. Id. at 128, 142, 151–152, 159. The State defended the districts as an effort to comply with Section 2 of the VRA. A three- judge panel rejected the State’s defense under the third Gingles precondition and declared the maps unconstitutional. Id. at 124. The panel reasoned that the General Assembly did not have “a strong basis in evidence” for thinking that racial bloc voting operated at such a level as “would enable the majority usually to defeat the minority group’s candidate of choice” in those districts. Id. at 167. The mere existence of racially polarized voting was not enough. Id. at 167–168. And evidence demonstrated that minority-preferred candidates “were already consistently winning” in the challenged areas without majority- minority districts. Id. at 172–173; see id. at 126.

Another lawsuit challenged two majority-minority congressional districts in the 2011 redistricting plan, one of which (CD1) included multiple northeastern counties at issue in this case. Harris v. McCrory, 159 F. Supp. 3d 600, 604 (M.D.N.C. 2016) (three- judge district court); see also Cooper, 137 S. Ct. at 1483–1484. It met the same fate as the suit challenging the state legislative districts. A three-judge panel concluded that “there [was] no evidence that the white majority votes sufficiently as a bloc to enable it . . . usually to defeat the minority’s preferred candidate” in CD1. Harris, 159 F. Supp. 3d at 624 (internal quotation marks omitted). In fact, the evidence “vividly demonstrate[d]” that “significant crossover voting by white voters” occurred in CD1. Id. at 625; see also id. at 606 (“For decades, African-Americans enjoyed tremendous success in electing their preferred candidates in former versions of CD 1 and CD 12 regardless of whether those

9

districts contained a majority black voting age population.”). The Supreme Court agreed. Cooper, 137 S. Ct. at 1468. Electoral history “provided no evidence that a § 2 plaintiff could demonstrate the third Gingles prerequisite” for CD1. Id. at 1470. Rather, the evidence showed that “the district’s white population did not vote sufficiently as a bloc to thwart black voters’ preference.” Id. (internal quotation marks and brackets omitted). As a result, the State had “no reason to think that the VRA required it” to create a majority- black district to avoid Section 2 liability. Id.; see also id. at 1471 (“North Carolina too far downplays the significance of a longtime pattern of white crossover voting in the area that would form the core of the redrawn District 1.”).

For the 2017 remedial redistricting after Covington, the General Assembly implemented a policy forbidding consideration of race. North Carolina v. Covington, 138 S. Ct. 2548, 2550 (2018) (per curiam). Yet the “dizzying succession of litigation” over North Carolina’s electoral districts continued—now on a partisan gerrymandering theory. Common Cause v. Lewis, No. 18 CVS 014001, 2019 WL 4569584, at *1 (N.C. Super. Ct. Sept. 3, 2019). A state court declared the partisan gerrymandering theory justiciable under the North Carolina Constitution and ordered the General Assembly to revise its state House and Senate districting maps. Id. at *124, *135–137. As part of the remedial phase of that litigation, the state court endorsed the analysis of the plaintiffs’ experts that, based on the minimum black voting-age population necessary for black voters to elect their candidates of choice, Gingles would not justify any majority-black district in any of the areas at issue there.

10

A few years later, the General Assembly’s 2021 redistricting plan for congressional, state House, and state Senate districts (which was also drawn without racial data) met with another partisan gerrymandering lawsuit. See N.C. League, of Conservation Voters, Inc. v. Hall, No. 21 CVS 015426, 2022 WL 124616, at *1–3 (N.C. Super. Ct. Jan. 11, 2022). The case made its way to the North Carolina Supreme Court, which affirmed the justiciability of the partisan gerrymandering claims under the State’s constitution. Harper v. Hall, 868 S.E.2d 499, 559 (N.C. 2022) (Harper I). The court ordered the maps intended for the 2022 election to be redrawn. Id.; see also Harper v. Hall, 881 S.E.2d 156, 161–162 (N.C. 2022) (Harper II) (reviewing constitutionality of remedial maps after 2022 elections).

Not long after, however, the North Carolina Supreme Court granted rehearing in Harper I and reversed course. Harper v. Hall, 886 S.E.2d 393, 416 (N.C. 2023) (Harper III). The court held that partisan gerrymandering claims “are nonjusticiable, political questions under the North Carolina Constitution.” Id. It further ordered that, because both the 2021 and 2022 maps were the byproduct of a “mistaken understanding of the North Carolina Constitution,” the General Assembly “shall have the opportunity to enact a new set of legislative and congressional redistricting plans” guided by federal and state law. Id. at 446–448 (referring to maps required by erroneous decisions in Lewis and Harper I).

C.

Taking that opportunity, the General Assembly went back to the drawing board and emerged with a new set of maps to be used for the 2024 elections. See 2023 N.C. Sess. Laws 145 (establishing congressional districts), 146 (establishing Senate districts), 149

11

(establishing House districts). This case concerns only the Senate district map, which began as SB 758. 3 The General Assembly “did not use racial data” when drawing the Senate map. J.A.

643; see Pierce v. N.C. State Bd. of Elections, --- F. Supp. 3d ---, 2024 WL 307643, at *4 (E.D.N.C. Jan. 26, 2024) (finding that “the General Assembly did not have race in the computer when it created the Senate, House, and Congressional redistricting plans in 2023”). Instead, the General Assembly considered other factors, including preserving communities of interest, following traditional redistricting principles, and adhering to the Whole County Provision.

This case concerns Senate District 1 (SD1) and Senate District 2 (SD2) in northeastern North Carolina. SD1 contains Bertie, Camden, Currituck, Dare, Gates, Hertford, Northampton, Pasquotank, Perquimans, and Tyrrell Counties. 2023 N.C. Sess. Laws 146. “SD1 kept together four of North Carolina’s five finger counties,” and “many SD1 residents in these counties work or travel frequently to the Virginia Tidewater region.” Pierce, 2024 WL 307643, at *4 (internal quotation marks omitted). Seven of SD1’s ten counties and 81 percent of its population “are in the Norfolk, Virginia media market.” Id. SD2 contains Carteret, Chowan, Halifax, Hyde, Martin, Pamlico, Warren, and Washington Counties. 2023 N.C. Sess. Laws 146. “SD2 follows the Roanoke River from Warren County to Washington County.” Pierce, 2024 WL 307643, at *4. “Five of SD2’s eight

3

This is not the only case challenging the maps adopted in 2023. See Complaint, Williams v. Hall, No. 1:23-cv-01057-TDS-JLW (M.D.N.C. Dec. 4, 2023); Complaint, N.C. State Conf. of the NAACP v. Berger, No. 1:23-cv-01104-WO-LPA (M.D.N.C. Dec. 19, 2023). However, this is the only case seeking preliminary injunctive relief.

12

counties are in the Greenville, North Carolina media market.” Id. And SD1 and SD2 “both include their respective incumbent senator’s residence.” Id.

“Before the General Assembly enacted SB 758 in October 2023, [it] held public hearings throughout North Carolina, including one in Elizabeth City in northeast North Carolina, to gather public input on the proposed Senate districts.” Id. The General Assembly “also accepted public comments through an online portal” on its website. Id. After legislators publicly filed the bill, but before enacting it, they directed legislative staff to “load racial data” into the map software, overlay it on the maps, and “make that information publicly available on the General Assembly website as soon as possible.” J.A. 645. The Senate Committee on Redistricting and Elections then set a public hearing for the following week to “consider any evidence that a member of [the] committee or a third party advocating altering plans for racial reasons brings forth that provides a strong basis in evidence that the Gingles preconditions are present in a particular area of the [S]tate” so as to warrant “amending the districts” to satisfy Section 2 of the VRA. J.A. 645–646.

Plaintiffs did not submit any evidence to the Committee. The Southern Coalition for Social Justice submitted a letter to the Committee stating that SD1 and SD2 would dilute the voting strength of black voters but did not request a majority-black district. See Letter from S. Coal. for Soc. Just., to Sen. Phil Berger et al. 2 (Oct. 22, 2023). Instead, the Coalition proposed that any potential vote dilution could be resolved by changing SD1 and SD2 to match former Senate Districts 1 and 3 under the 2022 remedial map. Id. at 3. Former Senate Districts 1 and 3 involved the same counties as SD1 and SD2, and, according to the Coalition, the former grouping provided black voters “with an opportunity

13

to elect a candidate of their choice” due to crossover voting. Id. at 3–4. SB 758 was not amended, and it passed both houses and became law on October 25, 2023. 4 2023 N.C. Sess. Laws 146.

The Senate district map remained on the books for roughly one month before being challenged. On November 20, Rodney Pierce and Matthew Moses, two black voters residing in SD2, filed this lawsuit alleging that SD1 and SD2 diluted the voting strength of black voters in northeastern North Carolina in violation of Section 2 of the VRA. Plaintiffs sought declaratory relief, preliminary and permanent injunctions barring the State from “enforcing or giving effect to” SD1 and SD2, and any “actions necessary to order the adoption of a valid state Senate plan . . . in time to use the remedial plan in the 2024 Senate elections.” J.A. 32. Two days later, Plaintiffs moved for a preliminary injunction barring enforcement of SD1 and SD2 and ordering “immediate use of Plaintiffs’ proposed remedial districts (labeled Demonstration Districts B-1 and B-2)” instead of SD1 and SD2, along with “use of Districts 3-50 from the 2023 enacted map” for the rest of the State. Dist. Ct. Dkt. ECF No. 16 at 2.

Legislative Defendants opposed the preliminary injunction motion, 5 and Plaintiffs filed a reply on December 26. On December 29, the district court indicated that it had questions for the advocates and scheduled a hearing on the motion for January 10, 2024.

4

Because the Governor cannot veto redistricting legislation, the bill took effect upon passage. See N.C. Const. art. II, § 22(5).

5

The North Carolina State Board of Elections and its members have consistently taken no position on the merits of Plaintiffs’ motion for preliminary injunctive relief or Plaintiffs’ appeal.

14

The same day, Plaintiffs filed an appeal to this Court, claiming that the district court had functionally denied their preliminary injunction motion by scheduling a hearing. We dismissed the appeal for lack of jurisdiction.

“[M]indful of the time sensitive nature of the issues in this case,” the district court proceeded with the preliminary injunction hearing while it waited for this Court’s mandate to issue. J.A. 852. About two weeks later, the district court issued a fulsome, sixty-nine- page opinion denying Plaintiffs preliminary injunctive relief. As an initial matter, the court observed that Plaintiffs were requesting the creation of “a racially gerrymandered majority- black Senate district in northeast North Carolina” even though: “the 2024 Senate elections are underway”; there was “no evidence . . . the General Assembly in 2023 [had] a strong basis in evidence to believe that Section 2 required [it] to create a majority-black Senate district in northeast North Carolina”; “insufficient evidence shows that Section 2 requires a majority-black Senate district in northeast North Carolina”; and “federal litigation from 2011 to 2016 helped to show that there was not legally significant racially polarized voting in North Carolina, including in the counties in northeast North Carolina at issue in this case.” Pierce, 2024 WL 307643, at *1.

Moving through the detailed layers of legal requirements that Plaintiffs must satisfy, the district court rejected Plaintiffs’ arguments at nearly every step of the analysis. At the outset, the court determined that Plaintiffs’ request for injunctive relief sought to alter rather than maintain the status quo and that they failed to justify such an extraordinary form of relief. Even so, the court went on to analyze Plaintiffs’ motion under the traditional test for preliminary injunctions. It concluded Plaintiffs failed to show a likelihood of success

15

on the merits and a likelihood of irreparable harm because they had not satisfied all three Gingles preconditions required to prevail under Section 2. The court assumed Plaintiffs’ Demonstration District A satisfied the majority-minority requirement under the first Gingles precondition but held that Demonstration District B-1 did not because it has a black voting-age population (BVAP) of less than 50 percent. Legislative Defendants did not contest the second Gingles precondition, so the court found it satisfied. As for the third Gingles precondition, the court held that Plaintiffs failed to demonstrate legally significant, as opposed to statistically significant, racially polarized voting. Moreover, even if Plaintiffs had satisfied all three Gingles preconditions, the court found that they had failed to show under the totality of circumstances that the political process is not equally open to minority voters. Most of the applicable factors in the totality of circumstances inquiry either weighed against or did not support Plaintiffs, and the court found that “partisanship better explains polarized voting in North Carolina than race.” Id. at *26.

Despite concluding that Plaintiffs were unlikely to succeed on the merits or suffer irreparable harm, the district court went on to balance the equities, which, again, disfavored preliminary injunctive relief. Plaintiffs had delayed filing a complaint. The State would suffer irreparable harm if it were precluded from enforcing the Senate district map while the litigation proceeded. And, most importantly, preliminary injunctive relief would violate the Purcell principle, which cautions courts against enjoining state election laws in the period close to an election. See Purcell v. Gonzalez, 549 U.S. 1, 4–5 (2006) (per curiam); Merrill v. Milligan, 142 S. Ct. 879, 880–881 (2022) (Kavanaugh, J., concurring).

Plaintiffs appealed, and we have jurisdiction under 28 U.S.C. § 1292(a)(1).

16

II.

In challenging the district court’s denial of preliminary injunctive relief, Plaintiffs shoulder a heavy burden on appeal. A preliminary injunction is “an extraordinary remedy” that “may only be awarded upon a clear showing that the plaintiff is entitled to such relief.” Winter v. Nat. Res. Def. Council, Inc., 555 U.S. 7, 22 (2008). Thus, Plaintiffs can prevail only if it is clear that they are “likely to succeed on the merits,” that they are “likely to suffer irreparable harm in the absence of preliminary relief, that the balance of equities tips in [their] favor, and that an injunction is in the public interest.” Id. at 20.

Moreover, Plaintiffs seek a particularly aggressive form of preliminary injunction, one that is “‘disfavored’” in “‘any circumstance.’” League of Women Voters of N.C. v. North Carolina, 769 F.3d 224, 235 (4th Cir. 2014) (quoting Taylor v. Freeman, 34 F.3d 266, 270 n.2 (4th Cir. 1994)). Plaintiffs are not asking the Court to maintain the status quo until after a trial and final judgment, which is the traditional function of a preliminary injunction. See Pashby v. Delia, 709 F.3d 307, 319 (4th Cir. 2013). Rather, Plaintiffs seek an order altering the status quo before the case even begins, what we have called a “mandatory” injunction. League of Women Voters, 769 F.3d at 236. Mandatory preliminary injunctions are “warranted only in the most extraordinary circumstances.” Taylor, 34 F.3d at 270 n.2; see also Wetzel v. Edwards, 635 F.2d 283, 286 (4th Cir. 1980) (“Mandatory preliminary injunctions do not preserve the status quo and normally should be granted only in those circumstances when the exigencies of the situation demand such relief.”).

17

Plaintiffs dispute that they seek a mandatory injunction. But plainly the Senate district map adopted in SB 758 is already in place and constitutes the status quo. The General Assembly adopted that map in October 2023, and it was uncontested for a month before Plaintiffs sued. See League of Women Voters, 769 F.3d at 236 (defining the status quo as “the last uncontested status between the parties which preceded the controversy” (internal quotation marks omitted)); Wise v. Circosta, 978 F.3d 93, 98 (4th Cir. 2020) (en banc) (explaining that “the state’s action” in an election procedures case “establishes the status quo”). This case is not like League of Women Voters, where the plaintiffs filed suit “the very same day [the challenged law] was signed.” 769 F.3d at 236.

Plaintiffs’ assertion that maps used in 2020 (two revisions ago) are the status quo cannot be taken seriously. Even if that were the case—and it obviously is not—Plaintiffs don’t seek to preserve that supposed status quo. They do not ask the Court to reinstate the 2020 Senate district map (or even the 2022 map) but instead seek to impose an entirely different map: one that implements their Demonstration Districts B-1 and B-2. Thus, even under Plaintiffs’ own mistaken conception of the status quo, they clearly request court action to alter it while this case is pending. Simply put, Plaintiffs seek an (extra) extraordinary remedy that raises the stakes of an erroneous decision and erects a high bar for relief.

On top of the high standard for obtaining a mandatory preliminary injunction, Plaintiffs must also overcome a deferential standard of appellate review. We review the denial of a preliminary injunction for “whether the record shows an abuse of discretion by the district court, not whether [we] would have granted or denied the injunction.” Wetzel,

18

635 F.2d at 286. Pursuant to this standard, we review legal conclusions de novo and factual findings for clear error. Pashby, 709 F.3d at 319. Whether vote dilution has occurred is “peculiarly dependent upon the facts of each case.” Gingles, 478 U.S. at 79. So, like all other factual findings, “[t]he [d]istrict [c]ourt’s determination whether the § 2 requirements are satisfied must be upheld unless clearly erroneous.” League of United Latin Am. Citizens v. Perry (LULAC), 548 U.S. 399, 427 (2006). Our mere disagreement with the district court does not make its findings clearly erroneous. “If the district court’s account of the evidence is plausible in light of the record viewed in its entirety, the court of appeals may not reverse it even though convinced that had it been sitting as the trier of fact, it would have weighed the evidence differently.” Williams v. Martorello, 59 F.4th 68, 86 (4th Cir. 2023) (quoting United States v. Hall, 664 F.3d 456, 462 (4th Cir. 2012)). Indeed, “[w]here there are two permissible views of the evidence, the [district court’s] choice between them cannot be clearly erroneous.” Anderson v. City of Bessemer City, 470 U.S. 564, 574 (1985).

III.

The district court determined that Plaintiffs had failed to demonstrate the extraordinary circumstances necessary for a preliminary injunction that would disrupt the status quo and compel the race-based sorting of voters for the 2024 Senate elections in North Carolina while this case remains pending. Because the court ruled against Plaintiffs on all four preliminary injunction requirements, Plaintiffs necessarily dispute the court’s assessment of each one on appeal. We will address the requirements in pairs: first the

19

likelihood of success and irreparable harm, and then the balance of hardships and the public interest.

A.

A plaintiff seeking a preliminary injunction must make a clear showing that he is likely to succeed at trial and to suffer irreparable harm in the absence of preliminary relief. Winter, 555 U.S. at 20, 22; Pashby, 709 F.3d at 321. Plaintiffs contend that “they will be irreparably harmed if they are forced to vote in a district that dilutes their votes in violation of the VRA.” Dist. Ct. Dkt. ECF No. 17 at 27. The district court determined that, because Plaintiffs were unlikely to succeed on the merits of their Section 2 claim, they also had failed to demonstrate they will suffer irreparable harm absent an order preliminarily enjoining SB 758 for the 2024 Senate elections. The parties agree that, in this case, these two requirements rise or fall together. Therefore, like the district court and the parties, we focus on whether Plaintiffs have made a clear showing that they are likely to succeed on the merits of their Section 2 vote dilution claim. 6 To make that showing, Plaintiffs were required to demonstrate a likelihood of success on all three Gingles preconditions as well as the ultimate totality of circumstances inquiry. After lengthy analysis, the district court assumed without deciding that Plaintiffs’

6

Legislative Defendants briefly contend that Plaintiffs lack a private cause of action to enforce Section 2 and urge us to affirm on that alternative basis. See Ark. State Conf. NAACP v. Ark. Bd. of Apportionment, 86 F.4th 1204, 1206–1207 (8th Cir. 2023). The district court did not address this argument, and we decline to resolve it in the first instance on the thin briefing provided.

20

Demonstration District A 7—which Plaintiffs offered “solely for illustrative purposes to satisfy Gingles One” and not “for use in any election,” Dist. Ct. Dkt. ECF No. 42 at 5— satisfied the first Gingles precondition that black voters be a sufficiently large and geographically compact group to constitute a majority in a reasonably configured district. 8 As for the second precondition, the district court found Plaintiffs likely to succeed in

7

Plaintiffs’ Demonstration District A consists of some counties from SD1, some from SD2, and Vance County, which is in Senate District 11. Plaintiffs proposed Demonstration District A in isolation, without accounting for changes that would be necessary in the remainder of the Senate map to accommodate that hypothetical district. Legislative Defendants argued that Demonstration District A was not a reasonably configured district under the first Gingles precondition. According to Legislative Defendants’ expert, moving Vance County out of Senate District 11 would leave that district without sufficient population to support a single Senate district and lead to “a cascade of changes that are difficult to sort out,” requiring a new statewide Senate districting plan. Pierce, 2024 WL 307643, at *14 (internal quotation marks omitted). Doing so would disrupt North Carolina’s traditional redistricting principles, including respect for county groupings throughout the State, and potentially dismantle crossover districts neighboring existing SD1 and SD2. Although the district court did not resolve these disputes at this juncture, it found that Legislative Defendants’ arguments “ha[d] force.” Id. at *15.

8

By contrast, the district court concluded that Plaintiffs’ Demonstration District B-

1 was unlikely to satisfy the first Gingles precondition because black voters were not more than 50 percent of the voting-age population of that hypothetical district. See Strickland, 556 U.S. at 19–20 (“It remains the rule . . . that a party asserting § 2 liability must show by a preponderance of the evidence that the minority population in the potential election district is greater than 50 percent.”). According to Plaintiffs, the BVAP of Demonstration District B-1 was 48.41 percent and the black citizen voting-age population (black CVAP) of that proposed district was 50.19 percent. The district court declined to use the CVAP statistic “[a]t this preliminary stage” because of uncertainty about its reliability and accuracy. Pierce, 2024 WL 307643, at *17. As the court explained, it had unanswered questions about the significant margins of error in the CVAP data set, and Plaintiffs had not explained how they arrived at their black CVAP figures, why BVAP and black CVAP differed in the proposed district, or why CVAP would be more accurate for this population. But the court left open the possibility that, “[a]fter discovery, [P]laintiffs may be able to demonstrate why the court should use black CVAP and why black CVAP is higher than BVAP in Demonstration District B-1.” Id.

21

showing that black voters in North Carolina are a “‘politically cohesive’” group whose members “‘usually vote for the same candidates.’” Pierce, 2024 WL 307643, at *17 (quoting Gingles, 478 U.S. at 56). Plaintiffs’ evidence faltered, in the district court’s estimation, on the third Gingles precondition and the totality of circumstances inquiry. We turn to those requirements now.

1.

The third Gingles precondition requires Plaintiffs “‘to demonstrate that the white majority votes sufficiently as a bloc to enable it . . . to defeat the minority’s preferred candidate.’” Milligan, 143 S. Ct. at 1503 (quoting Gingles, 478 U.S. at 51). This requires “racial bloc voting that is ‘legally significant,’” not merely statistically significant. Covington, 316 F.R.D. at 167 (quoting Gingles, 478 U.S. at 56). “‘[I]n the absence of significant white bloc voting it cannot be said that the ability of minority voters to elect their chosen representatives is inferior to that of white voters.’” Voinovich v. Quilter, 507 U.S. 146, 158 (1993) (quoting Gingles, 478 U.S. at 49 n.15). And “[i]n areas with substantial crossover voting”—that is, white voting for minority-preferred candidates—“it is unlikely that the plaintiffs would be able to establish the third Gingles precondition.” Strickland, 556 U.S. at 24. “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.” Covington, 316 F.R.D. at 168 (quoting Gingles, 478 U.S. at 56).

22

a.

The district court concluded that Plaintiffs, “[a]t this stage of the case,” had failed to make a clear showing they are likely to succeed on the third Gingles precondition. Pierce, 2024 WL 307643, at *22. Plaintiffs presented a report from one expert, Dr. Matt Barreto, assessing racially polarized voting for purposes of Gingles’ third precondition. But the district court found Barreto’s report unreliable, incomplete, and contradicted by other evidence.

Unreliable. Barreto performed ecological inference statistical models across 31 recent elections in 2020 and 2022, including statewide elections and elections in the “Northeast region” of North Carolina, which he defined to include some but not all counties in SD1 and SD2 as well as other counties not within those districts. J.A. 273. Based on those models, Barreto reported that between 80 and 88 percent of white voters in this “Northeast region” vote against black voters’ candidate of choice. Barreto also reported that black voters’ candidates of choice would have lost SD1 and SD2 in all but one election if those districts were in place in 2020 and 2022.

The one election a black-preferred candidate would have won according to Barreto was the 2022 state Senate race in SD2. The district court found this fact significant, noting that Barreto’s analysis of the 2022 Senate race in SD2 was one of just four elections he analyzed that most directly concerned whether black voters could elect Senate candidates of choice in SD1 and SD2—i.e., the reconstituted 2020 SD1 and SD2 Senate elections and the reconstituted 2022 SD1 and SD2 Senate elections.

23

At the hearing, the district court asked Plaintiffs about “this startling piece of [P]laintiffs’ evidence,” and “Plaintiffs responded that this figure must have been a typo.” Pierce, 2024 WL 307643, at *18. The court granted Plaintiffs’ request to supplement Barreto’s report. In his supplemental declaration, however, Barreto stated that this figure was not a typo. Instead, he opined that this figure resulted from a methodological flaw in his original analysis. As Barreto explained, his model excluded consideration of votes cast in uncontested elections, which allegedly skewed his results. Barretto claimed that, if votes in uncontested elections had been included in his analysis, his model would have predicted a loss for the black-preferred candidate in SD2 in 2022. Barreto did not discuss what effect, if any, this methodological shift would have on his other electoral predictions about SD1 and SD2, particularly the endogenous state House and Senate elections.

The district court was deeply troubled by Barreto’s changing methods and outcomes. From the court’s perspective, Barreto’s “belated explanation undercuts all of [his] conclusions by demonstrating that fuller data sets could change his estimated outcomes.” Id. at *19. Finding “profound discrepancies” between the methods of analysis performed in Barreto’s initial report and supplemental declaration, the district court was left wondering “why the court should credit any of his estimated outcomes for elections in SD2.” Id. At a minimum, the district court reasoned that the questions raised by Barretto’s supplemental declaration demonstrated “that this case would greatly benefit from discovery,” including Barreto’s deposition and complete data files. Id. As the court put it, “[t]his hotly contested factual issue weighs in favor of the court preserving the status quo ante litem.” Id.

24

Incomplete. The district court also found Barreto’s attempt to show legally significant racially polarized voting incomplete without a “‘district effectiveness analysis,’ which is ‘a district-specific evaluation used to determine the minority voting-age population level at which a district becomes effective in providing a realistic opportunity for . . . voters of that minority group to elect candidates of their choice.’” Id. (quoting Covington, 316 F.R.D. at 168 n.46). As the court explained, legally significant white bloc voting does not exist where crossover white voting enables black-preferred candidates to succeed. See Cooper, 137 S. Ct. at 1471–1472; Strickland, 556 U.S. at 15–16; Gingles, 478 U.S. at 56. And Section 2 “does not require crossover districts.” Strickland, 556 U.S. at 23. Thus, a district effectiveness analysis can support the third Gingles precondition by showing that “black voters’ candidates of choice cannot win elections” “‘without a VRA remedy,’” that is, a district in which the “BVAP . . . exceeds 50% plus one vote.” Pierce, 2024 WL 307643, at *19 (quoting Covington, 316 F.R.D. at 168).

The district court observed that, in Covington and Lewis, courts held that the North Carolina General Assembly had failed to justify using race to create majority-black House and Senate districts because it presented no evidence that majority-black districts were necessary for black-preferred candidates usually to win—i.e., a district effectiveness analysis. See, e.g., Covington, 316 F.R.D. at 168 (faulting map designer for “not conduct[ing] any district effectiveness analysis prior to drawing the districts”); Lewis, 2019 WL 4569584, at *100 (faulting legislators for not producing evidence “to establish the minimum African-American percentage of the voting age population (‘BVAP’) needed in any particular area of the State for the African American community to be able to elect the

25

candidate of its choice”). Without a proper district effectiveness analysis, the district court reasoned, the General Assembly could not use Section 2 to justify creating a majority-black district, whether on its own initiative or at the insistence of Plaintiffs.

Looking at the analysis Barreto did provide, the district court found it to “suggest[]

that a proper district effectiveness analysis for [P]laintiffs’ demonstration districts likely would yield a [BVAP] below 50% which provides ‘a realistic opportunity for . . . voters of that minority group to elect candidates of their choice.’” Pierce, 2024 WL 307643, at *20 (quoting Covington, 316 F.R.D. at 168 n.46). For example, Barreto estimated that black voters’ candidates of choice would have won every endogenous and exogenous election in Demonstration District B-1 in 2020 and 2022, even though that district had a BVAP of 48.41 percent.

Contradicted by other evidence. Finally, the district court found that other evidence of white crossover voting cast doubt on Plaintiffs’ attempt to satisfy the third Gingles precondition. For example, for the 2022 congressional elections, North Carolina’s first congressional district contained all the counties in Demonstration District A and Demonstration District B-1. Congressional District 1 had a BVAP of approximately 40 percent and in 2022 elected Congressman Don Davis, a black Democrat. This evidence of white crossover voting in northeastern North Carolina, including in the counties at issue in Plaintiffs’ proposed remedial districts, was relevant to the court’s assessment of Plaintiffs’ evidence. Further, the court noted Plaintiffs’ implicit concession that legally significant racially polarized voting does not exist in neighboring Senate District 5. And scatterplots aggregating electoral outcomes in Barreto’s report showed that black voters’ candidates of

26

choice begin winning precincts in the North Carolina counties at issue here when BVAP meets or exceeds 30 to 40 percent.

In conclusion, “[o]n the current record,” the district court found:

(1) “the black voting-age population in the counties at issue in this case live and work in” “‘communities in which [they] 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’”; and

(2) “the white voting-age population in the communities at issue do not vote as a bloc against black-preferred candidates to enable the white bloc usually to defeat the black-preferred candidates.”

Id. at *21 (quoting Johnson v. De Grandy, 512 U.S. 997, 1020 (1994)).

b.

On appeal, Plaintiffs largely argue as though we may decide the facts anew—they urge us to find certain facts more persuasive than others, attempt to rehabilitate Barreto’s report, and present new evidence to bolster their case. But we are legally obligated to decline their invitation to “reweigh the evidence presented to the district court,” as that is not our function. United States v. Charleston County, 365 F.3d 341, 349 (4th Cir. 2004). The Supreme Court has instructed that we must uphold the district court’s “determination whether the § 2 requirements are satisfied . . . unless [it is] clearly erroneous.” LULAC, 548 U.S. at 427; Milligan, 143 S. Ct. at 1506 (same, at preliminary injunction stage). A district court’s Section 2 findings may be “certainly disputable” without being “clearly mistaken.” Charleston County, 365 F.3d at 349.

After careful review, we conclude Plaintiffs have not shown reversible legal error or clear factual error in the district court’s assessment of the third Gingles precondition.

27

Many of Plaintiffs’ arguments on appeal are better directed to the district court on remand after the parties develop the facts in discovery. Nevertheless, we briefly address each of Plaintiffs’ principal points.

Barreto’s performance analysis. Recognizing the centrality of the district court’s concerns about the reliability of Barreto’s predicted outcomes for elections in SD1 and SD2, Plaintiffs emphasize that the statewide elections he analyzed were not affected by the alleged methodological flaw because each of those elections was contested. Plaintiffs imply the district court should not have been so concerned about the reconstituted 2020 and 2022 Senate elections for SD1 and SD2 because Barreto asserted that statewide elections were more probative than endogenous ones. At the same time, Plaintiffs double down on the accuracy of Barreto’s report that the black-preferred candidate would have won SD2 in 2022 by emphasizing that his supplemental declaration “did not suggest that he was wrong to have excluded votes in uncontested contests” or “that uncontested elections should be included.” Pls. Opening Br. 36.

The district court did not clearly err in finding Barreto’s analysis of how SD1 and SD2 would have performed in the 2020 and 2022 state Senate elections to be important for proving the third Gingles precondition, including Barreto’s assessment—which Plaintiffs stand by on appeal—that the black-preferred candidate would have won SD2 in 2022. Nor was it clear error for the district court to consider those results more probative than exogenous statewide elections. See, e.g., Johnson v. Hamrick, 196 F.3d 1216, 1222 (11th Cir. 1999) (concluding that “viewing endogenous elections as more probative than exogenous elections” was not clear error); Bone Shirt v. Hazeltine, 461 F.3d 1011, 1021

28

(8th Cir. 2006) (stating that exogenous elections “hold some probative value” but “are not as probative as endogenous elections”). And although Barreto’s supplemental declaration mentioned the statewide contests, it did not address whether his alternative analysis of uncontested elections, based on a different data set than was first provided, would change the results for other reconstituted 2020 and 2022 endogenous elections in SD1 and SD2, like the other state Senate and House races. That legitimately raised the question whether other changes might result from further examination of Barreto’s methods and opinions during his anticipated deposition and cross-examination as this case progresses.

Furthermore, we accord special deference to a district court’s valuation of expert opinion and credibility. See Hendricks v. Cent. Rsrv. Life Ins., 39 F.3d 507, 513 (4th Cir. 1994). A district court is not obligated to credit the opinions of an expert witness when it has serious doubts about the expert’s methodology, other evidence contradicts the expert’s opinions, and the expert’s response to questioning raises more questions than it answers. Cf. United States v. Hasson, 26 F.4th 610, 626 (4th Cir. 2022). Plaintiffs’ response to the district court’s question about Barreto’s reconstituted 2022 Senate election result in SD2— first claiming it was a typo, then retracting that explanation and claiming that a different data set produces the opposite result—raised an inference that, at best, Plaintiffs didn’t understand their own data and, at worst, Plaintiffs’ expert could selectively change his predicted outcomes to suit the exigencies of the moment. At bottom, Barreto’s response to the district court’s question created enough uncertainty that the court was quite reasonably unwilling to rely exclusively on Barreto’s analysis (as the only expert Plaintiffs offered to opine on white bloc voting) to find the third Gingles precondition satisfied. See

29

Perez, 138 S. Ct. at 2333 (“Courts cannot find § 2 effects violations on the basis of uncertainty.”).

Evidence not presented to the district court. Plaintiffs attempt to bolster Barreto’s performance analysis by reference to a “StatPack” on the General Assembly’s website that reports partisan election results for the 2023 Senate districts using 2020 and 2022 statewide races. This evidence was not before the district court, so it could not have clearly erred by failing to consider it. 9 Plaintiffs do not explain why they chose not to present this evidence to the district court, so we will not consider it for the first time on appeal.

Barreto’s racial polarization analysis. As an alternative to Barreto’s performance analysis, Plaintiffs claim that his “statistically significant finding of racially polarized voting in North Carolina statewide as well as within the Northeast region” is sufficient, by itself, to prove the third Gingles precondition. J.A. 280. Plaintiffs are wrong. While “evidence of especially severe racially polarized voting . . . can help support finding the existence of Gingles’ third factor,” a “general finding regarding the existence of any racially polarized voting, no matter the level, is not enough.” Covington, 316 F.R.D. at 167 (emphasis added) (internal quotation marks omitted). As the Supreme Court has explained, “statistically significant racially polarized voting” in past elections can demonstrate “that in North Carolina, as in most States, there are discernible, non-random

9

Plaintiffs cite United States v. Gregory, 871 F.2d 1239, 1245 (4th Cir. 1989), for the proposition that the Court there relied on “judicially-noticeable data ‘not suppl[ied] below’ to find clear error.” Pls. Reply 14. We did nothing of the sort. Instead, we judged the district court in clear error for refusing to consider “statistical data presented [to it] by the Government.” Gregory, 871 F.2d at 1245.

30

relationships between race and voting,” but “that generalized conclusion fails to meaningfully (or indeed, at all) address the relevant local question” whether, in the new district, “black voters would encounter sufficient white bloc-voting to cancel their ability to elect representatives of their choice.” Cooper, 137 S. Ct. at 1471 n.5 (internal quotation marks, brackets, and alterations omitted). Even Plaintiffs acknowledge elsewhere in their briefing that something further is necessary to “transform[] statistically significant racially polarized voting . . . into legally significant racially polarized voting.” Pls. Reply 13.

Plaintiffs claim that Barreto’s racial polarization statistics are similar to those in Milligan and Charleston County, but that comparison fails for multiple reasons. Most importantly, in neither appeal was it disputed that white bloc voting usually defeated the election of minority-preferred candidates, so the Court did not opine on the evidence necessary to carry the plaintiffs’ burden. See Milligan, 143 S. Ct. at 1505–1506; Charleston County, 365 F.3d at 349. Moreover, the expert evidence in the cases is quite different. Barreto analyzed two election years, focused largely on exogenous elections, and reported the results in purely partisan terms. To the extent he reported data for the “Northeast region,” that grouping included 11 of the 18 counties in SD1 and SD2 and one (for Northeast-1) to three (for Northeast-2) counties that are not in either district. By comparison, the expert in Milligan analyzed seven biracial endogenous elections in the districts at issue and six biracial exogenous elections, over four election cycles. See Singleton v. Merrill, 582 F. Supp. 3d 924, 967–968 (N.D. Ala. 2022) (three-judge district court); see also United States v. Charleston County, 316 F. Supp. 2d 268, 277 (D.S.C. 2003) (discussing expert analysis of 31 contested county council elections over sixteen

31

years in the county at issue in assessing the extent of racial polarization). Significant quantitative and qualitative differences between the evidence in these cases render Plaintiffs’ comparative argument for clear error unpersuasive.

Perhaps recognizing that Barreto’s racial polarization findings, by themselves, do not answer the relevant question for the third Gingles precondition, Plaintiffs proffer new calculations—apparently conducted by counsel—to make their case. Obviously, we cannot find new facts on appeal, much less uncritically accept attorney argument as though it were expert opinion. See Long v. Hooks, 972 F.3d 442, 463 (4th Cir. 2020) (en banc) (“[I]t is elemental that counsel’s arguments are not evidence in a case.”); Columbus- America Discovery Grp. v. Atl. Mut. Ins. Co., 56 F.3d 556, 575–576 (4th Cir. 1995) (“It is a basic tenet of our legal system that . . . [appellate courts] do not make [factual] findings in the first instance.”). If Plaintiffs have new evidence, they can present it to the district court at the proper time as this litigation progresses.

Absence of district effectiveness analysis. Plaintiffs next contend that the district court asked too much of their expert by faulting him for not conducting a district effectiveness analysis. As the district court observed, Section 2 does not “require the creation of crossover districts in the first instance.” Strickland, 556 U.S. at 24; see id. at 6–9 (rejecting North Carolina’s defense that the VRA required it to draw a crossover district). Further, “‘[i]n areas with substantial crossover voting,’ § 2 plaintiffs would not ‘be able to establish the third Gingles precondition’ and so ‘majority-minority districts would not be required.’” Cooper, 137 S. Ct. at 1472 (quoting Strickland, 556 U.S. at 24). Consequently, courts have repeatedly rejected arguments that Section 2 required North

32

Carolina to draw majority-minority districts when the State lacked evidence that such districts were necessary for black-preferred candidates to win. See id.; Covington, 316 F.R.D. at 168–169; Lewis, 2019 WL 4569584, at *100. A district effectiveness analysis, as envisioned by the district court, is designed to answer that question by determining the BVAP at which a district provides a realistic opportunity for black voters to elect their candidates of choice, thereby providing insight into whether black voters’ candidates of choice “would usually be defeated without a VRA remedy.” Covington, 316 F.R.D. at 168. As an example of the usefulness of this evaluation, the district court referenced a district effectiveness analysis from the Lewis litigation in North Carolina. The court found Barreto’s failure to conduct such an analysis to be “another deep flaw” in his report. Pierce, 2024 WL 307643, at *20.

Plaintiffs argue that the district court legally erred by stating that they could not show legally significant racially polarized voting without a district effectiveness analysis. As Plaintiffs point out, courts have found VRA violations in other cases without a district effectiveness analysis, so it is hardly an across-the-board requirement. Yet Plaintiffs have not shown that the district court misunderstood the requirements of the third Gingles precondition or its application here. Courts in North Carolina have previously faulted experts for concluding that sufficient white bloc voting exists to usually defeat black- preferred candidates without first conducting this type of analysis. See, e.g., Covington, 316 F.R.D. at 168; Lewis, 2019 WL 4569584, at *100. The results of such an assessment do not cease being probative for the third Gingles precondition simply because the

33

litigation roles are reversed—i.e., here it is Plaintiffs, not the State, who advocate for a majority-minority district drawn on the basis of race.

The district court’s inaccurate implication that a district effectiveness analysis is required for proving a VRA violation in every Section 2 case is not a basis for reversing its denial of a preliminary injunction. Plaintiffs have not shown that the court’s case-specific assessment—that the absence of a district effectiveness analysis affected the persuasiveness of Barreto’s opinions—was erroneous. And that determination was not the sole, or even primary, reason for the district court’s skepticism of Barreto’s opinions. Tweaking the district court’s statement to acknowledge that a district effectiveness analysis is probative, but not required in all cases, does not make Plaintiffs’ evidence more likely to succeed in proving the third Gingles precondition.

Demonstration District B-1. Somewhat relatedly, Plaintiffs briefly take issue with the district court’s observation that their claim is undermined by Barreto’s conclusion that black voters’ candidates of choice would have won every election in 2020 and 2022 in Demonstration District B-1, a crossover district. As Plaintiffs put it, “[t]hat a 48% BVAP district would perform does not save the 30% BVAP districts” currently in place. Pls. Reply 21. We think Plaintiffs have missed the court’s point. The district court did not suggest that Demonstration District B-1’s performance foreclosed Plaintiffs’ claim. Rather, the court made the significantly milder point that crossover voting in Demonstration District B-1 was probative of crossover voting in the counties at issue.

Lack of white bloc voting in surrounding counties and overlapping districts.

Finally, Plaintiffs make arguments about the relative strength of inferences to be drawn

34

from other evidence of crossover voting, like Barreto’s scatterplots, Senate District 5, and the historical performance of Congressional District 1. None of this shows clear error. Plaintiffs also fault the district court for not mentioning the 2022 Senate race in former Senate District 3, which covered seven of the same counties as current SD1 (with the other counties falling in SD2). The black-preferred candidate lost that election even though, Plaintiffs claim, Senate District 3’s BVAP was 42.33 percent. This single historical data point does not show that the district court clearly erred in finding that Plaintiffs were not likely to succeed in proving Gingles’ third precondition. Of course, the district court will consider this evidence, along with all the other evidence, as the case proceeds on the merits.

2.

The district court’s ruling on the third Gingles precondition was sufficient to deny relief. Yet the court also ruled, in the alternative, that even if Plaintiffs were likely to succeed on all three Gingles preconditions, they nevertheless had not shown a likelihood of success on the ultimate totality of circumstances inquiry.

“[S]imply clearing the Gingles hurdles, while necessary to prove a possible violation of § 2, is not sufficient to establish an actual violation.” Charleston County, 365 F.3d at 348. To prove an actual violation, a plaintiff who demonstrates the three preconditions must also show, “based on the totality of circumstances, . . . that the political processes leading to nomination or election . . . are not equally open to participation by” minority voters. 52 U.S.C. § 10301(b); see also Perez, 138 S. Ct. at 2331 (“If a plaintiff makes that showing [under the three preconditions], it must then go on to prove that, under the totality of the circumstances, the district lines dilute the votes of the members of the

35

minority group.”). To guide the totality of circumstances analysis, the Supreme Court has referred to factors identified in the Senate Report on the 1982 amendments to the VRA. Gingles, 478 U.S. at 36–37, 43–46. These so-called Senate factors are “neither comprehensive nor exclusive.” Id. at 45. “[A]ny circumstance that has a logical bearing on whether voting is ‘equally open’ and affords equal ‘opportunity’ may be considered.” Brnovich v. Democratic Nat’l Comm., 141 S. Ct. 2321, 2338 (2021); see also LULAC, 548 U.S. at 426, 437 (considering proportionality between number of minority districts and minority share of statewide population); Charleston County, 365 F.3d at 347, 352–353 (considering whether partisanship rather than race caused racially polarized voting).

There is “no requirement that any particular number of factors be proved, or that a majority of them point one way or the other.” Gingles, 478 U.S. at 45 (internal quotation marks omitted). The inquiry is “flexible” and “fact-intensive” and “requires an intensely local appraisal” that benefits from “the trial court’s particular familiarity with the indigenous political reality.” Id. at 46, 79 (internal quotation marks omitted); see Milligan, 143 S. Ct. at 1503 (“[T]he totality of circumstances inquiry recognizes that application of the Gingles factors is peculiarly dependent upon the facts of each case.” (internal quotation marks omitted)). It is therefore the province of the district court.

“[O]ur function is not to reweigh the evidence presented to the district court.”

Charleston County, 365 F.3d at 349. Indeed, “[t]he Supreme Court has been explicit that . . . we may set aside a trial court’s finding of vote dilution only if it is clearly erroneous,” id., or rests on an error of law, see Gingles, 468 U.S. at 79. The district court here thoroughly examined the “current preliminary and hotly contested record” and found that

36

Plaintiffs had not made a clear showing they were likely to succeed in the totality of circumstances inquiry. Pierce, 2024 WL 307643, at *28. Plaintiffs dispute this finding on appeal, and we consider their arguments in three groups below.

a.

We begin with the Senate factors, which the district court outlined as follows:

(1) the extent of the state’s historical discrimination concerning the right to vote against plaintiffs’ minority group; (2) the extent of racially polarized voting; (3) the extent to which the state has adopted other voting practices that may exacerbate discrimination against the minority group; (4) whether members of plaintiffs’ minority group have been denied access to a candidate slating process; (5) whether members of plaintiffs’ minority group in the state “bear the effects of discrimination” in education, employment, or health, hindering their ability to participate in the political process; (6) whether political campaigns have been characterized by overt or subtle racial appeals;

(7) the extent to which members of plaintiffs’ minority group have been elected to public office in the jurisdiction; (8) whether there is a significant lack of responsiveness by the state’s elected officials to the “particularized needs” of plaintiffs’ minority group; and (9) whether the state’s policy underlying its use of the challenged voting procedure is tenuous.

Id. at *22 (quoting Gingles, 478 U.S. at 36–38). The court then analyzed Plaintiffs’ evidence concerning each one.

Under the first factor, the court gave “little weight” to Plaintiffs’ evidence of historically discriminatory practices and cases because it was “very old” and “overwhelmingly outdated”—Plaintiffs cited “just one case from the last 30 years in which a court found the General Assembly acted with discriminatory intent.” Id. at *23. On the second factor, Plaintiffs repeated their arguments about the third Gingles precondition, which, like the first time around, the district court found unpersuasive. The district court found the third factor to favor Legislative Defendants because Plaintiffs did not cite any

37

evidence that North Carolina “presently employs other voting practices that may enhance the opportunity for discrimination against black voters.” Id. Plaintiffs conceded the fourth factor did not apply. The fifth factor “d[id] not help” Plaintiffs because, while Plaintiffs’ expert identified socioeconomic disparities, she did not demonstrate “that race discrimination by North Carolina caused the socioeconomic disparities.” Id. Regarding the sixth factor, racial appeals in political campaigns, the court gave “little weight” to examples from 1984 and 1990. Id. at *24. Of the two modern examples Plaintiffs offered, the court found one “was not a racial appeal” and the other, assuming it was a racial appeal, did not “characterize” North Carolina campaigns. Id.

As for the seventh factor—the extent to which members of the minority group have been elected to public office—the district court acknowledged the opinion of Plaintiffs’ own expert that black members of the North Carolina General Assembly “are ‘close to parity’ with the share of black people in North Carolina’s population.” Id. (quoting expert report); compare J.A. 429–430 (reporting “26 Black House members, or 21.6% of the chamber,” and “9 Black senators, making up 18.0% of the chamber”), with J.A. 162 (reporting a statewide BVAP of “20.10%”). The court also observed that black North Carolinians from both political parties occupy significant positions in state government, including Lieutenant Governor, minority leader of the state Senate, minority leader of the state House of Representatives, and state appellate judgeships. Moving to the eighth factor, the court found Plaintiffs offered “no evidence of elected officials’ responsiveness or unresponsiveness to black voters” but merely “ask[ed] the court to infer” unresponsiveness based on socioeconomic inequality. Pierce, 2024 WL 307643, at *24. And, finally, the

38

court was unpersuaded by Plaintiffs’ argument on the ninth factor, concluding instead that the policies underlying the challenged map—compliance with federal law, the state constitution, and traditional redistricting principles—were not tenuous.

On appeal, Plaintiffs contest the district court’s findings on every single Senate factor. More often than not, they simply disagree with the weight the district court accorded their evidence—a textbook factual dispute. See Charleston County, 365 F.3d at 349; Wright v. Sumter Cnty. Bd. of Elections & Registration, 979 F.3d 1282, 1301 (11th Cir. 2020) (“The clearly-erroneous standard extends . . . to [the district court’s] finding that different pieces of evidence carry different probative values in the overall section 2 investigation.” (internal quotation marks omitted)). As for the two legal issues they raise, we find no error in the district court’s analysis.

First, Plaintiffs contend the third Senate factor examines past, not present, voting practices that enhance the opportunity for voting discrimination. That contravenes Gingles itself, which looked to other voting procedures in operation at the time of the suit. 478 U.S. at 39–40, 56; see also Charleston County, 365 F.3d at 351 (assessing contemporaneous electoral structure). Considering current practices makes sense, as the third factor is concerned with whether other voting practices or procedures amplify the discriminatory effect of the challenged voting procedure. And, as the district court reasoned, a purely backward-looking analysis would replicate the first factor, with its focus on historical discrimination concerning the right to vote.

Second, we reject Plaintiffs’ argument that the district court legally erred by considering under the seventh factor the successful election of black candidates statewide

39

and not merely in the challenged districts. See Gingles, 478 U.S. at 40, 74–75 (assessing “the extent to which blacks have been elected to office in North Carolina, both statewide and in the challenged districts”). Indeed, Plaintiffs themselves focused exclusively on evidence of statewide underrepresentation in their briefing to the district court on this point, noting that North Carolina has never had a black governor or black United States Senator. 10 The remainder of Plaintiffs’ arguments are an attempt to relitigate the significance of each piece of evidence they put before the district court. But it is emphatically not our duty “to duplicate the role of the lower court.” Anderson, 470 U.S. at 573. That admonition is especially salutary at this preliminary stage, when the case will return to the district court for final fact finding on a more fulsome record after discovery and trial. As always, it matters not how we would have evaluated the evidence in the first instance but only whether the district court’s assessment was “plausible.” Cooper, 137 S. Ct. at 1474, 1478 (internal quotation marks omitted). “‘Where there are two permissible views of the evidence, the factfinder’s choice between them cannot be clearly erroneous.’” Brnovich, 141 S. Ct. at 2349 (quoting Anderson, 470 U.S. at 574). That remains true “even when the district court’s findings do not rest on credibility determinations, but are based instead on . . . documentary evidence or inferences from other facts.” Anderson, 470 U.S. at 574. Plaintiffs fall well short of identifying “clearly mistaken” factual findings by the district

10

Equally baseless is Plaintiffs’ assertion, under the ninth factor, that North Carolina’s interests in complying with federal constitutional prohibitions against racial gerrymandering and state constitutional prohibitions against splitting counties are illegitimate considerations. Cf. Brnovich, 141 S. Ct. at 2339–2340 (considering “the strength of the state interests served by a challenged voting rule” and “the reason for the rule”).

40

court, much less error sufficient to call into question the court’s ultimate conclusion. Charleston County, 365 F.3d at 349. We nevertheless briefly address Plaintiffs’ factual arguments.

On the first Senate factor, Plaintiffs take issue with the district court calling their evidence of historical discrimination “outdated” and giving it “little weight.” It is imminently reasonable to conclude that recent history is more persuasive of whether “the political process is . . . equally open to minority voters” than history far more removed. Milligan, 143 S. Ct. at 1503 (internal quotation marks omitted); cf. Shelby County v. Holder, 570 U.S. 529, 535 (2013) (striking down Section 4(b) of the VRA because “the conditions that originally justified these measures no longer characterize voting in the covered jurisdictions”).

As for more recent history of voting-related discrimination, the evidence Plaintiffs highlight on appeal is a mixed bag. Plaintiffs flag Department of Justice objection letters “[f]rom 1980 to 2013” to election law changes in North Carolina and note that “[s]ome” of those objections asserted the General Assembly acted with discriminatory intent. Pls. Opening Br. 43–44. Plaintiffs stress this Court’s finding that the General Assembly enacted certain voting changes in 2013 with “discriminatory intent” after requesting racial data, but the Court was also careful to disclaim any suggestion that legislators acted with “racial hatred or animosity.” N.C. State Conf. of NAACP v. McCrory, 831 F.3d 204, 215, 233 (4th Cir. 2016); see also Covington, 316 F.R.D. at 124 n.1 (“[W]e make no finding that the General Assembly acted in bad faith or with discriminatory intent.”); Harris, 159 F. Supp. 3d at 604 (making “no finding” regarding legislators’ “good faith”). The district

41

court acknowledged that decision, and Plaintiffs have not shown clear error in the court’s weighing of their evidence on this factor as a whole.

Similarly for the second Senate factor: the record at this preliminary stage shows that the extent of racially polarized voting, in North Carolina and in the northeastern counties at issue, is “certainly disputable.” Charleston County, 365 F.3d at 349. As the district court observed, Plaintiffs’ own evidence demonstrates notable crossover voting statewide and locally.

As for the fifth factor (the extent to which minority group members “bear the effects of discrimination” in education, employment, or health, hindering their ability to participate in the political process) and eighth factor (unresponsiveness by elected officials to the “particularized needs” of the minority group), on appeal Plaintiffs rely on ipse dixit. Gingles, 478 U.S. at 45. The district court found that Plaintiffs presented no evidence connecting the disparities they reported between black and white North Carolinians to official race discrimination or unresponsive elected officials. Plaintiffs reply not with evidence but by asserting this is “an obvious reality.” Pls. Opening Br. 49; see id. at 46 (asserting that “of course” race discrimination by North Carolina caused the disparities). On this record, we cannot conclude the district court clearly erred.

Finally, on the sixth factor, Plaintiffs have not shown that the district court clearly erred in finding that political campaigns in North Carolina were not “‘characterized by’” racial appeals based on the evidence Plaintiffs presented. Gingles, 478 U.S. at 37 (quoting S. Rep. No. 97-417, 97th Cong. 2d Sess. 28, 28–29 (1982)).

42

b.

In addition to the Senate factors, the district court also considered whether partisanship, rather than race, drove polarization in North Carolina. On this point, Legislative Defendants presented a report from their expert Dr. John Alford. Using the same election and voter data and employing the same ecological inference method as Barreto, Alford assessed the effect of race on electoral outcomes.

As the district court recounted, Alford began with two United States Senate elections, one featuring a white Republican against a white Democrat and another featuring a white Republican against a black Democrat. Black voters statewide and regionally supported both Democratic candidates at essentially identical rates. And white voters were not more likely to oppose a black Democrat compared to a white Democrat—in fact, white voters were slightly more supportive of the black Democrat in 2022 compared to the white Democrat in 2020. Alford reached similar results when analyzing state supreme court and appellate court elections. Again black voters’ support for black and white Democrats was essentially identical, while white voters were not more likely to oppose a black Democrat than a white Democrat. And the pattern persisted when Alford evaluated an election featuring a black Republican against a white Democrat: the black Republican candidate received no more black voter support and no less white voter support than the average white Republican candidate. Alford’s analysis of all 2020 and 2022 elections yielded the same results. White voters supported Republican candidates at essentially the same rate regardless of their race, and black voters supported Democratic candidates at essentially the same rate, regardless of their race. Alford opined that, “[i]n contrast to the strong

43

impact of candidate party affiliation, . . . the race of the candidates does not appear to have a polarizing impact on vote choice” but was “essentially indetectable.” J.A. 682. Thus, Alford concluded, and the district court accepted, “that Dr. Barreto’s analysis ‘clearly demonstrates that the party affiliation of the candidates is sufficient to fully explain the divergent voting preferences of Black and White voters in the 2020 and 2022 North Carolina elections.’” Pierce, 2024 WL 307643, at *26 (quoting J.A. 687).

Plaintiffs contend that Alford’s analysis is irrelevant, but that argument contradicts our precedent. “Certainly the reason for polarized voting is a critical factor in the totality analysis,” including evidence that “partisanship [is] the cause of the racially divergent voting.” Charleston County, 365 F.3d at 347, 349; see also id. at 348 (reasoning that the totality of circumstances examination “is tailor-made for considering why voting patterns differ along racial lines”); Lewis v. Alamance County, 99 F.3d 600, 615 n.12 (4th Cir. 1996) (concluding that “causation . . . [is] relevant in the totality of circumstances inquiry”); cf. Gingles, 478 U.S. at 100–101 (O’Connor, J., concurring in judgment) (explaining that causation evidence should be part of “the overall vote dilution inquiry”); League of United Latin Am. Citizens v. Clements, 999 F.2d 831, 861 (5th Cir. 1993) (concluding that plaintiffs failed to establish a Section 2 claim where the evidence “unmistakably shows that divergent voting patterns among white and minority voters are best explained by partisan affiliation”).

For example, in Charleston County, the district court considered evidence that minority-preferred minority candidates were defeated more often than minority-preferred white candidates, and that white voters offered less cohesive support to minority

44

Democratic candidates than to white Democratic candidates. 365 F.3d at 353. It was not clearly erroneous, we explained, for the district court to conclude this evidence cut against a finding that partisanship was the cause of racially polarized voting. Id. Although caselaw does not require the minority-preferred candidate to be a member of the minority group, a model that accounts for the candidate’s race can provide probative evidence about causation.

Plaintiffs also dispute what conclusions can be drawn from Alford’s analysis and fault him for failing to further isolate and measure other potential race-based reasons why black voters prefer Democrats. These arguments go to the persuasiveness of the evidence and the weight it should receive. Plaintiffs do not identify any legal error or clear factual error in the district court’s consideration of this evidence; indeed, they aver that Alford’s analysis supports multiple equally plausible conclusions. It is possible that further expert discovery will limit the plausible conclusions that can be drawn from the partisanship data, but the absence of such certainty now simply demonstrates the district court did not clearly err. See Brnovich, 141 S. Ct. at 2349.

c.

As a final consideration in its totality of circumstances analysis, the district court reviewed the extended litigation about North Carolina’s congressional and legislative maps over the years. Plaintiffs do not dispute that the district court could account for this broader context when examining whether the totality of circumstances proved vote dilution. See id. at 2338 (permitting consideration of “any circumstance that has a logical bearing” on vote dilution).

45

As the district court explained, “[w]hen the General Assembly enacted SB 758 in October 2023, state and federal courts had repeatedly affirmed that the General Assembly must draw legislative districts without reference to race because legally significant racially polarized voting did not exist in North Carolina,” including the northeastern counties at issue here. Pierce, 2024 WL 307643, at *28. In case after case, courts struck down North Carolina’s redistricting efforts premised on VRA compliance. In Strickland, when the General Assembly disregarded the state constitution’s Whole County Provision to create a minority crossover district, the Supreme Court held that Section 2 did not justify that departure. 566 U.S. at 23. In subsequent cases, when the State used race to draw majority- minority congressional and legislative districts, courts struck down those maps as unconstitutional, in part because the absence of legally significant racially polarized voting showed that Section 2 did not justify using race to draw majority-black districts. E.g., Cooper, 137 S. Ct. at 1470–1471; Covington, 316 F.R.D. at 167–174; Harris, 159 F. Supp. 3d at 624–625. State courts also limited the General Assembly’s consideration of race during redistricting. E.g., Harper II, 881 S.E.2d at 180; Lewis, 2019 WL 4569584, at *133.

Plaintiffs quibble with the district court’s statement that Cooper and Harris involved the same portion of northeastern North Carolina at issue in this case. While the congressional district featured in Cooper and Harris certainly included parts of counties not at issue here, the core of that challenged district consisted of the entirety of Bertie, Halifax, Hertford, Northampton, and Warren Counties (the only whole counties in that district) and parts of other counties involved here. See Cooper, 137 S. Ct. at 1484. Those counties are the heart of Plaintiffs’ dispute. Plaintiffs’ argument also obscures a broader

46

point. The fact that prior cases involved other counties in addition to the ones at issue here simply supports an inference that crossover voting occurs in many parts of the State, including northeastern North Carolina.

* * *

In sum, Plaintiffs have not identified any misunderstanding of law or clearly erroneous fact undergirding the district court’s alternative conclusion that, based on the evidence presented thus far, Plaintiffs are unlikely to succeed in proving vote dilution under the totality of circumstances. The court engaged in an “intensely local appraisal” of the practical and functional political realities in North Carolina, and specifically northeastern North Carolina, and avoided giving any one factor conclusive weight. Charleston County, 365 F.3d at 349, 351 (internal quotation marks omitted). The district court judged this “preliminary and hotly contested record,” Pierce, 2024 WL 307643, at *28, insufficient to justify ordering the North Carolina General Assembly to make “the drastic decision to draw lines” on the basis of race for the 2024 state Senate elections, Perez, 138 S. Ct. at 2334; see Strickland, 556 U.S. at 21 (“[R]acial classifications are permitted only as a last resort.” (internal quotation marks omitted)). We see no reversible error.

B.

We turn now to the remaining requirements for a preliminary injunction. Plaintiffs must show “that the balance of equities tips in [their] favor” and “that an injunction is in the public interest.” Winter, 555 U.S. at 20. These “factors ‘merge when the Government is the opposing party.’” Miranda v. Garland, 34 F.4th 338, 365 (4th Cir. 2022) (quoting Nken v. Holder, 556 U.S. 418, 435 (2009)). The district court concluded that the equities

47

favor Legislative Defendants and that “the Purcell principle”—that “‘federal district courts ordinarily should not enjoin state election laws in the period close to an election’”— “teaches that a federal court should not issue the requested mandatory federal preliminary injunction of North Carolina’s 2024 Senate elections.” Pierce, 2024 WL 307643, at *29 (quoting Milligan, 142 S. Ct. at 879 (Kavanaugh, J., concurring)). We see no abuse of discretion in the district court’s conclusions.

Redistricting “is primarily the duty and responsibility of the State,” and “[f]ederal-

court review of districting legislation represents a serious intrusion on the most vital of local functions.” Miller, 515 U.S. at 915 (internal quotation marks omitted). Enjoining North Carolina from enforcing its duly enacted redistricting law in the 2024 state Senate elections would inflict “a form of irreparable injury.” Maryland v. King, 567 U.S. 1301, 1303 (2012) (Roberts, C.J., in chambers) (“Any time a State is enjoined by a court from effectuating statutes enacted by representatives of its people, it suffers a form of irreparable injury.” (internal quotation marks and brackets omitted)). Not to mention the practical effects of an injunction on the State’s sound and orderly administration of the 2024 Senate election, which we will discuss momentarily.

On the other side of the balance, Plaintiffs rightly emphasize that the public interest favors protecting federally guaranteed voting rights. But Plaintiffs have not shown that they are likely to prevail on their Section 2 claim or suffer irreparable harm.

By contrast, the district court found it significant that, at the time the General Assembly voted SB 758 into law, it did not have a “strong basis in evidence” to conclude that “Section 2 required a majority-black Senate district in northeast North Carolina.”

48

Pierce, 2024 WL 307643, at *29; see also Cooper, 137 S. Ct. at 1471. And without a strong basis for concluding that the VRA required race-based districting, the General Assembly was forbidden by the Equal Protection Clause from taking such action. Cooper, 137 S. Ct. at 1463–1464. If a federal court were to intervene and order the General Assembly to create districts hitting a racial quota, it would create the real risk of imposing racially gerrymandered districts for the 2024 Senate election, should Plaintiffs’ Section 2 claim ultimately fail on the merits. Such a result is obviously not in the public interest. See Leaders of a Beautiful Struggle v. Balt. Police Dep’t, 2 F.4th 330, 346 (4th Cir. 2021) (en banc) (“[I]t is well-established that the public interest favors protecting constitutional rights.”).

Plaintiffs assert that the six months it took the General Assembly to enact new maps after the North Carolina Supreme Court authorized it to do so was “an effort to thwart review” that should be counted against the State in the balance of equities. Pls. Opening Br. 60. We recognize that the timing of SB 758’s passage in October 2023 set a tight timeline for any pre-election challenge. But there has been no finding that the General Assembly acted in bad faith by taking six months to finalize new congressional districts, state Senate districts, and state House districts. So at this point “the good faith of [the] state legislature must be presumed.” Miller, 515 U.S. at 915. As for Plaintiffs, the district court noted that they did not present their views about SB 758 to the legislature while the bill was under consideration. Instead, Plaintiffs waited 26 days after the General Assembly enacted SB 758 to file suit and 28 days to seek a preliminary injunction. Weighing all the

49

facts, the district court did not abuse its discretion in concluding that inequity would result if the court enjoined the use of SB 758 in the 2024 Senate elections.

Then there’s Purcell. That name stands for the principle that “federal courts ordinarily should not enjoin a state’s election laws in the period close to an election,” and that when “lower federal courts contravene that principle,” the Supreme Court will stop them. Milligan, 142 S. Ct. at 879–880 (Kavanaugh, J., concurring). “When an election is close at hand, the rules of the road must be clear and settled. Late judicial tinkering with election laws can lead to disruption and to unanticipated and unfair consequences for candidates, political parties, and voters, among others.” Id. at 880–881. Indeed, “[c]ourt orders affecting elections . . . can themselves result in voter confusion and consequent incentive to remain away from the polls,” a risk that increases “[a]s an election draws closer.” Purcell, 549 U.S. at 4–5.

For example, in Merrill v. Milligan, a federal district court enjoined Alabama from using its newly drawn congressional districts for the 2022 elections and ordered that the districts be redrawn when absentee voting in the primary elections was set to begin nine weeks later. The Supreme Court stayed the injunction, with Justices Kavanaugh and Alito explaining that “the Purcell principle require[d]” a stay, “‘[g]iven the imminence of the election and the inadequate time to resolve the factual disputes.’” Milligan, 142 S. Ct. at 882 (Kavanaugh, J., concurring) (quoting Purcell, 549 U.S. at 5–6). In numerous other cases, the Court has similarly stayed lower court injunctions of state election laws in the period close to an election. See, e.g., Merrill v. People First of Ala., 141 S. Ct. 190 (2020); Merrill v. People First of Ala., 141 S. Ct. 25 (2020); Andino v. Middleton, 141 S. Ct. 9

50

(2020); Republican Nat’l Comm. v. Democratic Nat’l Comm., 140 S. Ct. 1205 (2020) (per curiam); Democratic Nat’l Comm. v. Wis. State Legislature, 141 S. Ct. 28 (2020) (declining to vacate stay); Miller v. Johnson, 512 U.S. 1283 (1994); see also Wise, 978 F.3d at 96, 98–99 (declining to enjoin state election procedure based on Purcell).

The 2024 North Carolina Senate election is well underway. The statewide primary election is scheduled for March 5, 2024. 11 Candidate filing ended on December 15, 2023. Absentee ballots were distributed on January 19, 2024. In-person early voting began on February 15, 2024. The election is not merely “close[],” or even “imminen[t]”—it is happening right now. Purcell, 549 U.S. at 5.

If a federal court were to enjoin North Carolina from enforcing SB 758 for the 2024 Senate elections, then the court must afford the General Assembly a reasonable opportunity to redraw the Senate districts, as “it is the domain of the States, and not the federal courts, to conduct apportionment.” Voinovich, 507 U.S. at 156; see also Covington, 138 S. Ct. at 2554 (“[A] legislature’s freedom of choice to devise substitutes for an apportionment plan found [unlawful], either as a whole or in part, should not be restricted beyond the clear commands of federal law.” (internal quotation marks omitted)). As the district court explained, in this counterfactual scenario where a court accepted Plaintiffs’ argument, the General Assembly “would first have to draw a majority-black VRA district in northeast North Carolina before drawing non-VRA districts using other state-law redistricting principles and rules, including county grouping or clustering requirements under”

11

In fact, as of the date this opinion is publicly released, the March 5 primary is over and done. The boards of elections have certified final results.

51

Stephenson v. Bartlett, 562 S.E.2d 377 (N.C. 2002) (Stephenson I). Pierce, 2024 WL 307643, at *30 (internal quotation marks omitted). For example, if the General Assembly chose to enact Plaintiffs’ Demonstration District A as a VRA-required majority-black Senate district, it would “then have to regroup the remaining 92 counties under Stephenson I and its progeny and redraw all other Senate districts.” Id.

Then the county boards of elections would have to discard completed ballots—both in-person ballots and absentee ballots, including “the ballots of the numerous North Carolina citizens in the United States military who are deployed overseas.” Id. at *31. The North Carolina Board of Elections would have to conduct its “geocoding process” again to reassign voters to the proper districts. Id. New district boundaries would require a fresh opportunity for candidate filing. The Board would then have to generate, proof, and distribute new ballots. It would have to move the March primary elections to May or later and create a new runoff date in July or August. 12 All of this would result in the voter confusion and disruptive consequences the Purcell principle is designed to avoid. See Purcell, 549 U.S. at 4–5; Milligan, 142 S. Ct. at 880–881 (Kavanaugh, J., concurring). The district court rightly heeded the Supreme

12

As previously noted, by now the March 5 primary has already occurred.

Of course, Plaintiffs’ complaint seeks permanent injunctive relief barring use of SD1 and SD2 as drawn in SB 758 for “any Senate elections,” not just for the 2024 elections. J.A. 32. Even absent a preliminary injunction, therefore, the district court retains its ability “to render a meaningful judgment on the merits.” In re Microsoft Corp. Antitrust Litig., 333 F.3d 517, 526 (4th Cir. 2003).

52

Court’s warning against triggering this cascade of election chaos with a preliminary injunction during the ongoing 2024 Senate elections.

In response, Plaintiffs’ only argument is that the General Assembly could choose to change the boundaries of only SD1 and SD2 by enacting Plaintiffs’ Demonstration Districts B-1 and B-2. Plaintiffs note that the Senate candidates currently running for election in SD1 and SD2 do not face challengers in the primary election, and so, according to Plaintiffs, this limited remedy would be feasible. We see at least three problems with Plaintiffs’ argument.

First, the General Assembly does not have to choose Plaintiffs’ proposed remedial districts, and a federal court should not dictate that it do so when other revisions the legislature might choose would comply with federal law. See Covington, 138 S. Ct. at 2554–2555; Voinovich, 507 U.S. at 156; Wise v. Lipscomb, 437 U.S. 535, 540 (1978). As discussed, the General Assembly could choose to enact Demonstration District A or some other remedial majority-black Senate district in northeastern North Carolina, which would necessarily affect the boundaries of other districts beyond SD1 and SD2. And, as the district court observed, “the root cause of any ensuing upheaval would be the federal court injunction prohibiting the use of SB 758 in the 2024 Senate elections and requiring the General Assembly to remedy an alleged Section 2 violation.” Pierce, 2024 WL 307643, at *32. Because of the domino effect of changing the Senate districting map, we think the district court was correct that Purcell directs us not to preliminarily enjoin SB 758 at this late hour. And even if the Purcell principle did not independently bar relief, it is enough to conclude, as we do, that it counsels against federal court intervention at this point in

53

North Carolina’s 2024 Senate election, providing yet another reason why a preliminary injunction would not serve the public interest.

Second, it may be the case, as Legislative Defendants argue, that the General Assembly is legally forbidden from adopting Demonstration District B-1. As previously noted, the North Carolina Constitution prohibits the General Assembly from dividing counties when drawing legislative districts for the state House and Senate. See N.C. Const. art. II, §§ 3(3), 5(3). Redistricting plans can “depart from strict compliance” with the Whole County Provision “only to the extent necessary to comply with federal law,” like the VRA. Stephenson I, 562 S.E.2d at 397; see also Pender County, 649 S.E.2d at 366, aff’d sub nom. Strickland, 556 U.S. 1; Stephenson v. Bartlett, 582 S.E.2d 247, 251–252 (N.C. 2003) (Stephenson II). In Strickland, the Supreme Court held that the VRA “does not mandate creating or preserving crossover districts.” 556 U.S. at 23. Consequently, federal law did not require North Carolina to “override” the Whole County Provision in that case to retain a crossover district. Id. at 14.

Demonstration Districts B-1 and B-2 split Pasquotank County, in violation of the Whole County Provision. Plaintiffs claim the VRA requires that departure. But the district court found, at least on the current record, that Demonstration District B-1 is a crossover district, not a majority-minority district. In that case, VRA compliance would not require (and therefore would not justify) departure from the Whole County Provision. See Stephenson I, 562 S.E.2d at 397. At this stage of the case, we express no opinion on Demonstration District B-1’s legality but must acknowledge this potentially fatal flaw in Plaintiffs’ sole argument for avoiding Purcell.

54

Third, even if the General Assembly could enact Demonstration Districts B-1 and B-2 or choose some other reconfiguration that affects only the eighteen counties in SD1 and SD2 in response to a federal court injunction, the Purcell principle would still counsel against a mandatory preliminary injunction interfering with the 2024 Senate elections. The Board of Elections would have to reassign voters. Candidates would have to refile, which could result in contested primaries in the two new Senate districts. The Board of Elections would have to generate, proof, and distribute new ballots, schedule any contested primary elections in these districts for May 2024 or later, and, if a runoff were needed, schedule and conduct that special election before August 6, 2024, in order to prepare ballots for the general election. All this would disrupt the orderly election process and sow voter confusion, with the “consequent incentive to remain away from the polls.” Purcell, 549 U.S. at 5. Even feasible feats can cause the “chaos and confusion” with which Purcell is concerned. Milligan, 142 S. Ct. at 880 (Kavanaugh, J., concurring).

Finally, Plaintiffs contend that the Purcell principle “might be overcome . . . if a plaintiff establishes at least the following: (i) the underlying merits are entirely clearcut in favor of the plaintiff; (ii) the plaintiff would suffer irreparable harm absent the injunction; (iii) the plaintiff has not unduly delayed bringing the complaint to court; and (iv) the changes in question are at least feasible before the election without significant cost, confusion, or hardship.” Id. at 881. Those prerequisites have not been shown here. The merits of Plaintiffs’ Section 2 claim are not clearcut in their favor, to say nothing of the other requirements for this potential exception to Purcell’s mandate.

55

“North Carolina voters deserve clarity” about their elections. Wise, 978 F.3d at 96.

With the statewide 2024 Senate election underway, candidates and voters alike are now entitled to the stability and sense of repose that engender trust and confidence in our elections. The district court did not abuse its discretion by heeding the Supreme Court’s warnings against federal courts enjoining state election laws in the period close to an election.

IV.

The denial of preliminary relief is just that: preliminary. It may be that with discovery and further factual development, Plaintiffs can prove that these two Senate districts violate Section 2 of the VRA and they are entitled to a majority-minority district in northeastern North Carolina. But the standard for winning relief before trial and obtaining concomitant federal court interference with state redistricting decisions while elections are underway is high indeed, and Plaintiffs have not satisfied it with the record they have developed thus far. Instead, the legal principles that must govern our decision direct us not to intervene and order North Carolina to create race-based electoral districts while this litigation remains pending.

AFFIRMED

56

GREGORY, Circuit Judge, dissenting:

In October of last year, the North Carolina General Assembly enacted a map that cracked the state’s Black Belt 1 right down the middle. Yet the district court concluded that this new map was unlikely to dilute Black voters’ power. In doing so, it misconstrued the standard Appellants must meet under Thornburg v. Gingles, 478 U.S. 30, 51 (1986), and improperly concluded that Appellants had to present a specific type of analysis that neither this Court nor the Supreme Court has ever required. Further, in balancing the equities of granting the preliminary injunction, the district court made much of Appellants’ 28-day delay in bringing the case, without so much as mentioning Appellees’ six-month delay in enacting the map in the first place. Therefore, I cannot agree with my colleagues that the district court’s denial of a preliminary injunction was appropriate.

I.

We evaluate a district court’s decision to deny a preliminary injunction for abuse of discretion, reviewing legal conclusions de novo and the factual findings for clear error. Although we must accord great deference to the district court under this standard, such deference is overcome where the district court “based its ruling on an erroneous view of the law or on a clearly erroneous assessment of the evidence.” Cooter & Gell v. Hartmarx Corp., 496 U.S. 384, 405 (1990); see United States v. Jenkins, 22 F.4th 162, 167 (4th Cir.

1

The Black Belt refers to a region stretching across the South characterized by its “thick, dark, and naturally rich soil.” Booker T. Washington, Up from Slavery: An Autobiography 108 (1st elec. ed. 1997), https://perma.cc/49Z8-NEGN. The counties in the challenged districts are part of this Black Belt region.

57

2021). Thus, a district court abuses its discretion where, as here, it misapprehends or misapplies the applicable law. League of Women Voters of N.C. v. North Carolina, 769 F.3d 224, 235 (4th Cir. 2014); see also Centro Tepeyac v. Montgomery County, 722 F.3d 184, 188 (4th Cir. 2013) (en banc).

II.

As my colleagues explain, the third Gingles precondition requires Appellants to show that the white majority votes sufficiently as a bloc such that it will usually defeat the minority-preferred candidate, absent a remedial district. Gingles, 478 U.S. at 51. This showing, combined with satisfaction of the first two preconditions, establishes that “the minority [group] has the potential to elect a representative of its own choice in a possible district, but that racially polarized voting prevents it from doing so in the district as actually drawn because it is submerg[ed] in a larger white voting population.” Cooper v. Harris, 581 U.S. 285, 302 (2017) (internal quotation omitted) (brackets in original). If a legislature has reason to believe that all three preconditions are met, “then so too it has good reason to believe that § 2 requires drawing a majority-minority district.” Id.

A.

For context, it is helpful to compare the challenged map to the map in place before.

Former Senate Districts 1 and 3 (Map A) are made up of the same counties that now constitute challenged Senate Districts 1 and 2 (Map B). Map A shows that almost all of the northeast counties with a high Black voting age population are within former Senate District 3. Map B shows that the district boundary line divides the Black belt into Senate Districts 1 and 2.

58

Map A (2022):

J.A. 43.

Map B (2023):

J.A. 20.

59

Before addressing the merits, I begin with the legislature’s argument that it was free not to adopt a majority-minority district because it lacked a “strong basis in evidence” that Section 2 required one. Leg. Def. Opening Br. at 43.

The district court was convinced by this argument: “Without a contemporaneous strong basis in evidence in 2023 that Section 2 required the General Assembly to create a VRA district by grouping citizens by race in order to form a majority-black Senate district, the General Assembly would have violated the Fourteenth Amendment.” Dist. Ct. Op. at 6. According to the district court, had the General Assembly enacted a map with a majority-minority district in the region at issue, it would have “committed the same mistake” as it did when it enacted the 2011 maps. Id. Those maps were struck down as impermissible racial gerrymanders in Covington v. North Carolina, 316 F.R.D. 117, 130– 41, 167–74 (M.D.N.C. 2016), and Harris v. McCrory, 159 F. Supp. 3d 600, 604 (M.D.N.C. 2016). See Dist. Ct. Op. at 7. In the district court’s view, because there was no evidence (or, at least Appellants cited none) that “anyone submitted information to the General Assembly[,] before the General Assembly enacted SB 758 in 2023[,] that Section 2 required a majority-black Senate district in northeast North Carolina,” the General Assembly wasn’t required to adopt a majority-minority district. Dist. Ct. Op. at 8 (emphasis added).

This misunderstands the standard. It assumes that a legislature cannot violate Section 2 unless it knows that majority-minority districts are required but fails to enact them. But Section 2 speaks only of effects—not knowledge. See 52 U.S.C. § 10301(a) (“No voting . . . procedure shall be imposed or applied by any State or political subdivision

60

in a manner which results in a denial or abridgement of the right of any citizen of the United States to vote on account of race or color . . .” (emphasis added)). So legislatures cannot claim that they declined to enact majority-minority districts because they received no evidence that declining to enact them would violate Section 2. If they could, legislatures could insulate themselves from VRA compliance by remaining ignorant or, more perniciously, feigning ignorance.

It therefore does not matter that the General Assembly held public hearings to gather input on the proposed districts or “accepted public comments through an online portal on the General Assembly’s website.” Dist. Ct. Op. at 8. The legislature cannot outsource its evidence-gathering duty and then argue that it lacked evidence that a majority-minority district was required. There is no head-in-the-sand defense to VRA compliance.

Even if there were, the legislature would not be entitled to it here. Legislative Appellees say they received no evidence “that the Gingles preconditions are present in a particular area of the state.” Leg. Def. Opening Br. at 10. This is patently false. The Southern Coalition for Social Justice submitted a 27-page letter, including an expert report with data analysis of racially polarized voting. See J.A. 280 n.17 (citing Southern Coalition for Social Justice, Letter re: Racially Polarized Voting in North Carolina and its Effect on the 2023 Redistricting Plans (Oct. 22, 2023), https://perma.cc/2AQK-SEB7 (last accessed Mar. 18, 2024)).

That report stated:

Appendix A confirms that the second and third Gingles preconditions are satisfied in Proposed Senate Districts 1 & 2. As discussed above, it is possible to draw reasonably configured Gingles demonstrative districts in

61

several areas of North Carolina, each of which would satisfy the first Gingles precondition. This includes the areas covered by Proposed Senate Districts 1 & 2. When combined with the analysis laid out in Appendix A, this shows that all three Gingles preconditions are established in the area covered by Proposed Senate Districts 1 & 2, and when combined with North Carolina’s pervasive history of discrimination in voting, makes clear that enacting Proposed Senate Districts 1 & 2 would violate the VRA.

Southern Coalition Letter at 2–3 (first emphasis added). In other words, the Southern Coalition for Justice explicitly told the legislature that the proposed map “would violate the VRA.” Id. at 3.

The district court also seemed to find it meaningful that the Southern Coalition for Social Justice “did not request any majority-minority districts,” but instead just “asked that the county grouping for SD1 and 2 be changed to the alternate county grouping used in 2022.” Dist. Ct. Op. at 9. But majority-minority districts are not the only way to remedy Section 2 violations. See Cooper, 581 U.S. at 305. More to the point, it does not matter what remedy a third party requests. Just as the legislature cannot outsource determining whether its map would violate Section 2, it cannot outsource the task of choosing which remedial map to adopt.

B.

The third Gingles precondition asks whether white voters in the challenged district vote sufficiently as a bloc to “thwart[] a distinctive minority vote.” Allen v. Milligan, 599 U.S. 1, 19 (2023) (internal quotation omitted). There is no per se rule for what counts as a racial bloc. Rather, “bloc voting is a matter of degree.” James Buchwalter, et al., Corpus Juris Secundum Elections § 96 (2024). Whether statistically significant racially polarized voting rises to the level of legally significant racially polarized voting hinges on the

62

particulars of the evidence in the record. See Gingles, 478 U.S. at 57–58 (“[T]he degree of racial bloc voting that is cognizable as an element of a § 2 vote dilution claim will vary according to a variety of factual circumstances. Consequently, there is no simple doctrinal test for the existence of legally significant racial bloc voting.”). A review of the record here makes clear that voters in the districts at issue vote in legally significant racial blocs.

Appellants’ expert, Dr. Matt Barreto, conducted two analyses: “a racially polarized voting analysis using the ecological inference regression technique and a performance analysis of election outcomes in current Senate Districts 1 and 2 based on the results of past elections (since Districts 1 and 2 are new).” Pierce Opening Br. at 27.

i.

Ecological Inference Regressions. Ecological inference regressions involve “compiling data on the percentage of each racial group in a precinct and merging that with precinct-level vote choice from relevant election results.” J.A. 279 (Barreto report). Barreto used official election result data, and voter file data obtained from the North Carolina State Board of Elections. J.A. 273, 279. Voter file data contains information about the voter’s self-reported race or ethnicity. J.A. 279. Barreto used that data to create percentages of voter race/ethnicity by voting precinct. He then merged that precinct-level race/ethnicity data with precinct-level election results from 31 recent North Carolina elections. J.A. 279.

The results showed that Black voters in the Northeast region “demonstrate unified and cohesive voting, siding for the same candidates of choice with clear support in the 95%

63

range.” J.A. 280. White bloc voting has rates as high as 85% in opposition to minority- preferred candidates in some instances. J.A. 280.

Specifically, in the elections that most closely resemble “endogenous” elections, 2 State House and State Senate elections in Northeast North Carolina in 2020 and 2022, 98– 99% of Black voters were cohesive in voting for their candidates of choice. J.A. 281. 3 There is no truly endogenous election data here, of course, because this map creates new districts. In contrast to Black voters, white voters voted against minority-preferred candidates at rates between 80 and 88% in State House and State Senate elections. J.A. 281.

In statewide elections, 97–99% of Black voters were unified in their support of their candidate of choice, while white voters “vote[d] in the exact opposite direction in every one of these elections.” J.A. 281. In fact, voting was so racially polarized that Barreto’s scatter plots look like this:

2

An election is endogenous when it was in the same district and for the same office as the election at issue.

3

Barreto’s non-statewide analysis focused only on precincts in certain counties.

The “Northeast-1” analysis included all precincts in Bertie, Chowan, Gates, Halifax, Hertford, Martin, Northampton, Pasquotank, Perquimans, Warren, Washington, and Vance. The “Northeast-2” analysis included all those precincts and added all precincts in Pitt and Edgecombe counties. J.A. 280 n.18. Neither Pitt nor Edgecombe is part of the current or proposed maps here.

64

J.A. 282–83. These graphs show that each candidate’s percent vote within each precinct (Y axis) changed almost perfectly in step with change in BVAP (X axis).

Reconstituted Election Analysis. Barreto next drew on precinct data from past elections to determine whether Appellants’ proposed remedial districts would solve the white bloc voting issues that exist in the challenged districts. J.A. 291–93. This is called a “reconstituted election analysis.” It involves “extract[ing] actual election results from a variety of statewide

65

and local races that subsume the area being analyzed and determines, precinct-by-precinct within the proposed district, the racial composition of the vote and the ‘winner’ within the proposed district.” Buchwalter, Corpus Juris Secundum Elections § 96. It “allows a researcher to determine how an individual candidate performed within the boundaries of the target district even though the actual election covered a different geographical area.” Id.

Barreto analyzed precinct data from the precincts in what are now Senate Districts 1 and 2. That data covered 31 elections from 2020 and 2022. See Appendix A. Of those, 27 were for statewide office (e.g., Governor). 4 J.A. 291–93. The remaining four elections were for State Senate and State House races in current Senate Districts 1 and 2 (two in 2020 and two in 2022). J.A. 291–93. Barreto found that in every election conducted in Demonstration District A, with a Black CVAP (BCVAP) of 53.1%, J.A. 291, the minority- preferred candidate would win. 5 J.A. 291–93. The same was true of Demonstration District B1, with a BCVAP of 50.2%. J.A. 291–93. In Demonstration District B2, where BCVAP was 12.6%, the minority-preferred candidate lost every time (as would be expected). J.A. 291–93. And had current Senate Districts 1 and 2 been in place during the 2020 and 2022 elections, the minority-preferred candidate in every single statewide race would have lost both districts. J.A. 291–93. Of the four non-statewide races, if current

4

Though statewide elections are generally less probative of district elections, litigants “of course, can present statewide evidence in order to prove racial gerrymandering in a particular district.” Ala. Leg. Black Caucus v. Alabama, 575 U.S. 254, 263 (2015).

5

The district court took issue with Appellants’ use of BCVAP (as opposed to BVAP) to show that Demonstration District B-1 satisfied the first Gingles precondition. Dist. Ct. Op. at 31–32. But that is a separate issue from whether BCVAP can be used to demonstrate racial polarization, and the district court raised no concerns with the latter.

66

Senate Districts 1 and 2 had been in place, minority-preferred candidates would have lost three (SD1 in 2020 and 2022 and SD2 in 2020). J.A. 291–93.

ii.

Appellees’ expert, Dr. John Alford, used the same ecological inference technique that Barreto used (Barreto’s first analysis). J.A. 675. Alford attempted to replicate Barreto’s ecological inference results by using the election and voter data sources that Barreto cited. J.A. 678. Alford reported that his “initial replication results [were] substantively similar to those reported by Dr. Barreto, but [did] not match as precisely as would be expected based on [his] experience in multiple similar cases.” J.A. 678. However, Alford explained that these inconsistencies were not surprising because, he claimed, Barreto did not disclose “input data files or any details of the [ecological inference] analytical options used” for his report. J.A. 678. 6 iii.

The district court took issue with Appellants’ arguments about the third Gingles precondition for two main reasons. First, the district court said that Appellants failed to satisfy the third precondition because Barreto’s analysis did not show that Black voters would be unable to elect their candidates of choice unless BVAP exceeded 50% plus one. Second, the district court said, Barreto’s finding and explanation about the results of a hypothetical election in reconstituted Senate District 2 suggested that there was not racial bloc voting in that district and that all of Barreto’s other findings were unreliable.

6

The remainder of Alford’s report was mostly about the cause of racial polarization.

J.A. 679–85. This is not relevant to the third Gingles precondition.

67

Failure to Include District Effectiveness Analysis. The district court legally erred by assuming that, to prevail on the third Gingles precondition, Appellants must show that “black voters’ candidates of choice cannot win elections unless BVAP in the contested districts exceeds 50% plus one vote.” Dist. Ct. Op. at 40–41. The third precondition asks whether white voting patterns thwart minority voters’ preferences in the contested district as its demographics currently stand. See Milligan, 599 U.S. at 17–19. The district court formulated the standard to mean that, if minority voters could elect candidates of their choice in a counterfactual present district, then minority voters can elect candidates of their choice in the actual present district. This standard would mean that districts where minority voters are consistently unable to elect their preferred candidates could stay in place just because, if the district’s demographics were different, minority voters would be able to elect their preferred candidates. But the whole point here is that the district’s demographics aren’t different. And as a result, minority voters cannot elect their candidates of choice.

The district court’s belief that the third Gingles precondition requires showing “black voters’ candidates of choice cannot win elections unless BVAP in the contested districts exceeds 50% plus one vote” rested on a misunderstanding of Bartlett v. Strickland. 556 U.S. 1 (2009). Strickland dealt with the first Gingles precondition, which requires that a minority group show that it is “sufficiently large and geographically compact to constitute a majority in a single-member district.” Id. at 11 (quoting Gingles, 478 U.S. at 50–51). The first Gingles precondition, unlike the third, is about the demographics of a differently drawn, potentially remedial district. Strickland held that meeting this precondition requires showing a potential majority-minority district; a district with sufficient white crossover

68

voting to elect the minority-preferred candidate will not meet the first precondition. Id. at 12–14. Strickland also held that, although a legislature may implement a crossover district as a Section 2 remedy, a legislature cannot be required to do so. Id. at 23 (“[Section] 2 allows States to choose their own method of complying with the Voting Rights Act, and we have said that may include drawing crossover districts”); see also Cooper, 581 U.S. at 305 (rejecting North Carolina legislature’s argument that because § 2 “does not require crossover districts (for groups insufficiently large under Gingles), then § 2 also cannot be satisfied by crossover districts (for groups in fact meeting Gingles’ size condition)”).

The district court took Strickland to mean that if a minority-preferred candidate could win elections in the contested district when the contested district’s BVAP was less than 50% plus one vote, the contested district was a crossover district. Dist. Ct. Op. at 40– 41 (“Section 2 does not require crossover districts. Thus, a proper district effectiveness analysis supporting plaintiffs’ challenge must show that black voters’ candidates of choice cannot win elections unless BVAP in the contested districts exceeds 50% plus one vote.”) (citation omitted). But a crossover district is not a district with crossover potential, contingent on demographic change. It is a district that actually crosses over (at least sometimes), with its current demographics.

The district court’s misunderstanding of the legal standard led to an insurmountable roadblock for Appellants. In sum, the district court believed that showing racial bloc voting requires showing that Black voters can elect their candidates of choice only when BVAP exceeds 50%. The way to show this is through a “district effectiveness analysis.” Dist. Ct. Op. at 40 (“To demonstrate legally significant racially polarized voting, an expert must

69

engage in a ‘district effectiveness analysis.’”). A district effectiveness analysis is a “district-specific evaluation used to determine the minority voting-age population level at which a district becomes effective in providing a realistic opportunity for . . . voters of that minority group to elect candidates of their choice.” Covington, 316 F.R.D. at 148 n.46 (internal quotations omitted) (ellipses in original). So, the district court reasoned, Appellants could not demonstrate racial bloc voting unless they provided a district effectiveness analysis. Unsurprisingly, because the premise of this reasoning was legally erroneous (explained above), its conclusion was, too.

Though the majority briefly addresses the district court’s erroneous conclusion that a district effectiveness analysis is required, it elides the premise of that conclusion. On the one hand, the majority describes the “district court’s inaccurate implication that a district effectiveness analysis is required for proving a VRA violation in every Section 2 case,” seemingly acknowledging that such a requirement is legally erroneous. Maj. Op. at 33. On the other hand, the majority claims that “[Appellants] have not shown that the district court misunderstood the requirements of the third Gingles precondition or its application here.” Maj. Op. at 32. It is hard to reconcile these two statements. The very reason the district court erroneously thought a district effectiveness analysis was necessary is because it did not understand what the third Gingles precondition required.

Because of the district court’s legal error—a per se rule that plaintiffs cannot show legally significant racial bloc voting, and thus cannot meet the third Gingles precondition, without a district effectiveness analysis—Appellants’ case was doomed from the start, regardless of the quantity, strength, and probativeness of their other evidence.

70

The majority says Appellants “have not shown that the court’s case-specific assessment—that the absence of a district effectiveness analysis affected the persuasiveness of Barreto’s opinions—was erroneous.” Maj. Op. at 33. But by requiring a district effectiveness analysis, the district court made Barreto’s opinions (all based on data from other types of analyses) irrelevant. They weren’t simply less persuasive; they had no bearing on the legal analysis.

For that reason, I cannot agree with the majority’s statement that “[t]weaking the district court’s statement to acknowledge that a district effectiveness analysis is probative, but not required in all cases, does not make [Appellants’] evidence more likely to succeed in proving the third Gingles precondition.” Maj. Op. at 33. First, what the majority does is far more than tweaking; more accurately, it inverts the district court’s statement. Second, inverting the district court’s statement does “make [Appellants’] evidence more likely to succeed” by making Appellants’ other evidence relevant in the first place. If, as the district court believed, Appellants’ other data was wholly irrelevant absent a district effectiveness analysis, the district court’s thoughts on the reliability and accuracy of the data that Barreto did present is neither here nor there.

Reliability of Barreto’s Analysis. Nonetheless, because the majority treats Barreto’s alleged unreliability as dispositive, I explain why the district court abused its discretion in discounting Barreto’s analysis. In addition to finding that Barreto could not show legally significant racial bloc voting without conducting a district effective analysis, the district court concluded that the analyses that Barreto did conduct were unreliable. It based this

71

finding on Barreto’s predictions of how current Senate District 2 would have performed had the district been in place during the 2022 elections.

The table in Barreto’s original report shows that the minority-preferred candidate would have won by a margin of 8.2 percentage points (54.1 to 45.9) in that district had it been in place in 2022. J.A. 291. This finding was not helpful for Appellants’ case, and the district court asked about it at its January 10 hearing on the preliminary injunction. Dist. Ct. Op. at 38. Though Appellants’ lawyer first stated that the number was likely a typo, Barreto later said in a supplemental declaration that it wasn’t. Id. In the declaration, he explained why the 8.2 percentage points—though an accurate finding based on the data he included—were not an indication that the minority-preferred candidate could be viable in the district. See J.A. 853–54.

He explained that current Senate District 2 contains portions of districts from the previous State Senate map, specifically former Senate Districts 1 and 3. J.A. 853. That map was last used in 2022. That year, the Senate District 3 seat was contested, but the Senate District 1 seat was uncontested. Barreto explained that his analysis included only contested elections because the results of uncontested elections are not probative. J.A. 853. Therefore, all his Senate District 2 analysis showed, he said, was “that a hypothetical district containing only [the counties in previous Senate District 3] would perform for Black-preferred candidates based on the 2022 State Senate elections.” J.A. 853.

Former Senate District 3, however, had an unrepresentatively high BVAP of 48.4%.

J.A. 853. Adding in the populations from the counties in former Senate District 1 (to create what is now Senate District 2) decreases the overall BVAP to 30%. J.A. 853. While

72

minority-preferred candidates would prevail in a 48.4% BVAP district, Barreto explained, they wouldn’t prevail in a 30% BVAP district, i.e. in current Senate District 2. J.A. 854.

“When all 2022 State Senate elections are counted across both contested and uncontested races in the counties now within current Senate District 2, 51,019 ballots were cast for white-preferred (Republican) State Senate candidates, while only 16,877 ballots were cast for Black-preferred (Democratic) State Senate candidates,” Barreto explained. J.A. 853–54. That’s three-quarters of votes cast for the white-preferred (Republican) candidate and only one-quarter cast for the Black-preferred (Democratic) candidate.

The district court took issue with Barreto’s explanation. Barreto’s admission that his initial analysis did not use data from uncontested elections “shows that [he] is doing an unusual form of reconstituted election analysis,” the court said. Dist. Ct. Op. at 38.

The purpose of a reconstituted election analysis is to show whether minority-

preferred candidates could prevail in a new district. If the district court believed that excluding uncontested elections would not reliably predict minority-preferred candidates’ chances (a conclusion I agree with), then it presumably believed that including uncontested elections would reliably predict minority-preferred candidates’ chances. Yet when Barreto did what the district court would have had him do in the first place, the district court continued to be skeptical of that finding. But either skepticism was warranted for Barreto’s first figure or it was warranted for Barreto’s second figure—it cannot be warranted for both at the same time.

What’s more, this skepticism bled into all of Barreto’s remaining figures, even though every other election was contested. For instance, the district court agreed with Appellees

73

that Barreto’s “new representation ‘cannot seem to be cabined to the contests he would prefer the Court ignore.’” Dist. Ct. Op. at 38 (quoting Def’s. Resp. to Supp. Dec. at 2). It reasoned that “Dr. Barreto’s belated explanation undercuts all of Dr. Barreto’s conclusions by demonstrating that fuller data sets could change his estimated outcomes.” Id. at 39. But there was no relevant fuller data set—no other election in the data set was uncontested. See J.A. 285–90 (listing out names of candidates from each party in every 2020 and 2022 statewide election); State Board of Elections, 2020 Candidate List Grouped by Contest, at 6, https://perma.cc/P3WB-9FBR (last accessed Mar. 18, 2024) (showing that, in 2020, then- Senate District 1 and then-Senate District 3 were both contested); State Board of Elections, 2022 Candidate List Grouped by Contest, at 4, https://perma.cc/5PRS-THT6 (last accessed Mar. 18, 2024) (showing that, in 2022, then-Senate District 1 was uncontested but then- Senate District 3 was contested). So while the district court may have been justified in discounting Barreto’s analysis of that one reconstituted election in current Senate District 2, it was an abuse of discretion to disregard his analysis as to every other election.

***

In sum, the district court misunderstood what the third Gingles precondition requires. As a result, it imposed a standard that made all of Appellants’ evidence irrelevant. The district court then went on to find that, because one piece of that evidence may have been unreliable, Appellants’ entire expert analysis must have been flawed. These collective errors are enough to conclude that the district court abused its discretion in finding that Appellants were unlikely to succeed on the merits of their case.

74

III.

A.

After considering the Gingles preconditions, courts ask whether “based on the totality of circumstances,” voters of color “have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice.” See 42 U.S.C. § 10301(b). “[I]t will be only the very unusual case in which the plaintiffs can establish the existence of the three Gingles [preconditions] but still have failed to establish a violation of § 2 under the totality of the circumstances.” Jenkins v. Red Clay Consol. Sch. Dist. Bd. of Educ., 4 F.3d 1103, 1135 (3d Cir. 1993), aff’d, 581 U.S. 285. In the Senate Report accompanying the 1982 amendments to the VRA, the Senate identified several factors that guide a court’s totality of circumstances analysis. Gingles, 478 U.S. at 36–37. These Senate factors assist the court in conducting “an intensely local appraisal of the electoral mechanism at issue, as well as a searching practical evaluation of the past and present reality.” Milligan, 599 U.S. at 19.

The most important factors in the totality of the circumstances inquiry are the “extent to which minority members have been elected to public office in the jurisdiction” and “the extent to which voting in the elections of the state or political subdivision is racially polarized.” Gingles, 478 U.S. at 48–49. Congress did not intend for these factors to be comprehensive or exclusive, nor did Congress require that a particular number of factors be satisfied to establish a § 2 violation. Id. at 45.

The district court gave the Senate factors short shrift. The law requires that courts conduct “a searching practical evaluation of the past and present reality” to assess “whether

75

the political process is equally open to minority voters.” Id. at 79 (internal citation and quotation marks omitted). And the district court’s analysis overlooked much of North Carolina’s past and present reality.

The first Senate factor examines “the extent of any history of official discrimination in the state or political subdivision that touched the right of the members of the minority group to register, to vote, or otherwise participate in the democratic process.” Id. at 36–37 (quoting S. Rep. No. 97-417, at 28–29 (1982)). Appellants presented ample evidence of such history, spanning from 1900 to 2016. The district court gave little weight to Appellants’ evidence, which it described as “overwhelmingly outdated.” J.A. 942. But history is necessarily in the past and “[a]n eye toward past practices is part and parcel of the totality of the circumstances.” League of Women Voters, 769 F.3d at 241.

Gingles itself, which arrived at the Supreme Court from a challenge to a North Carolina General Assembly redistricting plan, outlined the history of discriminatory voting practices within the state. To do so, it pointed to discrimination against Black citizens from “approximately 1900 to 1970 by employing at different times a poll tax, a literacy test, a prohibition against bullet (single-shot) voting and designated seat plans.” Id. at 38–39. Between 1982 and 2006, private plaintiffs brought at least fifty-five successful cases challenging North Carolina’s election related practices. N.C. State Conference of NAACP v. McCrory, 831 F.3d 204, 224 (4th Cir. 2016).

In 2016, this Court elaborated on the history recounted in Gingles, noting that the record is “replete with evidence of instances since the 1980s in which the North Carolina legislature . . . attempted to suppress and dilute the voting rights of African Americans.”

76

McCrory, 831 F.3d at 223 (4th Cir. 2016). In fact, we observed, the Department of Justice and federal courts have determined that the legislature at times even acted with discriminatory intent, engaging in “a series of official actions taken for invidious purposes.” Id. (quoting Arlington Heights v. Metropolitan Housing Development Corp., 429 U.S. 252, 267 (1977)). This includes a 2013 omnibus election law that “target[ed] African-Americans with almost surgical precision.” Id. at 214.

Even though the district court found that events from as recently as 2016 are “overwhelmingly outdated,” it was still required to grapple with them. As Appellants observe, the first Senate factor focuses on “the state’s historical discrimination” regarding voting rights—and most history is old. Pierce Opening Br. at 44. Though this Court defers to the district court on factual determinations, this deference does not permit the district court to move the historical goalposts. And courts have concluded time and time again that this same history satisfies the first Senate factor. See, e.g., Harris, 159 F. Supp. 3d 600, aff’d sub nom. Cooper v. Harris, 581 U.S. 285 (2017) (invalidating 2011 congressional district lines).

The second Senate factor considers “the extent to which voting in the elections of the state or political subdivision is racially polarized.” Gingles, 478 U.S. at 37. Barreto’s racially polarized voting analysis showed that approximately 85% of white voters in and around the contested region voted against Black-preferred candidates. In that same region, Black voters voted together 95% of the time. The district court discounted this racial polarization for the same reasons that it found that Appellants could not meet the third Gingles precondition. But as discussed above, the district court abused its discretion in

77

evaluating the third Gingles precondition. It therefore erred in relying on that reasoning in assessing the second Senate factor.

The district judge likewise rejected Appellants’ arguments under the third Senate factor. That factor analyzes “the extent to which the state or political subdivision has used . . . voting practices that may enhance the opportunity for discrimination.” Gingles, 478 U.S. at 36–37. Senate factors one and three overlap significantly. Legislative Appellees claim that the third factor requires “a present tense inquiry because only existing election features can combine with the challenged feature to enhance dilution.” Leg. Def. Reply Br. at 17. But courts regularly look to history to evaluate this factor. See League of Women Voters, 769 F.3d at 241 (“Neither the Supreme Court nor this Court has ever held that, in determining whether an abridgement has occurred, courts are categorically barred from considering past practices . . . .”). The district court erred by adopting Legislative Appellees’ interpretation of this factor.

The parties agree that the fourth Senate factor is not relevant to this case.

The fifth Senate factor analyzes “the extent to which members of the minority group in the state or political subdivision bear the effects of discrimination in areas such as education, employment and health, which hinder their ability to effectively participate in the political process.” Gingles, 478 U.S. at 37. The district court found that this factor weighed against Appellants because they presented “no statistical analysis demonstrating that race discrimination by North Carolina caused the socioeconomic disparities.” J.A. 943.

But this Court has never required Appellants to show that racial discrimination caused the socioeconomic disparities at issue, and requiring Appellants to show that link

78

was legal error. See, e.g., League of Women Voters, 769 F.3d at 246 (presenting evidence that Black North Carolinians lagged behind whites in several important socioeconomic indicators but not requiring proof that aforementioned disparities were the product of discrimination); League of United Latin American Citizens v. Perry (LULAC), 548 U.S. 399, 440 (2006) (pointing out that “the political, social, and economic legacy of past discrimination . . . may well hinder [a minority group’s] ability to effectively participate in the political process” (internal quotation omitted)).

Factor six examines “whether political campaigns have been characterized by overt or subtle racial appeals.” Gingles, 478 U.S. at 37. The district court gave little weight to Appellants’ examples, finding that they were either too old, not clear racial appeals, or not characteristic of recent campaigns within the state. J.A. 944. But in Gingles itself, the Supreme Court affirmed the district court’s reliance on evidence of political campaigns nearly a century earlier, and Appellants’ evidence here is far more recent. See Gingles, 478 U.S. at 40.

For instance, United States Senator Jesse Helms, who represented the state of North Carolina until 2003, famously deployed racial appeals in state-wide campaigns. In a 1990 ad for Helms’s campaign, white hands crumpled a job rejection letter while a voiceover blamed that rejection on a minority applicant. Mtn. Prelim. Inj. at 426. The district court said this ad was too old. Dist. Ct. Op. at 49. More recently, now-Senator Ted Budd successfully deployed an ad against his Black opponent that was reminiscent of the infamous 1988 Willie Horton ad. Id. The district court said that this ad was “not a racial appeal” because it “never explicitly mention[ed] race.” Dist. Ct. Op. at 49 (internal quotation omitted). But that is of no consequence; as the text of this Senate factor explains,

79

political campaigns can be “characterized by . . . subtle racial appeals.” Gingles, at 37; see also id. at 40 (affirming the district court when it found Senate factor six satisfied by racial appeals “ranging in style from overt and blatant to subtle and furtive”). Appellants also offer evidence from former U.S. Representative Madison Cawthorn’s 2020 campaign, in which Cawthorn accused his opponent of associating with people who want to “ruin white males.” Mtn. Prelim. Inj. at 427. The district court, “without deciding” whether Cawthorn’s statements were “an overt or subtle racial appeal,” dismissed this example because it didn’t “characterize” today’s North Carolina campaigns. Dist. Ct. Op. at 49. It went out of its way to minimize the racial appeal in each ad, defying common sense.

The seventh factor considers “the extent to which members of the minority group have been elected to public office in the jurisdiction.” Gingles, 478 U.S. at 37. The district court correctly observes that Black candidates have found electoral success in other parts of the state and in statewide elections. J.A. 945. But the success of a few minority candidates does not “necessarily foreclose the possibility of dilution of the black vote.” S. Rep. No. 97-417, at 29 n.115.

More importantly, whether Black candidates can win in Durham, for example, has no bearing on whether Black candidates can win in the challenged districts. See Dist. Ct. Op. 50. For the same reason, whether a Black candidate can win a statewide election says nothing about whether they can prevail in the challenged districts. Indeed, “the extent to which minority candidates have been elected to office in the jurisdiction, obviously must be measured by elections in the political jurisdiction in question.” City of Carrollton Branch of the NAACP v. Stallings, 829 F.2d 1547, 1560 (11th Cir. 1987); see Sanchez v.

80

Colorado, 97 F.3d 1303, 1324–25 (10th Cir. 1996) (concluding that “it is probative” that no member of the minority group had been elected to the challenged office since 1940, “notwithstanding” that members of the minority group had been elected to other offices in the geographic region). The district court’s failure to even consider Black candidates’ performance in the challenged districts was an abuse of discretion.

The eighth Senate factor asks, “whether there is a significant lack of responsiveness on the part of elected officials to the particularized needs of the members of the minority group.” Gingles, 478 U.S. at 37. Appellants assert that the General Assembly’s enactment of discriminatory voting laws, as well as persistent socioeconomic disparities within the region, render this factor self-evident. Pierce Opening Br. at 48. The district court declined to make that inference. Yet the Supreme Court made a similar inference in LULAC. There, the appellants demonstrated a combination of circumstances, including declining incumbent popularity, a history of official discrimination within the state, and persistent sociological disparity. LULAC, 548 U.S. at 439–40. The Court agreed that these circumstances showed that the appellants’ representative was unresponsive to their needs. Id. Here, as in LULAC, Appellants have demonstrated historical discrimination and sociological disparity. Combined with present attempts to weaken their voting power, see Pierce Opening Br. at 48, these circumstances evince their representatives’ nonresponsiveness.

Finally, the ninth Senate factor evaluates “whether the policy underlying the state or political subdivision’s use of such . . . practice or procedure is tenuous”—in other words, whether the government actually had a legitimate interest in enacting the challenged map. Gingles, 478 U.S. at 37. Legislative Appellees assert that they adopted the map to comply

81

with federal law, a North Carolina state redistricting principle known as the Whole County Provision, and traditional districting principles. Leg. Def. Opening Br. at 43. The district court shared this reasoning. Dist. Ct. Op. at 51. But the serpentine shape of Senate District 2 is far from compact (a traditional districting principle), requiring several hours and multiple ferry rides to span. Gov. Amici Br. 13–14. And the Whole County Provision applies only after any VRA districts are drawn. See Stephenson v. Bartlett, 562 S.E.2d 377, 396–97 (N.C. 2002); see also Pender County v. Bartlett, 649 S.E.2d 364, 367–68 (N.C. 2007) (explaining that a district required by § 2 trumps the need to comply with the Whole County Provision).

My colleagues in the majority are correct in pointing out that our review of the district court’s decision requires deference. But affording deference does not mean abdicating our responsibility to ensure that district courts adhere to existing guideposts. And because there are few cases expounding on the meaning of individual Senate factors, the best guideposts remain Gingles and its progeny. Here, the district court strayed from the four corners of this jurisprudence throughout its totality analysis, planting additional obstacles for Appellants to surmount along the way. If the totality analysis has any meaningful role in identifying Section 2 violations, this must be error.

The district court erred throughout its analysis. The court failed to meaningfully contend with the role of history in contravention of Gingles and subsequent case law. It concluded that Senate factor three mandates a present-tense inquiry when no precedent supports that requirement. It mandated never-before-required statistical proof of causation between discrimination and socioeconomic disparities on Senate factor five and overlooked areas in Appellants’ expert report where a connection is drawn between

82

discrimination and disparate outcomes. And in its evaluation of the seventh factor, it erroneously assumed that the presence of minority office holders in other parts of the state is nearly dispositive of the ability of minority candidates in the challenged area to succeed.

B.

Both the district court and the majority opinion recognize the importance of the second Senate factor (“the extent to which voting in the elections of the state or political subdivision is racially polarized,” Gingles, 478 U.S. at 37) and dedicate significant space to its discussion. Specifically, they reason that because the racially polarized voting in the challenged districts can be explained by partisan alignment, not race, the second Senate factor cuts against Appellants. 7 See Maj. Op. at 43 (“Thus Alford concluded, and the district court accepted, that [the] analysis clearly demonstrates that the party affiliation of the candidates is sufficient to fully explain the divergent voting preferences of Black and White voters in the 2020 and 2022 North Carolina elections.” (internal quotations omitted)). Implicitly, if not expressly, they take this point one step further, contending that because there is evidence that partisan motivation underlies the divergent voting patterns among Black and white voters, Appellants cannot establish a Section 2 violation under the totality of the circumstances. See id. This reasoning threatens the viability of any Section 2 claim because it permits courts to dismiss every Section 2 case in which voting patterns

7

Though the majority treats this as a separate consideration, see Maj. Op. at 42, it is part of the second Senate factor. See United States v. Charleston County, 365 F.3d 341, 347 (4th Cir. 2004).

83

could be explained on partisan grounds. But a conclusion that polarization is explained by partisan affiliation is not a conclusion that polarization is not explained by race.

In support of their argument, the majority and the district court cite United States v.

Charleston County. 365 F.3d 341 (4th Cir. 2004). But Charleston County does not say that evidence of partisan motivation undercuts a finding that the challenged districts violate Section 2.

In Charleston County, the district court considered evidence that voting patterns shifted when a candidate’s race changed but their party affiliation remained the same. Id. at 353. We said that was evidence that racial motivation underlay the racially divergent voting patterns. And, we said, that racial motivation was relevant to (though not dispositive of 8) a finding that the second Senate factor favored the party challenging the districts. Id. We did not say the inverse. That is, we did not say that if there was evidence of a non- racial motivation underlying voting patterns, then the second Senate factor does not favor the party challenging the relevant districts. The majority conflates these two distinct ideas.

The district court reasoned that the presence of partisan motivation in voting means the absence of racial motivation. See Dist. Ct. Op. at 53–54. But partisan motivations and racial motivations are not mutually exclusive. That is, the same voter can be motivated to vote for a particular candidate due to both partisanship and race. For that reason, evidence

8

The district court in Charleston County found that changing the candidate’s race changed voting patterns. Charleston County, 365 F.3d at 353. Based in part on this finding, this Court concluded that the district court did not clearly err by saying that race appeared to influence voting patterns. Id. But its holding was not contingent on that finding. Id. at 352. Charleston County’s holding primarily hinged on the district court’s more robust evidence of racial polarization. Id.

84

of a partisan motivation says nothing about the existence or extent of a racial motivation. While it may be true that affirmative evidence of a racial motivation underlying voting patterns strengthens a party’s showing on the second Senate factor, see Charleston County, 365 F.3d at 353, evidence of a partisan motivation does not weaken that showing.

In Gingles itself, a plurality of the Supreme Court addressed a similar argument that the second Senate factor is not satisfied when the racially divergent voting patterns can be explained by voters’ socioeconomic characteristics. See Gingles, 478 U.S. at 64. The plurality rejected this argument, explaining that

it ignores the fact that members of geographically insular racial and ethnic groups frequently share socioeconomic characteristics, such as income level, employment status, amount of education, housing and other living conditions, religion, language, and so forth. Where such characteristics are shared, race or ethnic group not only denotes color or place of origin, it also functions as a shorthand notation for common social and economic characteristics.

Id.

In the same way, the argument that voting is not racially motivated because it is motivated by partisanship “ignores the fact that members of geographically insular racial and ethnic groups” frequently share the same partisan affiliation. See id. This is because partisan motivation and racial motivation among minority voters regularly coexist. Accordingly, whether voting patterns in northeast North Carolina can be explained on partisan grounds says nothing about whether they can be explained only on partisan grounds. Permitting evidence of partisan motivation to explain away racial polarization converts Black voters’ motivations for voting into an impediment to Black voting power.

85

This cannot be squared with the purpose of Section 2: ensuring that minority voters can participate equally in the political process.

IV.

Appellants argue that the remaining factors favor imposing a preliminary injunction.

They explain that failing to do so would disenfranchise thousands of Black voters in the affected districts. In contrast, Appellants proposed remedy would impact only the two challenged districts (where there are currently no contested primaries), and the State Board of Elections has confirmed that it is administratively feasible to enact a new map and hold a later primary in the new districts. 9 Appellees disagree. First, they argue that Appellants’ proposed remedy (altering only the line between the two challenged districts) is inappropriate. Second, they argue that because all other potential remedies impact districts where primaries are already underway, imposing an appropriate remedy is not administratively feasible. This is particularly true because, they contend, the Purcell principle prohibits federal courts from “alter[ing] the election rules on the eve of an election.” RNC v. DNC, 140 S. Ct. 1205, 1207 (2020) (per curiam) (citing Purcell v. Gonzalez, 549 U.S. 1 (2006) (per curiam)).

9

Though the Board’s proposed schedule envisioned a ruling by February 15 to allow enough time for a May 14 primary, the Board’s affidavit explains that it is possible to schedule a primary as late as the end of July, for which there is still time as of the date of this ruling. See J.A. 826–27; Pl.-Appellants’ Emergency Mot. to Expedite Briefing & Decision at 5. Because the races in Senate Districts 1 and 2 were uncontested and thus no primaries were held, the fact that the March primary date has come and gone does not affect the continued viability of scheduling later primary elections.

86

The district court and the majority rely heavily on Purcell. They believe that because primaries are ongoing in North Carolina, any federal court intervention is bound to “result in voter confusion and [a] consequent incentive to remain away from the polls,” and Purcell instructs courts to avoid such intervention. Purcell v. Gonzalez, 549 U.S. 1, 4–5 (2006) (per curiam). That application of Purcell is improperly narrow.

Some background is helpful here. Purcell itself did not prevent federal court intervention in an election following thorough, judicious proceedings. Instead, it prevented federal court intervention based solely on a barebones, unexplained order from the court of appeals, which had granted an injunction pending appeal after the district court had denied a functionally equivalent preliminary injunction. See Purcell, 549 U.S. at 3; see also League of Women Voters, 769 F.3d at 248 n.6 (explaining that the Purcell Court “seemed troubled by the fact that a two-judge motions panel of the Ninth Circuit entered a factless, groundless ‘bare order’ enjoining a new voter identification provision in an impending election”). The Purcell Court said that the need for a full and thorough evaluation of the district court’s decision outweighed concerns about voter confusion (which are central to the Purcell principle today). See Purcell, 549 U.S. at 5. Accordingly, the Supreme Court’s admonition to the court of appeals to stay its hand is better understood as an instruction not to sacrifice the integrity of its judicial proceedings to the urgency of the moment. It is not a mandate that courts sit on their hands in the weeks before the election, when they still have time to engage in reasoned decision-making, solely because an election is impending.

Initially, lower courts incorporated Purcell’s considerations into their weighing of the equitable preliminary injunction factors. See, e.g., Ne. Ohio Coalition for Homeless &

87

Serv. Emps. Union v. Blackwell, 467 F.3d 999, 1012 (6th Cir. 2006); Ray v. Texas, No. 2:06-cv-385, 2006 WL 8441630, at *5 (E.D. Tex. Oct. 31, 2006); United States v. City of Philadelphia, No. 2:06-cv-4592, 2006 WL 3922115, at *2 (E.D. Pa. Nov. 7, 2006); State ex rel. Applegate v. Franklin Cnty. Bd. of Elections, No. C2-08-092, 2008 WL 341300, at *6 (S.D. Ohio Feb. 6, 2008); Hunter v. Hamilton Cnty. Bd. of Elections, 635 F.3d 219, 244–45 (6th Cir. 2011); Democratic-Republican Org. of N.J. v. Guadagno, 900 F. Supp. 2d 447, 461 n.8 (D.N.J. 2012); see also Hall v. Merrill, 2012 F. Supp. 3d 1148, 1157 (M.D. Ala. 2016) (“[T]he Purcell principle should be considered along with all the other factors that courts use in determining whether to grant a temporary restraining order or a preliminary injunction.”). The Purcell principle in its modern form did not emerge until 2014, eight years after Purcell. See, e.g., N.C. State Conf. of the NAACP v. McCrory, 182 F. Supp. 3d 320, 526 (M.D.N.C. 2016); Feldman v. Arizona, 841 F.3d 791, 795 (Mem) (9th Cir. 2016) (O’Scannlain, J., dissenting from the grant of rehearing en banc); Republican Party of Pa. v. Cortes, 218 F. Supp. 3d 396, 404–05 (E.D. Pa. 2016).

In 2014, the Fifth Circuit gave Purcell a broader scope. Dissenting from the Supreme Court’s one-line, unsigned order denying a motion to vacate the Fifth Circuit’s stay, Justice Ginsburg criticized the Fifth Circuit for misreading Purcell. See Veasey v. Perry, 574 U.S. 951 (2014). Joined by Justices Sotomayor and Kagan, she expressed her view that the Fifth Circuit, which had stayed a district court’s permanent injunction of voter identification procedures, erred when it stayed the injunction based “exclusively on the potential disruption of Texas’ electoral processes” while “[r]efusing to evaluate the defendants’ likelihood of success on the merits.” Id. at 10. As she explained, “Purcell

88

held only that courts must take careful account of considerations specific to election cases, not that election cases are exempt from traditional stay standards.” Id.

Similarly, in Frank v. Walker, the petitioners asked the Supreme Court to vacate a court of appeals’ stay in the lead up to an election, which would have put the challenged election procedure back in place. 574 U.S. 929 (2014). The dissenting justices argued, without referencing Purcell, that the fact that the election was imminent could not by itself justify vacating the stay. Id. (Alito, J., dissenting). Today, by contrast, Purcell is regularly invoked as a standalone principle.

When the Supreme Court next cited Purcell in 2018, the Purcell principle as we know it had mostly taken shape as applied to district court intervention in elections. See Benisek v. Lamone, 138 S. Ct. 1942, 1945 (2018); North Carolina v. Covington, 138 S. Ct. 2548, 2554 (2018). Around the same time, members of the Supreme Court writing separately began to cite Purcell in support of blocking district and appellate court rulings that the authoring justice believed came too close to an election. See Brakebill v. Jaeger, 139 S. Ct. 10 (Mem) (Ginsburg, J., dissenting); RNC v. DNC, 140 S. Ct. 1205, 1207 (2020); id. at 1210–11 (Ginsburg, J., dissenting); Andino v. Middleton, 141 S. Ct. 9 (Mem) (2020) (Kavanaugh, J., concurring); DNC v. Wis. State Legislature, 141 S. Ct. 28 (Mem) (2020) (Kavanaugh J., concurring).

However, no precedential Supreme Court opinion has ever addressed Purcell’s proper scope. Left to decipher conflicting separate writings by individual justices, inconsistent lower court applications of the doctrine come as no surprise. But in the

89

absence of a Supreme Court majority opinion sufficient to clarify Purcell’s proper application, we must make the most of these separate writings.

The most thorough explanation of the modern Purcell principle is Justice Kavanaugh’s concurrence in Merrill v. Milligan in 2022 (16 years after Purcell). 142 S. Ct. 879 (Mem) (2022). There, the Supreme Court stayed the district court’s injunction on Alabama’s use of its electoral maps. See id. Justice Kavanaugh expressly rejected the idea that “the [Purcell] principle is absolute and that a district court may never enjoin a State’s election laws in the period close to an election.” Id. at 881. Instead, he explained, “the Purcell principle is probably best understood as a sensible refinement of ordinary stay principles for the election context—a principle that is not absolute but instead simply heightens the showing necessary for a plaintiff to overcome the State’s extraordinarily strong interest in avoiding late, judicially imposed changes to its election laws and procedures.” Id. Justice Kavanaugh also acknowledged that “the Court has not yet had occasion to fully spell out all of [the Purcell principle’s] contours.” Id.

Justice Kavanaugh’s concurrence thus confirms what Purcell’s jurisprudential history implies. Purcell emerged from equity balancing; it would make little sense for it to be a rigid bar against relief. Courts applying Purcell must balance the equities (the underlying interests of the parties and the public) in each case to determine whether federal court intervention in state election procedures is appropriate. While this balancing requires evaluation of “considerations specific to election cases,” Purcell, 549 U.S. at 5, the analysis does not start and end with the amount of time left until the next election. See Moore v. Harper, 142 S. Ct. 1089, 1091 (Mem) (2022) (Alito, J., dissenting) (arguing that it would

90

have been appropriate for the Court to grant relief, despite the candidate filing deadline being only seven days away, because “promptly granting a stay would have been only minimally disruptive”); DNC v. Wisc. State Legislature, 141 S. Ct. 28, 42 (Mem) (2020) (Kagan, J., dissenting) (“At its core, Purcell tells courts to apply, not depart from, the usual rules of equity.”).

Properly applied, the Purcell principle should be incorporated into a court’s consideration of the equitable preliminary injunction factors. In this case, the parties dispute whether the requested injunctive relief is mandatory or prohibitory. That distinction is not meaningless. Because a prohibitory injunction preserves the status quo that existed at the time litigation was initiated, while a mandatory injunction changes the status quo, the standard to impose a mandatory injunction is much higher. But resolving the parties’ dispute over whether the requested injunction is mandatory or prohibitory is unnecessary because Appellants meet the heightened standard for a mandatory injunction.

A court is justified in issuing a mandatory injunction only if such relief is necessary both (1) “to protect against irreparable harm in deteriorating circumstances created by the defendant,” and (2) “to preserve the court’s ability to enter ultimate relief on the merits of the same kind.” In re Microsoft Corp. Antitrust Litig., 333 F.3d 517, 526 (4th Cir. 2003). Appellants have met their burden on the first prong. Absent relief, thousands of Black voters in North Carolina’s Black Belt counties will be forced to vote in districts that unlawfully dilute their vote. See id. And the urgency in this case is of Legislative Appellees’ making. Though Appellees attempt to blame Appellants for the urgency of this litigation, noting

91

repeatedly that Appellants waited 28 days after the map was enacted to file suit, it was Appellees’ delay in enacting the map that created this condensed litigation schedule.

In April of 2023, the North Carolina Supreme Court struck down the map in place at the time and allowed the General Assembly to enact a new map. Harper v. Hall, 886 S.E.2d 393, 446–48 (N.C. 2023). But the General Assembly did not act immediately. Rather, the General Assembly did not enact a map until six months later. At that point, candidate filing was just six weeks away and the primary ballots were scheduled to be sent out in less than three months. Thus, the need for extraordinary relief, the urgency of this litigation, and the “deteriorating circumstances” that underlie this case have been “created by the defendant[s].” See In re Microsoft Corp., 333 F.3d at 526.

Further, relying on a standalone Purcell principle to leave maps in place under circumstances like this permits legislatures to insulate themselves from judicial review— and subvert federal courts’ role in ensuring that states comply with the Voting Rights Act— by waiting until the last minute to enact new maps. We cannot let one of our country’s most important pieces of civil rights legislation be nullified by clever timing.

Appellants have also satisfied the second prong of the analysis: whether the injunction is necessary “to preserve the court’s ability to enter ultimate relief on the merits of the same kind.” In re Microsoft Corp., 333 F.3d at 526. Because the election is imminent, the court will soon lose “its ability to enter ultimate relief on the merits” if an injunction is not entered now. See id. “[O]nce the election occurs, there can be no do-over and no redress.” League of Women Voters, 769 F.3d at 247.

92

That brings us to the central reason why Purcell does not bar relief here: there is still time. Unlike Purcell this is not a case of rushed, unexplained, unreasoned decision- making. The district court conducted an evidentiary hearing and issued a detailed 69-page opinion. This Court received extensive briefing and heard oral argument. And these opinions are far more than the “four-sentence order” at issue in Purcell. See Purcell, 549 U.S. at 3. In fact, to ensure full consideration and appropriate development of the record, we did not take up Appellants’ first appeal. That appeal would have required this Court to examine the issues without the benefit of fact finding by the district court. Rather than engage in the same kind of rushed decision-making for which the Supreme Court faulted the court of appeals in Purcell, we sent the case back to the district court to allow for an evidentiary hearing and subsequent written opinion.

Appellees make much of the fact that Appellants waited too long after the challenged map was enacted to file suit. Indeed, one of the motivating principles behind Purcell is that it “discourages last-minute litigation and instead encourages litigants to bring any substantial challenges to election rules ahead of time, in the ordinary litigation process.” DNC, 141 S. Ct. at 31. But if there is a Purcell problem here, Appellants could not have brought the litigation early enough to avoid it.

Here, the district court’s opinion came on January 26, one week after absentee voting began. That’s two months later than the relevant date in Merrill v. Milligan. See Milligan, 142 S. Ct. at 879 (Kavanaugh, J., concurring) (invoking Purcell to stay a district court’s final injunction because absentee voting in the primaries would begin seven weeks from the date of the Supreme Court’s opinion). There is no reason to think that the district

93

court’s ruling here—let alone a ruling from this Court—could have come more than seven weeks before the start of absentee voting if Appellants had filed suit 28 days earlier.

In sum, even if a 28-day delay in filing suit did amount to dilatory conduct by Appellants, that delay is not the cause of the urgency of this litigation or the fact that any federal court ruling will necessarily come on the eve of the election. The cause of the timing of this litigation is instead the General Assembly’s failure to take any action on redistricting until six months after the North Carolina Supreme Court struck down the 2022 maps in Harper v. Hall. Appellees should not be allowed to benefit from their delay. And North Carolina’s voters should not be forced to vote under illegal maps for an entire election cycle because the General Assembly delayed the enactment of the map. At the very least, the record demonstrates that the district court abused its discretion by completely ignoring how the legislature’s late enactment of the map affects the equitable considerations in this case.

Appellees contend that they have only two remedial options. They say that because neither can be imposed here, the current map cannot be enjoined. But courts have a range of options when selecting an appropriate remedy. Here, those options are not limited to imposing Demonstration District A (incompatible with Purcell) or Demonstration Districts B1 and B2 (incompatible with Strickland) by judicial fiat. There is at least one more option: this Court can order the General Assembly to adopt a new map that complies with Section 2. The legislature can then choose whether to reconfigure the entire state map, adopt crossover districts, or solve the problem some other way entirely.

94

In deciding between these options, the legislature does not face the same constraints as the federal judiciary. Because Purcell only prohibits federal courts from changing election procedures too close to an election, the legislature is empowered to redraw the entire state map if it so chooses. See Wisc. State Legislature, 141 S. Ct. at 31 (Kavanaugh, J., concurring) (“It is one thing for state legislatures to alter their own election rules in the late innings and bear the responsibility for any unintended consequences. It is quite another thing for a federal district court to swoop in and alter carefully considered and democratically enacted state election rules when an election is imminent.”). And while Appellees have repeatedly argued that the General Assembly cannot possibly be required to comply with Section 2 because it could only do so by violating the state’s Whole County Provision, compliance with federal law cannot be subordinated to state law requirements. See Stephenson, 562 S.E.2d at 389 (“[T]he State retains significant discretion when formulating legislative districts, so long as the ‘effect’ of districts created pursuant to a ‘whole-county’ criterion or other constitutional requirement does not dilute minority voting strength in violation of federal law.”). The legislature still has time to remedy the illegality of the existing map.

V.

The district court erroneously construed the standard Appellants must meet under Gingles. And it failed to consider in its Purcell analysis the deleterious impact of the timeline on which Legislative Appellees chose to enact the new map. The VRA is “the Nation’s signal piece of civil-rights legislation.” Shelby County v. Holder, 570 U.S. 529,

95

580 (2013) (Ginsburg, J., dissenting). Section 2 is essential to ensuring the continuation of the VRA’s noble purpose; its enforcement cannot be left solely to the will or whim of state legislatures. I would reverse and therefore dissent.

96

Appendix A

Table A1: Performance Analysis (Reconstituted Election Analysis) of 2022 Elections (J.A. 291)

97

Table A2: Performance Analysis (Reconstituted Election Analysis) of 2020 Elections (J.A. 292–93)